1
Clientelism-Augmented Political Budget Cycles. Evidence from Argentina
Osvaldo Meloni* Universidad Nacional de Tucumán, Argentina
“Cuando yo era chico Perón era nuestro Rey Mago: el 6 de enero bastaba con ir al correo para que nos dieran un
oso de felpa, una pelota o una muñeca para las chicas. Para mi padre esa era una vergüenza: hacer cola delante
de una ventanilla que decía "Perón cumple, Evita dignifica", era confesarse pobre y peronista.”
Osvaldo Soriano Aquel Peronismo de juguete Cuentos de los años felices
Editorial Norma, 1999
Abstract
This paper incorporates Clientelism, a distinctive feature of Argentine politics, into the analysis of Political Budget Cycles (PBC) across provinces. We argue that organizing and keeping clienteles modify the pattern of expenditures through time of the typical PBC. Likewise, clientelism, by favoring spending in private over public goods, alters the expected PBC expenditure mix in the proximity of elections. Our estimations of a dynamic panel data reveal that clientelism keeps total public expenditure high one and two years before election years. We found no evidence of changes in the expenditure mix in election years: incumbent parties benefit clienteles through current expenditures.
Key Words: Political Budget Cycle, Clientelism, expenditure mix.
JEL Classification Codes: D72, P16.
* I thank Víctor Elías and participants of the Economics Workshop at the University of Tucumán for comments and suggestions to an earlier version of this paper. Álvaro Domínguez, Ana Lucía Montes and Martín Perdomo provided superb research assistance. All remaining errors are mine. I gratefully acknowledge the support of the CIUNT Grant F 308 and PICT 2004, Project 21226. The usual disclaimer applies.
2
I. Introduction
Political Budget Cycles (henceforth, PBC) are an emblematic feature of young and emergent
democracies. This is one of the undisputable facts ensuing from the copious empirical
investigation. Whether fostered by low transparency fiscal policy, which enlarged the country
list with some developed ones, as shown by Alt and Lessen (2005) in a recent study; or
encouraged by inexperienced voters from young democracies as in Brender and Drazen
(2005), the result remains the same: opportunistic incumbents increase public outlays or
fiscal deficit or even change the expenditure mix to improve their chances of continuing in
office.
Another unquestionable fact steaming mostly from the comparative politics literature is that
clientelism dominates the political scene in young and emergent democracies. There are
abundant case studies, with particular focus on Latin America, reporting a variety of
clientelistic practices aimed at acquiring and preserving political power. Clientelism is carried
out in general, but not exclusively, by incumbents. Opposition parties also take clientelistic
actions to get votes from targeted constituencies. Typically, incumbents underprovide public
goods (secure property rights, access to information, universal education, health, etc.) and
overprovide private transfers (jobs, subsidies, goods and services, etc.) to narrow groups in
exchange for support in ballots and in various political activities.
Remarkably, the literature on comparative politics relates clientelism to elections but rarely
connects it with PBC. Similarly, PBC papers usually leave aside clientelistic considerations.
Our paper attempts to link both literatures by examining the costs of organizing, strengthen
and keeping a clientele (supply side) and the factors determining the demand for clientelism.
We center our attention on Argentina, where clientelism is perceived as a prevalent feature in
local politics.
We claim that the demand for clientelism is encouraged, among other factors, by the
inefficiency of the public sector (either national, provincial or municipal) to deliver public
goods, in particular to poor people, and by the misguided incentives of the argentine tax-
sharing system. In some provinces the electoral law (chiefly the double simultaneous voting
system) also helps the proliferation of clienteles. The inefficiency of the public sector gives
birth to private intermediaries that funnel goods and services “produced” by the government
mainly to the poor people. The intermediaries, called punteros políticos”, trade support for
themselves and/or the incumbent in elections and various political events for the access to
streams of goods and services.
3
In Argentine provinces there are quasi-permanent clientelistic organizations with heavy
operations during election years but also active during non-election years. Actually, it takes
time and abundant resources to generate and preserve a patron- client relationship. We
hypothesize that clientelism modifies the usual PBC pattern described by the main theories
at stake. As a general rule, PBC entail “normal” outlays before elections and either a peak in
expenditure and fiscal deficit or a change in the expenditure mix, favoring “visible” outlays, in
election years. We conjecture that clientelism induces increases in public expenditures and
reduces balance budget long before the election year. The reason is simple: keeping a
clientele requires a continuous stream of public expenditure. Moreover, contrary to the
prescriptions of the Rogoff (1990) model and the findings of Eslava and Drazen (2005) for
Colombia, we speculate that incumbent governments will secure their clientele by “investing”
in private transfers such as jobs and subsidies rather than “spending” in infrastructure.
The objective of the paper is twofold. Firstly, we test the existence of a clientelistic-
augmented PBC in argentine provinces by estimating dynamic panel data for several fiscal
policy variables. The panel spans 24 districts and 21 years, from 1984 to 2004. Secondly, we
examine the fiscal policy instrument used by governors of a federal emergent democracy to
enhance their reelection prospects.
This paper proceeds as follows. The next section presents a theoretical discussion on PBC
and clientelism. Section III summarizes the empirical literature, particular the emergent
democracies experience, while section IV looks briefly at the fiscal behavior of Argentina.
Section V describes the empirical investigation as well as the results. Section V concludes.
II. Demand and Supply of Clientelism and PBC
Juan Carlos Mamaní is a Council member of the city of San Miguel de Tucumán, the largest
city of the poor North West region of Argentina. In April 2008 Mamaní hit the front cover of
local newspapers after recognizing that he operates a “small parallel municipality”1. Mamaní
employs 48 “political advisors” to provide some basic services to targeted poor neighbors
such as weeding out sidewalks and streets and changing streets light bulbs. This was a
small scale scandal compared to the 5,000 “advisors of the forty-member local Legislature or
the 2,000 temporary workers of local communities (“comunas”) during 2003-2007 period2. In
local politics jargon “advisor” is a euphemism for client3.
The market for clientelism features the exchange of valuable goods and services for
politicians, (such as votes and support in political rallies and street propaganda and even
1 See La Gaceta, April 11, 12 and 13th, 2008. 2 See La Nación, February 3, 2008. 3 Total provincial public sector employees are 53,000 (estimated in December 2007).
4
acting as shock-troops), for a large variety of goods valuable for poor voters, for instance
food, medicines, scholarship for education purposes: The trade also involves services like
helping the poor circumventing laws and regulations, or progressing in waiting lists and
skipping procedures to get public certifications.
Clientelism is positively associated with poverty, income inequality and inefficiencies of the
public sector, and negatively related to political competition and transparent fiscal policies.
The usual assumption of diminishing marginal utility of income explains the link between
clientelism and poverty: lower income voters are more sensitive to clientelistic transfers than
higher income voters. The relationship can also be accounted for by relying on risk aversion:
poor people value more a private transfer today than the promise of redistributive public
policy tomorrow (See Stokes, 2007). Some authors argue (see Robinson and Verdier, 2002)
that income inequality rather poverty promotes clientelism. They sustain that, given the
income level, clientelistic redistribution gets the society’s support as income distribution
impairs.
