OPPORTUNISTIC POLITICAL CYCLES TEST IN AYOUNG DEMOCRACY SETTING

AKHMED AKHMEDOV AND EKATERINA ZHURAVSKAYA

This paper tests the theory of opportunistic cycles in a decade-old democracymdashRussiamdashfinds strong evidence of cycles and provides an explanation for whyprevious literature often found weaker evidence Using regional monthly paneldata we find that (1) the budget cycle is sizable and short-lived public spendingshifts toward direct monetary transfers to voters (2) the magnitude of the cycledecreases with democracy government transparency media freedom voterawareness and over time and (3) preelectoral manipulation increases incum-bentsrsquo chances for reelection The short length of the cycle explains underestima-tion of its size by previous literature because of low frequency data used inprevious studies

You can fool some of the people all of the timeand all of the people some of the time but you

cannot fool all of the people all of the time[Attributed to Abraham Lincoln 1809ndash1865]

I INTRODUCTION

Despite the fact that theoretical research on opportunisticpolitical cycles is very intuitive and well developed empiricalliterature produced mixed results in attempts to find convincingevidence of opportunistic cycles for almost a quarter of a centuryafter the pioneering work of Nordhaus [1975] The evidence fromdeveloped countries is particularly weak1 The apparent contra-diction between the theory and evidence created an intellectualpuzzle Why did many tests fail Should the theory or the empir-ics be held responsible Motivated by this gap several recenttheoretical works argued that opportunistic cycles should be most

We thank Alberto Alesina Alessandra Casella Sergei Guriev Eric MaskinViktor Polterovich Gerard Roland Thomas Romer Paul Seabright ClaudiaSenik Andrei Shleifer two anonymous referees seminar participants at theInstitute for Advanced Study in Princeton New Jersey New Economic SchoolCEFIR Ohio State University and the participants of the CEPR-WDI 2002Transition Conference and the EEA 2002 Meetings for useful comments Wethank Alexander Andriakov Yury Andrienko and Konstantin Kozlov for helpwith data collection We are grateful to the New Economic School for financialsupport and Alexei Ravitchev for excellent research assistance in the early stagesof this project A part of the work on this paper took place when EkaterinaZhuravskaya was on leave at the Institute for Advanced Study The hospitalityand congenial environment of the Institute are gratefully acknowledged

1 Little empirical support for opportunistic cycles in developed countries wasfound in Klein [1996] Berger and Woitek [1997] Reid [1998] and Alesina andRoubini [1992] Alesina Roubini and Cohen [1997] Drazen [2000] and Franzese[2002] provide detailed surveys of theoretical and empirical literatures

copy 2004 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics November 2004

1301

sizable in countries with immature democratic regimes [Gonzalez2000 Shi and Svensson 2002a] The evidence has been stronglysupportive of this view studies of country-level panels with alarge presence of young democracies and within-country studiesof imperfect democracies have shown robust significant fiscalcycles almost exclusively2 Many of these tests however sufferfrom severe data limitations in particular insufficient frequencyand often too high a level of aggregation of fiscal spending Thispaper sheds further light on the puzzle we provide strong evi-dence of opportunistic cycles using a regional monthly panel froma decade-old democracymdashRussiamdashdemonstrate that use of lowerfrequency data leads to underestimation of cycles explaining theweakness of results in previous studies and document the linkbetween cycles on the one hand and democracy and transpar-ency on the other

Russiarsquos regional elections of executives provide an ideal casefor an empirical test of opportunistic political cycle theory FirstRussia is a typical immature democracy it is characterized bydependent media nontransparent government and a large frac-tion of an uninformed and myopic electorate Since the theorypredicts sizable cycles in immature democracies one should ex-pect to find evidence of cycles on Russian data Simple plots ofactual series of regional spending in many cases suggest vividincreases in spending prior to elections (examples are given inFigure I) Second high uniformity in electoral rules and scope forpolicy-making among regions combined with high variation in thelevel of democracy and government transparency help to identifythe effect of the latter on the magnitude of political cycles Thirddetailed monthly regional panel data allow more powerful tests ofthe theory than the ones done so far

We find sizable and short-lived cycles in public expenditureand its composition and no cycles in growth or inflation control-ling for region-specific characteristics federal trend seasonalityand ideology Monthly panel data allow us careful measurementof even very short cycles This turned out to be important becausemost sizable shifts in spending happen within a month or two ofelections We show that use of quarterly data leads to severeunderestimation of cycles because the opposite-sign shifts in pub-

2 See evidence and discussion in Brender and Drazen [2003] as well asevidence produced by Ames [1987] Gonzalez [2002] Krueger and Turan [1993]Drazen and Eslava [2003] Block [2001 2002] Block Ferree and Singh [2003]Shi and Svensson [2002a 2002b] and Schuknecht [2000]

1302 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EI

Ele

ctio

ns

and

Act

ual

Ser

ies

ofR

egio

nal

per

Cap

ita

Soc

ial

Exp

endi

ture

s(S

moo

thed

bya

Mov

ing

Ave

rage

)

1303OPPORTUNISTIC POLITICAL CYCLES

lic expenditure around elections cancel out in low frequency dataPrevious empirical studies of cycles in developing countries usedquarterly or lower frequency data3 Short length of the cyclehowever does not undermine its economic significance First thecycle is primarily targeted at poor voters who have the worstpossibilities for consumption smoothing the largest fluctuationsare observed in repayment of wage arrears to public workers andspending on welfare and other public assistance programs Sec-ond irrespective of the driving force of the cyclemdashasymmetricinformation or voter myopiamdashit indicates the absence of checksand balances on politicians that make politicians accountable inmature democracies (ie separation of powers free press activeNGOs etc) These institutions should provide voters with fullerinformation and longer-term control over politicians and there-fore limit possibilities for misuse of public office for private gainLarge and short-lived cycles therefore provide evidence of poorlong-term accountability of politicians

In line with recent theoretical results we find that proxies forinformational symmetry voter awareness and the level of regionaldemocracy significantly reduce cycles In addition cycles get smallerover time The latter could be an indication of a general phenome-non as democracy matures voters learn and independent mediaand civil society develop or of a Russia-specific effect of an informalchange in federal control over regional budgets from one Russianpresident to the other (data are insufficient to distinguish betweenthese two explanations of the time effect) Finally we find thatcycles in fiscal policy instruments significantly increase the popu-larity of incumbents and help them win

The remainder of the paper is organized as follows In thenext section we formulate testable hypotheses motivated bytheoretical literature Section III describes the data Section IVpresents results Conclusions follow in Section V

II HYPOTHESES

First we test predictions of the classic opportunistic politicalcycle models Nordhaus [1975] built the first formal model ofopportunistic political cycles based on the Phillips curve withadaptive expectations of voters In the model naıve voters get

3 To the best of our knowledge the only paper that uses monthly data isBerger and Woitek [1997] it rejects hypothesis of opportunistic cycles for theGerman developed democracy

1304 QUARTERLY JOURNAL OF ECONOMICS

consistently fooled by a preelectoral inflationary boom despite therecession following election The model predicts political cycles ingrowth unemployment and inflation Rogoff and Sibert [1988]Rogoff [1990] and Persson and Tabellini [1990] reconciled ratio-nal expectations with opportunistic political cycles in dynamicsignaling models with asymmetric information about politiciansrsquocompetence In particular Rogoff and Sibert [1988] and Rogoff[1990] focus on a political budget cycle in which voters rewardpreelectoral distortion in public spending because it signals aboutan incumbentrsquos ability to provide public goods Rogoff and Sibertrsquosmodel predicts a preelectoral fall in tax collection along withincreases in deficit and inflation Rogoffrsquos model focuses on thepreelectoral shift in composition of government spending awayfrom investment observed with a lag toward more visible publicconsumption We look at the dynamics of a wide range of publicexpenditure items composition of the budget growth and infla-tion to test predictions of these theories

Second we study determinants of the magnitude of the bud-get cycle Asymmetry of information about politiciansrsquo compe-tence plays a central role in rational opportunistic cycles Recentliterature extends the basic setup of Rogoffrsquos model to illustratethat transparency modeled as the probability that voters get acorrect exogenous signal about an incumbentrsquos competence[Gonzalez 2000] and awareness of voters defined as a share ofperfectly informed voters [Shi and Svensson 2002a] reduce themagnitude of the cycle Furthermore Gonzalez [2000] introducesthe level of democracy directly into the model Democracy has anonmonotone effect on the cycle first if the cost of enforcingpolitical turnover is prohibitively high for voters (as is the case indictatorships) then politicians do not have an incentive to engagein costly signaling and therefore there are no cycles Secondwhen change in office can be enforced the cycle decreases withdemocracy because it ensures institutional checks and balanceson government that allow voters to observe politiciansrsquo compe-tence directly with higher probability (ie free media) We testwhether these theoretical predictions are consistent with the databy looking at how cross-sectional variation in voter awareness(measured by education and urbanization) transparency (mea-sured by indices of government transparency and media free-dom) and democracy influence the cycles

In young democracies like Russia emerging NGOs and inde-pendent media accumulate experience in collecting packaging and

1305OPPORTUNISTIC POLITICAL CYCLES

disseminating information about politicians and their policies whilevoters learn how to process this information Thus cycles should getsmaller over time as learning proceeds and civil society develops[Brender and Drazen 2003] We test for the effect of time on thecycle Apart from learning however the time effect may also beattributed to the difference in scope for fiscal manipulation by re-gional governments under Yeltsin and Putinrsquos administrations (Ro-goff [1990] discusses the effects of restraining fiscal policies on po-litical cycles) We attempt to separate these hypotheses

All opportunistic cycle theories predict that preelectoral ma-nipulations are rewarded by voters In rational signaling modelsa cycle has costs (policy distortions) and benefits (transmittinginformation about politiciansrsquo competence) If voters learn politi-ciansrsquo type directly for instance through well-functioning inde-pendent media signaling has no value to voters In fact in de-veloped democracies voters reward politicians for restrained fis-cal policies [Alesina Perotti and Tavares 1998 Brender 2003]As a last step of our analysis we investigate whether it pays topursue preelectoral fiscal expansion in a maturing democracy

III THE DATA

The comprehensive list of regional governor elections that tookplace in Russia between August 1995 and December 2003 consists of194 electoral events4 Data on most policy instruments and out-comes are available for 159 elections between September 1996 andJuly 2003 Four regions had three rounds of elections 65 regionshad two rounds of elections and 17 regions had just one round ofelections during this period The source of the data on elections isTsentrizbirkom the Central Elections Committee of the RussianFederation

Regional monthly series of fiscal instruments and outcomescome from two sources Goskomstat the State Committee of Statis-tics provided data on wages and income wage arrears from theregional budgets price level and industrial output between 1995and 2003 the Ministry of Finance of the Russian Federation pro-vided detailed data on the execution of regional budgets for theperiod between 1996 and 2003

4 This list covers all the regions but Dagestan the only region where therehave not been any governor elections We excluded Chechnya and Ingushetia fromthe sample because fluctuations in fiscal policies of these regions have been drivenby war rather than elections

1306 QUARTERLY JOURNAL OF ECONOMICS

To test for the determinants of cycle magnitude we use cross-section data Data on urbanization and education come from Gos-komstat The data on freedom of media in the regions were providedby the Institute of Free Media (wwwfreepressru) Data on trans-parency of the regional government come from ldquoMedia-Soyuzrdquo anindependent professional association of Russian journalists Data onthe scope of regional democracy are from the Carnegie MoscowCenter Carnegie Endowment for International Peace [Petrov 2001]The Appendix presents descriptive statistics of the data

IV RESULTS

IVA Test for Opportunistic Cycles

Three groups of policy instruments and outcomes are consid-ered budgetary expenditures (total spending as well as levels andshares of expenditures on social programs education culturehealth care mass media and industrial subsidies) budget reve-nues and deficit (total revenues tax revenues deficit and federaltransfers) and such outcomes as growth inflation regional bud-getary wage arrears wage level and income We test for politicalcycles in these variables treating election time as exogenous5 Thefollowing equation is estimated on regional monthly panel data

(1) yit j1212

jmjit Lyit1

1Termit 2Leftit t fis εit

where i identifies regions t is real time in months and y stands fora logarithm of instrument or outcome of regional policy (all mone-tary variables are expressed in real terms per capita) To control forthe federal trend and macroeconomic shocks we include the com-plete set of time fixed effects t one for each month t To control forregion-specific fixed effects and region-specific seasonality we in-clude fixed effects fis for each of the twelve calendar months s(t) in

5 Ito [1990] Reid [1998] and Heckelman and Berument [1998] amongothers pointed out that opportunistic cycles can occur as a result of setting anelection date at the time of a boom Although almost 19 percent of Russiarsquosregional elections happened a month or more from their expected date in the vastmajority of these cases the time was shifted for exogenous reasons There were afew cases however when there was no exogenous reason for the shift of electiontime even though it is illegal to shift the date of regional elections To make surethat our results are not driven by the presence of endogenous elections werepeated all tests on the subsample of elections that had exogenously predeter-mined timing and got virtually identical results

1307OPPORTUNISTIC POLITICAL CYCLES

sizable in countries with immature democratic regimes [Gonzalez2000 Shi and Svensson 2002a] The evidence has been stronglysupportive of this view studies of country-level panels with alarge presence of young democracies and within-country studiesof imperfect democracies have shown robust significant fiscalcycles almost exclusively2 Many of these tests however sufferfrom severe data limitations in particular insufficient frequencyand often too high a level of aggregation of fiscal spending Thispaper sheds further light on the puzzle we provide strong evi-dence of opportunistic cycles using a regional monthly panel froma decade-old democracymdashRussiamdashdemonstrate that use of lowerfrequency data leads to underestimation of cycles explaining theweakness of results in previous studies and document the linkbetween cycles on the one hand and democracy and transpar-ency on the other

Russiarsquos regional elections of executives provide an ideal casefor an empirical test of opportunistic political cycle theory FirstRussia is a typical immature democracy it is characterized bydependent media nontransparent government and a large frac-tion of an uninformed and myopic electorate Since the theorypredicts sizable cycles in immature democracies one should ex-pect to find evidence of cycles on Russian data Simple plots ofactual series of regional spending in many cases suggest vividincreases in spending prior to elections (examples are given inFigure I) Second high uniformity in electoral rules and scope forpolicy-making among regions combined with high variation in thelevel of democracy and government transparency help to identifythe effect of the latter on the magnitude of political cycles Thirddetailed monthly regional panel data allow more powerful tests ofthe theory than the ones done so far

We find sizable and short-lived cycles in public expenditureand its composition and no cycles in growth or inflation control-ling for region-specific characteristics federal trend seasonalityand ideology Monthly panel data allow us careful measurementof even very short cycles This turned out to be important becausemost sizable shifts in spending happen within a month or two ofelections We show that use of quarterly data leads to severeunderestimation of cycles because the opposite-sign shifts in pub-

2 See evidence and discussion in Brender and Drazen [2003] as well asevidence produced by Ames [1987] Gonzalez [2002] Krueger and Turan [1993]Drazen and Eslava [2003] Block [2001 2002] Block Ferree and Singh [2003]Shi and Svensson [2002a 2002b] and Schuknecht [2000]

1302 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EI

Ele

ctio

ns

and

Act

ual

Ser

ies

ofR

egio

nal

per

Cap

ita

Soc

ial

Exp

endi

ture

s(S

moo

thed

bya

Mov

ing

Ave

rage

)

1303OPPORTUNISTIC POLITICAL CYCLES

lic expenditure around elections cancel out in low frequency dataPrevious empirical studies of cycles in developing countries usedquarterly or lower frequency data3 Short length of the cyclehowever does not undermine its economic significance First thecycle is primarily targeted at poor voters who have the worstpossibilities for consumption smoothing the largest fluctuationsare observed in repayment of wage arrears to public workers andspending on welfare and other public assistance programs Sec-ond irrespective of the driving force of the cyclemdashasymmetricinformation or voter myopiamdashit indicates the absence of checksand balances on politicians that make politicians accountable inmature democracies (ie separation of powers free press activeNGOs etc) These institutions should provide voters with fullerinformation and longer-term control over politicians and there-fore limit possibilities for misuse of public office for private gainLarge and short-lived cycles therefore provide evidence of poorlong-term accountability of politicians

In line with recent theoretical results we find that proxies forinformational symmetry voter awareness and the level of regionaldemocracy significantly reduce cycles In addition cycles get smallerover time The latter could be an indication of a general phenome-non as democracy matures voters learn and independent mediaand civil society develop or of a Russia-specific effect of an informalchange in federal control over regional budgets from one Russianpresident to the other (data are insufficient to distinguish betweenthese two explanations of the time effect) Finally we find thatcycles in fiscal policy instruments significantly increase the popu-larity of incumbents and help them win

The remainder of the paper is organized as follows In thenext section we formulate testable hypotheses motivated bytheoretical literature Section III describes the data Section IVpresents results Conclusions follow in Section V

II HYPOTHESES

First we test predictions of the classic opportunistic politicalcycle models Nordhaus [1975] built the first formal model ofopportunistic political cycles based on the Phillips curve withadaptive expectations of voters In the model naıve voters get

3 To the best of our knowledge the only paper that uses monthly data isBerger and Woitek [1997] it rejects hypothesis of opportunistic cycles for theGerman developed democracy

1304 QUARTERLY JOURNAL OF ECONOMICS

consistently fooled by a preelectoral inflationary boom despite therecession following election The model predicts political cycles ingrowth unemployment and inflation Rogoff and Sibert [1988]Rogoff [1990] and Persson and Tabellini [1990] reconciled ratio-nal expectations with opportunistic political cycles in dynamicsignaling models with asymmetric information about politiciansrsquocompetence In particular Rogoff and Sibert [1988] and Rogoff[1990] focus on a political budget cycle in which voters rewardpreelectoral distortion in public spending because it signals aboutan incumbentrsquos ability to provide public goods Rogoff and Sibertrsquosmodel predicts a preelectoral fall in tax collection along withincreases in deficit and inflation Rogoffrsquos model focuses on thepreelectoral shift in composition of government spending awayfrom investment observed with a lag toward more visible publicconsumption We look at the dynamics of a wide range of publicexpenditure items composition of the budget growth and infla-tion to test predictions of these theories

Second we study determinants of the magnitude of the bud-get cycle Asymmetry of information about politiciansrsquo compe-tence plays a central role in rational opportunistic cycles Recentliterature extends the basic setup of Rogoffrsquos model to illustratethat transparency modeled as the probability that voters get acorrect exogenous signal about an incumbentrsquos competence[Gonzalez 2000] and awareness of voters defined as a share ofperfectly informed voters [Shi and Svensson 2002a] reduce themagnitude of the cycle Furthermore Gonzalez [2000] introducesthe level of democracy directly into the model Democracy has anonmonotone effect on the cycle first if the cost of enforcingpolitical turnover is prohibitively high for voters (as is the case indictatorships) then politicians do not have an incentive to engagein costly signaling and therefore there are no cycles Secondwhen change in office can be enforced the cycle decreases withdemocracy because it ensures institutional checks and balanceson government that allow voters to observe politiciansrsquo compe-tence directly with higher probability (ie free media) We testwhether these theoretical predictions are consistent with the databy looking at how cross-sectional variation in voter awareness(measured by education and urbanization) transparency (mea-sured by indices of government transparency and media free-dom) and democracy influence the cycles

In young democracies like Russia emerging NGOs and inde-pendent media accumulate experience in collecting packaging and

1305OPPORTUNISTIC POLITICAL CYCLES

disseminating information about politicians and their policies whilevoters learn how to process this information Thus cycles should getsmaller over time as learning proceeds and civil society develops[Brender and Drazen 2003] We test for the effect of time on thecycle Apart from learning however the time effect may also beattributed to the difference in scope for fiscal manipulation by re-gional governments under Yeltsin and Putinrsquos administrations (Ro-goff [1990] discusses the effects of restraining fiscal policies on po-litical cycles) We attempt to separate these hypotheses

