C E N T R E D E R E C E R C AE N E C O N O M I A I N T E R N A C I O N A L

Why does thePublic Sector

Grow?The Role of

EconomicDevelopment,

Trade andDemocracy

Els Opuscles del CREI

num. 5november 99

Carles Boix

1

Why Does the PublicSector Grow?

The Role of EconomicDevelopment, Trade and

DemocracyCarles Boix

Two stylized facts describe the evolution ofthe public sector across the world during the lastcentury: first, its steady growth; second, thepresence of persistent crossnational differences inits size. The public sector has grown substantiallysince the turn of the century. Excluding wartimes, government expenditure remained constantaround 10 per cent of GDP during the 19thcentury in the advanced world. After 1914,however, the size of the public sector expandedsubstantially. As shown in Figure 1, in the early1950s, total current public revenue averaged 24per cent of GDP in the OECD. By the mid-1970s,it had risen to 36 per cent. By the early 1980s ithad stabilized at around 44 per cent. Althoughthe pace of change has been less dramatic, thepublic sector has also grown in the developingword. Among non-OECD countries, current publicrevenue averaged 14 per cent of GDP in 1950,reached 20 per cent of GDP by the late 1960s andthen hovered around 27 per cent from the late1970s onward1.

The Center for Research in InternationalEconomics (CREI) is a research centersponsored by Universitat Pompeu Fabra (UPF)and the Generalitat de Catalunya. It is locatedat the campus of the UPF in Barcelona.

CREI arises as the result of two developmentsin today’s economy and in economic science.First, the present evolution of the internationaleconomy and the increasing role that regionssuch as Catalonia may have in a united Europe,show the need for promoting research andeducational centers in fields such asinternational economic policy. Second, therecent advances of economic theory in fieldssuch as international economics, the theoryof general equilibrium, game theory, growththeory, development economics,macroeconomics in open economies andinternational finances, which were until recentlyvery segmented, call for an increasinglyinterrelated and integrated approach to theanalysis of the economy.

Els Opuscles del CREI are the instruments for the diffusion ofCREI’s research beyond the academic world. Each Opuscle,aimed at a general public, gathers together conclusions andobservations of specialized papers published, or about to bepublished, in professional journals. The opinions stated in ElsOpuscles del CREI are the sole responsibility of the authors.

Published by: CREIUniversitat Pompeu Fabra.Ramon Trias Fargas, 25-27 08005 BarcelonaTel. 93 542 24 98© 1999, CREI© of this edition: Carles BoixISSN: 1137 - 7844Design: Fons GràficPrinted by: Masanas GràfiquesLegal register: B-47861-99

Despite the steady growth of the public sector,differences across nations have remainedsubstantial. In the mid-1980s, for example, publicrevenue ranged from less than 10 per cent ofGDP in Sierra Leone and Paraguay to over 60 percent in Botswana, Kuwait, Reunion and Sweden.Figure 2 shows the mean and dispersion ofpublic revenue among OECD nations. In the early1950s, public revenue went from 19 to 32 percent of GDP. In 1985 it ranged from 31 to 60 percent of GDP. As shown in Figure 3, crossnationalvariation has become even sharper in thedeveloping world. In the mid-1980s, publicrevenue went from 6 percent in Sierra Leone toalmost 83 percent of GDP in Reunion.

It is interesting to notice, however, that, inspite of this growing divergence across nations,there has been a remarkable stability in therelative ranking of nations regarding the size oftheir public sector. As is apparent in Figures 4and 5, which show the relationship betweenaverage public revenue in 1950-59 and 1970-74and between average public revenue in 1970-74and 1985-89 respectively, those countries that hada considerable public sector in 1950 continue tohave today a large public sector. Similarly, mostof those countries with a small state forty yearsago still rank the lowest in terms of publicsector2.

32

Fig

ure

1T

he

evo

luti

on

of

pu

bli

c re

ven

ue,

195

0-93

45 40 35 30 25 20 15 10

Percent of GDP

1950

1952

1954

1956

1958

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

OECD

Nat

ions

Non O

ECD

Nat

ions

54

Fig

ure

2P

ub

lic

reve

nu

e in

OE

CD

Nat

ion

s. M

ean

an

d d

isp

ersi

on

, 195

0-93

Fig

ure

3P

ub

lic

reve

nu

e in

No

n-O

EC

D N

atio

ns.