We claim that the demand for clientelism also depends on the inefficiency of the public
sector. Emergent democracies feature inefficient (and sometimes ineffective and corrupt)
public sector, particularly in redistributive matters. This is one reason why income distribution
usually takes the inefficient form of clientelism. Simply, government is more inefficient and
barely effective. The excess demand for good and services to alleviate poverty generated by
public sector incompetence creates a market opportunity for intermediaries, called
“punteros”, which funnel goods and services in exchange for support for themselves (acting
as dealers) or local party bosses in elections and various political events4. Intermediaries
transform a non accessible or scarce public good into a private good for their clients.
The supply of clientelism depends on the costs of providing private goods financed with
public funds. Hence rules and practices inhibiting accountability and transparency will favor
clientelistic exchanges. Stokes (2005) coined the term “perverse accountability” to describe
the situation where machine politics influences how people vote by threatening to punish
them for voting to another party. She claims that Argentina at the beginning of the 20th
century fits the description of that term.
On the other hand, some rules and institutions such as the tax-sharing agreement and the
electoral law (chiefly the double simultaneous voting system) also help the proliferation of
4 Stringent civil service laws regarding work conditions, red tape and extended licenses are some of the factors explaining public sector inefficiency. For example, floods during summertime when most of the public sector employees have their annual mandatory license creates an opportunity for “punteros” to deliver goods and services to the victims.
5
clienteles. The tax-sharing agreement (called “coparticipación”) generates a common pool of
tax revenues that is distributed among the federal government and the provinces according
to fixed percentages, encouraging governors to extract revenues from the federal
government rather than taxing local activities and local asset owners5. The tax-sharing
system decouples provincial income and expenditures, thus hurting accountability. Instead of
controlling how the incumbent use public funds, key local economic actors and the electorate
reward the amount of resources pulled out from the common pool because their prosperity
depends more on them than on austere fiscal policy6.
The double simultaneous voting system, in force in some provinces during the period under
study, fosters the proliferation of brokers that sells the votes of “their” own constituency to
major candidates. It is in each candidate best interest to encourage as many brokers as
possible to sum votes for her/his appointment as governor.
Similarly, fragile political competition also encourages the supply of clientelism. Magaloni et
al. (2001) find that, controlling for levels of development, clientelism is less prevalent where
there is more political competition. The authors develop a portfolio theory of electoral
investment to explain under what conditions politicians attempt to buy votes through the
provision of private goods rather than non-excludable public goods. They view clientelism as
a political investment strategy to deter voter exit and hedging electoral risk when more
investment in public goods is required to win elections. In an empirical study with data from
selected argentine municipalities, Weitz-Shapiro (2003) tests and confirms the hypothesis
that increases in political competition decrease the likelihood of clientelism in cities with low
levels of poverty. In contrast, where poverty is high, high levels of political competition are
associated with a high probability of clientelism.
The impact of clientelism on PBC
An important part of the literature on clientelism focuses on exchanges around elections
between local party bosses and clienteles. Studies on vote buying such as Hicken (2007)
and Schaffer (2002) have contributed to understand the rules, institutions and political
environment that facilitate this kind of trade. Likewise, empirical applications to the Argentine
case as in Brusco et al. (2004) shed light on the extent of clientelism and the characteristics
of the “vote sellers”. Though centering the attention on vote buying is insightful, it misses a
central point: the patron-clientele exchanges largely exceeded the election year. Vote buying
is just one dimension of clientelism concentrated in a limited time span, usually few days
5 Notice that this mechanism adds uncertainty to provincial public finances, since it is not easy to project future allocations. 6 For an extensive analysis of the federal tax sharing agreement see Spiller and Tommasi (2007)
6
before or on the Election Day. “Selling” votes is possibly the most important clientelistic
transaction but it is not the only. A large array of goods and services are traded long before
elections. This basic feature of the relationship has implications on the costs and timing of
organizing and maintaining a clientele.7 Organizing a clientelistic-based winning electoral
machine takes time and fiscal resources8. Leaving the preparation of a clientele to the
election year would be a serious risk that no boss or broker would dare to run. Obtaining the
loyalty of clienteles requires active regular trade. For example, there are several events and
rallies involving clientelistic exchange throughout the year such as Epiphany day (January),
Labor day (May), National Day (May), Father’s day (June), Independence day (July),
Children Day (August), Mother’s day (October). Notice that the clientelistic structure devised
by most of political parties in Argentina solves the problem faced by any political organization
of keeping alive the party in off-election periods.
Uncertainty about the availability of resources is another reason that induces bosses and
brokers to gain clientele loyalty as soon as they can, even if far from election year. Recurrent
crises and recessions have taught politicians to invest in advance in organizing and
consolidating clienteles to avoid being surprised by shocks that may deprive them from
“enough” resources during elections years. Hence, depending on the phase of the business
cycle, opportunity rather than opportunism may be the driving force of public expenditure.
Interestingly, during crises and recessions opportunism may come from the demand side:
poor people claim for higher amounts of goods and services as approaching elections,
leading to a peak in expenditures during election years.
Supply of clientelism: upstream and downstream actors
A frequent and useful assumption when discussing clientelism is that local party bosses
supply clienteles directly with no intermediaries. Although this simplifying assumption clarifies
the nature of the exchanges involved, it hides costs associated with brokerage, coordination
and competition among brokers. In Argentina the supply of clientelism habitually features
vertical relationships with upstream (bosses) and downstream actors (brokers) which imply a
more complex and costly structure. Some brokers are mere agents or distributors of the party
boss but some other have a certain degree of independence based on their accessibility of
resources to exchange, thus they act as brokers of brokers. Generally, the availability of
resources is linked to elected positions. Local party bosses and brokers from the incumbent
7 Actually, this is what the word client means in business usage. Client comes from the Latin word cliens, which in turn descend from clinare "to incline, bend," from suffixed form of PIE base *klei- "to lean" (see lean (v.)). The ground sense is of one who leans on another for protection. In ancient Rome, a plebian under protection of a patrician (in this relationship called patronus-Clientele). 8 Kydland and Prescott (1982) in their celebrated time-to-built model assume that investment projects take four periods to complete.
7
party are typically public officials, legislators or city council members with some power to get
around laws and regulations, to distribute largesse or to favor their clientele with public jobs
and temporary contracts. We claim that large and vertical related clientelistic apparatus entail
intermediation and coordination costs with unequivocal influence on total expenditures and
expenditure mix. Exerting control over a growing vertical related structure requires increasing
amounts of fiscal revenues. Moreover, bosses may find expensive allocate exclusive
territories among brokers which generates costs associated with competition for a given
territory.
Summarizing, we argue that organizing, strengthen and keeping a clientele requires large
amount of resources spent in private transactions various periods in advance the election
year, hence modifying the usual PBC pattern.
III. Previous literature
The PBC and the clientelism literature arise basically from observations on the behavior of
incumbents in the proximity of elections. While the former tries to explain their fiscal
performance the latter their redistribute policy. Theoretical PBC models, whether driven by
adverse selection à la Rogoff and Siebert (1988) or spurred by moral hazard considerations
as in Persson and Tabellini (2000), conclude that incumbents engage in pre-election
profligacy to influence voters and maximize chances to remain in office. There are also
ideological reasons that could explain budget manipulation before ballots. Polarized political
environments result in political budget cycles as well.
The empirical investigations from cross countries as well as from country studies support the
existence of PBC. Schuknecht (2000) analyzing the fiscal performance of 24 developing
countries for the period 1973-1992 finds that incumbents try to influence the electorate
augmenting public expenditures rather than lowering taxes. Moreover, his estimations
support the very well hypothesis that government’s favorite instrument to expand fiscal policy
is capital spending (public works programs).