All opportunistic cycle theories predict that preelectoral ma-nipulations are rewarded by voters In rational signaling modelsa cycle has costs (policy distortions) and benefits (transmittinginformation about politiciansrsquo competence) If voters learn politi-ciansrsquo type directly for instance through well-functioning inde-pendent media signaling has no value to voters In fact in de-veloped democracies voters reward politicians for restrained fis-cal policies [Alesina Perotti and Tavares 1998 Brender 2003]As a last step of our analysis we investigate whether it pays topursue preelectoral fiscal expansion in a maturing democracy

III THE DATA

The comprehensive list of regional governor elections that tookplace in Russia between August 1995 and December 2003 consists of194 electoral events4 Data on most policy instruments and out-comes are available for 159 elections between September 1996 andJuly 2003 Four regions had three rounds of elections 65 regionshad two rounds of elections and 17 regions had just one round ofelections during this period The source of the data on elections isTsentrizbirkom the Central Elections Committee of the RussianFederation

Regional monthly series of fiscal instruments and outcomescome from two sources Goskomstat the State Committee of Statis-tics provided data on wages and income wage arrears from theregional budgets price level and industrial output between 1995and 2003 the Ministry of Finance of the Russian Federation pro-vided detailed data on the execution of regional budgets for theperiod between 1996 and 2003

4 This list covers all the regions but Dagestan the only region where therehave not been any governor elections We excluded Chechnya and Ingushetia fromthe sample because fluctuations in fiscal policies of these regions have been drivenby war rather than elections

1306 QUARTERLY JOURNAL OF ECONOMICS

To test for the determinants of cycle magnitude we use cross-section data Data on urbanization and education come from Gos-komstat The data on freedom of media in the regions were providedby the Institute of Free Media (wwwfreepressru) Data on trans-parency of the regional government come from ldquoMedia-Soyuzrdquo anindependent professional association of Russian journalists Data onthe scope of regional democracy are from the Carnegie MoscowCenter Carnegie Endowment for International Peace [Petrov 2001]The Appendix presents descriptive statistics of the data

IV RESULTS

IVA Test for Opportunistic Cycles

Three groups of policy instruments and outcomes are consid-ered budgetary expenditures (total spending as well as levels andshares of expenditures on social programs education culturehealth care mass media and industrial subsidies) budget reve-nues and deficit (total revenues tax revenues deficit and federaltransfers) and such outcomes as growth inflation regional bud-getary wage arrears wage level and income We test for politicalcycles in these variables treating election time as exogenous5 Thefollowing equation is estimated on regional monthly panel data

(1) yit j1212

jmjit Lyit1

1Termit 2Leftit t fis εit

where i identifies regions t is real time in months and y stands fora logarithm of instrument or outcome of regional policy (all mone-tary variables are expressed in real terms per capita) To control forthe federal trend and macroeconomic shocks we include the com-plete set of time fixed effects t one for each month t To control forregion-specific fixed effects and region-specific seasonality we in-clude fixed effects fis for each of the twelve calendar months s(t) in

5 Ito [1990] Reid [1998] and Heckelman and Berument [1998] amongothers pointed out that opportunistic cycles can occur as a result of setting anelection date at the time of a boom Although almost 19 percent of Russiarsquosregional elections happened a month or more from their expected date in the vastmajority of these cases the time was shifted for exogenous reasons There were afew cases however when there was no exogenous reason for the shift of electiontime even though it is illegal to shift the date of regional elections To make surethat our results are not driven by the presence of endogenous elections werepeated all tests on the subsample of elections that had exogenously predeter-mined timing and got virtually identical results

1307OPPORTUNISTIC POLITICAL CYCLES

FIG

UR

EI

Ele

ctio

ns

and

Act

ual

Ser

ies

ofR

egio

nal

per

Cap

ita

Soc

ial

Exp

endi

ture

s(S

moo

thed

bya

Mov

ing

Ave

rage

)

1303OPPORTUNISTIC POLITICAL CYCLES

lic expenditure around elections cancel out in low frequency dataPrevious empirical studies of cycles in developing countries usedquarterly or lower frequency data3 Short length of the cyclehowever does not undermine its economic significance First thecycle is primarily targeted at poor voters who have the worstpossibilities for consumption smoothing the largest fluctuationsare observed in repayment of wage arrears to public workers andspending on welfare and other public assistance programs Sec-ond irrespective of the driving force of the cyclemdashasymmetricinformation or voter myopiamdashit indicates the absence of checksand balances on politicians that make politicians accountable inmature democracies (ie separation of powers free press activeNGOs etc) These institutions should provide voters with fullerinformation and longer-term control over politicians and there-fore limit possibilities for misuse of public office for private gainLarge and short-lived cycles therefore provide evidence of poorlong-term accountability of politicians

In line with recent theoretical results we find that proxies forinformational symmetry voter awareness and the level of regionaldemocracy significantly reduce cycles In addition cycles get smallerover time The latter could be an indication of a general phenome-non as democracy matures voters learn and independent mediaand civil society develop or of a Russia-specific effect of an informalchange in federal control over regional budgets from one Russianpresident to the other (data are insufficient to distinguish betweenthese two explanations of the time effect) Finally we find thatcycles in fiscal policy instruments significantly increase the popu-larity of incumbents and help them win

The remainder of the paper is organized as follows In thenext section we formulate testable hypotheses motivated bytheoretical literature Section III describes the data Section IVpresents results Conclusions follow in Section V

II HYPOTHESES

First we test predictions of the classic opportunistic politicalcycle models Nordhaus [1975] built the first formal model ofopportunistic political cycles based on the Phillips curve withadaptive expectations of voters In the model naıve voters get

3 To the best of our knowledge the only paper that uses monthly data isBerger and Woitek [1997] it rejects hypothesis of opportunistic cycles for theGerman developed democracy

1304 QUARTERLY JOURNAL OF ECONOMICS

consistently fooled by a preelectoral inflationary boom despite therecession following election The model predicts political cycles ingrowth unemployment and inflation Rogoff and Sibert [1988]Rogoff [1990] and Persson and Tabellini [1990] reconciled ratio-nal expectations with opportunistic political cycles in dynamicsignaling models with asymmetric information about politiciansrsquocompetence In particular Rogoff and Sibert [1988] and Rogoff[1990] focus on a political budget cycle in which voters rewardpreelectoral distortion in public spending because it signals aboutan incumbentrsquos ability to provide public goods Rogoff and Sibertrsquosmodel predicts a preelectoral fall in tax collection along withincreases in deficit and inflation Rogoffrsquos model focuses on thepreelectoral shift in composition of government spending awayfrom investment observed with a lag toward more visible publicconsumption We look at the dynamics of a wide range of publicexpenditure items composition of the budget growth and infla-tion to test predictions of these theories

Second we study determinants of the magnitude of the bud-get cycle Asymmetry of information about politiciansrsquo compe-tence plays a central role in rational opportunistic cycles Recentliterature extends the basic setup of Rogoffrsquos model to illustratethat transparency modeled as the probability that voters get acorrect exogenous signal about an incumbentrsquos competence[Gonzalez 2000] and awareness of voters defined as a share ofperfectly informed voters [Shi and Svensson 2002a] reduce themagnitude of the cycle Furthermore Gonzalez [2000] introducesthe level of democracy directly into the model Democracy has anonmonotone effect on the cycle first if the cost of enforcingpolitical turnover is prohibitively high for voters (as is the case indictatorships) then politicians do not have an incentive to engagein costly signaling and therefore there are no cycles Secondwhen change in office can be enforced the cycle decreases withdemocracy because it ensures institutional checks and balanceson government that allow voters to observe politiciansrsquo compe-tence directly with higher probability (ie free media) We testwhether these theoretical predictions are consistent with the databy looking at how cross-sectional variation in voter awareness(measured by education and urbanization) transparency (mea-sured by indices of government transparency and media free-dom) and democracy influence the cycles

In young democracies like Russia emerging NGOs and inde-pendent media accumulate experience in collecting packaging and

1305OPPORTUNISTIC POLITICAL CYCLES

disseminating information about politicians and their policies whilevoters learn how to process this information Thus cycles should getsmaller over time as learning proceeds and civil society develops[Brender and Drazen 2003] We test for the effect of time on thecycle Apart from learning however the time effect may also beattributed to the difference in scope for fiscal manipulation by re-gional governments under Yeltsin and Putinrsquos administrations (Ro-goff [1990] discusses the effects of restraining fiscal policies on po-litical cycles) We attempt to separate these hypotheses

All opportunistic cycle theories predict that preelectoral ma-nipulations are rewarded by voters In rational signaling modelsa cycle has costs (policy distortions) and benefits (transmittinginformation about politiciansrsquo competence) If voters learn politi-ciansrsquo type directly for instance through well-functioning inde-pendent media signaling has no value to voters In fact in de-veloped democracies voters reward politicians for restrained fis-cal policies [Alesina Perotti and Tavares 1998 Brender 2003]As a last step of our analysis we investigate whether it pays topursue preelectoral fiscal expansion in a maturing democracy

III THE DATA

The comprehensive list of regional governor elections that tookplace in Russia between August 1995 and December 2003 consists of194 electoral events4 Data on most policy instruments and out-comes are available for 159 elections between September 1996 andJuly 2003 Four regions had three rounds of elections 65 regionshad two rounds of elections and 17 regions had just one round ofelections during this period The source of the data on elections isTsentrizbirkom the Central Elections Committee of the RussianFederation

Regional monthly series of fiscal instruments and outcomescome from two sources Goskomstat the State Committee of Statis-tics provided data on wages and income wage arrears from theregional budgets price level and industrial output between 1995and 2003 the Ministry of Finance of the Russian Federation pro-vided detailed data on the execution of regional budgets for theperiod between 1996 and 2003

4 This list covers all the regions but Dagestan the only region where therehave not been any governor elections We excluded Chechnya and Ingushetia fromthe sample because fluctuations in fiscal policies of these regions have been drivenby war rather than elections

1306 QUARTERLY JOURNAL OF ECONOMICS

To test for the determinants of cycle magnitude we use cross-section data Data on urbanization and education come from Gos-komstat The data on freedom of media in the regions were providedby the Institute of Free Media (wwwfreepressru) Data on trans-parency of the regional government come from ldquoMedia-Soyuzrdquo anindependent professional association of Russian journalists Data onthe scope of regional democracy are from the Carnegie MoscowCenter Carnegie Endowment for International Peace [Petrov 2001]The Appendix presents descriptive statistics of the data

IV RESULTS

IVA Test for Opportunistic Cycles

Three groups of policy instruments and outcomes are consid-ered budgetary expenditures (total spending as well as levels andshares of expenditures on social programs education culturehealth care mass media and industrial subsidies) budget reve-nues and deficit (total revenues tax revenues deficit and federaltransfers) and such outcomes as growth inflation regional bud-getary wage arrears wage level and income We test for politicalcycles in these variables treating election time as exogenous5 Thefollowing equation is estimated on regional monthly panel data

(1) yit j1212

jmjit Lyit1

1Termit 2Leftit t fis εit

where i identifies regions t is real time in months and y stands fora logarithm of instrument or outcome of regional policy (all mone-tary variables are expressed in real terms per capita) To control forthe federal trend and macroeconomic shocks we include the com-plete set of time fixed effects t one for each month t To control forregion-specific fixed effects and region-specific seasonality we in-clude fixed effects fis for each of the twelve calendar months s(t) in

5 Ito [1990] Reid [1998] and Heckelman and Berument [1998] amongothers pointed out that opportunistic cycles can occur as a result of setting anelection date at the time of a boom Although almost 19 percent of Russiarsquosregional elections happened a month or more from their expected date in the vastmajority of these cases the time was shifted for exogenous reasons There were afew cases however when there was no exogenous reason for the shift of electiontime even though it is illegal to shift the date of regional elections To make surethat our results are not driven by the presence of endogenous elections werepeated all tests on the subsample of elections that had exogenously predeter-mined timing and got virtually identical results

1307OPPORTUNISTIC POLITICAL CYCLES

lic expenditure around elections cancel out in low frequency dataPrevious empirical studies of cycles in developing countries usedquarterly or lower frequency data3 Short length of the cyclehowever does not undermine its economic significance First thecycle is primarily targeted at poor voters who have the worstpossibilities for consumption smoothing the largest fluctuationsare observed in repayment of wage arrears to public workers andspending on welfare and other public assistance programs Sec-ond irrespective of the driving force of the cyclemdashasymmetricinformation or voter myopiamdashit indicates the absence of checksand balances on politicians that make politicians accountable inmature democracies (ie separation of powers free press activeNGOs etc) These institutions should provide voters with fullerinformation and longer-term control over politicians and there-fore limit possibilities for misuse of public office for private gainLarge and short-lived cycles therefore provide evidence of poorlong-term accountability of politicians

In line with recent theoretical results we find that proxies forinformational symmetry voter awareness and the level of regionaldemocracy significantly reduce cycles In addition cycles get smallerover time The latter could be an indication of a general phenome-non as democracy matures voters learn and independent mediaand civil society develop or of a Russia-specific effect of an informalchange in federal control over regional budgets from one Russianpresident to the other (data are insufficient to distinguish betweenthese two explanations of the time effect) Finally we find thatcycles in fiscal policy instruments significantly increase the popu-larity of incumbents and help them win

The remainder of the paper is organized as follows In thenext section we formulate testable hypotheses motivated bytheoretical literature Section III describes the data Section IVpresents results Conclusions follow in Section V

II HYPOTHESES

First we test predictions of the classic opportunistic politicalcycle models Nordhaus [1975] built the first formal model ofopportunistic political cycles based on the Phillips curve withadaptive expectations of voters In the model naıve voters get

3 To the best of our knowledge the only paper that uses monthly data isBerger and Woitek [1997] it rejects hypothesis of opportunistic cycles for theGerman developed democracy

1304 QUARTERLY JOURNAL OF ECONOMICS

consistently fooled by a preelectoral inflationary boom despite therecession following election The model predicts political cycles ingrowth unemployment and inflation Rogoff and Sibert [1988]Rogoff [1990] and Persson and Tabellini [1990] reconciled ratio-nal expectations with opportunistic political cycles in dynamicsignaling models with asymmetric information about politiciansrsquocompetence In particular Rogoff and Sibert [1988] and Rogoff[1990] focus on a political budget cycle in which voters rewardpreelectoral distortion in public spending because it signals aboutan incumbentrsquos ability to provide public goods Rogoff and Sibertrsquosmodel predicts a preelectoral fall in tax collection along withincreases in deficit and inflation Rogoffrsquos model focuses on thepreelectoral shift in composition of government spending awayfrom investment observed with a lag toward more visible publicconsumption We look at the dynamics of a wide range of publicexpenditure items composition of the budget growth and infla-tion to test predictions of these theories

Second we study determinants of the magnitude of the bud-get cycle Asymmetry of information about politiciansrsquo compe-tence plays a central role in rational opportunistic cycles Recentliterature extends the basic setup of Rogoffrsquos model to illustratethat transparency modeled as the probability that voters get acorrect exogenous signal about an incumbentrsquos competence[Gonzalez 2000] and awareness of voters defined as a share ofperfectly informed voters [Shi and Svensson 2002a] reduce themagnitude of the cycle Furthermore Gonzalez [2000] introducesthe level of democracy directly into the model Democracy has anonmonotone effect on the cycle first if the cost of enforcingpolitical turnover is prohibitively high for voters (as is the case indictatorships) then politicians do not have an incentive to engagein costly signaling and therefore there are no cycles Secondwhen change in office can be enforced the cycle decreases withdemocracy because it ensures institutional checks and balanceson government that allow voters to observe politiciansrsquo compe-tence directly with higher probability (ie free media) We testwhether these theoretical predictions are consistent with the databy looking at how cross-sectional variation in voter awareness(measured by education and urbanization) transparency (mea-sured by indices of government transparency and media free-dom) and democracy influence the cycles

In young democracies like Russia emerging NGOs and inde-pendent media accumulate experience in collecting packaging and

1305OPPORTUNISTIC POLITICAL CYCLES

disseminating information about politicians and their policies whilevoters learn how to process this information Thus cycles should getsmaller over time as learning proceeds and civil society develops[Brender and Drazen 2003] We test for the effect of time on thecycle Apart from learning however the time effect may also beattributed to the difference in scope for fiscal manipulation by re-gional governments under Yeltsin and Putinrsquos administrations (Ro-goff [1990] discusses the effects of restraining fiscal policies on po-litical cycles) We attempt to separate these hypotheses

All opportunistic cycle theories predict that preelectoral ma-nipulations are rewarded by voters In rational signaling modelsa cycle has costs (policy distortions) and benefits (transmittinginformation about politiciansrsquo competence) If voters learn politi-ciansrsquo type directly for instance through well-functioning inde-pendent media signaling has no value to voters In fact in de-veloped democracies voters reward politicians for restrained fis-cal policies [Alesina Perotti and Tavares 1998 Brender 2003]As a last step of our analysis we investigate whether it pays topursue preelectoral fiscal expansion in a maturing democracy

III THE DATA

The comprehensive list of regional governor elections that tookplace in Russia between August 1995 and December 2003 consists of194 electoral events4 Data on most policy instruments and out-comes are available for 159 elections between September 1996 andJuly 2003 Four regions had three rounds of elections 65 regionshad two rounds of elections and 17 regions had just one round ofelections during this period The source of the data on elections isTsentrizbirkom the Central Elections Committee of the RussianFederation

Regional monthly series of fiscal instruments and outcomescome from two sources Goskomstat the State Committee of Statis-tics provided data on wages and income wage arrears from theregional budgets price level and industrial output between 1995and 2003 the Ministry of Finance of the Russian Federation pro-vided detailed data on the execution of regional budgets for theperiod between 1996 and 2003

4 This list covers all the regions but Dagestan the only region where therehave not been any governor elections We excluded Chechnya and Ingushetia fromthe sample because fluctuations in fiscal policies of these regions have been drivenby war rather than elections

1306 QUARTERLY JOURNAL OF ECONOMICS

To test for the determinants of cycle magnitude we use cross-section data Data on urbanization and education come from Gos-komstat The data on freedom of media in the regions were providedby the Institute of Free Media (wwwfreepressru) Data on trans-parency of the regional government come from ldquoMedia-Soyuzrdquo anindependent professional association of Russian journalists Data onthe scope of regional democracy are from the Carnegie MoscowCenter Carnegie Endowment for International Peace [Petrov 2001]The Appendix presents descriptive statistics of the data

IV RESULTS

IVA Test for Opportunistic Cycles

Three groups of policy instruments and outcomes are consid-ered budgetary expenditures (total spending as well as levels andshares of expenditures on social programs education culturehealth care mass media and industrial subsidies) budget reve-nues and deficit (total revenues tax revenues deficit and federaltransfers) and such outcomes as growth inflation regional bud-getary wage arrears wage level and income We test for politicalcycles in these variables treating election time as exogenous5 Thefollowing equation is estimated on regional monthly panel data

(1) yit j1212

jmjit Lyit1

1Termit 2Leftit t fis εit

where i identifies regions t is real time in months and y stands fora logarithm of instrument or outcome of regional policy (all mone-tary variables are expressed in real terms per capita) To control forthe federal trend and macroeconomic shocks we include the com-plete set of time fixed effects t one for each month t To control forregion-specific fixed effects and region-specific seasonality we in-clude fixed effects fis for each of the twelve calendar months s(t) in

5 Ito [1990] Reid [1998] and Heckelman and Berument [1998] amongothers pointed out that opportunistic cycles can occur as a result of setting anelection date at the time of a boom Although almost 19 percent of Russiarsquosregional elections happened a month or more from their expected date in the vastmajority of these cases the time was shifted for exogenous reasons There were afew cases however when there was no exogenous reason for the shift of electiontime even though it is illegal to shift the date of regional elections To make surethat our results are not driven by the presence of endogenous elections werepeated all tests on the subsample of elections that had exogenously predeter-mined timing and got virtually identical results