Mea

n a

nd

dis

per

sio

n, 1

950-

93

55 50 45 40 35 30 25 20 15

Percent of GDP

1950

1952

1954

1956

1958

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

One

stan

dar

d d

evia

tion

Mea

n

50 45 40 35 30 25 20 15 10

Percent of GDP

1950

1952

1954

1956

1958

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

One

stan

dar

d d

evia

tion

Mea

n

76

Fig

ure

4St

abil

ity

in r

even

ue,

195

0s-1

970s

Fig

ure

5St

abil

ity

in r

even

ue,

197

0s-1

980s

50 40 30 20 10 0

Public Revenue / GDP in 1970-74

Pu

bli

c R

eve

nu

e /

GD

P i

n 1

95

0-5

90

1020

3040

50

60 50 40 30 20 10 0

Public Revenue / GDP in 1985-89

Pu

bli

c R

eve

nu

e /

GD

P i

n 1

97

0-7

40

1020

3040

50

OEC

D c

ount

ries

in u

pper

cas

e;ot

her

coun

trie

s in

low

er c

ase

OEC

D c

ount

ries

in u

pper

cas

e;ot

her

coun

trie

s in

low

er c

ase

In modern societies, technological breakthroughsand the expansion of manufacturing and service-oriented jobs transform the old economic structurewith the following consequences. In the firstplace, the distribution of economic risk changes,concentrating in specific segments of thepopulation. More precisely, unemployment spellsand work-related accidents, which emerge as thedownside of manufacturing-led productivityincreases, become important among industrialworkers and, particularly, among those mostunskilled. In other words, the process ofindustrialization and the formation of a broadclass of wage-earners results in stronger pressuresfor intra-generational transfers. In the secondplace, a general improvement in materialconditions in general and in health technologies inparticular prolong life expectancy and eventuallylead to a shift in the demographic structure. Asthe profile of the population matures and theproportion of old cohorts expands, pressure forinter-generational transfers, in the form ofpensions and health care programs, goes up4.

1. 2. Redistribution and the role of democracy

If economic modernization models explain thegrowth of the public sector as a result of the newfunctional needs of industrializing societies, asecond set of theories link the expansion of thestate to its redistributive consequences – and thusexplained it as an outcome of particular politicalcoalitions attempting to redistribute incomebetween social groups. Among the many formsthis approach takes, Meltzer and Richards (1981)argue, in a rather influential piece, that, in ademocracy, as long as the mean income of agiven political community exceeds the income ofthe median voter, politicians will implementredistributive policies in favor of the lower half of

9

1. The debate over the sources of growth of the public sector

The growth of the public sector has spawneda wealth of theoretical models to account for itscauses3. Consider, very briefly, the following fourfamilies of explanations.

1. 1.The impact of modernization

According to a first set of models, generallyfavored by early sociological studies and by manyeconomists, the broad process of economicmodernization has had at least two effects on thestructure of society and on the role the state playsin the economy. On the one hand, a moderneconomy imposes new functional requirementsupon the state, such as setting up a regulatoryframework, paying for infrastructures andgenerating skilled workers. To fully reap thebenefits of technological shocks and growingcapital flows, that are associated with the processof development, policy-makers are increasinglypushed to employ the state to generate minimumlevels of public goods.

On the other hand, the process ofmodernization transforms the underlying structureof income flows as well as the channels throughwhich welfare is provided. Although economicand property arrangements vary substantially intraditional societies, most individuals holdagricultural jobs. In agricultural economies, boththe source of income (the exploitation of land)and the volatility of rents (basically linked toweather conditions) are broadly common to mostindividuals. Even though they are not universal,communal arrangements to share risk – such ascommon lands or church-distributed benefits –and the use of extended families for the provisionof food, shelter and care may be fairly extended.

8

incentives7. In short, in a highly internationalizedeconomy, reformism and redistributive policies areseverely curtailed by the mobility of factors.

An interesting variant of this thesis relatesfederalism and the level of public expenditure.According to this approach, in a federal system,that is, an economically integrated yet politicallyfragmented area, the state's monopoly power isbroken by both factor mobility and competitionbetween levels of governments. The capacity ofeconomic agents, and, in particular, capital tomove to lightly-taxed political subunits leads to alower tax share across the whole country(Brennan and Buchanan 1980, Przeworski andWallerstein 1988, Weingast 1995).