Brender and Drazen (2005) using a broader data set of 106 countries, including developing
as well as developed countries, for the period 1960-2001 find a political deficit cycle mainly
driven by the behavior new democracies. Actually, when new democracies are removed from
the sample the PBC disappears. The authors conjecture that voters in new democracies
have neither the fiscal information nor the ability to process it correctly. Conversely, in
established democracies relevant data to evaluate economic policy are readily available to
experienced voters that interpret that information correctly. However, a recent study by Alt
and Lessen (2005) using a sample of 19 OECD countries in the 1990s, shows that rather
8
than an intrinsic characteristic of specific group of countries, PBC is determined by the
degree of transparency of the fiscal policy. Alt and Lassen show that PBC also come out in
developed countries showing low transparency fiscal policy.
Country studies, particularly cases of developing countries, verify PBC too. The analysis of
Gonzalez (2000) on Mexico’s fiscal policy between 1957 and 1997 revealed that the federal
government made systematic use of public spending in infrastructure and current transfers
as a means to earn votes. Moreover, Gonzalez sustains that the magnitude of the election
cycle was exacerbated during the country’s most democratic episodes. In their work on the
Russian Federation, Ahkmedov et al. (2004) find strong evidence of opportunistic PBC in
regional elections. Russian cycles have some remarkably features: (a) they are short-lived:
most sizable cyclical changes in expenditures happen within a month away from elections.
(b) Their magnitude decrease with voter’s rationality and awareness. Supposedly, the
electorate learns with time.
Similarly, Drazen and Eslava (2005) obtain PBC in their study of Colombian municipalities.
They classify government expenditures in targeted (mainly infrastructure) and non-targeted
(purchases of supplies and services, payment to other government entities, etc.). They find
evidence that voters reward incumbents who increase investment expenditures but only to
the extent that they do so without running large election year deficits. In their model, voters
may dislike fiscal deficits even though they value government spending on some specific
types of goods. Politicians are aware of voters’ preferences and hence manipulate the
budget favoring targeting spending instead of increasing the overall size of the budget.
As reported by Goncalves Veiga & Veiga (2007), majors of Portuguese mainland
municipalities also behave opportunistically in pre-electoral periods increasing total
expenditures and changing their composition favoring items that are highly visible to the
electorate.
There is also evidence of PBC in developed countries. In their study on the fiscal policy of
German lánder, Galli and Rossi (2002) conclude that upcoming elections affect not only total
local spending and various categories of spending but also fiscal deficit. Interestingly, they
show an alternative empirical specification including dummy variables for post electoral and
pre electoral years with estimated negative coefficients indicating that spending is not
systematically cut in the post election year and progressively in the following year. As will be
apparent in section V, this result is relevant for our paper.
Leaving aside ideological reasons, it is agreed that politicians all over the world are prone to
expand fiscal policy to enhance their re-election prospects. It is equally accepted that
9
incumbents from low transparency countries will actually end up expanding fiscal policy while
incumbents facing strong accountability will not incur in such practices to avoid being
punished at the ballots. PBC flourish only if the electorate cannot observe fiscal policy (or
observe with some delay). The evidence suggest that political budget cycles appear to be a
general pattern in all countries with elections and that the size of election –induced budget
cycles are greater in developing countries.
Clientelism
The literature on clientelism is plagued with case studies reporting patron-client transactions
in emerging and young democracies. Most of them concentrate on the objects of exchange.
For instance, Robinson and Verdier (2002) analyze the trade of public jobs for votes. They
view clientelism as the result of self-enforcing exchange relationship between politicians and
voters. In this context, a job, by tying the continuation utility of a voter to the political success
of a particular politician, is an optimal object of exchange. They sustain that providing jobs for
votes becomes a relatively attractive political strategy in situations with high inequality and
low productivity.
There are also numerous studies focusing on the factors explaining the demand and supply
for clientelism. Papers emphasizing demand variables generally investigate how differences
in income levels and diverse socioeconomic characteristics result in different demands for
clientelism. In particular, the relationship between poverty and clientelism has extensively
and closely examined. Stokes (2007) provides a useful survey. Alternatively, papers
accentuating supply factors predominantly study the connection between political competition
and clientelism. As an example, Ortega and Penfold-Becerra (2006) look into the case of
Venezuela under Chavez’s rule, finding that clientelism is more ubiquitous in districts with
lower levels of political competition. They sustain that incumbent politicians invest important
resources to monitor those targeted voters in order to solve the dilemmas inherent to the
incomplete contracts underlying vote-buying exchanges.
Are there parties more inclined to offer clientelistic exchange than others? Calvo and Murillo
(2004) putting their spotlight on Argentina, demonstrate that political parties whose
constituents are more dependent on public largesse have higher electoral returns. They
show the existence of partisan effects on both the access to, and the returns from, the
distribution of patronage. Interestingly, they affirm that these partisan effects are linked to the
country’s fiscal and electoral subnational institutions (supply side) and the sociodemographic
characteristics of their partisan constituencies (demand side).
10
Argentina’s politics has been widely studied in the last decade or so. Papers by Stokes
(2005), Gibson (2004), Stokes and Medina (2004) to give a few references, document
different aspects of the peculiar argentine’s clientelism Nonetheless, to the best of our
knowledge, there is no paper assessing the influence of clientelism on the PBC.
IV. A brief look at fiscal behavior in Argentine Provinces
How did provinces fiscally perform in the neighborhood of elections? Is there any regularity in
the behavior of fiscal variables around these regular events? In order to approach to the
answer of these questions we carried out Graphic Analysis for public expenditures and the
ratio of current to capital expenditures. That is, we put a magnifier on the behavior of
selected fiscal variables from two years before to one year after each election. That is, for
each of the 5 elections considered, we normalize all the variables in the dataset in four
periods: t-2; t-1; t; t+1, with t standing for the election year.
Figure 1, panels A and B displays the conduct of total expenditures per capita9. If we leave
aside some outliers, we can conclude that there are distinctive differences among elections.
Only the 1987 and 1999 elections show a PBC pattern, with peaks in election years and
slowdowns the following periods. The 1995 and 2003 elections present a trough in both
election years, perhaps due to the so-called “tequila” crisis, starting in December 1994, and
to the “tango” crisis, during 2002. Conversely, the election year 1991 shows high
expenditures but to a lesser extent than 1992, which is related to the stabilization plan
christened “convertibility”.
Similarly, Figure 2, panel A and B exhibits the ratio of Current to Capital expenditures in each
election. Again, leaving aside some outliers, we can conclude that only the 1987 election fits
the pattern claimed by the “expenditure mix manipulation” version of the PBC. The 2003
election also shows markedly falls in the ratio but the following year (2004) the ratio still
diminishes, which contradicts the opportunism story. On the contrary, the election years 1991
and 1999 displays hiking ratios while 1995 presents no clear pattern.
Notice that graphical approach has advantages and disadvantages. Among the first, it
imposes no parametric structures on the data, it makes only a few assumptions that are
necessary in inference statistics, and it is often more accessible and informative than tables
with estimations of coefficients. On the other hand, the graphic analysis is informal, and
intrinsically univariate.