1307OPPORTUNISTIC POLITICAL CYCLES

consistently fooled by a preelectoral inflationary boom despite therecession following election The model predicts political cycles ingrowth unemployment and inflation Rogoff and Sibert [1988]Rogoff [1990] and Persson and Tabellini [1990] reconciled ratio-nal expectations with opportunistic political cycles in dynamicsignaling models with asymmetric information about politiciansrsquocompetence In particular Rogoff and Sibert [1988] and Rogoff[1990] focus on a political budget cycle in which voters rewardpreelectoral distortion in public spending because it signals aboutan incumbentrsquos ability to provide public goods Rogoff and Sibertrsquosmodel predicts a preelectoral fall in tax collection along withincreases in deficit and inflation Rogoffrsquos model focuses on thepreelectoral shift in composition of government spending awayfrom investment observed with a lag toward more visible publicconsumption We look at the dynamics of a wide range of publicexpenditure items composition of the budget growth and infla-tion to test predictions of these theories

Second we study determinants of the magnitude of the bud-get cycle Asymmetry of information about politiciansrsquo compe-tence plays a central role in rational opportunistic cycles Recentliterature extends the basic setup of Rogoffrsquos model to illustratethat transparency modeled as the probability that voters get acorrect exogenous signal about an incumbentrsquos competence[Gonzalez 2000] and awareness of voters defined as a share ofperfectly informed voters [Shi and Svensson 2002a] reduce themagnitude of the cycle Furthermore Gonzalez [2000] introducesthe level of democracy directly into the model Democracy has anonmonotone effect on the cycle first if the cost of enforcingpolitical turnover is prohibitively high for voters (as is the case indictatorships) then politicians do not have an incentive to engagein costly signaling and therefore there are no cycles Secondwhen change in office can be enforced the cycle decreases withdemocracy because it ensures institutional checks and balanceson government that allow voters to observe politiciansrsquo compe-tence directly with higher probability (ie free media) We testwhether these theoretical predictions are consistent with the databy looking at how cross-sectional variation in voter awareness(measured by education and urbanization) transparency (mea-sured by indices of government transparency and media free-dom) and democracy influence the cycles

In young democracies like Russia emerging NGOs and inde-pendent media accumulate experience in collecting packaging and

1305OPPORTUNISTIC POLITICAL CYCLES

disseminating information about politicians and their policies whilevoters learn how to process this information Thus cycles should getsmaller over time as learning proceeds and civil society develops[Brender and Drazen 2003] We test for the effect of time on thecycle Apart from learning however the time effect may also beattributed to the difference in scope for fiscal manipulation by re-gional governments under Yeltsin and Putinrsquos administrations (Ro-goff [1990] discusses the effects of restraining fiscal policies on po-litical cycles) We attempt to separate these hypotheses

All opportunistic cycle theories predict that preelectoral ma-nipulations are rewarded by voters In rational signaling modelsa cycle has costs (policy distortions) and benefits (transmittinginformation about politiciansrsquo competence) If voters learn politi-ciansrsquo type directly for instance through well-functioning inde-pendent media signaling has no value to voters In fact in de-veloped democracies voters reward politicians for restrained fis-cal policies [Alesina Perotti and Tavares 1998 Brender 2003]As a last step of our analysis we investigate whether it pays topursue preelectoral fiscal expansion in a maturing democracy

III THE DATA

The comprehensive list of regional governor elections that tookplace in Russia between August 1995 and December 2003 consists of194 electoral events4 Data on most policy instruments and out-comes are available for 159 elections between September 1996 andJuly 2003 Four regions had three rounds of elections 65 regionshad two rounds of elections and 17 regions had just one round ofelections during this period The source of the data on elections isTsentrizbirkom the Central Elections Committee of the RussianFederation

Regional monthly series of fiscal instruments and outcomescome from two sources Goskomstat the State Committee of Statis-tics provided data on wages and income wage arrears from theregional budgets price level and industrial output between 1995and 2003 the Ministry of Finance of the Russian Federation pro-vided detailed data on the execution of regional budgets for theperiod between 1996 and 2003

4 This list covers all the regions but Dagestan the only region where therehave not been any governor elections We excluded Chechnya and Ingushetia fromthe sample because fluctuations in fiscal policies of these regions have been drivenby war rather than elections

1306 QUARTERLY JOURNAL OF ECONOMICS

To test for the determinants of cycle magnitude we use cross-section data Data on urbanization and education come from Gos-komstat The data on freedom of media in the regions were providedby the Institute of Free Media (wwwfreepressru) Data on trans-parency of the regional government come from ldquoMedia-Soyuzrdquo anindependent professional association of Russian journalists Data onthe scope of regional democracy are from the Carnegie MoscowCenter Carnegie Endowment for International Peace [Petrov 2001]The Appendix presents descriptive statistics of the data

IV RESULTS

IVA Test for Opportunistic Cycles

Three groups of policy instruments and outcomes are consid-ered budgetary expenditures (total spending as well as levels andshares of expenditures on social programs education culturehealth care mass media and industrial subsidies) budget reve-nues and deficit (total revenues tax revenues deficit and federaltransfers) and such outcomes as growth inflation regional bud-getary wage arrears wage level and income We test for politicalcycles in these variables treating election time as exogenous5 Thefollowing equation is estimated on regional monthly panel data

(1) yit j1212

jmjit Lyit1

1Termit 2Leftit t fis εit

where i identifies regions t is real time in months and y stands fora logarithm of instrument or outcome of regional policy (all mone-tary variables are expressed in real terms per capita) To control forthe federal trend and macroeconomic shocks we include the com-plete set of time fixed effects t one for each month t To control forregion-specific fixed effects and region-specific seasonality we in-clude fixed effects fis for each of the twelve calendar months s(t) in

5 Ito [1990] Reid [1998] and Heckelman and Berument [1998] amongothers pointed out that opportunistic cycles can occur as a result of setting anelection date at the time of a boom Although almost 19 percent of Russiarsquosregional elections happened a month or more from their expected date in the vastmajority of these cases the time was shifted for exogenous reasons There were afew cases however when there was no exogenous reason for the shift of electiontime even though it is illegal to shift the date of regional elections To make surethat our results are not driven by the presence of endogenous elections werepeated all tests on the subsample of elections that had exogenously predeter-mined timing and got virtually identical results

1307OPPORTUNISTIC POLITICAL CYCLES

disseminating information about politicians and their policies whilevoters learn how to process this information Thus cycles should getsmaller over time as learning proceeds and civil society develops[Brender and Drazen 2003] We test for the effect of time on thecycle Apart from learning however the time effect may also beattributed to the difference in scope for fiscal manipulation by re-gional governments under Yeltsin and Putinrsquos administrations (Ro-goff [1990] discusses the effects of restraining fiscal policies on po-litical cycles) We attempt to separate these hypotheses

All opportunistic cycle theories predict that preelectoral ma-nipulations are rewarded by voters In rational signaling modelsa cycle has costs (policy distortions) and benefits (transmittinginformation about politiciansrsquo competence) If voters learn politi-ciansrsquo type directly for instance through well-functioning inde-pendent media signaling has no value to voters In fact in de-veloped democracies voters reward politicians for restrained fis-cal policies [Alesina Perotti and Tavares 1998 Brender 2003]As a last step of our analysis we investigate whether it pays topursue preelectoral fiscal expansion in a maturing democracy

III THE DATA

The comprehensive list of regional governor elections that tookplace in Russia between August 1995 and December 2003 consists of194 electoral events4 Data on most policy instruments and out-comes are available for 159 elections between September 1996 andJuly 2003 Four regions had three rounds of elections 65 regionshad two rounds of elections and 17 regions had just one round ofelections during this period The source of the data on elections isTsentrizbirkom the Central Elections Committee of the RussianFederation

Regional monthly series of fiscal instruments and outcomescome from two sources Goskomstat the State Committee of Statis-tics provided data on wages and income wage arrears from theregional budgets price level and industrial output between 1995and 2003 the Ministry of Finance of the Russian Federation pro-vided detailed data on the execution of regional budgets for theperiod between 1996 and 2003

4 This list covers all the regions but Dagestan the only region where therehave not been any governor elections We excluded Chechnya and Ingushetia fromthe sample because fluctuations in fiscal policies of these regions have been drivenby war rather than elections

1306 QUARTERLY JOURNAL OF ECONOMICS

To test for the determinants of cycle magnitude we use cross-section data Data on urbanization and education come from Gos-komstat The data on freedom of media in the regions were providedby the Institute of Free Media (wwwfreepressru) Data on trans-parency of the regional government come from ldquoMedia-Soyuzrdquo anindependent professional association of Russian journalists Data onthe scope of regional democracy are from the Carnegie MoscowCenter Carnegie Endowment for International Peace [Petrov 2001]The Appendix presents descriptive statistics of the data

IV RESULTS

IVA Test for Opportunistic Cycles

Three groups of policy instruments and outcomes are consid-ered budgetary expenditures (total spending as well as levels andshares of expenditures on social programs education culturehealth care mass media and industrial subsidies) budget reve-nues and deficit (total revenues tax revenues deficit and federaltransfers) and such outcomes as growth inflation regional bud-getary wage arrears wage level and income We test for politicalcycles in these variables treating election time as exogenous5 Thefollowing equation is estimated on regional monthly panel data

(1) yit j1212

jmjit Lyit1

1Termit 2Leftit t fis εit

where i identifies regions t is real time in months and y stands fora logarithm of instrument or outcome of regional policy (all mone-tary variables are expressed in real terms per capita) To control forthe federal trend and macroeconomic shocks we include the com-plete set of time fixed effects t one for each month t To control forregion-specific fixed effects and region-specific seasonality we in-clude fixed effects fis for each of the twelve calendar months s(t) in

5 Ito [1990] Reid [1998] and Heckelman and Berument [1998] amongothers pointed out that opportunistic cycles can occur as a result of setting anelection date at the time of a boom Although almost 19 percent of Russiarsquosregional elections happened a month or more from their expected date in the vastmajority of these cases the time was shifted for exogenous reasons There were afew cases however when there was no exogenous reason for the shift of electiontime even though it is illegal to shift the date of regional elections To make surethat our results are not driven by the presence of endogenous elections werepeated all tests on the subsample of elections that had exogenously predeter-mined timing and got virtually identical results

1307OPPORTUNISTIC POLITICAL CYCLES

To test for the determinants of cycle magnitude we use cross-section data Data on urbanization and education come from Gos-komstat The data on freedom of media in the regions were providedby the Institute of Free Media (wwwfreepressru) Data on trans-parency of the regional government come from ldquoMedia-Soyuzrdquo anindependent professional association of Russian journalists Data onthe scope of regional democracy are from the Carnegie MoscowCenter Carnegie Endowment for International Peace [Petrov 2001]The Appendix presents descriptive statistics of the data

IV RESULTS

IVA Test for Opportunistic Cycles

Three groups of policy instruments and outcomes are consid-ered budgetary expenditures (total spending as well as levels andshares of expenditures on social programs education culturehealth care mass media and industrial subsidies) budget reve-nues and deficit (total revenues tax revenues deficit and federaltransfers) and such outcomes as growth inflation regional bud-getary wage arrears wage level and income We test for politicalcycles in these variables treating election time as exogenous5 Thefollowing equation is estimated on regional monthly panel data

(1) yit j1212

jmjit Lyit1

1Termit 2Leftit t fis εit

where i identifies regions t is real time in months and y stands fora logarithm of instrument or outcome of regional policy (all mone-tary variables are expressed in real terms per capita) To control forthe federal trend and macroeconomic shocks we include the com-plete set of time fixed effects t one for each month t To control forregion-specific fixed effects and region-specific seasonality we in-clude fixed effects fis for each of the twelve calendar months s(t) in

5 Ito [1990] Reid [1998] and Heckelman and Berument [1998] amongothers pointed out that opportunistic cycles can occur as a result of setting anelection date at the time of a boom Although almost 19 percent of Russiarsquosregional elections happened a month or more from their expected date in the vastmajority of these cases the time was shifted for exogenous reasons There were afew cases however when there was no exogenous reason for the shift of electiontime even though it is illegal to shift the date of regional elections To make surethat our results are not driven by the presence of endogenous elections werepeated all tests on the subsample of elections that had exogenously predeter-mined timing and got virtually identical results

1307OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EI

CY

CL

ES

INB

UD

GE

TS

PE

ND

ING

AN

DIT

SC

OM

PO

SIT

ION

Totalexpenditures

Socialexpenditures

Educationexpenditures

Expendituresonculture

Healthcareexpenditures

Mediaexpenditures

Expendituresonindustry

Shareofsocialexp

Shareofmediaexp

mon

th

12

003

4

001

8

000

10

007

0

013

001

90

003

004

50

070

(00

26)

(00

43)

(00

26)

(00

32)

(00

28)

(00

59)

(01

11)

(00

40)

(00

56)

mon

th

110

000

000

3

003

8

005

1

001

30

013

008

1

000

7

000

2(0

026

)(0

041

)(0

025

)(0

031

)(0

027

)(0

057

)(0

106

)(0

039

)(0

054

)m

onth

10

0

013

002

80

011

0

011

001

90

033

001

20

029

002

6(0

026

)(0

041

)(0

024

)(0

031

)(0

027

)(0

055

)(0

105

)(0

039

)(0

053

)m

onth

9

004

70

053

002

70

037

001

70

077

018

80

004

001

1(0

026

)(0

041

)(0

025

)(0

031

)(0

027

)(0

055

)(0

107

)(0

039

)(0

053

)m

onth

8

001

50

041

000

10

005

002

70

036

012

20

064

001

3(0

026

)(0

041

)(0

024

)(0

031

)(0

027

)(0

055

)(0

104

)(0

039

)(0

054

)m

onth

7

002

0

004

4

001

5

003

30

003

0

034

0

028

0

020

0

047

(00

25)

(00

39)

(00

23)

(00

29)

(00

25)

(00

55)

(00

98)

(00

39)

(00

53)

mon

th

60

007

0

026

003

10

028

002

20

002

012

5

001

10

008

(00

25)

(00

39)

(00

23)

(00

29)

(00

25)

(00

55)

(01

00)

(00

38)

(00

51)

mon

th

50

025

007

40

033

000

60

003

006

2

011

40

016

001

1(0

025

)(0

038

)(0

023

)(0

029

)(0

025

)(0

055

)(0

099

)(0

038

)(0

051

)

1308 QUARTERLY JOURNAL OF ECONOMICS

mon

th

40

009

005

9

004

6

002

0

004

10

081

0

122

003

80

088

(00

26)

(00

40)

(00

24)

(00

30)

(00

26)

(00

57)

(01

05)

(00

40)

(00

54)

mon

th

30

024

003

80

001

0

001

001

00

032

0

096

0

001

001

8(0

025

)(0

040

)(0

024

)(0

030

)(0

026

)(0

056

)(0

105

)(0

038

)(0

052

)m

onth

2

0

001

003

00

031

002

00

045

017

0

001

70

049

015

9

(00

25)

(00

40)

(00

24)

(00

30)

(00

26)

(00

55)

(01

05)

(00

38)

(00

52)

mon

th

10

117

0

226

0

121

0

140

0

130

0

106

020

00

103

004

9(0

025

)(0

040

)(0

024

)(0

030

)(0

026

)(0

054

)(0

105

)(0

037

)(0

051

)m

onth

0el

ecti

ons

0

053

0

030

0

077

006

1

0

081

012

7

0

008

008

0

0

091

(00

26)

(00

41)

(00

25)

(00

31)

(00

27)

(00

57)

(01

07)

(00

39)

(00

53)

mon

th

1

004

6

009

2

0

039

0

116

003

1

016

4

0

147

0

042

0

122

(0

024

)(0

038

)(0

023

)(0

029

)(0

025

)(0

053

)(0

099

)(0

036

)(0

049

)m

onth

2

0

043

0

062

0

016

0

034

0

019

0

170

009

5

003

1

014

0

(00

24)

(00

38)

(00

23)

(00

28)

(00

25)

(00

52)

(00

98)

(00

35)

(00

49)

mon

th

30

004

001

6

000

5

003

70

006

000

40

029

002

8

001

4(0

023

)(0

038

)(0

023

)(0

029

)(0

025

)(0

052

)(0

097

)(0

035

)(0

049

)m

onth

4

0

005

0

006

002

2

000

70

038

006

00

108

0

011

005

0(0

023

)(0

038

)(0

023

)(0

028

)(0

025

)(0

052

)(0

096

)(0

035

)(0

048

)m

onth

5

001

10

019

002

20

013

001

20

046

0

039

000

90

038

(00

23)

(00

37)

(00

22)

(00

27)

(00

24)

(00

51)

(00

92)

(00

35)

(00

48)

mon

th

6

000

70

061

003

3

000

5

001

9

000

20

211

0

067

0

016

(00

23)

(00

37)

(00

22)

(00

27)

(00

24)

(00

51)

(00

94)

(00

35)

(00

49)

1309OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EI

(CO

NT

INU

ED

)

Totalexpenditures

Socialexpenditures

Educationexpenditures

Expendituresonculture

Healthcareexpenditures

Mediaexpenditures

Expendituresonindustry

Shareofsocialexp

Shareofmediaexp

mon

th

70

039

0

012

001

30

014

002

00

086

0

147

0

053

003

3(0

023

)(0

037

)(0

022

)(0

028

)(0

024

)(0

052

)(0

095

)(0

036

)(0

049

)m

onth

8

001

90

014

0

007

002

60

002

001

5

002

20

007

0

006

(00

24)

(00

38)

(00

22)

(00

28)

(00

25)

(00

52)

(00

97)

(00

36)

(00

49)

mon

th

9

001

8

000

50

031

0

010

001

20

047

0

056

002

30

065

(00

23)

(00

38)

(00

23)

(00

29)

(00

25)

(00

52)

(00

97)

(00

36)

(00

49)

mon

th

10

000

9

000

20

007

004

0

000

4

003

60

017

000

8

001

3(0

023

)(0

038

)(0

022

)(0

028

)(0

025

)(0

052

)(0

098

)(0

035

)(0

049

)m

onth

11

004

20

016

002

30

075

0

031

000

4

001

8

000

7

001

5(0

023

)(0

038

)(0

022

)(0

028

)(0

025

)(0

051

)(0

100

)(0

035

)(0

048

)m

onth

12

0

004

0

016

001

50

004

0

019

002

40

059

0

003

005

0(0

024

)(0

039

)(0

023

)(0

029

)(0

025

)(0

054

)(0

102

)(0

036

)(0

050

)T

erm

inpo

wer

0

012

0

006

0

010

0

017

0

001

0

026

0

037

0

002

0

028

(00

10)

(00

16)

(00

10)

(00

12)

(00

10)

(00

22)

(00

41)

(00

15)

(00

21)

Lef

t0

030

0

003

002

8

000

10

005

0

073

0

011

0

009

0

101

(00

34)

(00

48)

(00

28)

(00

36)

(00

31)

(00

77)

(01

22)

(00

52)

(00

70)

Obs

67

6770

0469

8969

2169

6666

0058

2563

8862

41R

20

574

056

10

548

048

10

511

023

50

611

034

90

131

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

gres

sion

for

fede

ralt

ran

sfer

sis

run

onth

esu

bsam

ple

wh

ere

regi

ons

rece

ived

stri

ctly

posi

tive

fede

ral

aid

1310 QUARTERLY JOURNAL OF ECONOMICS

each region i Results are robust to using alternative methods ofaccounting for seasonality and trend [Alesina and Roubini 1992]

mjit is a dummy that equals 1 if t is j months away fromelections ( j 0 in the month of elections negative j means thatt is before and positivemdashthat t is after the election month)Henceforth we refer to the mjits as cycle dummies Positiveestimates of j before and negative estimates of j after electionsserve as evidence of opportunistic political cycles

Lag polynomial (L)yit1 accounts for autocorrelation in y Wetested for optimal lag structure using Akaike criterion Four lagsturned out to be optimal for the vast majority of the series For thesake of uniformity we report regressions with four lags for all theseries but results do not change if we include an optimal number oflags separately for each series Lags in panel regressions with fixedeffects produce biases that converge to zero when the time dimen-sion of a panel goes to infinity [Nickell 1981 Hansen 1982 White1982] Our panel covers more than 80 months therefore asymptoticproperties apply Nonetheless we verified that results are robust tousing the Arellano-Bond [1991] procedure6