1. 4. Economic internationalization and an expanding public sector

The impact of the process of internationalizationis, however, uncertain. Low taxes do notautomatically attract private investors. Althoughmany Third World countries have low taxes,capital inflows from the industrialized, high-taxnations have been modest. This is because, inorder to maximize profits, capital holders investin those countries (or economic sectors) that offerthe highest net rate of return, that is, the widestwedge between the gross rate of return to capitaland taxes (Lucas 1990). By lowering taxes andholding wages down, it might be possible toboost profits and therefore encourage privateinvestment. But the rate of return of capital isequally dependent on the productivity of theinput factors that enter the production process.Since what mainly determines gross profitability isthe productivity of the input factors, taxes can behigh insofar as productivity remains high enoughto deliver the highest net profits (compared toother places) to capital. Capital will always prefer

11

the income distribution. In other words,whenever the distribution of income is skewedtoward the rich, politicians will tax the latter andtransfer the revenue to the poorest to obtain amajority of the votes5. This would explain whypublic expenditure started to increase with thegeneralization of universal suffrage after WWI. Ifthis theory were right, we should also expect,ceteris paribus, lower levels of expenditure innon-democratic regimes. The Meltzer andRichards' approach should be equally valid toexplain intergenerational (rather thanintragenerational redistribution): as old-agedcohorts grow in size in the population, publicexpenditure on pensions should rise6.

1. 3. Economic internationalization and a shrinking public sector

The decision to tax capital and labor to financea growing public sector cannot be isolated fromthe reaction of these factors to growing fiscalpressure. Accordingly, for an important strand ofthe literature, the mobility of factors (or, in morejournalistic terms, the progressive ‘globalization’ ofthe economy) constrains the ability ofgovernments to tax and spend. The logic of theargument, which is mostly applied to capital(mostly because of its reportedly higher mobilityvis-à-vis labor), is straightforward. Since economicgrowth depends on investment and investment inturn depends on profits, states and politicians areultimately constrained by the rational calculationsof the holders of capital, who are always in searchof the highest rate of return for their assets. Toavoid capital from moving to the most profitablecountries, and thus facing decreasing investmentrates and economic stagnation, all states arepushed to maximize the rate of return of privateinvestors. To lure them, states will outbid eachother through low taxes and by offering significant

10

weaknesses: the first one, empirical; the otherone, strictly theoretical.

To date most empirical studies areinconclusive (Alt and Chrystal 1983, chapter 8;Lybeck 1988; Mueller 1988, chapter 17; Holseyand Borcherding 1997). Most scholars use limitedsamples, such as one-country time-series analysisor a cross-section of countries, and focus onsingle policy measures, such as social securitytransfers or public consumption8. Generaltheoretical claims are hence difficult to derive.Some recent studies have developed pooled time-series cross-sectional samples for (most) OECDnations (Pampel and Williamson 1988; Hicks andSwank 1992; Huber, Ragin and Stephens 1993).Although these broader quantitative studies go along way in determining the forces behind thegrowth of the public sector, several explanatoryfactors, such as left-wing rule, corporatism,openness and the proportion of old population,are so well correlated that it is impossible toascertain, first, which variable actually mattersand, second, through what specific mechanisms itdoes. Their focus on OECD nations makes themlimited in their applicability. Broadly speaking,this sample of advanced democracies can onlyvery weakly test for the effects of economicgrowth and corresponding social change, theimpact of democratic (vs. authoritarian) regimesand the influence of an unequal distribution ofresources (leading to differing pressures forredistribution).

To remedy these problems, I have gathered abroad sample of developed and developingnations. This sample includes all the countries forwhich comparable data on public revenue(current receipts) of the general government areavailable from 1950 to 1990. The data are takenfrom the United Nations National Accounts (UN,

13

a country where taxes are high, as long asproductivity is very high, to a country where bothtaxation and productivity are low.