9 Table 1A in the Appendix shows the code for each the province.
11
Figure 1. Panel A. Total Public Expenditures: 1987, 1991 and 1995
Total Expenditure per capita: 1987 Election
0.0
2000.0
4000.0
6000.0
8000.0
10000.0
12000.0
1984 1985 1986 1987 1988
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC
Total Expenditure per Capita: the 1991 Election
0.0
1000.0
2000.0
3000.0
4000.0
5000.0
6000.0
7000.0
8000.0
1989 1990 1991 1992
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC
Total Expenditure per Capita: the 1995 Election
0.0
1000.0
2000.0
3000.0
4000.0
5000.0
6000.0
7000.0
8000.0
9000.0
10000.0
1993 1994 1995 1996
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC CABA
12
Figure 1. Panel B. Total Public Expenditures: 1999 and 2003
Total Expenditure per capita: the 1999 Election
0.0
1000.0
2000.0
3000.0
4000.0
5000.0
6000.0
7000.0
8000.0
9000.0
1997 1998 1999 2000
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC CABA
Total Expenditure per capita: the 2003 Election
0.0
1000.0
2000.0
3000.0
4000.0
5000.0
6000.0
7000.0
8000.0
9000.0
2001 2002 2003 2004
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC CABA
13
Figure 2. Panel A. Ratio of Current to Capital Expenditures: 1987, 1991 and 1995
Ratio Current to Capital Expenditures: the 1987 Election
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
1984 1985 1986 1987 1988
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC
Ratio Current to Capital Expenditures: the 1991 Election
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
1989 1990 1991 1992
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC
Ratio Current to Capital Expenditures: the 1995 Election
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
1993 1994 1995 1996
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC CABA
14
Figure 2. Panel B. Ratio of Current to Capital Expenditures: 1999 and 2003
Ratio Current to Capital Expenditures: the 1999 Election
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
20.0
1997 1998 1999 2000
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC CABA
Ratio Current to Capital Expenditure: the 2003 Election
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
2001 2002 2003 2004
BA Cat Cha Chu Cba Corr ER For Juj LP LR Mza
Mis NEU RN SAL SJ SL SC SF Sgo TF TUC CABA
15
Non-parametric tests
Another way to answer to the questions posed at the beginning of this section is performing
non-parametric tests to compare elections and non-elections episodes in each district. Non-
parametric tests provide useful information on the behavior of individual districts regarding
policy variables. Table 1 shows a summary of two tests of equality of population
(Wilcoxon/Mann-Whitney and Kruskal-Wallis) and the two-sample Kolmogorv-Smirnov test
for equality of distributions. The null hypothesis in all tests is that the selected policy
variables do not differ significantly during elections and non-elections periods.
This set of tests provides information of individual districts on the behavior policy variables in
election years. We consider the following fiscal policy variables: (a) Real Total Expenditure
per capita (TEPC), (b) Real Expenditures in public employment per capita (WAGEBILL) (c)
Real Provincial Revenue per capita (PRPC) and (d) Ratio of current to capital expenditures
(RATIOCK). A more detailed definition of these variables is given next section10.
Table 1. Wilcoxon-Mann-Whitney, Kruskal-Wallis and Kolmorov-Smirnov Tests. Provinces for which the null Hypothesis is accepted (Number of provinces in parenthesis)
Test TEPC WAGEBILL PRPC RATIOCK
Wilcoxon/ Mann-Whitney
Chubut, La Pampa, Mendoza, Misiones,
Salta, San Juan
(6)
Buenos Aires, Catamarca, Chaco, Chubut, Córdoba, Mendoza, Salta, San Luis,
Santa Cruz, Santa Fe, Santiago
(11)
Entre Ríos, Jujuy, Mendoza, Río
Negro, Santa FE, Tierra del Fuego
(6)
Buenos Aires, Chaco, Córdoba,
Jujuy, Salta, Santa Fe
(6)
Kruskal-Wallis Chubut, La Pampa, Mendoza, Misiones,
Salta, San Juan
(6)
Buenos Aires, Catamarca, Chaco, Chubut, Córdoba, Mendoza, Salta, San Luis,
Santa Cruz, Santa Fe, Santiago
(11)
Entre Ríos, Jujuy, Mendoza, Río
Negro, Santa Fe, Tierra del Fuego
(6)
Buenos Aires, Chaco, Córdoba,
Jujuy, Salta, Santa Fe
(6)
Kolmogorov Smirnov
Chubut, Córdoba, Jujuy, Mendoza,
Misiones, Salta, San Juan, Santa Cruz
(8)
Catamarca, Chaco, La Pampa, Mis, Río Negro, Salta, San Juan, San Luis, Santa Fe
(9)
Buenos Aires, Entre Ríos
(2)
Buenos Aires, Chaco, Corrientes,
Jujuy,
(4)
The body of table 1 contains the number of districts -and their names- for which the null
hypothesis is accepted at the 10% level. Interestingly, the null hypothesis of the
Wilcoxon/Mann-Whitney and Kruskal-Wallis tests is accepted for wage bill in more than 40%
of the jurisdictions and for expenditures, local revenues and the ratio of current to capital
expenditure in 25% of the provinces. Similarly, the null of Kolmogorov-Smirnov test is
accepted for 8 and 9 provinces for expenditures and the wage bill (33% and 38%
respectively). These results indicate that for an important number of provinces there is
statistically difference between election and non election years regarding the population and
10 Table 2A in the Appendix shows the results of non parametric test of Real Fiscal Deficit per capita.
16
distribution of chosen variables. However we have to be careful about concluding from them
for two reasons. Firstly, non-parametric tests have asymptotic properties and we are working
with a small sample. Secondly, rejection of the null hypothesis could be provoked by other
factors rather than the PBC that we include in the estimation of the dynamic panel data.
V. Empirical investigation
To test clientelistic-augmented political budget cycles in Argentina we added two dummy
variables to the usual baseline equation estimated in PBC studies. Equation (1) has fiscal
variables as function of the timing of elections, the lagged fiscal variables and several
socioeconomic and political controls to account for variability in the data due to factors other
than elections. The dummies YERABEFORE and TWOYEARSBEFORE are intended to
capture the role of clientelism.
εββββββ itjitjititititit controlselecyearyearbeforeforetwoyearsbeFF +++++= +− 4432110
(1)
Where i indexes the 24 districts (23 provinces plus the federal district, the city of Buenos
Aires) and t represents time and goes from 1984 to 2004. The subnational analysis allows us
to control for historic and cultural variables at the country level that may affect our
explanatory variables.
Dependent variables
We consider five policy instruments: Total expenditures, Wage bill, local taxes collection, the
ratio of current to capital expenditures and the ratio of current expenditures to direct
investment. In Argentine public accounts, Capital Expenditure has three main components:
Direct Investment, including mostly investment in infrastructure and machinery; Capital
Transfers, consisting of transfers to the private sector and other entities and Financial
Investments containing the acquisitions of financial assets. The coding of the dependent
variables is the following:
TEPCit: total expenditure per capita for province i in year t.
WAGEBILLPCit: Wage bill per capita for province i in year t.
PRPCit: provincial revenue from taxes per capita for province i in year t.
RATIOCKit: ratio of current to Capital expenditures for province i in year t.
RATIOCDIit: ratio of current expenditures to Direct Investment for province i in year t.
All variables are expressed in pesos of 2004.