Term controls for an incumbentrsquos political horizon that can bean important determinant of fiscal policies [Besley and Case 1995]It equals 0 1 2 or 3 depending on the term that the incumbentserves in office 0 means that the governor is appointed and has notbeen elected before 1 indicates that he was elected for the first timeetc Left is a dummy that equals 1 if the incumbent governor issupported by the Communist coalition It controls for partisancycles7

The results of estimation of equation (1) are presented inTables I and II Figure II plots the predicted political budget cycle

6 We tested for residual autocorrelation in the panel (1) The results showedthat the null hypothesis of no serial correlation in residuals cannot be rejected forall dependent variables P-values for this test are presented in the Appendix Thenull cannot be rejected for each dependent variable at the 5 percent significancelevel and for each dependent variable but wages at the 25 percent significancelevel In addition we tested for unit roots in each series for each region using anaugmented Dickey-Fuller test The null hypothesis of a unit root was rejected forall series in all regions with the exception of wage arrears Thus we run regres-sion (1) for the log change rather than level of wage arrears because growth ofwage arrears is stationary

7 Alesina Roubini and Cohen [1997] review the partisan theory and evidenceElectoral campaign platforms of Russiarsquos governors are polarized into the ldquocommunistleftrdquo and the ldquoliberal-democratic ideologyrdquo In practice the variation in the data isinsufficient to have a proper test of partisan theory because there are only a few caseswhen a new governor of the opposite ideological platform replaced an incumbenttherefore ideology is almost perfectly collinear with fixed effects Thus we do not putemphasis on interpretation of coefficients of the ldquoleft-wingrdquo dummy

1311OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

CY

CL

ES

INB

UD

GE

TR

EV

EN

UE

S

INC

OM

E

WA

GE

AR

RE

AR

S

GR

OW

TH

A

ND

INF

LA

TIO

N

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

120

039

001

7

005

3

0

057

000

7

000

6

000

30

003

0

060

(00

28)

(00

24)

(00

21)

(00

85)

(00

14)

(00

03)

(00

05)

(00

07)

(00

71)

mon

th

110

003

0

008

0

019

001

2

000

6

000

7

0

011

0

003

006

3(0

027

)(0

023

)(0

020

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

10

0

038

0

046

0

009

005

5

000

40

001

0

001

0

004

002

2(0

026

)(0

023

)(0

020

)(0

078

)(0

013

)(0

002

)(0

005

)(0

007

)(0

063

)m

onth

9

001

90

036

001

10

161

000

5

000

7

001

0

000

40

084

(00

26)

(00

23)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

51)

mon

th

80

014

002

80

002

009

4

000

8

000

4

000

6

000

7

006

7(0

026

)(0

023

)(0

020

)(0

075

)(0

013

)(0

002

)(0

005

)(0

007

)(0

050

)m

onth

7

0

014

0

023

001

7

010

10

024

000

00

004

0

003

0

056

(00

25)

(00

22)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

47)

mon

th

6

001

40

016

002

8

006

20

010

0

001

0

005

0

008

004

6(0

025

)(0

022

)(0

020

)(0

074

)(0

013

)(0

003

)(0

005

)(0

007

)(0

047

)m

onth

5

003

60

053

001

40

053

000

50

002

0

013

000

6

021

7

(00

24)

(00

22)

(00

20)

(00

75)

(00

13)

(00

03)

(00

05)

(00

07)

(00

47)

mon

th

4

001

4

000

40

000

0

137

0

003

0

004

000

80

002

002

3(0

026

)(0

023

)(0

021

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

055

)

1312 QUARTERLY JOURNAL OF ECONOMICS

mon

th

30

008

003

50

042

003

60

016

0

007

0

008

001

4

005

2(0

026

)(0

022

)(0

020

)(0

078

)(0

014

)(0

003

)(0

005

)(0

007

)(0

058

)m

onth

2

000

80

027

000

3

007

20

007

0

002

000

50

018

001

2(0

026

)(0

022

)(0

020

)(0

078

)(0

013

)(0

003

)(0

005

)(0

007

)(0

060

)m

onth

1

008

0

000

00

052

0

145

0

013

0

005

0

017

0

036

020

7

(00

26)

(00

22)

(00

19)

(00

79)

(00

13)

(00

03)

(00

05)

(00

07)

(00

66)

mon

th0

elec

tion

s

005

0

002

40

010

0

148

000

0

000

50

006

000

6

007

3(0

026

)(0

023

)(0

020

)(0

084

)(0

014

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

1

0

063

000

8

000

7

003

5

000

90

001

0

012

002

2

004

3(0

025

)(0

022

)(0

019

)(0

080

)(0

014

)(0

003

)(0

005

)(0

007

)(0

059

)m

onth

2

0

032

0

047

000

90

042

0

003

000

10

003

0

028

0

057

(00

24)

(00

22)

(00

19)

(00

77)

(00

15)

(00

03)

(00

05)

(00

07)

(00

56)

mon

th

30

010

0

001

000

00

028

001

20

004

000

4

000

50

098

(00

24)

(00

22)

(00

18)

(00

78)

(00

15)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

4

000

50

015

002

9

017

0

000

40

004

0

007

000

6

002

4(0

024

)(0

021

)(0

018

)(0

079

)(0

015

)(0

003

)(0

005

)(0

007

)(0

050

)m

onth

5

0

023

0

013

002

0

013

80

016

000

10

001

001

00

050

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

6

001

20

009

001

30

055

000

3

000

2

000

4

000

80

055

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

70

038

006

0

0

015

001

80

007

0

002

0

013

0

003

0

114

(0

024

)(0

021

)(0

019

)(0

077

)(0

014

)(0

003

)(0

005

)(0

007

)(0

049

)

1313OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

(CO

NT

INU

ED

)

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

80

005

0

004

000

50

031

0

022

0

003

000

00

002

004

4(0

025

)(0

021

)(0

019

)(0

076

)(0

014

)(0

003

)(0

005

)(0

007

)(0

051

)m

onth

9

0

009

0

021

0

014

0

024

000

00

000

000

3

000

10

064

(00

25)

(00

21)

(00

19)

(00

76)

(00

14)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

10

002

8

001

80

011

0

055

0

013

000

20

004

0

001

0

015

(00

25)

(00

21)

(00

19)

(00

76)

(00

15)

(00

03)

(00

05)

(00

07)

(00

57)

mon

th

110

055

0

008

0

004

0

125

0

009

000

5

000

6

000

9

010

1(0

024

)(0

021

)(0

018

)(0

074

)(0

014

)(0

003

)(0

005

)(0

007

)(0

061

)m

onth

12

0

008

002

0

000

5

008

7

001

8

000

1

000

3

000

3

006

6(0

025

)(0

022

)(0

019

)(0

075

)(0

014

)(0

003

)(0

005

)(0

007

)(0

072

)T

erm

inpo

wer

0

001

0

002

000

2

002

90

005

0

004

0

001

000

30

013

(00

11)

(00

09)

(00

08)

(00

33)

(00

05)

(00

01)

(00

02)

(00

03)

(00

24)

Lef

t0

025

007

1

0

038

022

7

000

9

000

1

000

3

001

20

000

(00

30)

(00

27)

(00

26)

(01

60)

(00

12)

(00

02)

(00

05)

(00

06)

(00

00)

Obs

64

9870

6059

1641

1569

4610

731

7970

8002

2588

R2

065

60

584

033

40

631

035

10

818

092

80

851

013

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1314 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

mon

th

40

009

005

9

004

6

002

0

004

10

081

0

122

003

80

088

(00

26)

(00

40)

(00

24)

(00

30)

(00

26)

(00

57)

(01

05)

(00

40)

(00

54)

mon

th

30

024

003

80

001

0

001

001

00

032

0

096

0

001

001

8(0

025

)(0

040

)(0

024

)(0

030

)(0

026

)(0

056

)(0

105

)(0

038

)(0

052

)m

onth

2

0

001

003

00

031

002

00

045

017

0

001

70

049

015

9

(00

25)

(00

40)

(00

24)

(00

30)

(00

26)

(00

55)

(01

05)

(00

38)

(00

52)

mon

th

10

117

0

226

0

121

0

140

0

130

0

106

020

00

103

004

9(0

025

)(0

040

)(0

024

)(0

030

)(0

026

)(0

054

)(0

105

)(0

037

)(0

051

)m

onth

0el

ecti

ons

0

053

0

030

0

077

006

1

0

081

012

7

0

008

008

0

0

091

(00

26)

(00

41)

(00

25)

(00

31)

(00

27)

(00

57)

(01

07)

(00

39)

(00

53)

mon

th

1

004

6

009

2

0

039

0

116

003

1

016

4

0

147

0

042

0

122

(0

024

)(0

038

)(0

023

)(0

029

)(0

025

)(0

053

)(0

099

)(0

036

)(0

049

)m

onth

2

0

043

0

062

0

016

0

034

0

019

0

170

009

5

003

1

014

0

(00

24)

(00

38)

(00

23)

(00

28)

(00

25)

(00

52)

(00

98)

(00

35)

(00

49)

mon

th

30

004

001

6

000

5

003

70

006

000

40

029

002

8

001

4(0

023

)(0

038

)(0

023

)(0

029

)(0

025

)(0

052

)(0

097

)(0

035

)(0

049

)m

onth

4

0

005

0

006

002

2

000

70

038

006

00

108

0

011

005

0(0

023

)(0

038

)(0

023

)(0

028

)(0

025

)(0

052

)(0

096

)(0

035

)(0

048

)m

onth

5

001

10

019

002

20

013

001

20

046

0

039

000

90

038

(00

23)

(00

37)

(00

22)

(00

27)

(00

24)

(00

51)

(00

92)

(00

35)

(00

48)

mon

th

6

000

70

061

003

3

000

5

001

9

000

20

211

0

067

0

016

(00

23)

(00

37)

(00

22)

(00

27)

(00

24)

(00

51)

(00

94)

(00

35)

(00

49)

1309OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EI

(CO

NT

INU

ED

)

Totalexpenditures

Socialexpenditures

Educationexpenditures

Expendituresonculture

Healthcareexpenditures

Mediaexpenditures

Expendituresonindustry

Shareofsocialexp

Shareofmediaexp

mon

th

70

039

0

012

001

30

014

002

00

086

0

147

0

053

003

3(0

023

)(0

037

)(0

022

)(0

028

)(0

024

)(0

052

)(0

095

)(0

036

)(0

049

)m

onth

8

001

90

014

0

007

002

60

002

001

5

002

20

007

0

006

(00

24)

(00

38)

(00

22)

(00

28)

(00

25)

(00

52)

(00

97)

(00

36)

(00

49)

mon

th

9

001

8

000

50

031

0

010

001

20

047

0

056

002

30

065

(00

23)

(00

38)

(00

23)

(00

29)

(00

25)

(00

52)

(00

97)

(00

36)

(00

49)

mon

th

10

000

9

000

20

007

004

0

000

4

003

60

017

000

8

001

3(0

023

)(0

038

)(0

022

)(0

028

)(0

025

)(0

052

)(0

098

)(0

035

)(0

049

)m

onth

11

004

20

016

002

30

075

0

031

000

4

001

8

000

7

001

5(0

023

)(0

038

)(0

022

)(0

028

)(0

025

)(0

051

)(0

100

)(0

035

)(0

048

)m

onth

12

0

004

0

016

001

50

004

0

019

002

40

059

0

003

005

0(0

024

)(0

039

)(0

023

)(0

029

)(0

025

)(0

054

)(0

102

)(0

036

)(0

050

)T

erm

inpo

wer

0

012

0

006

0

010

0

017

0

001

0

026

0

037

0

002

0

028

(00

10)

(00

16)

(00

10)

(00

12)

(00

10)

(00

22)

(00

41)

(00

15)

(00

21)

Lef

t0

030

0

003

002

8

000

10

005

0

073

0

011

0

009

0

101

(00

34)

(00

48)

(00

28)

(00

36)

(00

31)

(00

77)

(01

22)

(00

52)

(00

70)

Obs

67

6770

0469

8969

2169

6666

0058

2563

8862

41R

20

574

056

10

548

048

10

511

023

50

611

034

90

131

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

gres

sion

for

fede

ralt

ran

sfer

sis

run

onth

esu

bsam

ple

wh

ere

regi

ons

rece

ived

stri

ctly

posi

tive

fede

ral

aid

1310 QUARTERLY JOURNAL OF ECONOMICS

each region i Results are robust to using alternative methods ofaccounting for seasonality and trend [Alesina and Roubini 1992]

mjit is a dummy that equals 1 if t is j months away fromelections ( j 0 in the month of elections negative j means thatt is before and positivemdashthat t is after the election month)Henceforth we refer to the mjits as cycle dummies Positiveestimates of j before and negative estimates of j after electionsserve as evidence of opportunistic political cycles

Lag polynomial (L)yit1 accounts for autocorrelation in y Wetested for optimal lag structure using Akaike criterion Four lagsturned out to be optimal for the vast majority of the series For thesake of uniformity we report regressions with four lags for all theseries but results do not change if we include an optimal number oflags separately for each series Lags in panel regressions with fixedeffects produce biases that converge to zero when the time dimen-sion of a panel goes to infinity [Nickell 1981 Hansen 1982 White1982] Our panel covers more than 80 months therefore asymptoticproperties apply Nonetheless we verified that results are robust tousing the Arellano-Bond [1991] procedure6

Term controls for an incumbentrsquos political horizon that can bean important determinant of fiscal policies [Besley and Case 1995]It equals 0 1 2 or 3 depending on the term that the incumbentserves in office 0 means that the governor is appointed and has notbeen elected before 1 indicates that he was elected for the first timeetc Left is a dummy that equals 1 if the incumbent governor issupported by the Communist coalition It controls for partisancycles7

The results of estimation of equation (1) are presented inTables I and II Figure II plots the predicted political budget cycle

6 We tested for residual autocorrelation in the panel (1) The results showedthat the null hypothesis of no serial correlation in residuals cannot be rejected forall dependent variables P-values for this test are presented in the Appendix Thenull cannot be rejected for each dependent variable at the 5 percent significancelevel and for each dependent variable but wages at the 25 percent significancelevel In addition we tested for unit roots in each series for each region using anaugmented Dickey-Fuller test The null hypothesis of a unit root was rejected forall series in all regions with the exception of wage arrears Thus we run regres-sion (1) for the log change rather than level of wage arrears because growth ofwage arrears is stationary

7 Alesina Roubini and Cohen [1997] review the partisan theory and evidenceElectoral campaign platforms of Russiarsquos governors are polarized into the ldquocommunistleftrdquo and the ldquoliberal-democratic ideologyrdquo In practice the variation in the data isinsufficient to have a proper test of partisan theory because there are only a few caseswhen a new governor of the opposite ideological platform replaced an incumbenttherefore ideology is almost perfectly collinear with fixed effects Thus we do not putemphasis on interpretation of coefficients of the ldquoleft-wingrdquo dummy

1311OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

CY

CL

ES

INB

UD

GE

TR

EV

EN

UE

S

INC

OM

E

WA

GE

AR

RE

AR

S

GR

OW

TH

A

ND

INF

LA

TIO

N

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

120

039

001

7

005

3

0

057

000

7

000

6

000

30

003

0

060

(00

28)

(00

24)

(00

21)

(00

85)

(00

14)

(00

03)

(00

05)

(00

07)

(00

71)

mon

th

110

003

0

008

0

019

001

2

000

6

000

7

0

011

0

003

006

3(0

027

)(0

023

)(0

020

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

10

0

038

0

046

0

009

005

5

000

40

001

0

001

0

004

002

2(0

026

)(0

023

)(0

020

)(0

078

)(0

013

)(0

002

)(0

005

)(0

007

)(0

063

)m

onth

9

001

90

036

001

10

161

000

5

000

7

001

0

000

40

084

(00

26)

(00

23)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

51)

mon

th

80

014

002

80

002

009

4

000

8

000

4

000

6

000

7

006

7(0

026

)(0

023

)(0

020

)(0

075

)(0

013

)(0

002

)(0

005

)(0

007

)(0

050

)m

onth

7

0

014

0

023

001

7

010

10

024

000

00

004

0

003

0

056

(00

25)

(00

22)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

47)

mon

th

6

001

40

016

002

8

006

20

010

0

001

0

005

0

008

004

6(0

025

)(0

022

)(0

020

)(0

074

)(0

013

)(0

003

)(0

005

)(0

007

)(0

047

)m

onth

5

003

60

053

001

40

053

000

50

002

0

013

000

6

021

7

(00

24)

(00

22)

(00

20)

(00

75)

(00

13)

(00

03)

(00

05)

(00

07)

(00

47)

mon

th

4

001

4

000

40

000

0

137

0

003

0

004

000

80

002

002

3(0

026

)(0

023

)(0

021

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

055

)

1312 QUARTERLY JOURNAL OF ECONOMICS

mon

th

30

008

003

50

042

003

60

016

0

007

0

008

001

4

005

2(0

026

)(0

022

)(0

020

)(0

078

)(0

014

)(0

003

)(0

005

)(0

007

)(0

058

)m

onth

2

000

80

027

000

3

007

20

007

0

002

000

50

018

001

2(0

026

)(0

022

)(0

020

)(0

078

)(0

013

)(0

003

)(0

005

)(0

007

)(0

060

)m

onth

1

008

0

000

00

052

0

145

0

013

0

005

0

017

0

036

020

7

(00

26)

(00

22)

(00

19)

(00

79)

(00

13)

(00

03)

(00

05)

(00

07)

(00

66)

mon

th0

elec

tion

s

005

0

002

40

010

0

148

000

0

000

50

006

000

6

007

3(0

026

)(0

023

)(0

020

)(0

084

)(0

014

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

1

0

063

000

8

000

7

003

5

000

90

001

0

012

002

2

004

3(0

025

)(0

022

)(0

019

)(0

080

)(0

014

)(0

003

)(0

005

)(0

007

)(0

059

)m

onth

2

0

032

0

047

000

90

042

0

003

000

10

003

0

028

0

057

(00

24)

(00

22)

(00

19)

(00

77)

(00

15)

(00

03)

(00

05)

(00

07)

(00

56)

mon

th

30

010

0

001

000

00

028

001

20

004

000

4

000

50

098

(00

24)

(00

22)

(00

18)

(00

78)

(00

15)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

4

000

50

015

002

9

017

0

000

40

004

0

007

000

6

002

4(0

024

)(0

021

)(0

018

)(0

079

)(0

015

)(0

003

)(0

005

)(0

007

)(0

050

)m

onth

5

0

023

0

013

002

0

013

80

016

000

10

001

001

00

050

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

6

001

20

009

001

30

055

000

3

000

2

000

4

000

80

055

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

70

038

006

0

0

015

001

80

007

0

002

0

013

0

003

0

114

(0

024

)(0

021

)(0

019

)(0

077

)(0

014

)(0

003

)(0

005

)(0

007

)(0

049

)

1313OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

(CO

NT

INU

ED

)

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

80

005

0

004

000

50

031

0

022

0

003

000

00

002

004

4(0

025

)(0

021

)(0

019

)(0

076

)(0

014

)(0

003

)(0

005

)(0

007

)(0

051

)m

onth

9

0

009

0

021

0

014

0

024

000

00

000

000

3

000

10

064

(00

25)

(00

21)

(00

19)

(00

76)

(00

14)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

10

002

8

001

80

011

0

055

0

013

000

20

004

0

001

0

015

(00

25)

(00

21)

(00

19)

(00

76)

(00

15)

(00

03)

(00

05)

(00

07)

(00

57)

mon

th

110

055

0

008

0

004

0

125

0

009

000

5

000

6

000

9

010

1(0

024

)(0

021

)(0

018

)(0

074

)(0

014

)(0

003

)(0

005

)(0

007

)(0

061

)m

onth

12

0

008

002

0

000

5

008

7

001

8

000

1

000

3

000

3

006

6(0

025

)(0

022

)(0

019

)(0

075

)(0

014

)(0

003

)(0

005

)(0

007

)(0

072

)T

erm

inpo

wer

0

001

0

002

000

2

002

90

005

0

004

0

001

000

30

013

(00

11)