As a matter of fact, and in sharp contrast tothe prediction that higher levels ofinternationalization lead to lower levels of publicspending, a highly influential model of publicexpenditure in political economy has positivelyrelated the level of government intervention tothe degree of openness in the economy. Higherlevels of trade integration have been shown toincrease the size of the public sector in advanceddemocracies (in political science, Cameron 1978,Katzenstein 1985; in economics, Rodrik 1998).Government spending grows in open economiesas a mechanism to compensate for the adjustmentcosts of trade openness. Small and opencountries, which are affected intensely by worldbusiness cycles from which they cannot easilyinsulate themselves through standard(expansionary) macroeconomic policies, chooseto maintain high levels of public consumptionand transfers to protect workers in their losingeconomic sectors. High levels of spending aretherefore understood nearly as a functionalrequirement for the maintenance of internalstability and peace. Moreover, high levels ofgovernment intervention are seen as a way ofovercoming market failures in the provision ofskills and infrastructures: they ease the transitionof the unemployed to the more dynamic areas ofproduction in countries that need to compete inworld markets to survive.

2. Empirical weaknesses in previous research

The contemporary research on the causes ofpublic spending suffers from two fundamental

12

has on the tax rate to the extent of disregardinghow economic development alters the underlyingstructure of preferences in the electorate. As aresult, they cannot explain why per capita incomeis so well correlated with the size of the publicsector.

In turn, modernization models, which relyheavily on the idea that politicians mechanicallyrespond to the (changing) tastes of the medianvoter, discount the political and institutional arenain which policy is made. That is, they assumepoliticians to be benevolent planners that,interested in maximizing the national income,automatically use the state to provide for thosepublic goods (such as infrastructures, educationand regulatory agencies) that will in turn let thecountry reap the benefits of modernization. Yet itcannot be taken for granted that policymakers willalways behave as benign planners and pursue thecollective welfare over short-term personal gains.Implementing the optimal policies will onlyhappen under the presence of those political orlegal institutions that effectively restrain rent-seeking behavior among politicians. Democraticinstitutions, by easing the task of monitoringpolicy-makers, should, on average, lead to a fullerprovision of public goods11. Similarly, the extent towhich politicians will develop pension programsand a public health system will be eventuallydependent on the existence of institutionalchannels that make politicians responsive tocitizens’ demands. In short, economic developmentis a necessary but not sufficient condition for thepublic sector to grow: the institutional and politicalmechanisms through which politicians makedecisions shape the extent to which the process ofmodernization affects the size of the state.

A similar problem has affected the literatureon the internationalization of the economy and

15

several years) and from the GovernmentalFinancial Statistics Yearbook (IMF, 1971-90)9. Thesample includes about 80 countries (22 are OECDmembers), with some fluctuations depending onthe time period, and over 2,300 observations. Thestatistical analysis, discussed in sections 4 and 5below, relies on a panel data of both cross-sectional and yearly information.

The dependent variable is current receipts ofthe general government, rather than publicexpenditure, to maximize the sample underanalysis. The United Nations National Accountsoffer less comprehensive data on current publicdisbursements. Although two data bases offerlarger samples for parts of public expenditure,they are not well suited for the purposes of thisarticle. The Penn World Tables report the share ofgovernment consumption of over a hundredcountries – but government consumptionrepresents a fraction of all government spending.The World Bank’s World Data 1995 reports levelsof overall government spending for over 80countries. Still, the World Bank’s World Data (aswell as the IMF data) reports spending only at thecentral government level – which leads toextremely biased values for countries such asArgentina, India or the USA10.

3. Unanswered theoretical issues

The inconclusiveness of the research on thegrowth of the public sector is not merely due,however, to the limitations of the existingempirical work. It stems as well from aninadequate theoretical specification of the models.

Purely political models, like those described insubsection 1.2, concentrate too much on theeffect that an unequal distribution of resources

14

the size of the state. Why some economies aremore open than others is left unanswered and thepresence of a sizable public sector is merelythought of as a functional requirement of havinga free trade policy regime. Yet, as discussed inmore detail in section 5, the selection of bothtrade (and fiscal) policy can only be understoodand modeled as a result determined by politicalstruggles between different sets of agents(politicians and voters) over both the traderegime and the level of taxes.

4. Results (I). The interaction of economic development andpolitical regime

To overcome the pitfalls of the existingtheoretical approaches, we need a joint modelthat integrates both the impact of economicvariables and the underlying structure of politicalchoice. Let me consider, in rather broad terms, inthis section how development and politics interactto shape the size of the public sector12. Idiscussed the role of trade separately in section 5.