17
Election and clientelism variables
The key variables in our analysis are the election year dummy (ELECYEAR) and the
clientelism dummies (YEARBEFORE and TWOYEARSBEFORE). According to the PBC
theory, we expect high expenditures, wage bill and local taxes collection in election years. If
the Eslava and Drazen (2005) expenditure manipulation argument is to be supported, we
also expect ELECYEAR to be negative correlated to the ratio of current to capital
expenditures and to the ratio of current expenditure to Direct Investment.
The clientelistic hypothesis claims for a positive correlation between our dependent variables
and the dummy variables one year (YEARBEFORE) and two years before the ballot
(TWOYEARSBEFORE). The definitions of our variables of interest are the following:
ELECYEARit: is a dummy variable that equals 1 in governor’s election year and 0 otherwise. The usual problem in PBC studies caused by the lack of coincidence between the fiscal year and the election year is ameliorated in the case of Argentina since two thirds of the elections in our data set were carried out from September to December11. We consider that t is an election year if the voting ballot was carried out from May to December.
YEARBEFOREit: is a dummy variable that equals 1 in the year before the election date (if voting ballot was carried out from April to December), and 0 otherwise.
TWOYEARSBEFOREit: is a dummy variable that equals 1 in two years before the election date (if voting ballot was carried out from April to December), and 0 otherwise.
Notice that gubernatorial elections take place every four years in all districts, so our set of
election and clientelistic dummies cover three out of four years.
Controlling for political influences
Our empirical study control for two sources of political influences on fiscal policy variables:
the erosion of power due to uninterrupted administration and the affiliation of governors to
the President’s party. It is usually argued that time erodes the governor’s popularity. Hence, it
is possible that the incumbent tries to neutralize that effect by spending more, taxing less and
incurring in more deficits. To evaluate this argument, we include a variable (coded GOV) that
counts number of years the governor remains in office12.
Jones, Sanguinetti and Tommasi (2000) in their study on Argentina’s fiscal federalism found
that provinces where the governor is from the same party as the President have lower
spending. They conjecture that provincial governors politically affiliated with the President are
more likely to internalize the effect of spending an additional unit of national resources due to
11 On election dates, see table 3A in the Appendix. 12 In their study on opportunism in Portugal, Goncalves Veiga and Veiga (2006), include a similar variable.
18
internal party discipline. Even in a scenario of weak party obedience, it can be though that
governors allied to the President find advantageous, in terms of electoral results, to support
national policies aimed at controlling spending and fiscal deficit. The codification of this
variable, called SAME, is not straightforward. In the years following the 2001/2002 crisis,
there was a major break in the peronist party. One of the factions, leaded by San Luis
Governor, Rodríguez Saa, became the opposition of President Kirchner. The other main
party, the Alianza UCR/FEPASO also shattered. The fracture of the two most important
parties resulted in some atypical alliances that we consider in our variable SAME. We also
account for the agreements between some provincial parties and the incumbent President
during the 90’s13.
The definitions of political variables are the following:
SAMEit: dummy variable that takes the value 1 if the governor is affiliated to the same party as the President and 0 otherwise.
GOVit: number of years of uninterrupted administration in province i in year t.
Controlling for socioeconomic variables
Socioeconomic and demographic conditions may also affect subnational fiscal performance,
thus we include a set of control variables in several specifications of equation (1): the level of
GDP per capita and the change in rate of unemployment from one year to another.
We expect richer districts (measured by the GDP per capita) to have higher levels of
expenditure and revenues per capita and also a higher proportion of capital expenditures
than poorer districts.
Additionally, we suppose that provinces with severe labor market problems spend more
resources to alleviate unemployment and its consequences. From mid nineties to 2003, the
unemployment rate hiked from one-digit figures to more than 20% in almost all districts and
consequently the population under the poverty line soared.
We also control for the level of fiscal result per capita. We understand that high fiscal deficit
in period t-1 may prompt government not only to lessen expenditures and increase local
taxes at time t, but also may affect the expenditure mix favoring reductions in current
expenditures relative to capital expenditures.
13 In years 2003 and 2004, we coded 1 the provinces of Mendoza, Río Negro and Catamarca, administered by UCR governors allied with the Peronist President Kirchner. In contrast, the province of San Luis was coded 0 despite being administered by the peronist governor Rodríguez Saa. During the 1996-1999 we coded 1 the provinces of Tucumán and Tierra del Fuego to account for the alliances of Fuerza Republicana and Partido Popular Fueguino with former peronist president Menem.
19
The codes and definitions for above mentioned variables are:
GDPPCit: rate of growth of the GDP per capita of province i in year t.
UDIFit: difference between the unemployment rates in year t and t-1 in province i.
DEFPCit-1: real deficit per capita of province i in period t-1.
Controlling for Federal Transfers and Oil and Gas Grants
When working with subnational units is crucial to control for federal transfers to the
provinces. In the case of Argentina, we also take into consideration the oil and gas grants,
received by some provinces, which vary with international prices and local regulations. We
distinguish two types of transfers, the automatic ones coming from the tax-sharing
agreement that are computed as a fixed percentage on “shared resources” and discretionary
transfers. We expect, as in Jones, Sanguinetti and Tommasi (2000), that provinces with
growing federal transfers per capita have higher spending. This result comes from the de-
coupling of tax and spending decisions in provinces that are more favored by federal
transfers. Equally, grants would presumably contribute to higher expenditures, particularly
capital outlays.
A priori, it is not clear whether increasing amounts of transfers and grants would augment
relatively more the current expenditures than the capital expenditures. It could be expected
that governors dedicate “unexpected” increases in grants and discretionary transfers to
capital rather than current expenditures. Uncertainty about the future stream of federal
money makes unadvisable to devote transfers and grants in salaries and other current
expenditures.
The codes for the above mentioned variables are:
TAXSHAREit: per capita federal government automatic transfers to province i in year t.
TRANSFERSit: per capita total federal government discretionary transfers to province i in year t.
KTRANSFERSit: per capita federal government discretionary capital transfers to province i in year t.
GRANTSPCit: per capita oil and gas grants received by province i between years t and t-1.
Testing Partisan Theory
Partisan theory argues that ideology influences policy variables. Typically, left-wing parties
are assumed to assign greater value to expansionary fiscal policies than right-wing parties.
Nonetheless, it is frequently affirmed that in Argentina political parties disputing gubernatorial
20
and presidential elections are very much alike from the ideological point of view. On the other
hand, right-wing and left-wing factions coexist in the main parties. During the period under
study, 1984-2004, there were two major national parties that dominates argentine politics:
the Peronist party (known as PJ, that stands for Partido Justicialista) and the UCR/Alianza.
However, the political system is far from being characterized as a two-party system, like in
the U.S. This statement is truer in some provinces with strong local parties like the still
unbeaten in gubernatorial elections, Movimiento Popular Neuquino in the province of
Neuquén. Other provinces where local parties won gubernatorial elections at least once are
Chaco, Corrientes, Salta, San Juan, Tierra del Fuego, and Tucumán14.
We tested partisan theory in our model by interacting partisan variables with the election and
clientelistic dummies:
PJit: dummy variable that takes the value 1 if the incumbent party in province i in period t is the Peronist party and 0 otherwise.
PROVPARTYit: dummy variable that takes the value 1 if the incumbent party in province i in period t is any local party and 0 otherwise.
The resulting interaction variables are:
ELECYEAR*PJit: is a dummy variable that equals 1 if the incumbent party in province i, in governor’s election year is PJ and 0 otherwise.