(00

09)

(00

08)

(00

33)

(00

05)

(00

01)

(00

02)

(00

03)

(00

24)

Lef

t0

025

007

1

0

038

022

7

000

9

000

1

000

3

001

20

000

(00

30)

(00

27)

(00

26)

(01

60)

(00

12)

(00

02)

(00

05)

(00

06)

(00

00)

Obs

64

9870

6059

1641

1569

4610

731

7970

8002

2588

R2

065

60

584

033

40

631

035

10

818

092

80

851

013

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1314 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EI

(CO

NT

INU

ED

)

Totalexpenditures

Socialexpenditures

Educationexpenditures

Expendituresonculture

Healthcareexpenditures

Mediaexpenditures

Expendituresonindustry

Shareofsocialexp

Shareofmediaexp

mon

th

70

039

0

012

001

30

014

002

00

086

0

147

0

053

003

3(0

023

)(0

037

)(0

022

)(0

028

)(0

024

)(0

052

)(0

095

)(0

036

)(0

049

)m

onth

8

001

90

014

0

007

002

60

002

001

5

002

20

007

0

006

(00

24)

(00

38)

(00

22)

(00

28)

(00

25)

(00

52)

(00

97)

(00

36)

(00

49)

mon

th

9

001

8

000

50

031

0

010

001

20

047

0

056

002

30

065

(00

23)

(00

38)

(00

23)

(00

29)

(00

25)

(00

52)

(00

97)

(00

36)

(00

49)

mon

th

10

000

9

000

20

007

004

0

000

4

003

60

017

000

8

001

3(0

023

)(0

038

)(0

022

)(0

028

)(0

025

)(0

052

)(0

098

)(0

035

)(0

049

)m

onth

11

004

20

016

002

30

075

0

031

000

4

001

8

000

7

001

5(0

023

)(0

038

)(0

022

)(0

028

)(0

025

)(0

051

)(0

100

)(0

035

)(0

048

)m

onth

12

0

004

0

016

001

50

004

0

019

002

40

059

0

003

005

0(0

024

)(0

039

)(0

023

)(0

029

)(0

025

)(0

054

)(0

102

)(0

036

)(0

050

)T

erm

inpo

wer

0

012

0

006

0

010

0

017

0

001

0

026

0

037

0

002

0

028

(00

10)

(00

16)

(00

10)

(00

12)

(00

10)

(00

22)

(00

41)

(00

15)

(00

21)

Lef

t0

030

0

003

002

8

000

10

005

0

073

0

011

0

009

0

101

(00

34)

(00

48)

(00

28)

(00

36)

(00

31)

(00

77)

(01

22)

(00

52)

(00

70)

Obs

67

6770

0469

8969

2169

6666

0058

2563

8862

41R

20

574

056

10

548

048

10

511

023

50

611

034

90

131

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

gres

sion

for

fede

ralt

ran

sfer

sis

run

onth

esu

bsam

ple

wh

ere

regi

ons

rece

ived

stri

ctly

posi

tive

fede

ral

aid

1310 QUARTERLY JOURNAL OF ECONOMICS

each region i Results are robust to using alternative methods ofaccounting for seasonality and trend [Alesina and Roubini 1992]

mjit is a dummy that equals 1 if t is j months away fromelections ( j 0 in the month of elections negative j means thatt is before and positivemdashthat t is after the election month)Henceforth we refer to the mjits as cycle dummies Positiveestimates of j before and negative estimates of j after electionsserve as evidence of opportunistic political cycles

Lag polynomial (L)yit1 accounts for autocorrelation in y Wetested for optimal lag structure using Akaike criterion Four lagsturned out to be optimal for the vast majority of the series For thesake of uniformity we report regressions with four lags for all theseries but results do not change if we include an optimal number oflags separately for each series Lags in panel regressions with fixedeffects produce biases that converge to zero when the time dimen-sion of a panel goes to infinity [Nickell 1981 Hansen 1982 White1982] Our panel covers more than 80 months therefore asymptoticproperties apply Nonetheless we verified that results are robust tousing the Arellano-Bond [1991] procedure6

Term controls for an incumbentrsquos political horizon that can bean important determinant of fiscal policies [Besley and Case 1995]It equals 0 1 2 or 3 depending on the term that the incumbentserves in office 0 means that the governor is appointed and has notbeen elected before 1 indicates that he was elected for the first timeetc Left is a dummy that equals 1 if the incumbent governor issupported by the Communist coalition It controls for partisancycles7

The results of estimation of equation (1) are presented inTables I and II Figure II plots the predicted political budget cycle

6 We tested for residual autocorrelation in the panel (1) The results showedthat the null hypothesis of no serial correlation in residuals cannot be rejected forall dependent variables P-values for this test are presented in the Appendix Thenull cannot be rejected for each dependent variable at the 5 percent significancelevel and for each dependent variable but wages at the 25 percent significancelevel In addition we tested for unit roots in each series for each region using anaugmented Dickey-Fuller test The null hypothesis of a unit root was rejected forall series in all regions with the exception of wage arrears Thus we run regres-sion (1) for the log change rather than level of wage arrears because growth ofwage arrears is stationary

7 Alesina Roubini and Cohen [1997] review the partisan theory and evidenceElectoral campaign platforms of Russiarsquos governors are polarized into the ldquocommunistleftrdquo and the ldquoliberal-democratic ideologyrdquo In practice the variation in the data isinsufficient to have a proper test of partisan theory because there are only a few caseswhen a new governor of the opposite ideological platform replaced an incumbenttherefore ideology is almost perfectly collinear with fixed effects Thus we do not putemphasis on interpretation of coefficients of the ldquoleft-wingrdquo dummy

1311OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

CY

CL

ES

INB

UD

GE

TR

EV

EN

UE

S

INC

OM

E

WA

GE

AR

RE

AR

S

GR

OW

TH

A

ND

INF

LA

TIO

N

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

120

039

001

7

005

3

0

057

000

7

000

6

000

30

003

0

060

(00

28)

(00

24)

(00

21)

(00

85)

(00

14)

(00

03)

(00

05)

(00

07)

(00

71)

mon

th

110

003

0

008

0

019

001

2

000

6

000

7

0

011

0

003

006

3(0

027

)(0

023

)(0

020

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

10

0

038

0

046

0

009

005

5

000

40

001

0

001

0

004

002

2(0

026

)(0

023

)(0

020

)(0

078

)(0

013

)(0

002

)(0

005

)(0

007

)(0

063

)m

onth

9

001

90

036

001

10

161

000

5

000

7

001

0

000

40

084

(00

26)

(00

23)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

51)

mon

th

80

014

002

80

002

009

4

000

8

000

4

000

6

000

7

006

7(0

026

)(0

023

)(0

020

)(0

075

)(0

013

)(0

002

)(0

005

)(0

007

)(0

050

)m

onth

7

0

014

0

023

001

7

010

10

024

000

00

004

0

003

0

056

(00

25)

(00

22)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

47)

mon

th

6

001

40

016

002

8

006

20

010

0

001

0

005

0

008

004

6(0

025

)(0

022

)(0

020

)(0

074

)(0

013

)(0

003

)(0

005

)(0

007

)(0

047

)m

onth

5

003

60

053

001

40

053

000

50

002

0

013

000

6

021

7

(00

24)

(00

22)

(00

20)

(00

75)

(00

13)

(00

03)

(00

05)

(00

07)

(00

47)

mon

th

4

001

4

000

40

000

0

137

0

003

0

004

000

80

002

002

3(0

026

)(0

023

)(0

021

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

055

)

1312 QUARTERLY JOURNAL OF ECONOMICS

mon

th

30

008

003

50

042

003

60

016

0

007

0

008

001

4

005

2(0

026

)(0

022

)(0

020

)(0

078

)(0

014

)(0

003

)(0

005

)(0

007

)(0

058

)m

onth

2

000

80

027

000

3

007

20

007

0

002

000

50

018

001

2(0

026

)(0

022

)(0

020

)(0

078

)(0

013

)(0

003

)(0

005

)(0

007

)(0

060

)m

onth

1

008

0

000

00

052

0

145

0

013

0

005

0

017

0

036

020

7

(00

26)

(00

22)

(00

19)

(00

79)

(00

13)

(00

03)

(00

05)

(00

07)

(00

66)

mon

th0

elec

tion

s

005

0

002

40

010

0

148

000

0

000

50

006

000

6

007

3(0

026

)(0

023

)(0

020

)(0

084

)(0

014

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

1

0

063

000

8

000

7

003

5

000

90

001

0

012

002

2

004

3(0

025

)(0

022

)(0

019

)(0

080

)(0

014

)(0

003

)(0

005

)(0

007

)(0

059

)m

onth

2

0

032

0

047

000

90

042

0

003

000

10

003

0

028

0

057

(00

24)

(00

22)

(00

19)

(00

77)

(00

15)

(00

03)

(00

05)

(00

07)

(00

56)

mon

th

30

010

0

001

000

00

028

001

20

004

000

4

000

50

098

(00

24)

(00

22)

(00

18)

(00

78)

(00

15)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

4

000

50

015

002

9

017

0

000

40

004

0

007

000

6

002

4(0

024

)(0

021

)(0

018

)(0

079

)(0

015

)(0

003

)(0

005

)(0

007

)(0

050

)m

onth

5

0

023

0

013

002

0

013

80

016

000

10

001

001

00

050

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

6

001

20

009

001

30

055

000

3

000

2

000

4

000

80

055

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

70

038

006

0

0

015

001

80

007

0

002

0

013

0

003

0

114

(0

024

)(0

021

)(0

019

)(0

077

)(0

014

)(0

003

)(0

005

)(0

007

)(0

049

)

1313OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

(CO

NT

INU

ED

)

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

80

005

0

004

000

50

031

0

022

0

003

000

00

002

004

4(0

025

)(0

021

)(0

019

)(0

076

)(0

014

)(0

003

)(0

005

)(0

007

)(0

051

)m

onth

9

0

009

0

021

0

014

0

024

000

00

000

000

3

000

10

064

(00

25)

(00

21)

(00

19)

(00

76)

(00

14)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

10

002

8

001

80

011

0

055

0

013

000

20

004

0

001

0

015

(00

25)

(00

21)

(00

19)

(00

76)

(00

15)

(00

03)

(00

05)

(00

07)

(00

57)

mon

th

110

055

0

008

0

004

0

125

0

009

000

5

000

6

000

9

010

1(0

024

)(0

021

)(0

018

)(0

074

)(0

014

)(0

003

)(0

005

)(0

007

)(0

061

)m

onth

12

0

008

002

0

000

5

008

7

001

8

000

1

000

3

000

3

006

6(0

025

)(0

022

)(0

019

)(0

075

)(0

014

)(0

003

)(0

005

)(0

007

)(0

072

)T

erm

inpo

wer

0

001

0

002

000

2

002

90

005

0

004

0

001

000

30

013

(00

11)

(00

09)

(00

08)

(00

33)

(00

05)

(00

01)

(00

02)

(00

03)

(00

24)

Lef

t0

025

007

1

0

038

022

7

000

9

000

1

000

3

001

20

000

(00

30)

(00

27)

(00

26)

(01

60)

(00

12)

(00

02)

(00

05)

(00

06)

(00

00)

Obs

64

9870

6059

1641

1569

4610

731

7970

8002

2588

R2

065

60

584

033

40

631

035

10

818

092

80

851

013

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1314 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

each region i Results are robust to using alternative methods ofaccounting for seasonality and trend [Alesina and Roubini 1992]

mjit is a dummy that equals 1 if t is j months away fromelections ( j 0 in the month of elections negative j means thatt is before and positivemdashthat t is after the election month)Henceforth we refer to the mjits as cycle dummies Positiveestimates of j before and negative estimates of j after electionsserve as evidence of opportunistic political cycles

Lag polynomial (L)yit1 accounts for autocorrelation in y Wetested for optimal lag structure using Akaike criterion Four lagsturned out to be optimal for the vast majority of the series For thesake of uniformity we report regressions with four lags for all theseries but results do not change if we include an optimal number oflags separately for each series Lags in panel regressions with fixedeffects produce biases that converge to zero when the time dimen-sion of a panel goes to infinity [Nickell 1981 Hansen 1982 White1982] Our panel covers more than 80 months therefore asymptoticproperties apply Nonetheless we verified that results are robust tousing the Arellano-Bond [1991] procedure6

Term controls for an incumbentrsquos political horizon that can bean important determinant of fiscal policies [Besley and Case 1995]It equals 0 1 2 or 3 depending on the term that the incumbentserves in office 0 means that the governor is appointed and has notbeen elected before 1 indicates that he was elected for the first timeetc Left is a dummy that equals 1 if the incumbent governor issupported by the Communist coalition It controls for partisancycles7

The results of estimation of equation (1) are presented inTables I and II Figure II plots the predicted political budget cycle

6 We tested for residual autocorrelation in the panel (1) The results showedthat the null hypothesis of no serial correlation in residuals cannot be rejected forall dependent variables P-values for this test are presented in the Appendix Thenull cannot be rejected for each dependent variable at the 5 percent significancelevel and for each dependent variable but wages at the 25 percent significancelevel In addition we tested for unit roots in each series for each region using anaugmented Dickey-Fuller test The null hypothesis of a unit root was rejected forall series in all regions with the exception of wage arrears Thus we run regres-sion (1) for the log change rather than level of wage arrears because growth ofwage arrears is stationary

7 Alesina Roubini and Cohen [1997] review the partisan theory and evidenceElectoral campaign platforms of Russiarsquos governors are polarized into the ldquocommunistleftrdquo and the ldquoliberal-democratic ideologyrdquo In practice the variation in the data isinsufficient to have a proper test of partisan theory because there are only a few caseswhen a new governor of the opposite ideological platform replaced an incumbenttherefore ideology is almost perfectly collinear with fixed effects Thus we do not putemphasis on interpretation of coefficients of the ldquoleft-wingrdquo dummy

1311OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

CY

CL

ES

INB

UD

GE

TR

EV

EN

UE

S

INC

OM

E

WA

GE

AR

RE

AR

S

GR

OW

TH

A

ND

INF

LA

TIO

N

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

120

039

001

7

005

3

0

057

000

7

000

6

000

30

003

0

060

(00

28)

(00

24)

(00

21)

(00

85)

(00

14)

(00

03)

(00

05)

(00

07)

(00

71)

mon

th

110

003

0

008

0

019

001

2

000

6

000

7

0

011

0

003

006

3(0

027

)(0

023

)(0

020

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

10

0

038

0

046

0

009

005

5

000

40

001

0

001

0

004

002

2(0

026

)(0

023

)(0

020

)(0

078

)(0

013

)(0

002

)(0

005

)(0

007

)(0

063

)m

onth

9

001

90

036

001

10

161

000

5

000

7

001

0

000

40

084

(00

26)

(00

23)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

51)

mon

th

80

014

002

80

002

009

4

000

8

000

4

000

6

000

7

006

7(0

026

)(0

023

)(0

020

)(0

075

)(0

013

)(0

002

)(0

005

)(0

007

)(0

050

)m

onth

7

0

014

0

023

001

7

010

10

024

000

00

004

0

003

0

056

(00

25)

(00

22)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

47)

mon

th

6

001

40

016

002

8

006

20

010

0

001

0

005

0

008

004

6(0

025

)(0

022

)(0

020

)(0

074

)(0

013

)(0

003

)(0

005

)(0

007

)(0

047

)m

onth

5

003

60

053

001

40

053

000

50

002

0

013

000

6

021

7

(00

24)

(00

22)

(00

20)

(00

75)

(00

13)

(00

03)

(00

05)

(00

07)

(00

47)

mon

th

4

001

4

000

40

000

0

137

0

003

0

004

000

80

002

002

3(0

026

)(0

023

)(0

021

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

055

)

1312 QUARTERLY JOURNAL OF ECONOMICS

mon

th

30

008

003

50

042

003

60

016

0

007

0

008

001

4

005

2(0

026

)(0

022

)(0

020

)(0

078

)(0

014

)(0

003

)(0

005

)(0

007

)(0

058

)m

onth

2

000

80

027

000

3

007

20

007

0

002

000

50

018

001

2(0

026

)(0

022

)(0

020

)(0

078

)(0

013

)(0

003

)(0

005

)(0

007

)(0

060

)m

onth

1

008

0

000

00

052

0

145

0

013

0

005

0

017

0

036

020

7

(00

26)

(00

22)

(00

19)

(00

79)

(00

13)

(00

03)

(00

05)

(00

07)

(00

66)

mon

th0

elec

tion

s

005

0

002

40

010

0

148

000

0

000

50

006

000

6

007

3(0

026

)(0

023

)(0

020

)(0

084

)(0

014

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

1

0

063

000

8

000

7

003

5

000

90

001

0

012

002

2

004

3(0

025

)(0

022

)(0

019

)(0

080

)(0

014

)(0

003

)(0

005

)(0

007

)(0

059

)m

onth

2

0

032

0

047

000

90

042

0

003

000

10

003

0

028

0

057

(00

24)

(00

22)

(00

19)

(00

77)

(00

15)

(00

03)

(00

05)

(00

07)

(00

56)

mon

th

30

010

0

001

000

00

028

001

20

004

000

4

000

50

098

(00

24)

(00

22)

(00

18)

(00

78)

(00

15)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

4

000

50

015

002

9

017

0

000

40

004

0

007

000

6

002

4(0

024

)(0

021

)(0

018

)(0

079

)(0

015

)(0

003

)(0

005

)(0

007

)(0

050

)m

onth

5

0

023

0

013

002

0

013

80

016

000

10

001

001

00

050

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

6

001

20

009

001

30

055

000

3

000

2

000

4

000

80

055

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

70

038

006

0

0

015

001

80

007

0

002

0

013

0

003

0

114

(0

024

)(0

021

)(0

019

)(0

077

)(0

014

)(0

003

)(0

005

)(0

007

)(0

049

)

1313OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

(CO

NT

INU

ED

)

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

80

005

0

004

000

50

031

0

022

0

003

000

00

002

004

4(0

025

)(0

021

)(0

019

)(0

076

)(0

014

)(0

003

)(0

005

)(0

007

)(0

051

)m

onth

9

0

009

0

021

0

014

0

024

000

00

000

000

3

000

10

064

(00

25)

(00

21)

(00

19)

(00

76)

(00

14)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

10

002

8

001

80

011

0

055

0

013

000

20

004

0

001

0

015

(00

25)

(00

21)

(00

19)

(00

76)

(00

15)

(00

03)

(00

05)

(00

07)

(00

57)

mon

th

110

055

0

008

0

004

0

125

0

009

000

5

000

6

000

9

010

1(0

024

)(0

021

)(0

018

)(0

074

)(0

014

)(0

003

)(0

005

)(0

007

)(0

061

)m

onth

12

0

008

002

0

000

5

008

7

001

8

000

1

000

3

000

3

006

6(0

025

)(0

022

)(0

019

)(0

075

)(0

014

)(0

003

)(0

005

)(0

007

)(0

072

)T

erm

inpo

wer

0

001

0

002

000

2

002

90

005

0

004

0

001

000

30

013

(00

11)

(00

09)

(00

08)

(00

33)

(00

05)

(00

01)

(00

02)

(00

03)

(00

24)

Lef

t0

025

007

1

0

038

022

7

000

9

000

1

000

3

001

20

000

(00

30)

(00

27)

(00

26)

(01

60)

(00

12)

(00

02)

(00

05)

(00

06)

(00

00)

Obs

64

9870

6059

1641

1569

4610

731

7970

8002

2588

R2

065

60

584

033

40

631

035

10

818

092

80

851

013

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1314 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

CY

CL

ES

INB

UD

GE

TR

EV

EN

UE

S

INC

OM

E

WA

GE

AR

RE

AR

S

GR

OW

TH

A

ND

INF

LA

TIO

N

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

120

039

001

7

005

3

0

057

000

7

000

6

000

30

003

0

060

(00

28)