With economic development, pressures toenlarge the public sector increase for tworeasons. First, the processes of urbanization andindustrialization generate incentives for the stateprovision of certain collective goods such asregulatory agencies, infrastructures and skillformation. Second, both the emergence of anindustrial economy and an increasingly ageingpopulation shift the underlying incomedistribution in a way that results in strongerdemands for public expenditure. The process ofeconomic development constitutes, however, anecessary but not sufficient condition for theemergence of a large public sector. Policy-makers, who make policy through a political

16 17

mechanism, set the tax rate (to finance theprovision of services and transfers) to match thepreferences of the median voter. The identity ofthe latter varies conditional on the electoralfranchise in place (as well as on the extent towhich voters are mobilized). This variationshapes, in turn, the size of the public sector.Under a democratic regime, where politiciansrespond to the demands of all voters, the publicsector grows parallel to the structural changesaffected by the process of economicdevelopment. Instead, in authoritarian systems,where all or a substantial part of the electorate isexcluded from the decision-making process,precisely to avoid the redistributionalconsequences of democracy, the size of thepublic sector remains small.

Table 1 shows the results of estimating,through econometric techniques detailed in theAnnex, the impact of economic development(measured as real per capita income in prices of1985), trade openness (measured as a the logvalue of the ratio of the sum of imports andexports to GDP) and political regime (democraticor authoritarian), separately and jointly13.

In line with the predictions of themodernization theory, and for the period underanalysis (1950-90), column 1 shows thateconomic development positively affects the sizeof government. Holding trade openness constant(at the sample mean of 62 per cent of GDP), inunderdeveloped nations (a per capita income of$1,000), public revenue fluctuates around 17 percent of GDP. Above a per capita income of$3,500, public revenue climbs to over 30 percent. In a country with a per capita income of$15,000, current public revenues should beexpected to reach 47 per cent of GDP. Thepresence of democratic institutions has a positive,

19

but small, effect on the size of revenue – it is0.95 per cent of GDP higher if the country isdemocratic.

To test for the interaction of socioeconomicmodernization and political institutions, I add theinteractive term ‘Democratic Institutions x (Log of)Real per Capita Income’ in column 2 of Table 1.To interpret the results of column 2, whichconfirm the theoretical predictions, I simulate in

18

Table 1The interaction of political regime and economicdevelopment

Public revenue as per cent of GDP, 1950-1990

Independentvariables (1) (2) (3)

Constant -76.73* -60.44* -69.11*(2.29) (3.13) (2.59)

Per Capita Income 11.33* 8.88* 10.98*(Log)a (0.31) (0.44) (0.31)

Trade Openness (log ofsum of exports and 3.48* 3.93* 2.23*imports over GDP)b (0.41) (0.41) (0.45)

Democratic Institutionsc 0.95* -23.62* -11.45*(0.35) (3.25) (2.05)

Democratic Institutions x 3.21*Log of Real Per Capita Income (0.42)

Democratic Institutions x 3.24*Trade Openness (0.53)

Number of observations 2322 2322 2322

a Per Capita Income. Log of per capita GDP in $ in 1985constant prices. Source: World Penn Tables.b Trade Openness. Log of the sum of exports and imports overGDP. Source: World Penn Tables.c. Democratic Institutions. Democracy=1; Non-democracy=0.Source: Alvarez, Cheibub, Limongi and Przeworski (1996).

Estimation: Generalized Least Squares estimation of Random-Effects Model. Standard errors in parenthesis.* p<0.01

Fig

ure

6T

he

evo

luti

on

of

pu

bli

c re

ven

ue

as a

n i

nte

ract

ion

of

dev

elo

pm

ent

and

po

liti

cal

regi

me

50 45 40 35 30 25 20 15 10

Public revenue as per cent of GDP

Re

al p

er

cap

ita

inc

om

e0

4000

6000

1000

016

000

8000

1200

014

000

2000Dem

ocr

atic

Reg

ime

Auth

orita

rian

Reg

ime

governments. The latter generates a set ofdemands and needs that democratic politiciansneed to respond to. Once real per capita incomegoes over $1,000, the public sector expands at afaster rate under democratic regimes. With a percapita income of $4,000, public revenue is 3percentage points higher in a democratic country.For a per capita income of $10,000, publicrevenue would hypothetically be 6 percentagepoints higher in a democracy (about a sixth morein relative terms)14. The historical experience ofSpain tracks quite nicely these results. In 1974,Spain had a per capita income of $7,291 (in 1985prices) and current public revenues totaled 22.8per cent of GDP. Ten years later, although percapita income had hardly gone up (to $7,330),current public revenues had risen to 32.7% ofGDP. The transition to democracy hadtransformed the role of the public sector. Moregenerally, whereas in OECD countries, with bothhigh per capita incomes and stable democraticregimes, current public revenue averaged 42 percent of GDP in the late 1980s, in Singapore, or inKorea for medium levels of development, the lackof a democratic system led to a much smallerpublic sector that would be expected in purelyeconomic terms.