ELECYEAR*PROVPARTYit: is a dummy variable that equals 1 if the incumbent party in province I, in governor’s election year is any local party and 0 otherwise.
YEARBEFORE*PJit: is a dummy variable that equals 1 if the incumbent party in province I, in the year before the governor’s election is PJ and 0 otherwise.
YEARBEFORE *PROVPARTYit: is a dummy variable that equals 1 if the incumbent party in province i, in the year before the governor’s election is any local party and 0 otherwise.
TWOYEARSBEFORE *PJit: is a dummy variable that equals 1 if the incumbent party in province i, two years before the election is PJ and 0 otherwise.
TWOYEARSBEFORE *PROVPARTYit: is a dummy variable that equals 1 if the incumbent party in province i, two years before the election is any local party and 0 otherwise.
Data
The data set for this study was compiled from diverse sources. Fiscal variables at constant
2004 prices were drawn from the Secretaría de Hacienda, Ministerio de Economía de la
Nación. Provincial GDP at constant prices originated from the Bureau of Statistics of each
province and from Universidad Nacional de La Plata estimations, based on official figures.
The Population and unemployment data were taken from the INDEC (the Argentinean
Bureau of Statistics). Political and electoral data comes from Dirección Nacional Electoral.
14 In the province of Mendoza the Demócrata party has a numerous constituency never won a gubernatorial election.
21
This study covers five of the seven gubernatorial elections carried out in Argentina since
democracy was reestablished, in 1983. Gubernatorial elections took place regularly in most
of the 24 districts every four years, in 1983, 1987, 1991, 1995, 1999, 2003 and 200715. We
excluded two ballots, the 2007 due to missing data for some variables and the 1983 election
because there was no party allied with the military regime, so we could not treat any party as
incumbent in that election. Table 2 provides summary statistics for each of the variables
included in the analysis. A detailed description of the data sources appears in the Appendix.
Table 2. Descriptive Statistics. Argentina’s 24 districts. Period: 1984- 2004
Variable Obs Mean Std. Dev. Min Max
Dependent variables
Tepc 504 2368.78 1571.176 592 9987
Wagebillpc 504 1128.932 674.8568 319.6 4441
Ratiock 504 6.15119 4.432022 0.5 37.7
Ratiocdi 504 8.130159 8.013493 .5 118.7
Prpc 504 348.1804 295.9481 27.6 1800.3
Socioeconomic controls
Gdppc 504 372.0131 253.9265 100 1811.2
Udif 504 0.2422619 2.39556 -10.6 8.1
Political controls
Gov 504 3.31 2.73 1 19
Same 504 0.4980159 0.5004928 0 1
Transfers and Grants
Taxshare 504 1217.74 694.18 75.00 4373.68
Transferspc 504 193.89 313.26 0.1 1933.3
Ktransfers 504 23.51806 51.28285 0 559.3
Grantspc 504 268.07 650.35 0 4177.8
PBC and Clientelistic variables
Elecyear 504 0.2321429 0.4226191 0 1
Yearbefore 504 0.2321429 0.4226191 0 1
Twoyearsbefore 504 0.2301587 0.4213525 0 1
Elecyear*PJ 504 0.1428571 0.3502748 0 1
Yearbefore*PJ 504 0.1468254 0.3542834 0 1
Twoyearsbefore*PJ 504 .1448413 .3522904 0 1
Elecyear*Provparty 504 .031746 .1754974 0 1
Yearbefore*Provparty 504 .0297619 .1700986 0 1
Twoyearsbefore*Provparty 504 .0337302 .1807131 0 1
Results
Tables 3 and 4 show the estimations of our five fiscal policy variables. In Table 3 we test the
influence of election year and clientelism dummies, while table 4 includes interaction
dummies to examine partisan effects. As already remarked, our models include lagged
15 The President of the Nation appointed the governor of the Federal district (City of Buenos Aires) until 1996, so we have three gubernatorial ballots, in 1996, 2000 and 2003. Similarly, Tierra del Fuego reached the status of province in 1991, consequently we consider the 1991, 1995, 1999 and 2003 elections.
22
dependent variables to capture the rigidity of the budget from one year to another, thus we
estimate using the dynamic panel technique developed by Arellano and Bond.
Table 3. Is there clientelistic-augmented PBC in fiscal policy variables?
Expenditures per capita
(1)
Wage Bill per capita
(2)
Local Tax collection per capita
(3)
Ratio Current to Capital
Expenditures (4)
Ratio Current Expenditure to
Direct Investment (5)
Lagged tepc 0.16* (1.85)
Lagged Wagebillpc 0.48*** (5.74)
Lagged Prpc 0.53*** (9.31)
Lagged RatioCK 0.30*** (5.16)
Lagged RatioCDI 0.38*** (9.24)
GDPpc 0.96*** (4.17)
0.37** (2.06)
0.29*** (3.81)
-0.004 (-1.12)
-0.01 (1.12)
Udif 7.53 (1.39)
-2.09 (-0.63)
-7.24*** (-3.81)
0.21*** (2.62)
0.41*** (2.78)
Tax Share 1.32*** (26.23)
0.50*** (9.84)
0.15*** (4.07)
-0.001* (-1.70)
-0.0004 (-0.57)
Transfers 0.85*** (3.77)
0.24* (1.70)
-0.03 (-1.24)
-0.01*** (-4.99)
-0.01*** (-3.92)
Grants 0.60*** (7.09)
-0.10 (-0.75)
0.001* (1.63)
0.003* (1.66)
Defpc (t-1) -0.002*** (-3.16)
-0.004** (-2.39)
Gov 16.96** (2.00)
-2.21 (-0.68)
0.13 (0.08)
-0.13 (-1.16)
-0.38 (-1.24)
Same -89.67*** (-2.57)
-43.53 (-1.22)
17.05 (1.12)
-0.77** (-2.03)
-0.65* (-1.68)
Elecyear 120.62*** (3.66)
51.05*** (3.50)
4.39 (0.82)
0.19 (0.60)
0.31 (0.43)
Yearbefore 94.17*** (3.80)
31.04*** (2.59)
4.95 (0.95)
1.71*** (3.58)
2.60** (2.05)
Twoyearsbefore 71.24** (2.40)
20.50** (2.07)
-2.65 (-0.36)
0.97*** (3.47)
1.45** (2.28)
Constant -5.88 (-1.62)
-4.78** (-2.02)
1.41* (1.66)
0.30*** (6.58)
0.52*** (4.28)
Observations 456 456 456 456 456
Number of districts 24 24 24 24 24
Observations per districts
19 19 19 19 19
F Statistics F(11, 444)= 627.64
F(11, 444)= 373.58
F(10, 445)= 781.42
F(12, 443)= 14.21
F(11, 444)= 50.62
Arellano-Bond test that average autocovariance in residuals of order 1 is 0. H0: no autocorrelation Pr>z=
0.0010 0.0280 0.0096 0.0010 0.0699
Arellano-Bond test that average autocovariance in residuals of order 2 is 0. H0: no autocorrelation Pr>z=
0.1382 0.1735 0.1229 0.1405 0.1872
Our estimations show strong evidence of clientelistic-augmented PBC in total expenditures
and public sector wages. ELECYEAR, YEARBEFORE and TWOYEARSBEFORE present
positive and statistically significant coefficients. Interestingly, the coefficients augment as
23
approaching the election year indicating that incumbents carry out continuous and strong
clientelistic activity to preserve political power starting as far as two years before elections.