(00

24)

(00

21)

(00

85)

(00

14)

(00

03)

(00

05)

(00

07)

(00

71)

mon

th

110

003

0

008

0

019

001

2

000

6

000

7

0

011

0

003

006

3(0

027

)(0

023

)(0

020

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

10

0

038

0

046

0

009

005

5

000

40

001

0

001

0

004

002

2(0

026

)(0

023

)(0

020

)(0

078

)(0

013

)(0

002

)(0

005

)(0

007

)(0

063

)m

onth

9

001

90

036

001

10

161

000

5

000

7

001

0

000

40

084

(00

26)

(00

23)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

51)

mon

th

80

014

002

80

002

009

4

000

8

000

4

000

6

000

7

006

7(0

026

)(0

023

)(0

020

)(0

075

)(0

013

)(0

002

)(0

005

)(0

007

)(0

050

)m

onth

7

0

014

0

023

001

7

010

10

024

000

00

004

0

003

0

056

(00

25)

(00

22)

(00

20)

(00

75)

(00

13)

(00

02)

(00

05)

(00

07)

(00

47)

mon

th

6

001

40

016

002

8

006

20

010

0

001

0

005

0

008

004

6(0

025

)(0

022

)(0

020

)(0

074

)(0

013

)(0

003

)(0

005

)(0

007

)(0

047

)m

onth

5

003

60

053

001

40

053

000

50

002

0

013

000

6

021

7

(00

24)

(00

22)

(00

20)

(00

75)

(00

13)

(00

03)

(00

05)

(00

07)

(00

47)

mon

th

4

001

4

000

40

000

0

137

0

003

0

004

000

80

002

002

3(0

026

)(0

023

)(0

021

)(0

081

)(0

013

)(0

003

)(0

005

)(0

007

)(0

055

)

1312 QUARTERLY JOURNAL OF ECONOMICS

mon

th

30

008

003

50

042

003

60

016

0

007

0

008

001

4

005

2(0

026

)(0

022

)(0

020

)(0

078

)(0

014

)(0

003

)(0

005

)(0

007

)(0

058

)m

onth

2

000

80

027

000

3

007

20

007

0

002

000

50

018

001

2(0

026

)(0

022

)(0

020

)(0

078

)(0

013

)(0

003

)(0

005

)(0

007

)(0

060

)m

onth

1

008

0

000

00

052

0

145

0

013

0

005

0

017

0

036

020

7

(00

26)

(00

22)

(00

19)

(00

79)

(00

13)

(00

03)

(00

05)

(00

07)

(00

66)

mon

th0

elec

tion

s

005

0

002

40

010

0

148

000

0

000

50

006

000

6

007

3(0

026

)(0

023

)(0

020

)(0

084

)(0

014

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

1

0

063

000

8

000

7

003

5

000

90

001

0

012

002

2

004

3(0

025

)(0

022

)(0

019

)(0

080

)(0

014

)(0

003

)(0

005

)(0

007

)(0

059

)m

onth

2

0

032

0

047

000

90

042

0

003

000

10

003

0

028

0

057

(00

24)

(00

22)

(00

19)

(00

77)

(00

15)

(00

03)

(00

05)

(00

07)

(00

56)

mon

th

30

010

0

001

000

00

028

001

20

004

000

4

000

50

098

(00

24)

(00

22)

(00

18)

(00

78)

(00

15)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

4

000

50

015

002

9

017

0

000

40

004

0

007

000

6

002

4(0

024

)(0

021

)(0

018

)(0

079

)(0

015

)(0

003

)(0

005

)(0

007

)(0

050

)m

onth

5

0

023

0

013

002

0

013

80

016

000

10

001

001

00

050

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

6

001

20

009

001

30

055

000

3

000

2

000

4

000

80

055

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

70

038

006

0

0

015

001

80

007

0

002

0

013

0

003

0

114

(0

024

)(0

021

)(0

019

)(0

077

)(0

014

)(0

003

)(0

005

)(0

007

)(0

049

)

1313OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

(CO

NT

INU

ED

)

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

80

005

0

004

000

50

031

0

022

0

003

000

00

002

004

4(0

025

)(0

021

)(0

019

)(0

076

)(0

014

)(0

003

)(0

005

)(0

007

)(0

051

)m

onth

9

0

009

0

021

0

014

0

024

000

00

000

000

3

000

10

064

(00

25)

(00

21)

(00

19)

(00

76)

(00

14)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

10

002

8

001

80

011

0

055

0

013

000

20

004

0

001

0

015

(00

25)

(00

21)

(00

19)

(00

76)

(00

15)

(00

03)

(00

05)

(00

07)

(00

57)

mon

th

110

055

0

008

0

004

0

125

0

009

000

5

000

6

000

9

010

1(0

024

)(0

021

)(0

018

)(0

074

)(0

014

)(0

003

)(0

005

)(0

007

)(0

061

)m

onth

12

0

008

002

0

000

5

008

7

001

8

000

1

000

3

000

3

006

6(0

025

)(0

022

)(0

019

)(0

075

)(0

014

)(0

003

)(0

005

)(0

007

)(0

072

)T

erm

inpo

wer

0

001

0

002

000

2

002

90

005

0

004

0

001

000

30

013

(00

11)

(00

09)

(00

08)

(00

33)

(00

05)

(00

01)

(00

02)

(00

03)

(00

24)

Lef

t0

025

007

1

0

038

022

7

000

9

000

1

000

3

001

20

000

(00

30)

(00

27)

(00

26)

(01

60)

(00

12)

(00

02)

(00

05)

(00

06)

(00

00)

Obs

64

9870

6059

1641

1569

4610

731

7970

8002

2588

R2

065

60

584

033

40

631

035

10

818

092

80

851

013

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1314 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

mon

th

30

008

003

50

042

003

60

016

0

007

0

008

001

4

005

2(0

026

)(0

022

)(0

020

)(0

078

)(0

014

)(0

003

)(0

005

)(0

007

)(0

058

)m

onth

2

000

80

027

000

3

007

20

007

0

002

000

50

018

001

2(0

026

)(0

022

)(0

020

)(0

078

)(0

013

)(0

003

)(0

005

)(0

007

)(0

060

)m

onth

1

008

0

000

00

052

0

145

0

013

0

005

0

017

0

036

020

7

(00

26)

(00

22)

(00

19)

(00

79)

(00

13)

(00

03)

(00

05)

(00

07)

(00

66)

mon

th0

elec

tion

s

005

0

002

40

010

0

148

000

0

000

50

006

000

6

007

3(0

026

)(0

023

)(0

020

)(0

084

)(0

014

)(0

003

)(0

005

)(0

007

)(0

064

)m

onth

1

0

063

000

8

000

7

003

5

000

90

001

0

012

002

2

004

3(0

025

)(0

022

)(0

019

)(0

080

)(0

014

)(0

003

)(0

005

)(0

007

)(0

059

)m

onth

2

0

032

0

047

000

90

042

0

003

000

10

003

0

028

0

057

(00

24)

(00

22)

(00

19)

(00

77)

(00

15)

(00

03)

(00

05)

(00

07)

(00

56)

mon

th

30

010

0

001

000

00

028

001

20

004

000

4

000

50

098

(00

24)

(00

22)

(00

18)

(00

78)

(00

15)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

4

000

50

015

002

9

017

0

000

40

004

0

007

000

6

002

4(0

024

)(0

021

)(0

018

)(0

079

)(0

015

)(0

003

)(0

005

)(0

007

)(0

050

)m

onth

5

0

023

0

013

002

0

013

80

016

000

10

001

001

00

050

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

6

001

20

009

001

30

055

000

3

000

2

000

4

000

80

055

(00

24)

(00

21)

(00

19)

(00

79)

(00

14)

(00

03)

(00

05)

(00

07)

(00

48)

mon

th

70

038

006

0

0

015

001

80

007

0

002

0

013

0

003

0

114

(0

024

)(0

021

)(0

019

)(0

077

)(0

014

)(0

003

)(0

005

)(0

007

)(0

049

)

1313OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

(CO

NT

INU

ED

)

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

80

005

0

004

000

50

031

0

022

0

003

000

00

002

004

4(0

025

)(0

021

)(0

019

)(0

076

)(0

014

)(0

003

)(0

005

)(0

007

)(0

051

)m

onth

9

0

009

0

021

0

014

0

024

000

00

000

000

3

000

10

064

(00

25)

(00

21)

(00

19)

(00

76)

(00

14)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

10

002

8

001

80

011

0

055

0

013

000

20

004

0

001

0

015

(00

25)

(00

21)

(00

19)

(00

76)

(00

15)

(00

03)

(00

05)

(00

07)

(00

57)

mon

th

110

055

0

008

0

004

0

125

0

009

000

5

000

6

000

9

010

1(0

024

)(0

021

)(0

018

)(0

074

)(0

014

)(0

003

)(0

005

)(0

007

)(0

061

)m

onth

12

0

008

002

0

000

5

008

7

001

8

000

1

000

3

000

3

006

6(0

025

)(0

022

)(0

019

)(0

075

)(0

014

)(0

003

)(0

005

)(0

007

)(0

072

)T

erm

inpo

wer

0

001

0

002

000

2

002

90

005

0

004

0

001

000

30

013

(00

11)

(00

09)

(00

08)

(00

33)

(00

05)

(00

01)

(00

02)

(00

03)

(00

24)

Lef

t0

025

007

1

0

038

022

7

000

9

000

1

000

3

001

20

000

(00

30)

(00

27)

(00

26)

(01

60)

(00

12)

(00

02)

(00

05)

(00

06)

(00

00)

Obs

64

9870

6059

1641

1569

4610

731

7970

8002

2588

R2

065

60

584

033

40

631

035

10

818

092

80

851

013

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1314 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

(CO

NT

INU

ED

)

Totalrevenues

Taxrevenues

Deficit

Transfers

Growth

Inflation

Wage

Moneyincome

Growthinwagearrears

mon

th

80

005

0

004

000

50

031

0

022

0

003

000

00

002

004

4(0

025

)(0

021

)(0

019

)(0

076

)(0

014

)(0

003

)(0

005

)(0

007

)(0

051

)m

onth

9

0

009

0

021

0

014

0

024

000

00

000

000

3

000

10

064

(00

25)

(00

21)

(00

19)

(00

76)

(00

14)

(00

03)

(00

05)

(00

07)

(00

52)

mon

th

10

002

8

001

80

011

0

055

0

013

000

20

004

0

001

0

015

(00

25)

(00

21)

(00

19)

(00

76)

(00

15)

(00

03)

(00

05)

(00

07)

(00

57)

mon

th

110

055

0

008

0

004

0

125

0

009

000

5

000

6

000

9

010

1(0

024

)(0

021

)(0

018

)(0

074

)(0

014

)(0

003

)(0

005

)(0

007

)(0

061

)m

onth

12

0

008

002

0

000

5

008

7

001

8

000

1

000

3

000

3

006

6(0

025

)(0

022

)(0

019

)(0

075

)(0

014

)(0

003

)(0

005

)(0

007

)(0

072

)T

erm

inpo

wer

0

001

0

002

000

2

002

90

005

0

004

0

001

000

30

013

(00

11)

(00

09)

(00

08)

(00

33)

(00

05)

(00

01)

(00

02)

(00

03)

(00

24)

Lef

t0

025

007

1

0

038

022

7

000

9

000

1

000

3

001

20

000

(00

30)

(00

27)

(00

26)

(01

60)

(00

12)

(00

02)

(00

05)

(00

06)

(00

00)

Obs

64

9870

6059

1641

1569

4610

731

7970

8002

2588

R2

065

60

584

033

40

631

035

10

818

092

80

851

013

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1314 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

FIG

UR

EII

Est

imat

edP

olit

ical

Bu

dget

Cyc

le

1315OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

measured in percentage deviations from trend around electionsTotal budget expenditures experience the first significant jumpup of 7 percent nine months before elections8 After that there areno significant changes until one month before elections whenexpenditures rise significantly by 13 percent The cumulativeincrease in budget spending during the year prior to electionsamounts to 18 percent (5 percent of monthly regional product)The election month and two months right after elections arecharacterized by a significant fall in budget expenditures of 17percent Spending on education culture and health care exhibitsignificant increases during the two months prior to elections of14 percent on average and significant decreases during the twomonths after elections of about 18 percent Social expendituresrise five months and then one month before elections by 9 and 24percent respectively reaching the level of 31 percent above thetrend In the two months after elections social expenditures comeback to the trend level9 Cycles in public expenditure are sup-ported by intensive use of mass media Media spending increasesby 23 percent during six months preceding elections and drops by32 percent in the month of elections and two postelectoralmonths10 We also find significant budget composition effects of

8 Henceforth the changes in expenditures are calculated by comparisons offitted values net of the federal trend and seasonality (see Figure II for theillustration) This is because the estimates of coefficients of the 25 cycle dummiesindicate the shift in the underlying autoregressive process rather than the shift inexpenditures themselves

9 We verified that standard errors of electoral dummies are estimated con-sistently in the dynamic panel (1) and that the results are not driven by a specificseasonal structure of elections in our sample Following Bertrand Duflo andMullainathan [2004] we conduct a series of estimations of the effect of randomlygenerated placebo elections (with and without holding their seasonal structureequal to the seasonal structure of the actual elections) on total and social regionalexpenditures Significant cycles (ie increases in budget spending before electionsand decreases after) were found in less than 1 percent of the cases

10 To understand the kinds of publicity associated with the budget cycle weread local newspapers and press releases of regional authorities in a few regionsFor six months before elections most local newspapers actively praise incumbentgovernors When a few months are still left before elections they report promisesto pay out arrears in wages and child benefits in the near future (see for instanceldquoSeverInformrdquo [October 21 1999] for elections in Kirov Oblast in March 2000 andldquoAgenstvo Informatsii Udmurtiirdquo [March 24 2000] for elections in Udmurtiya inOctober 2000) Right before elections articles usually claim that all arrears arepaid off (see ldquoKrasnoyarsky Rabochyrdquo [August 20 2002] for elections in Krasno-yarsky Krai in September 2002 Press Release of Tomsk Oblast Administration[August 10 1999] for elections in Tomsk Oblast in September 1999) Occasionallyone comes across articles in the federal press that criticize incumbent governorsand point to sources of funds used for repayment of arrears (eg ldquoKommersantrdquo[March 24 2000] for elections in Pskovskaya Oblast in November 2000) The cyclein publicity around governors is vivid we counted the number of press releasesthat were issued by administration of Sverdlovskaya Oblast On average control-

1316 QUARTERLY JOURNAL OF ECONOMICS

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

elections the share of social expenditures in total spending risesby 14 percent and the share of media expenditures by 23 percentduring the six months preceding elections In the preelectoralmonth shares of spending on social programs and media reachlevels that are 14 and 19 percent higher than their levels in themiddle of the term

Table II presents the effect of elections on budget revenuesgrowth inflation income and wage arrears growth Repaymentsof wage arrears to public workers exhibit strong cyclical patternsRegression results for the growth of wage arrears imply thefollowing level dynamics wage arrears drop by 32 percent in thethree preelection months The cumulative decline in wage arrearsduring the nine months before elections amounts to 42 percentFor six months after elections wage arrears gradually rise but donot reach their initial level The preelectoral rise in public spend-ing is financed in part by increases in nontax and tax revenues inpart by an increase in the deficit and in part by increased federaltransfers (in regions-recipients of federal transfers) The budgetcycle and in particular sizable fluctuations in government wagearrears and social spending drive the dynamics of wages andincome During the four months prior to elections wages andincome rise significantly reaching the levels of 5 and 1 percentabove the trend respectively and return to the trend level in thethree subsequent months

Regional growth does not have a cyclical pattern Inflationsignificantly decreases for six months prior to elections and grad-ually (insignificantly) rises for six months after elections11

Therefore contrary to Nordhausrsquos prediction politicians do notexplore a growth-inflation trade-off to increase their chances ofreelection12

ling for trend and seasonality the number of press releases in the four monthsbefore elections is 40 percent higher than throughout the electoral term

11 The fall in inflation before elections is much smaller in magnitude thanfluctuations in fiscal policies or wages and income Thus our results hold irre-spective of whether we take real or nominal variables As a baseline we reportdynamics of real spending and income the results remain the same when we testfor cycles in nominal expenditures and income disregarding the price differencesbetween regions

12 Keller and May [1984] were the first to argue that one needs to look at thepolitical actions rather than the real economic outcomes to find evidence ofopportunistic cycles based on an analysis of President Nixonrsquos election campaignDrazen [2000] surveyed empirical literature to show that models with adaptiveexpectations a la Nordhaus [1975] are inconsistent with results of virtually allempirical tests cycles if found affect fiscal and monetary policies rather thangrowth or unemployment

1317OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

Budget cycles are short-lived the largest shifts in expendi-tures occur within a month or two of the election date The use ofquarterly and annual data by the previous literature most likelyled to underestimation of cycles because the opposite-sign shiftsin fiscal policies around elections cancel out in the data with lowfrequency To illustrate this point we collapsed monthly series atthe quarterly level and estimated equation analogous to (1) onquarterly data First statistical significance in quarterly paneldata is reduced few expenditure items show significant fluctua-tions around elections Second the magnitude of these fluctua-tions is considerably smaller Figure III presents cycles estimatedwith monthly and quarterly data For the vast majority of expen-diture items deviation from the trend near elections predictedusing quarterly data is less than one-third of the one predicted bymonthly panel data

Overall we find very strong evidence of sizable opportunisticcycles in fiscal policies and no evidence of cycles in economicgrowth The most vivid increases in budget spending occur amonth or two before elections This implies that instruments ofpreelectoral manipulation are observed by voters almost imme-diately This could be achieved only with direct monetary pay-ments to voters Indeed the two main instruments are repaymentof wage arrears and social expenditures comprised of welfarechild benefits veteran allowances social insurance and otherpublic assistance programs Our analysis of local and federalnewspapers and regional governmentsrsquo press releases providesnumerous anecdotes suggesting that cycles in total health careeducation and cultural spending also reflect manipulation withcompensation of medical workers teachers and other govern-ment employees The government wage bill is sufficiently large toshow cycles in aggregate spending For instance in education andhealth care spending it takes up about two-thirds of the totalTherefore we observe the shift of public expenditure (both overtime and across items) toward what is the most visible to votersas predicted by Rogoff rsquos model [1990]13

13 A shift of expenditure toward more visible items was also found by Block[2003] Using an annual panel of developing countries he found that budgetcomposition shifted away from public investment toward current consumption inthe face of elections In contrast Khemani [2000] and Gonzalez [2002] findsignificant preelectoral increases in public investment in Mexico and India Khe-mani however shows that it is road constructionmdashthe most visible type ofinvestmentmdashthat is cyclical

1318 QUARTERLY JOURNAL OF ECONOMICS

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

FIG

UR

EII

IU

nde

rest

imat

ion

inQ

uar

terl

yD

ata

1319OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EII

ID

ET

ER

MIN

AN

TS

OF

TH

EM

AG

NIT

UD

EO

FT

HE

BU

DG

ET

CY

CL

E

PA

NE

LE

ST

IMA

TIO

N

Dependentvariable

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Socialexpenditures

Expendituresonculture

Rst

ands

for

Edu

cati

onU

rban

izat

ion

Dem

ocra

cyT

ran

spar

ency

Med

iafr

eedo

mT

ime

R

Mon

th

30

092

0

107

012

10

051

0

035

000

3

004

80

004

000

20

000

0

019

000

0(0

155

)(0

115

)(0

136

)(0

106

)(0

042

)(0

032

)(0

039

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

2

0

263

0

154

0

031

0

086

0

002

000

1

002

0

002

8

000

6

0

004

0

025

0

031

(0

160

)(0

120

)(0

145

)(0

114

)(0

043

)(0

033

)(0

040

)(0

030

)(0

003

)(0

002

)(0

017

)(0

013

)R

M

onth

1

0

081

0

232

031

2

0

014

0

076

0

034

0

089

001

4

000

7

0

002

0

063

003

2

(01

45)

(01

09)