5. Results (II). The interaction of trade and democracy

The results in Table 1 (column 1) point towarda straight relationship between trade and the sizeof the public sector, thus confirming previouswork by Cameron (1978), Katzenstein (1985) andRodrik (1998). For these authors, as opennessincreases, the state, mainly acting as a benevolentdictator or a welfare maximizer, adopts a salientrole to minimize the risks of higher economicintegration and to compensate declining

21

Figure 6 the evolution of current public revenueas a proportion of GDP as real per capita incomerises under both a democratic polity and anauthoritarian regime (trade openness has beenset equal to the sample mean of 62 per cent ofGDP). The structure of the simulation in Figure 6suggests the following stylized facts. In the firstplace, the level of development has anunconditional impact on the size of the publicsector. Due to the incentives of providing certainpublic goods, the public sector always growswith per capita income. Regardless of thepolitical regime in place, the size of publicrevenues increases by around 15 percentagepoints from very low to medium levels ofdevelopment, and then another 10 percentagepoints from medium to high levels ofdevelopment.

In the second place, the nature of the politicalregime does not affect, on its own, the size of thegovernment. For that to be true, the public sectorshould always be larger under a democraticsystem at all income levels. The results show,instead, that democratic regimes in trulyunderdeveloped economies have no incentives tospend more than authoritarian regimes. Atextremely low levels of development, publiccurrent revenue is, in fact, somewhat higher innondemocratic regimes. At a per capita income of$500 (in 1985 prices), public revenue is 4 percent lower in democracies than in authoritarianregimes. This may be due to two factors. First,the demands for transfers associated withdevelopment have not affected democratic states.Second, it is likely that authoritarian states arelikely to impose higher taxes to finance theirrepressive apparatus.

Finally, as socioeconomic modernization takesoff, democratic institutions lead to larger

20

23

economic sectors. Politics is, however,prominently absent in this approach. Disregardingthe most recent research on the political sourcesof different tariff regimes (Keohane and Milner1996), trade is taken as an exogenouslydetermined variable. Neglecting the literature onthe redistributive consequences of publicspending (Esping-Andersen 1990; Holsey andBorcherding 1997), the growth of the publicsector is then regarded as a merely functionalresponse to the requirements of trade.

A more satisfactory understanding of therelationship between the international economyand domestic politics requires, by contrast, takinginto account the set of economic and politicaltrade-offs that simultaneously underlie the choiceof trade and fiscal policies. Once this is done,countries can be found to pursue three (stylized)alternative strategies15:

1. To insulate domestic actors frominternationally-induced changes in relative prices,national policymakers may choose to close thedomestic economy. Once domestic actors arerelatively isolated from the world business cycle,there are no incentives to resort to higher levelsof public expenditure to compensate voters for(temporary or permanent) employment losses16. Inshort, ceteris paribus, economic insulationdepresses the level of public expenditure.

2. Once free trade policies are embraced, andgiven that Keynesian demand management ishardly available to open economies (Alt 1985),policymakers can only ensure high levels ofsocial welfare (and therefore the support requiredto govern) by expanding the public sector toshore up declining economic sectors.

22

3. Since the combination of openness andcompensation requires higher taxes, policymakersmay consider favoring a third political strategy.Excluding in a systematic manner, that is, throughauthoritarian rule, those sectors that may losefrom increasing economic integration, they willavoid increasing public spending.

To test for the impact that political institutions,i.e. democracy, may have on the size of thepublic sector for different levels of trade, I add, tothe basic model of column 1, the interactive term‘Democratic Institutions x Trade Openness’. Theresults of the regression are presented in column3 in Table 1. Per capita income and tradeopenness continue to boost public revenue. Thepresence of democratic institutions slightlyreduces public expenditure. But this result has tobe set against the sign of the interactive term. Astrade grows, the public sector grows indemocratic regimes.