The proposed model fits very well the Expenditures per capita equation (column 1): all
explanatory variables have the expected sign and statistical significance. In particular, the
sign of GOV is positive meaning that extended incumbency results in higher expenditures.
We interpreted this coefficient as governors’ response to political erosion. The negative
coefficient of SAME implies that governors politically affiliated with the President spend less
than incumbent associated with the President’s opposition.
Conversely, we found no evidence of manipulation in provincial revenues per capita, which
indicate that expenditures rather than taxes are the incumbents’ favorite instruments to
improve their chances of remaining at office.
Regarding changes in the expenditure mix to influence electoral outcomes, our estimations
show that incumbents raise the ratio of current to capital expenditures one and two years
before elections but not during the election year. The dummy variables
TWOYEARSBEFORE and YEARBEFORE presented positive and statistical significant
coefficients while the ELECYEAR coefficient was non significant at customary levels.
Analogous results were obtained for the ratio of current expenditures to direct investment.
Clearly, incumbents feed their clientele with private goods (current expenditures) rather
public goods (capital expenditures, direct investment).
The first lags of explanatory variables were statistically significant in all five regressions of
both tables reflecting inertia in public spending, wage bill, local collection and expenditure
mix ratios.
Equations studying partisan influences on fiscal instruments show an unambiguous
clientelistic behavior of the Peronist party regarding expenditures which is consistent with
larger returns to clientelistic “investments” (See Calvo and Murillo, 2004). Public outlays
under peronist incumbents increase two years before elections, the year before and during
the election year. Conversely, provincial parties start spending timidly the year before
election and remarkably increase outlays in the election year. Regression 2 shows that
payroll goes up during election years and only under Peronist incumbents. Meanwhile,
column 3 demonstrates that neither the Peronist party nor provincial parties use local taxes
collection as electoral instrument.
The estimations in columns (4) and (5) illustrate the importance of the Peronist clientelistic
organization: current expenditures are larger than capital expenditures (and its sub category,
24
Direct Investment) the year before elections and two years before the ballots. There is no
evidence that political parties manipulate expenditure during the election year.
Table 4. Clientelistic-augmented PBC and partisan Effects
Expenditures per capita
(1)
Wage Bill per capita
(2)
Local Tax collection per capita
(3)
Ratio Current to Capital
Expenditures (4)
Ratio Current Expenditure to
Direct Investment (5)
Lagged tepc 0.17* (1.86)
Lagged Wagebillpc 0.49*** (5.70)
Lagged Prpc 0.53*** (8.97)
Lagged RatioCK 0.30*** (5.06)
Lagged RatioCDI 0.38*** (8.90)
GDPpc 1.00*** (4.30)
0.37** (2.09)
0.28*** (3.69)
-0.03 (-1.01)
-0.01 (-1.06)
Udif 7.1 (1.30)
-2.26 (-0.70)
-7.25*** (-3.78)
0.22*** (2.77)
0.42*** (2.88)
Tax Share 1.33*** (28.80)
0.50*** (9.24)
0.15*** (3.92)
-0.001* (-2.02)
-0.001 (-0.73)
Capital Transfers 0.85*** (3.89)
0.24* (1.79)
-0.03*** (-1.33)
-0.01*** (-4.66)
-0.01*** (-3.63)
Grants 0.60*** (7.23)
-0.10 (-0.74)
0.001 (1.32)
-0.003* (-1.62)
Defpc (t-1) -0.002*** (-2.98)
-0.004** (-2.41)
Gov 16.02** (2.03)
-2.05 (-0.67)
0.27 (0.19)
-0.12 (-1.11)
-0.3 (-1.28)
Same -115.66*** (-3.33)
-50.70 (-1.51)
15.09 (0.90)
-0.97** (-2.29)
-0.89* (-1.62)
Elecyear*PJ 164.58*** (5.34)
55.96*** (3.14)
6.17 (0.65)
0.14 (0.30)
0.32 (0.43)
Yearbefore*PJ 136.38*** (4.83)
29.90 (1.52)
5.84 (0.80)
1.54** (2.23)
2.60 (1.51)
Twoyearsbefore*PJ 81.92*** (2.89)
2.71 (0.21)
4.65 (-0.52)
0.66** (2.18)
1.49** (2.37)
Elecyear*Provparty 214.53* (1.76)
73.55 (1.52)
0.88 (0.05)
-0.19 (-0.28)
0.95 (0.89)
Yearbefore*Provparty 98.22* (1.85)
16.93 (0.34)
5.02 (0.25)
0.03 (0.04)
0.21 (0.25)
Twoyearsbefore*Provparty 56.11 (0.58)
49.57 (1.01)
7.14 (0.52)
0.71 (1.28)
1.08* (1.87)
Constant -5.04 (-1.39)
-6.86*** (-3.07)
1.48* (1.72)
0.30*** (6.60)
0.52*** (4.35)
Observations 456 456 456 456 456 Number of districts 24 24 24 24 24 Observations per districts 19 19 19 19 19 F Statistics F(14, 441)=
892.94 F(16, 438)= 582.85
F(13, 442)= 543.47
F(15, 440)= 12.04
F(15, 440)= 36.07
Arellano-Bond test that average autocovariance in residuals of order 1 is 0. H0: no autocorrelation Pr>z=
0.0010 0.0273 0.0114 0.0008 0.0648
Arellano-Bond test that average autocovariance in residuals of order 2 is 0. H0: no autocorrelation Pr>z=
0.1240 0.19.38 0.1212 0.2350 0.2481
25
VI. Concluding remarks
Differently from most of the literature on PBC, this paper analyzes the impact of clientelism
on the behavior of fiscal policy variables. We also distinguish from the mainstream literature
on clientelism in two aspects. Firstly, clientelism experts usually focus on exchanges around
elections between local party bosses and clienteles, while we also discuss the patron-
clientele exchanges that occur largely before the election year. Organizing a clientelistic-
based winning electoral machine takes time and fiscal resources. Obtaining the loyalty of
clienteles requires active regular trade.
Secondly, a large portion of the studies on clientelism concentrates on vote buying which is
possibly the most important clientelistic transaction but it is not the only one. A large array of
goods and services are traded long before elections which also affects provincial fiscal
behavior. Theses basic features of the patron- clientele relationship has implications on the
costs and timing of organizing and maintaining a clientele. This paper argues that the typical
PBC described by the theoretical literature and supported by the empirical evidence in
developed as well as in developing countries applies with some modifications in Argentina’s
regained democracy. We sustain that organizing and keeping clienteles modify the pattern of
expenditures through time of the typical PBC.
Using non parametric tests as well as a dynamic panel data estimations covering all 24
Argentine districts and the period 1983- 2004, we found that total expenditures per capita are
high not only in election years but also one and two years before elections. Moreover, total
expenditures display an increasing trend as approaching the voting appointment. Our
empirical investigation also reveals the importance of the clientelistic organization of the
Peronist party relative to other parties.
Furthermore, contrary to the prescriptions of the Rogoff (1990) model and the findings of
Eslava and Drazen (2005) for Colombia, we found no evidence of changes in the
expenditure mix (the decision between current and capital expenditures) in election years.
We speculate that incumbent governments attempt to secure their clientele by “investing” in
private transfers such as jobs and subsidies rather than “spending” in infrastructure.