(01

41)

(01

11)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

R

Mon

th0

0

085

0

193

006

0

007

30

008

001

60

038

001

5

000

1

000

2

000

30

012

(01

45)

(01

08)

(01

43)

(01

06)

(00

43)

(00

33)

(00

39)

(00

30)

(00

03)

(00

02)

(00

17)

(00

13)

Mon

th

30

022

0

002

003

30

002

002

50

008

002

70

008

002

90

002

003

50

006

(00

37)

(00

28)

(00

39)

(00

29)

(00

38)

(00

28)

(00

38)

(00

28)

(00

38)

(00

28)

(00

39)

(00

29)

Mon

th

20

024

001

20

028

001

10

026

001

20

028

001

20

027

001

40

026

000

3(0

037

)(0

028

)(0

039

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

038

)(0

029

)(0

039

)(0

029

)

1320 QUARTERLY JOURNAL OF ECONOMICS

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

Mon

th

10

223

0

127

0

227

0

121

0

216

0

121

0

218

0

121

0

227

0

126

0

197

0

107

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

028

)(0

039

)(0

029

)M

onth

00

038

0

067

0

036

0

063

0

035

0

066

0

036

0

065

0

036

0

067

0

040

0

059

(0

037

)(0

028

)(0

038

)(0

029

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

038

)(0

029

)M

onth

1

0

097

011

5

0

094

011

0

0

102

010

6

0

101

010

6

0

107

010

8

0

090

010

8

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Mon

th

2

006

9

003

4

006

9

003

4

006

9

002

5

006

9

002

5

007

1

002

2

006

9

003

4(0

037

)(0

027

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)(0

037

)(0

028

)M

onth

3

000

3

005

20

005

0

045

001

1

005

10

010

0

051

000

1

004

50

005

0

046

(00

37)

(00

27)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

(00

37)

(00

28)

Lef

t

000

30

000

0

003

000

0

000

3

000

1

000

4

000

1

000

30

000

0

003

0

001

(00

45)

(00

34)

(00

48)

(00

36)

(00

46)

(00

35)

(00

46)

(00

35)

(00

46)

(00

35)

(00

47)

(00

35)

Ter

min

pow

er

001

1

001

2

001

0

001

2

001

2

001

4

001

1

001

4

001

1

001

6

000

7

001

0(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)(0

015

)(0

011

)T

ime

005

8

004

1

(00

10)

(00

07)

Obs

erva

tion

s64

1063

5670

0469

2165

9365

6365

9365

6366

7166

3270

0469

21N

um

ber

ofre

gion

s76

7686

8680

8080

8081

8186

86R

20

596

049

70

560

047

90

584

047

90

584

047

90

581

048

20

561

048

0

All

regr

essi

ons

incl

ude

fou

rla

gsa

full

set

ofti

me

dum

mie

san

dfi

xed

effe

cts

for

each

regi

on-c

alen

dar

mon

thco

mbi

nat

ion

All

depe

nde

nt

vari

able

sar

ein

logs

and

mea

sure

din

real

term

spe

rca

pita

Rob

ust

stan

dard

erro

rsar

ein

pare

nth

eses

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

Th

ere

sult

sar

ero

bust

toth

en

um

ber

ofpr

eele

ctor

aldu

mm

ies

incl

ude

dan

dto

acco

un

tin

gfo

rR

and

Tim

ein

flu

ence

join

tly

orin

sepa

rate

spec

ifica

tion

s

1321OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

IVB Determinants of Opportunistic Cycles

Methodologically the best way to test for the determinants ofcycle magnitude is to run specification (1) with additional regres-sors proxies for possible determinants of the cycle and theirinteraction with the cycle dummies mjit The following potentialdeterminants are considered level of democracy (measured by aregional democracy index) voter awareness (measured by logs ofregional shares of population with higher education and of urbanpopulation) and transparency (measured by region-level indicesof media freedom and government transparency) all of which arepositively correlated These measures are available only as crosssection we denote them by Ri Proxies for voter awareness reflectboth the possibilities of the electorate to get access to information(ie the Internet and TV) which is easier in urban areas and theability of the electorate to process this information that comeswith education We also test how cycle magnitude changes withtime For ease of interpretation of coefficients we rescale time tobe measured in years (Time t12) As discussed below Time isa proxy for voter learning and development of independent mediaand civil society since the countryrsquos democratization but it couldalso reflect increased federal control over Russian regions underPutinrsquos presidency compared with Yeltsinrsquos For presentation pur-poses we report results of estimation of the short specification(1) that preserves the main results of estimating the specificationwith the full set of 25 cycle dummies

(1) yit j33

jmjit j30

13jmjitRiD

j30

jmjitTimetD

Timet L yit1 1Termit 2Leftit t fis εit

Here we look only at budgetary expenditures that exhibit cyclicaldynamics Superscript D indicates that we subtract means fromR and Time before taking cross-terms in this case the coefficientof the respective m is equal to the full effect evaluated at themean values of R and Time Negative significant coefficients atmjitRi and mjitTimet (given that coefficients at respective mjitare positive) serve as evidence that R and Time decrease thecycle magnitude

Table III presents the results of estimation of equation (1)for social expenditures and expenditures on culture The interac-tion terms of voter awareness democracy transparency andtime with cycle dummies two months before elections usually

1322 QUARTERLY JOURNAL OF ECONOMICS

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

have negative and often significant coefficients Thus these vari-ables reduce the magnitude of the cycle For instance the resultsimply that a one standard deviation increase in our measures ofeducation urbanization democracy government transparencyand media freedom leads to a significant reduction in the jump ofsocial expenditures prior to elections of 7 9 10 7 and 8 percent-age points respectively Results for the total budget expendituresand revenues share of social expenditures and spending onhealth care and education are very similar There are no resultsfor media spending and industrial subsidies

The fact that Ri does not vary across time and for two of thefive measures (viz education and urbanization) varies only alittle across regions potentially creates a problem in the estima-tion of this panel because regressors mjitRi and mjit are corre-lated To make sure that our results are not driven by thiscorrelation we carry out a cross-section test We construct thefollowing measure of cycle magnitude the cycle amplitude in aparticular policy instrument for a particular election is defined asthe residual corresponding to the last month before the electionfrom estimation of the following equation run separately for eachregion and each instrument

(2) yit L yit1 j112

jsjt t εit

sjt stands for dummies corresponding to twelve calendar monthst is the real time (L) yit1 is the lag polynomial of the sameorder as in (1) Summary statistics for constructed amplitudes arepresented in Table IV For all fiscal policy instruments meanamplitudes are positive and for eight out of nine they are sig-nificantly different from zero In addition we constructed anaggregate measure of the magnitude of the political budget cycleas the first principal component of the amplitudes in individualfiscal policies with the most profound cycle To test for the deter-minants of the cycle we run the following equation on the pooledcross section of elections

(3) Ai 0 1Ri 2Timei 3Budgeti 4Durationi εi

where i is the ordinal number of elections and Ai is a measure ofcycle amplitude As above Ri is a proxy for awareness democ-racy and transparency and Timei is real time measured in

1323OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

years14 Richer regions may have higher financial slack for ma-nipulation of fiscal policies because the necessary fixed expendi-ture takes up a smaller part of the whole budget We control forthis effect with the regional mean of the size of per capita budgetover the whole periodmdashBudgeti

15 Persson and Tabellini [2003]show that differences in electoral rules explain a part of cross-country variation in political cycles Generally regional electoralrules in Russia are uniform There are potentially significantdifferences only in the duration of a governorrsquos electoral term thatvaries from four to five years with the exception of one region witha seven-year term The duration of term may affect ldquoego-rentsrdquo

14 In cross-section regressions instead of the entire index of democracy weuse a dummy that indicates whether a region has a value of democracy that isabove the median (this proxy gives us a better fit possibly because of poor cardinalproperties of the index) In the estimation of equation (3) we excluded electionsthat took place in nine Autonomous Okrugs that are subdivisions of other largerregions because for the most part R data are unavailable for them elections witha single candidate and elections in which an incumbent did not run for reelectionand did not name his successor Equation (3) has regional-level regressors thuswe allow error terms to be clustered within regions

15 Including Term as a control in cross-section regressions does not changeany of the results It is highly correlated with Time when both are included asregressors Term comes out insignificant but the effect of Time does not changeThe cycle magnitude could also be affected by world oil price because in times ofhigh oil prices governors may be less constrained in preelectoral fiscal manipula-tions A large chunk of budgetary revenues at all levels of government depends ontaxation of oil rents As a robustness check we have controlled for the oil price andoil revenues in all regressions and for their interaction with the cycle dummies inpanel regressions and found that none of our baseline results are driven by thedynamics of oil prices

TABLE IVSUMMARY STATISTICS FOR THE CONSTRUCTED MEASURES

OF THE AMPLITUDE OF THE CYCLES

Obs Mean Median SE Min Max

Amplitude of the cycle inTotal budget expenditures 132 0075 0073 0018 0713 0653Social expenditures 136 0158 0100 0030 0624 1442Education expenditures 136 0088 0082 0017 0517 0581Expenditures on culture 134 0078 0093 0020 0669 0642Health care expenditures 135 0089 0091 0020 0734 0692Media expenditures 131 0089 0068 0039 1219 1346Expenditures on industry 113 0171 0240 0062 1662 1793Deficit 127 0035 0038 0014 0389 0440Negative of wage arrears growth 44 0035 0020 0037 0738 0631

Asterisks mark mean amplitudes that are significantly different from zero

1324 QUARTERLY JOURNAL OF ECONOMICS

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

[Rogoff 1990] increasing the politiciansrsquo incentives for cycles ormay reflect the fact that incumbents had more time for getting ridof political opposition decreasing the need for cycle We use adummy indicating regions with the term above four years as acontrol (Durationi)

16

Table V presents results of the estimation of equation (3) Across-section test confirms findings from the panel regressionsVoter awareness democracy transparency and time reduce cy-cles In all regressions coefficients of these measures are negativeand in more than two-thirds significant A 10 percent increase inthe share of educated population decreases cycle amplitude ineducation cultural and health care expenditures by about 2percentage points A 10 percent increase in urbanization leads toa decrease in cycles in total health care and cultural spending of2 percentage points and 5 in social spending The difference inmagnitudes of cycles in total social cultural and health carespending between regions with democracy above and below themedian is about ten percentage points A standard deviationincrease in the index of government transparency leads to adecrease of cycles in social and total spending of 9 and 3 percent-age points while a standard deviation increase in the mediafreedom index leads to a decrease in cycles in social and culturalexpenditure of 8 and 6 percentage points The first principalcomponent of cycle amplitudes is significantly affected by all themeasures

Panel and cross-section results show that time negativelyaffects the size of the budget cycle As shown in Table V cyclesfade away relatively fast each additional year on average de-creases cycle magnitude by about 3 percentage points We alsoestimated the effect of the number of previous elections as wasdone by Block Ferree and Singh [2003] an additional election ina region significantly reduces the cycle amplitude by 6 to 16percent depending on the expenditure item considered We con-sider two possible interpretations of the negative influence oftime on the magnitude of the cycle the emergence of civil societyand learning by voters as democracy matures and the disciplin-

16 There is one other difference in electoral rules across regions the vastmajority of regions have two-round elections ie a runoff follows the first roundin the case when none of the candidates receive more than half of the votes a fewregions however have single-round elections The number of single-round elec-tions is insufficient to test its influence on cycles Controlling for the number ofrounds does not have any effect on the results

1325OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

DE

TE

RM

INA

NT

SO

FT

HE

MA

GN

ITU

DE

OF

TH

EB

UD

GE

TC

YC

LE

CR

OS

SS

EC

TIO

N

Th

efi

rst

com

pon

ent

ofcy

cle

ampl

itu

des

Am

plit

ude

ofto

tal

expe

ndi

ture

cycl

e

Edu

cati

on

126

9

0

016

(06

21)

(00

81)

Urb

aniz

atio

n

170

6

0

175

(0

515

)(0

065

)D

emoc

racy

0

801

007

6

(02

52)

(00

38)

Gov

tr

ansp

aren

cy

027

2

003

8

(01

45)

(00

18)

Med

iafr

eedo

m

001

90

(00

10)

(00

01)

Tim

e

031

4

0

339

032

3

0

323

032

1

0

034

003

5

0

034

003

4

0

034

(0

068

)(0

067

)(0

066

)(0

066

)(0

067

)(0

008

)(0

009

)(0

008

)(0

008

)(0

009

)A

vre

gion

albu

dget

072

6

088

6

076

5

061

0

060

7

008

3

011

0

009

6

008

2

008

2

(02

68)

(02

79)

(02

75)

(02

98)

(02

71)

(00

37)

(00

37)

(00

35)

(00

37)

(00

35)

Len

gth

ofte

rm

026

1

028

1

024

1

031

1

031

9

001

7

002

9

002

3

003

7

001

4(0

338

)(0

348

)(0

367

)(0

352

)(0

369

)(0

042

)(0

041

)(0

043

)(0

040

)(0

047

)C

onst

ant

281

4

316

0

292

3

290

7

290

0

039

2

042

0

039

5

039

5

039

4

(05

92)

(05

81)

(05

84)

(05

85)

(05

94)

(00

75)

(00

76)

(00

75)

(00

74)

(00

76)

Obs

erva

tion

s11

711

711

711

711

712

112

112

112

112

1R

20

242

025

90

252

023

50

234

015

40

182

017

40

179

015

3

1326 QUARTERLY JOURNAL OF ECONOMICS

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

Am

plit

ude

ofso

cial

expe

ndi

ture

cycl

eA

mpl

itu

deof

edu

cati

onex

pen

ditu

recy

cle

Edu

cati

on

013

9

020

2

(01

17)

(00

79)

Urb

aniz

atio

n

050

8

0

108

(00

94)

(00

74)

Dem

ocra

cy

016

4

0

064

(00

57)

(00

34)

Gov

tr

ansp

aren

cy

009

8

0

017

(00

27)

(00

19)

Med

iafr

eedo

m

000

6

0

002

(00

02)

(00

01)

Tim

e

004

3

0

049

004

3

0

044

004

5

0

031

003

2

0

031

003

1

0

032

(0

016

)(0

017

)(0

016

)(0

016

)(0

016

)(0

008

)(0

008

)(0

008

)(0

008

)(0

008

)A

vre

gion

albu

dget

010

2

016

9

012

3

008

60

094

007

7

007

1

006

8

005

40

056

(00

50)

(00

52)

(00

52)

(00

58)

(00

47)

(00

30)

(00

35)

(00

32)

(00

33)

(00

31)

Len

gth

ofte

rm

000

5

003

3

001

9

005

1

005

0

103

010

2

0

102

010

4

0

109

(0

066

)(0

063

)(0

065

)(0

061

)(0

070

)(0

042

)(0

046

)(0

048

)(0

044

)(0

046

)C

onst

ant

055

7

064

8

056

6

057

0

058

5

038

8

040

2

038

8

038

5

039

1

(01

51)

(01

58)

(01

51)

(01

53)

(01

55)

(00

72)

(00

72)

(00

71)

(00

72)

(00

71)

Obs

erva

tion

s12

512

512

512

512

512

612

612

612

612

6R

20

086

015

50

109

013

30

121

023

60

186

019

10

180

186

1327OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

(CO

NT

INU

ED

)

Am

plit

ude

ofcy

cle

inex

pen

ditu

reon

cult

ure

Am

plit

ude

ofh

ealt

hca

resp

endi

ng

cycl

e

Edu

cati

on

021

3

0

164

(0

061

)(0

072

)U

rban

izat

ion

0

167

0

216

(0

086

)(0

078

)D

emoc

racy

0

117

008

1

(00

43)

(00

39)

Gov

tr

ansp

aren

cy

001

5

002

9(0

022

)(0

023

)M

edia

free

dom

0

004

000

2(0

001

)(0

001

)T

ime

0

042

004

3

0

042

004

2

0

042

003

5

0

037

003

5

0

035

003

6

(00

11)

(00

10)

(00

10)

(00

11)

(00

10)

(00

11)

(00

11)

(00

11)

(00

11)

(00

11)

Av

regi

onal

budg

et0

066

006

90

068

004

20

045

010

5

012

2

010

5

008

70

090

(0

047

)(0

051

)(0

051

)(0

052

)(0

048

)(0

046

)(0

043

)(0

047

)(0

048

)(0

044

)L

engt

hof

term

0

013

0

013

0

016

0

011

0

033

0

094

0

102

009

7

010

4

0

107

(0

049

)(0

052

)(0

050

)(0

055

)(0

051

)(0

048

)(0

046

)(0

049

)(0

046

)(0

052

)C

onst

ant

045

1

047

2

045

2

044

3

045

4

042

4

046

2

042

9

042

8

043

5

(00

90)

(00

85)

(00

88)

(00

90)

(00

88)

(00

96)

(00

94)

(00

95)

(00

93)

(00

96)

Obs

erva

tion

s12

312

312

312

312

312

512

512

512

512

5R

20

195

016

60

184

014

80

190

176

018

10

166

016

016

2

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1328 QUARTERLY JOURNAL OF ECONOMICS

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

ing role of increased central control over regions during Putinrsquosadministration compared with Yeltsinrsquos Although there were noformal changes in electoral rules or authority of regional govern-ments over spending from Yeltsinrsquos to Putinrsquos time one couldargue that Putin has monitored regional governments moreclosely than Yeltsin and increasingly closely throughout his termIn an attempt to separate the two hypotheses we repeat theanalysis separately for the two time periods when each of thepresidents was in power In each subsample the budget cycle issignificant The difference in magnitude of the cycle between thetwo subsamples is 25 times Under Putin the effect of time on thecycle is negative but for the majority of expenditure items insig-nificant under Yeltsin it is essentially zero Thus most of thevariation in the cycle magnitude over time comes from compari-son of the two waves of elections that took place under differentpresidents We cannot rule out the voter learning explanation ofthe effect of time however because insignificance of time withinsubsamples could be attributed to the reduction in the number ofobservations or clustering of election dates17

IVC Do Cycles Help Winning

To test whether cycles helped incumbent governors get re-elected we estimate how the share of votes for the incumbent andthe probability of winning depend on the cycle magnitude con-trolling for an incumbentrsquos ideology performance in the last termand differences in electoral rules The following equation is esti-mated on the pooled cross section of elections

(4) Pi 0 1Ai 2 AiDTimei

D 3Timei 4Lefti

5Urbani 6Performi 7Durationi εi

Pi is the popularity of the incumbent measured by the ratioof votes for the incumbent to the sum of votes for the incumbentand the most popular challenger A (cycle amplitude) TimeLeft and Duration are described above Again we subtractmeans before taking the cross-term to make interpretation of 1easier A positive coefficient at A is an indication that cyclesare associated with an increase in an incumbentrsquos popularity a

17 Clustering of election dates is particularly severe during Yeltsinrsquos time53 percent of governor elections covered by our sample during Yeltsinrsquos presidencyoccurred within four months of each other In the Putin subsample elections aremore spread out over time

1329OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

negative coefficient at the cross-term indicates that this rela-tionship weakens over time We control for the incumbentrsquospast performance (Perform) which can be important in the caseof retrospective voting [Alesina and Rosenthal 1995] In re-ported regressions we control for the relative share of socialexpenditures relative per capita regional industrial productand relative inflation measured as net-of-federal-trend meansover the last term relative to the overall regional means net offederal trend We verified that the main results do not dependon the presence or the choice of proxies for past performanceIn particular we also controlled for relative income relativetaxation level relative total public spending as well as rela-tive growth rates in all of these performance indicators Wealso control for the regional share of urban population Urbanwhich has a very strong direct effect on popularity Cookrsquos[1977] distance and DFITS statistic [Belsley Kuh and Welsch1980] indicate that two observations have excessively strongeffects on the estimates (these are elections with very popularincumbents) we exclude them from the sample18 In additionto OLS regressions for an incumbentrsquos popularity we estimatesimilar binary and ordered probit regressions with the follow-ing outcomes for an incumbent winslosses and winsis thefirst runner-upis below the second place