Figure 7 simulates the results of column 3. Thesimulation includes the evolution of currentpublic revenue when trade openness goes up fortwo different levels of development – a countrywith a per capita income of $4,000 and a countrywith a per capita income of $12,500. When thesum of exports and imports amount to 20 percent of GDP, public revenue totals around 27-28per cent of GDP. As the economy opens, the sizeof the government increases, yet at different rates.In democratic regimes it grows about 9percentage points of GDP when exports andimports represent 100 per cent of GDP. Inauthoritarian regimes, instead, it only goes up byabout 3.5 percentage points. In short, a closedeconomy pushes public revenue downward. But,it is the combination of political regime andopenness that really speeds up the formation oflarge governments.

6. Discussion and concluding remarks

The exploration of the forces that have shapedthe economic role of the state across developedand developing nations shows that two mainforces, modernization and trade openness,determine the size of the public sector.

Pre-modern societies have small governments,regardless of their political regime. DemocraticIndia, the authoritarian regimes of sub-SaharanAfrica or Central America or even the limiteddemocracies of 19th-century Europe fit into thispattern.

The process of economic modernization leadsto larger public sectors through two sequentialsteps. In the first stage, as mainly agriculture-based economies become urban andmanufacturing societies, two structural changesopen the way for a growing state sector. On theone hand, the processes of urbanization andindustrialization generate new demands: a skilledforce is required to take advantage of newproductivity gains; infrastructures are a must forthe proper development of the country. On theother hand, the distribution of risks and themechanisms to cope with them change. Inagrarian societies, risks are generally common tomost individuals. The provision of care takesplace through (extended) families. Technologicalshocks lead to the differentiation of thepopulation according to skills and risks, such asindustrial accidents and joblessness, in particularsegments of the population. With the decline ofextended families, the traditional means tosupport workers during the periods ofunemployment and economic downturns, that is,informal help from relatives, disappear. Collectiveinsurance schemes must then be developed toease the impact of unemployment. Finally,

2524

Fig

ure

7T

he

evo

luti

on

of

pu

bli

c re

ven

ue

as a

n i

nte

ract

ion

of

trad

e an

d p

oli

tica

l re

gim

e

55 50 45 40 35 30 25

Public revenue as per cent of GDP

Tra

de

op

en

ne

ss20

6080

120

200

100

140

160

4018

0

Dem

ocra

cy &

Per

Cap

ita In

com

e of

$4,

000

Dem

ocra

cy &

Per

Cap

ita In

com

e of

$12

,500

Auth

orita

rian

Regi

me

& P

er C

apita

Inco

me

of $

4,00

0

Auth

orita

rian

Regi

me

& P

er C

apita

Inco

me

of $

12,5

00

spending. According to the empirical results oftable 1, in closed economies, such as Argentina,India, Iran, Japan, Mexico or the USA, publicrevenue as a proportion of GDP is much lowerthan it should be given their respective level ofdevelopment.

To maintain trade openness in democraticsettings, policymakers have to develop publicly-funded compensatory schemes to muster thesupport of the losers of higher economicintegration. In response to the economic shock ofthe 1930s, political elites used welfare andinvestment spending to structure a pro-free tradecoalition in small European states17. That solutioncontrasted with the decision to set upprotectionist policies as a way of steadyingrelative prices at home without having to raisetaxes and redistributive income through thepublic sector in Latin America (and in NewZealand and Australia to some extent too) by thatsame period of time.

Finally, since the combination of opennessand compensation requires higher taxes,policymakers may consider excluding in asystematic manner, that is, through authoritarianrule, those sectors that may lose from increasingeconomic integration, to avoid increasing publicspending. In open authoritarian regimes, such asthe East Asian economies, the public sector isbetween a 10-15 percent smaller than in ademocratic system with similar levels of economicintegration.

The issues and results raised in this paper arerelevant to contemporary debates on the politicaland economic consequences of trade (andfinancial) integration on, at least, two counts: thesustainability of the welfare state, and thechanges that the international economy may

27

technological improvements, in the areas of food-production and health care, increase lifeexpectancy and lead to the emergence of healthinstitutions and pension systems.

In a second stage, the forces of modernizationcontinue to affect the more mature economies.Once the welfare state has been set up and thedemographic transition to a more old populationprofile has taken place, public expenditure keepsrising, now driven by the increasing costs ofhealth and pensions’ programs.