Finally, we examine the fiscal policy instrument used by Argentine’s governors to enhance
their reelection prospects. We found no trace of manipulation in local taxes collection. The
favorite instrument is public expenditure, particularly current expenditures.
26
References
Akhmedov, A. Ravichez, A. and Zhuravskaya, E. (2004) “Opportunistic Political Cycles: test in a young democracy setting”. Quarterly Journal of Economics. Vol. 119, Pp. 1301-38.
Alt, James and Lassen, David (2005) “The political budget cycle is where you can’t see it: transparency and fiscal manipulation”. Mimeo.
Brender, Adi and Drazen, Allan (2005) Political budget cycles in new versus established democracies. Journal of Monetary Economics 52: 1271-1295.
Brusco, Valeria, Nazareno, Marcelo and Stokes, Susan (2004) “Vote Buying in Argentina”. Latin America Research Review. Vol. 39, No. 2, June.
Calvo, Ernesto and Murillo, María Victoria (2004) “Who delivers? Partisan Clients in the Argentine Electoral Markets”. American Journal of Political Science, 48: 750-1.
Drazen, Allan and Eslava, Marcela (2005). “Electoral manipulation via Expenditure Composition: Theory and evidence” NBER Working Paper No. 11085. January.
Eslava, Marcela (2006). “The Political economy of Fiscal policy: survey” Inter-American Development Bank. Working paper No. 583.
Galli, Emma and Rossi, Stefania (2002) “Political Budget Cycles: the case of the Western German Länder”. Public Choice 110: 282-303.
Gélineau, Francois and Remmer, Karen (2005) “Political Descentralization and Electoral Accountability: The Argentine Experience, 1983-2001” British Journal of Political Science 36: 133-157.
Gibson, Edward (2004) Subnational authoritarianism: territorial strategies of political control in democratic regimes. Annual Meeting of the American Political Science association.
Goncalves Veiga, Linda and Veiga, José (2007) “Political Business cycles at municipal level”. Public Choice 131: 45-64.
Gonzalez, María (2002) “Do Changes in Democracy Affect the Political Budget Cycle? Evidence from Mexico”. Review of Development Economics, 6(2) Pp. 204-224.
Hicken, Alan D. (2007). "How Do Rules and Institutions Encourage Vote Buying?" in Frederic C. Schaffer (ed.), Elections for Sale: The Causes and Consequences of Vote Buying. Boulder, CO. Lynne Rienner Publishers.
Jones, Mark, Meloni, Osvaldo and Tommasi, Mariano (2007) “Voters as Fiscal Liberals”. Paper presented at the 12th LACEA Meeting, Bogota. .
Jones, Mark, Sanguinetti, Pablo and Tommasi, Mariano (2000) "Politics, Institutions, and Fiscal Performance in a Federal System: An Analysis of the Argentine Provinces" Journal of Development Economics. 61(2): 305-333.
Keefer, Philip (2007). “Clientelism, Credibility, and the Policy Choices of Young democracies” American Journal of Political Science, Vol. 51, No. 4, October, Pp. 804-821.
Khemani, Stuti (2004) “Political Cycles in a Developing Economy: Effect of Elections in the Indian States”. Journal of Development Economics 73:125-154.
27
Magaloni, Beatriz, Díaz-Cayeros, Alberto and Estévez, Federico (2007) “Clientelism and Portfolio Diversification: A model of Electoral Investment with Applications to Mexico.” In Herbert Kitschelt and Steven Wilkinson, eds., “Patrons. Clients, and Policies: Patterns of Democratic Accountability and Political Competition”, pp. 381-429. Cambridge University Press.
Perdomo, Martín (2007) “Ciclos Presupuestarios Políticos. Evidencia Empírica para las Provincias Argentinas (1983-2004)” Facultad de Ciencias Económicas Universidad Nacional de Tucumán.
Persson, Torsteen and Tabellini, Guido (2000) “Political Economics: Explaining Economic Policy”. MIT Press, Cambridge MA.
Robinson and Verdier (2002) “The Political Economy of Clientelism”. Working paper. Center for Economic Policy Research.
Rodrik, Dani (1998) “Why do more open economies have bigger governments?” Journal of Political Economy. Vol. 106 No. 51.
Rodden, Jonathan (2002). “The Dilemma of Fiscal Federalism: Grants and Fiscal Performance around the World”. American Journal of Political Science, Vol. 46, No. 3, July, Pp. 670-687.
Rogoff, Kenneth (1990) “Equilibrium Political Budget Cycles”. American Economic Review 80: 21-36.
Rogoff, Kenneth and Siebert, Ann (1988) “Elections and Macroeconomic Policy Cycles”. Review of Economic Studies 55: 1-16.
Schaffer, Frederic (2002) “What is Vote Buying?” Paper presented at the International Conference “Trading Political Rights: The comparative Politics of Vote Buying. Center for International Studies. MIT.
Schuknecht, Ludger (2000) “Fiscal Policy Cycles and Public Expenditures in developing Countries”. Public Choice 102: 115-130.
Spiller, Pablo and Tommasi, Mariano (2007) “The Institutional Foundations of Public Policy in Argentina. A Transactions Cost Approach”. Cambridge University Press.
Stokes, Susan (2005) “Perverse Accountability: A Formal Model of Machine Politics with Evidence from Argentina”. American Political Science Review. Vol. 99 No.3.
Stokes, Susan (2007) “Political Clientelism”. Handbook of Comparative Politics. Oxford University Press.
Weitz-Shapiro, Rebecca (2003) “Clientelism and Social Welfare Policy in Argentine Municipalities”. Unpublished Ph.D. Dissertation. University of Columbia.
28
Appendix
Table 1 A. Codes for Argentine Provinces
Number Province Code
1 Buenos Aires BA
2 Catamarca Cat
3 Chaco Cha
4 Chubut Chu
5 Córdoba Cba
6 Corrientes Corr
7 Entre Ríos ER
8 Formosa For
9 Jujuy Juj
10 La Pampa LP
11 La Rioja LR
12 Mendoza Mza
13 Misiones Mis
14 Neuquén NEU
15 Río Negro RN
16 Salta SAL
17 San Juan SJ
18 San Luis SL
19 Santa Cruz SC
20 Santa Fe SF
21 Santiago del Estero Sgo
22 Tierra del Fuego TF
23 Tucumán TUC
24 CABA CABA
29
Table 2 A. Wilcoxon-Mann-Whitney, Kruskal-Wallis and Kolmorov-Smirnov Tests. Provinces for which the null Hypothesis is accepted (Number of provinces in parenthesis)
Test Real Deficit per capita
(DEFPC)
Wilcoxon/ Mann-Whitney Chubut, Córdoba, San Luis
(3)
Kruskal-Wallis Chubut, Córdoba, San Luis
(3)
Kolmogorov Smirnov Buenos Aires, Chubut, Jujuy, Santa Fe
(4)
Table 3 A. The timing of election. Number of gubernatorial elections carried out in a given month
Month Number of Elections Percentage
January 0 0.0
February 0 0.0
March 1 0.9
April 2 1.7
May 19 16.2
June 7 6.0
July 1 0.9
August 8 6.8
September 52 44.4
October 21 17.9
November 3 2.6
December 3 2.6
Total number of elections 117 100.0
Two-thirds (67.5%) of the gubernatorial elections (24 districts) took place in the last four months of the year