There is an endogeneity problem in these regressions If anincumbent is certain about the results of an upcoming electionbecause he is either extremely popular or extremely unpopularhe has little incentive for preelectoral manipulations The tighterthe electoral competition the higher the cycle OLS underesti-mates the causal relationship between cycles and popularitywhen incumbents are confident of reelection and overestimatesthe relationship when incumbents are sure of losing We do nothave a good instrument for the cycle amplitude because all vari-ables that sufficiently strongly correlate with it have an indepen-dent-of-the-cycles effect on the popularity of incumbents In thevast majority of cases however incumbents were the most popularcandidates to win the next election and therefore overall we arelikely to underestimate the effect To make sure that we do notoverestimate the effect of cycles on popularity we rerun the regres-

18 As above we exclude from the sample elections with a single candidateelections in which an incumbent did not run for reelection and did not name hissuccessor and elections in nine Autonomous Okrugs Error terms are allowed tocluster within regions

1330 QUARTERLY JOURNAL OF ECONOMICS

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

sions on the subsample of elections in which the incumbent was atleast as popular as the main opponent (for OLS) and on the sub-sample where the opponent is at most 13 times as popular as theincumbent (for OLS and probits) and got very similar results

Table VI presents the results Cycles in social health careeducation and cultural spending generate significant politicalbenefits for incumbent governors while preelectoral expansion oftotal spending has no effect A one-standard-deviation increase inthe cycle amplitude in social education and cultural expendi-tures leads to growth in an incumbentrsquos popularity of 4 5 and 5percentage points respectively In the table we report coefficientsof the ordered probit regressions which translate into the follow-ing marginal effects A one-standard-deviation increase in thecycle in social and education spending leads to an increase in theprobability of the incumbentrsquos win of about 10 and 13 percentagepoints respectively Binary probit results are analogous We donot find any significant weakening of political benefits of cyclesover time in all regressions the coefficient of the interaction termof time and cycle magnitude is insignificant

V CONCLUSION

We tested for existence of opportunistic political cycles andstudied whether voter awareness democracy transparency andtime affect cycles We also examined whether cycles increasegovernorsrsquo chances of reelection The monthly regional panel dataallowed us to define timing of the cycle more precisely than it hasbeen done previously in the literature The key findings are asfollows 1) We find significant political cycles in budget spendingand its composition The average preelectoral increase in totalregional expenditure amounts to about 5 percent of monthly grossregional product and in welfare and other public assistance pro-grams to 043 percent of monthly gross regional product Consis-tent with Rogoff [1990] we observe a shift of public spendingtoward direct monetary payments to voters 2) Previous studieslikely underestimated the budget cycle because quarterly fre-quency of the data is insufficient to measure the cycle preciselyThe underestimation is particularly strong when cycles are short-lived Most sizable manipulations in our sample occur within amonth or two away from an election date We show that use ofquarterly data as opposed to monthly data results in estimatesthat are one-third of the actual deviations from the trend around

1331OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

TA

BL

EV

IE

FF

EC

TO

FC

YC

LE

SO

NP

OP

UL

AR

ITY

OF

INC

UM

BE

NT

SA

ND

PR

OB

AB

ILIT

YT

OW

IN

Incu

mbe

nts

rsquopop

ula

rity

O

LS

Incu

mbe

nt

win

sfi

rst

run

ner

up

lose

sor

dere

dpr

obit

Fir

stco

mpo

nen

tof

ampl

itu

des

003

1

015

5(0

015

)(0

088

)A

mpl

itu

deof

tota

lex

pcy

cle

006

50

087

(01

29)

(07

02)

Am

plit

ude

ofso

cial

exp

cycl

e0

125

0

760

(00

57)

(03

94)

Am

plit

ude

ofh

ealt

hca

reex

pcy

cle

011

20

804

(00

97)

(05

39)

Am

plit

ude

ofed

uca

tion

exp

cycl

e0

234

180

1

(01

27)

(07

02)

Am

plit

ude

ofcy

cle

inex

pon

cult

ure

020

1

081

4(0

074

)(0

505

)T

ime

004

5

003

4

003

5

003

1

003

6

004

0

025

8

019

2

023

5

021

7

024

7

023

4

(00

13)

(00

12)

(00

10)

(00

11)

(00

11)

(00

11)

(00

82)

(00

76)

(00

67)

(00

69)

(00

72)

(00

69)

1332 QUARTERLY JOURNAL OF ECONOMICS

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

Tim

e

resp

ecti

vecy

cle

am

pli

tud

e

000

30

010

024

001

20

015

0

036

0

007

006

50

241

0

013

015

6

019

8(0

007

)(0

057

)(0

026

)(0

050

)(0

064

)(0

043

)(0

034

)(0

257

)(0

169

)(0

249

)(0

263

)(0

256

)L

eft

0

073

0

078

0

099

008

1

008

0

009

1

0

398

0

395

0

499

039

1

037

3

044

4

(00

42)

(00

47)

(00

45)

(00

46)

(00

45)

(00

41)

(02

25)

(02

21)

(02

20)

(02

19)

(02

19)

(02

20)

Len

gth

ofte

rm0

152

0

110

116

012

9

014

6

015

0

087

4

061

0

073

0

080

6

097

3

090

8

(00

64)

(00

69)

(00

63)

(00

62)

(00

61)

(00

59)

(03

37)

(03

09)

(03

10)

(03

12)

(03

16)

(03

38)

Rel

ativ

epr

odu

ct0

366

019

90

318

016

30

307

030

90

843

018

095

082

21

138

071

9(0

218

)(0

221

)(0

213

)(0

214

)(0

218

)(0

195

)(1

192

)(1

118

)(1

107

)(1

265

)(1

219

)(1

107

)R

elat

ive

soci

alex

pen

ditu

resh

are

005

10

092

006

30

043

004

004

0

026

015

026

20

145

000

90

091

(01

09)

(01

15)

(01

04)

(01

07)

(01

02)

(01

04)

(06

17)

(06

13)

(05

44)

(05

74)

(06

00)

(05

51)

Rel

ativ

ein

flat

ion

10

826

1

495

2

754

2

828

4

437

13

678

106

341

6426

274

227

2810

952

19

452

(13

206)

(12

396)

(11

139)

(11

971)

(12

291)

(12

786)

(53

504)

(47

893)

(49

212)

(49

525)

(50

594)

(52

491)

Urb

aniz

atio

n0

306

0

286

0

321

0

309

0

310

0

305

1

685

1

553

1

725

1

680

1

670

1

625

(0

079

)(0

081

)(0

079

)(0

081

)(0

079

)(0

075

)(0

454

)(0

443

)(0

443

)(0

445

)(0

430

)(0

455

)C

onst

ant

019

40

297

0

288

0

320

0

263

0

230

(0

118

)(0

117

)(0

096

)(0

118

)(0

112

)(0

105

)O

bser

vati

ons

115

119

123

123

124

121

115

119

123

123

124

121

R2

026

90

197

023

40

20

232

027

6

Rob

ust

and

adju

sted

for

clu

ster

sw

ith

inre

gion

sst

anda

rder

rors

are

inpa

ren

thes

es

and

de

not

esi

gnifi

can

ceat

the

15

and

10pe

rcen

tle

vel

resp

ecti

vely

1333OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

elections 3) The magnitude of the cycle decreases with educationurbanization level of democracy transparency of the govern-ment and freedom of the media Thus information symmetryand development of democratic institutions are important factorsinfluencing cycles 4) Cycles have become smaller over time Thisis consistent with the view that voters and independent medialearn as democracy matures but could also be explained by achange in Russia-specific institutional factors 5) The scale ofpreelectoral manipulations increases the popularity of incumbentgovernors and probability of getting reelected

We show that maturity of democracy is an important factordetermining the scope for effective use of political cycles cyclesare smaller in more democratic regions This result contrastswith the finding of Gonzalez [2002] that Mexican political cyclesgot stronger with democratization The evidence suggests thatunlike in Mexico during the PRI domination democracy in Russiain 1996ndash2003 reached the point where enforcement of politicalturnover at the regional level was not prohibitively costly and insome regions it even gave rise to an embryo of free press andinstitutions of civil society allowing voters to get informationabout politicians in a less costly way than by means of politicalcycles

1334 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX

PA

NE

LA

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

PO

LIC

YIN

ST

RU

ME

NT

SA

ND

OU

TC

OM

ES

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

axS

ourc

eU

nit

sT

ime

span

P-v

alu

e

Tot

albu

dget

expe

ndi

ture

s73

7045

07

269

261

58

579

606

16M

F$

per

capi

ta03

199

6ndash07

200

30

42S

ocia

lex

pen

ditu

res

7522

290

210

305

018

293

6M

F$

per

capi

ta01

199

6ndash07

200

30

62S

har

eof

soci

alex

pen

ditu

res

7244

783

730

376

117

298

3M

FP

erce

nta

ge01

199

6ndash07

200

30

58E

duca

tion

expe

ndi

ture

s75

289

286

1210

24

129

894

0M

F$

per

capi

ta01

199

6ndash07

200

30

78E

xpen

ditu

res

oncu

ltu

re75

121

070

681

280

1211

31

MF

$pe

rca

pita

011

996ndash

072

003

065

Hea

lth

care

expe

ndi

ture

s75

245

954

126

050

8350

71

MF

$pe

rca

pita

011

996ndash

072

003

065

Med

iaex

pen

ditu

res

7332

018

008

035

000

351

MF

$pe

rca

pita

031

996ndash

072

003

096

Sh

are

ofm

edia

expe

ndi

ture

s72

100

370

300

260

021

76M

FP

erce

nta

ge03

199

6ndash07

200

30

40E

xpen

ditu

res

onin

dust

ry71

812

230

497

130

0090

90

MF

$pe

rca

pita

011

996ndash

072

003

093

Tot

albu

dget

reve

nu

es73

5545

35

273

060

60

585

519

08M

F$

per

capi

ta01

199

6ndash07

200

30

27T

axre

ven

ues

7542

257

616

72

340

81

5432

062

MF

$pe

rca

pita

011

996ndash

072

003

040

Rat

ioof

expe

ndi

ture

sto

reve

nu

es70

9410

010

984

723

83

434

726

637

MF

Per

cen

tage

031

996ndash

072

003

062

Gro

wth

7826

000

030

0005

002

56

017

015

GK

SP

erce

nta

ge02

199

5ndash07

200

30

56In

flat

ion

1124

60

190

120

23

018

168

GK

SP

erce

nta

ge02

199

2ndash07

200

30

26R

egio

nal

wag

ear

rear

s38

700

0064

000

120

0220

000

000

25G

KS

$pe

rca

pita

011

999ndash

072

003

076

Wag

ele

vel

8410

164

6113

109

104

0548

21

671

72G

KS

$pe

rca

pita

021

995ndash

072

003

009

Mon

eyin

com

e84

6312

524

102

6576

32

320

056

603

GK

S$

per

capi

ta01

199

5ndash07

200

30

38

ldquo$

per

capi

tardquo

stan

dsfo

rco

nst

ant

1997

US

doll

ars

per

capi

ta

P

-val

ues

are

for

the

test

ofre

sidu

alau

toco

rrel

atio

nin

equ

atio

n(1

)se

efo

otn

ote

6

1335OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

AP

PE

ND

IX(C

ON

TIN

UE

D))

PA

NE

LB

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

EL

EC

TIO

NS

Var

iabl

eN

oof

obs

Mea

nM

edia

nS

DM

inM

ax

Du

mm

yfo

rpa

rtic

ipat

ion

ofin

cum

ben

t21

40

901

031

01

Du

mm

yfo

rin

cum

ben

tsrsquow

in19

20

651

048

01

Du

mm

yfo

rin

cum

ben

tco

min

gin

thir

dor

wor

se19

20

060

023

01

Per

cen

tage

ofvo

tes

for

incu

mbe

nt

192

539

856

23

228

44

7699

9P

erce

nta

geof

vote

sfo

rm

ain

com

peti

tor

ofin

cum

ben

t18

132

30

285

020

87

071

82P

erce

nta

geof

vote

sfo

rw

inn

er21

462

28

592

815

06

235

999

Per

cen

tage

ofvo

tes

for

the

firs

tru

nn

er-u

p20

224

09

242

312

98

071

48N

um

ber

ofca

ndi

date

s21

45

695

296

116

PA

NE

LC

D

ES

CR

IPT

IVE

ST

AT

IST

ICS

OF

DE

TE

RM

INA

NT

SO

FC

YC

LE

MA

GN

ITU

DE

Var

iabl

eR

egio

ns

Mea

nM

edia

nS

DM

inM

ax

Log

ofed

uca

tion

in19

9576

278

275

026

219

368

Log

ofsh

are

ofu

rban

popu

lati

on86

417

423

028

291

461

Inde

xof

free

dom

ofm

edia

prod

uct

ion

8136

25

3714

31

075

Inde

xof

dem

ocra

cy77

296

30

871

5In

dex

ofgo

vern

men

ttr

ansp

aren

cy77

299

30

931

5D

um

my

for

len

gth

ofte

rmab

ove

4ye

ars

880

261

044

01

1336 QUARTERLY JOURNAL OF ECONOMICS

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH

CENTER FOR ECONOMIC AND FINANCIAL RESEARCH AND CEPR

REFERENCES

Alesina Alberto Nouriel Roubini and Gerald D Cohen Political Cycles and theMacroeconomy (Cambridge MA MIT Press 1997)

Alesina Alberto and Nouriel Roubini ldquoPolitical Cycles in OECD EconomiesrdquoReview of Economic Studies LXIX (1992) 663ndash688

Alesina Alberto Roberto Perotti and Jose Tavares ldquoThe Political Economy ofFiscal Adjustmentsrdquo Brookings Papers on Economic Activity 1 (1998)197ndash266

Alesina Alberto and Howard Rosenthal Partisan Politics Divided Governmentand the Economy (Cambridge UK Cambridge University Press 1995)

Ames Barry Political Survival Politicians and Public Policy in Latin America(Berkeley CA The University of California Press 1987)

Arellano Manuel and Stephen Bond ldquoSome Tests of Specification for Panel DataMonte-Carlo Evidence and an Application to Employment Equationsrdquo Reviewof Economic Studies LVIII (1991) 277ndash297

Belsley David A Edwin Kuh and Roy E Welsch Regression Diagnostics Iden-tifying Influential Data and Sources of Collinearity (New York NY Wiley1980)

Berger Helge and Ulrich Woitek ldquoSearching for Political Business Cycles inGermanyrdquo Public Choice XCI (1997) 179ndash197

Besley Timothy and Anne Case ldquoDoes Political Accountability Affect EconomicPolicy Choices Evidence from Gubernatorial Term Limitsrdquo Quarterly Jour-nal of Economics CX (1995) 769ndash798

Bertrand Marianne Esther Duflo and Sendhil Mullainathan ldquoHow MuchShould we Trust Differences-in-Differences Estimatesrdquo Quarterly Journal ofEconomics CXIX (2004) 249ndash275

Block Steven A ldquoElections Electoral Competitiveness and Political BudgetCycles in Developing Countriesrdquo Harvard University CID Working Paper No78 2001

mdashmdash ldquoPolitical Business Cycles Democratization and Economic Reform TheCase of Africardquo Journal of Development Economics LXVII (2002) 205ndash228

mdashmdash ldquoElections and the Composition of Public Spending in Developing Coun-triesrdquo Working paper Fletcher School of Law and Diplomacy Tufts Univer-sity 2003

Block Steven A Karen Ferree and Smita Singh ldquoMultiparty CompetitionFounding Elections and Political Business Cycles in Africardquo Journal of Afri-can Economics XII (2003) 444ndash468

Brender Adi ldquoThe Effect of Fiscal Performance on Local Government ElectionResults in Israel 1989ndash1998rdquo Journal of Public Economics LXXXVII (2003)2187ndash2205

Brender Adi and Allan Drazen ldquoWhere Does the Political Budget Cycle ReallyCome fromrdquo CEPR Discussion Paper No 4049 2003

Cook R Dennis ldquoDetection of Influential Observations in Linear InferencerdquoJournal of Statistical Planning and Inference XXXVII (1977) 51ndash68

Drazen Allan ldquoThe Political Business Cycle After 25 Yearsrdquo NBER Macroeco-nomic Annual (Cambridge MA MIT Press 2000)

Drazen Allan and Marcela Eslava ldquoThe Political Budget Cycle in ColumbiardquoWorking paper University of Maryland 2003

Franzese Robert ldquoElectoral and Partisan Cycles in Economic Policies and Out-comesrdquo Annual Review of Political Science V (2002) 369ndash421

Gonzalez Maria de los Angeles ldquoOn Elections Democracy and MacroeconomicPolicy Evidence from Mexicordquo Working Paper Department of EconomicsPrinceton University 2000

mdashmdash ldquoDo Changes in Democracy Affect the Political Budget Cycle Evidence fromMexicordquo Review of Development Economics VI (2002) 204ndash224

Hansen Lars P ldquoLarge Sample Properties of Generalized Method of MomentsEstimatorsrdquo Econometrica L (1982) 1029ndash1054

1337OPPORTUNISTIC POLITICAL CYCLES

Heckelman Jac C and Hakan Berument ldquoPolitical Business Cycles andEndogenous Electionsrdquo Southern Economic Journal LXIV (1998)987ndash1000

Ito Takatoshi ldquoThe Timing of Elections and Political Business Cycles in JapanrdquoJournal of Asian Economics I (1990) 135ndash156

Keller Robert A and Ann M May ldquoThe Presidential Political Business Cycle of1972rdquo The Journal of Economic History XLIV (1984) 265ndash271

Khemani Stuti ldquoPolitical Cycles in a Developing Economy Effect of Elections inthe Indian Statesrdquo Working paper World Bank 2000

Klein Michael W ldquoTiming Is All Elections and the Duration of United StatesBusiness Cyclesrdquo Journal of Money Credit and Banking XXVIII (1996)84ndash101

Krueger Anne O and Ilter Turan ldquoThe Politics and Economics of Turkish PolicyReform in the 1980rsquosrdquo in Political and Economic Interactions in EconomicPolicy Reform Evidence from Eight Countries Robert H Bates and Anne OKrueger eds (Oxford UK Basil Blackwell 1993)

Nickell Stephen ldquoBiases in Dynamic Models with Fixed Effectsrdquo EconometricaXLIX (1981) 1417ndash1426

Nordhaus William D ldquoThe Political Business Cyclerdquo Review of Economic Stud-ies XLII (1975) 169ndash190

Persson Torsten and Guido Tabellini Macroeconomic Policy Credibility andPolitics (Amsterdam The Netherlands Harwood Academic Publishers 1990)

Persson Torsten and Guido Tabellini ldquoDo Electoral Cycles Differ across PoliticalSystemsrdquo Universita Bocconi Working Paper No 232 2003

Petrov Nikolai Russiarsquos Regions in 1999 an Annual Supplement to RussiarsquosPolitical Almanac Nikolai Petrov and Michael McFaul eds (Moscow Car-negie Endowment for International Peace 2001)

Reid Bradford G ldquoEndogenous Elections Electoral Budget Cycles and CanadianProvincial Governmentsrdquo Public Choice XCVII (1998) 35ndash48

Rogoff Kenneth ldquoEquilibrium Political Budget Cyclesrdquo American Economic Re-view LXXX (1990) 21ndash36

Rogoff Kenneth and Anne Sibert ldquoElections and Macroeconomic Policy CyclesrdquoReview of Economic Studies LV (1988) 1ndash16

Schuknecht Ludger ldquoFiscal Policy Cycles and Public Expenditure in DevelopingCountriesrdquo Public Choice CII (2000) 115ndash130

Shi Min and Jakob Svensson ldquoConditional Political Budget Cyclesrdquo CEPRDiscussion Paper No 3352 2002a

Shi Min and Jakob Svensson ldquoPolitical Business Cycles in Developed and De-veloping Countriesrdquo mimeo 2002b

White Halbert ldquoInstrumental Variables Regression with Independent Observa-tionsrdquo Econometrica L (1982) 483ndash499

1338 QUARTERLY JOURNAL OF ECONOMICS