Still, modernization is only a pre-condition forlarger public sectors. Market failures may hauntthe provision of education, roads and sewers.Similarly, regulatory bodies and a well-managedpublic administration boost private investmentsubstantially. This should lead states to step inand increase tax collection. However, most of thegrowth of the public sector is related to thecreation of redistributive programs (mostlythrough transfers and, to some extent, throughpublic consumption). Redistributive programsemerge conditional on the political regime inplace and the level of political mobilization. Inauthoritarian regimes, generally imposed to blockredistribution, taxes remain low. Conversely, indemocratic regimes, taxes, reflecting the interestsof voters, become high as modernization shiftsthe underlying distribution of interests toward thedevelopment of both intragenerational and,especially, intergenerational transfers.

Again, trade affects the size of the publicsector. But it does conditional on the politicalregime in place. Those countries that eitherembrace protectionist policies to shore up thewelfare of key domestic sectors or benefit from aquasi-closed economy due to their size anddiversity, do not engage in substantial public

26

effect on the number and system of states.Consider the first question. Broadly speaking, thecurrent literature on the effects of economicopenness can be divided in two camps. On theone hand, the most extended (and popular)approach sees the process of economicglobalization as simply imposing increasingconstraints on the ability of states to govern theeconomy. On the other hand, a set of scholarspoints to the striking correlation betweenopenness and the size of the public sector to callinto question the former’s conclusions: the factthat the most open economies consistentlyespouse larger governments shows, in anunequivocal manner, that more trade does notrequire lower taxes – and that the opposite maybe actually more accurate. It is likely that, onceall the relevant variables are taken intoconsideration, both approaches can bereconciled. On the one hand, it is true thatopenness does not automatically constrain thespending capacity of states – although, again,why public sectors grow in trading nations ismainly a function of political decisions (andsecondarily of economic or structural needs). But,on the other hand, that public compensation mayrun into limits seems to be forgotten by the actualliterature on trade and government growth. Howsustainable a large public sector is over timedepends on the competitive advantage of theexporting sectors that pay for it. If thiscompetitive advantage erodes, the incentive tosustain a large government declines – andcountries start shifting toward either aprotectionist system (the South American path ofthe 1930s and 1940s) or an authoritarian freetrade regime.

The way in which fiscal policy and traderegimes are related suggests also that the work

28 29

that serves as the basis of this opuscle justpresented may be useful to shed light upon thecauses that explain the evolution of the statesystem and any historical variation in the numberof nations. The underlying assumption in themodel that underpins the empirical findings Ihave described consists of a policymakerinterested in maximizing the welfare function ofthe median voter in order to win elections (or,more generally, stay in power). Thepolicymaker’s first choice consists in eitherestablishing a relatively closed economy (wheresmoothing the business cycle is possible) oropening the economy (where demandmanagement is fraught with risks). But the manyways through which autarky may come abouthave been left unexplored. By assumption,autarky has been equated to raising domestictariffs. Nonetheless, a closed economy can bealso achieved by the integration of previouslyseparated countries. In the framework of thismodel, the process of European unification canbe basically understood as an alternative (andcheaper) response to globalization thanexpanding the welfare state in each Europeannation one step further.

30 31

(c) the proportion of nonfuel primary exportsover total exports, for 1970-90, taken from WorldBank tables.

3. ‘Political Institutions’ includes the followingset of political and institutional variables:

(a) a yearly variable that indicates whethereach country was a competitive democracy;

(b) a variable that indicates whether eachcountry was a ‘bureaucracy’ each year;

(c) a variable that indicates whether eachcountry was an ‘autocracy’ each year;

(d) a variable that indicates whether eachcountry was independent each year.

To measure the presence of a democratic,bureaucratic or autocratic regime, we follow theindex developed by Alvarez, Cheibub, Limongiand Przeworski (1996) and the classificationreported in appendix 1 of their paper. Democraticregimes are defined as those regimes “in whichsome governmental offices are filled as aconsequence of contested elections.” (p.4)Bureaucracies are those dictatorships that havelegislatures. Autocracies are those dictatorshipsthat do not and that therefore can be thought ofas not having any sort of institutionalized rule foroperating the government.

To compute the regression I have employed a‘variance-component’ generalized least squares(GLS) technique to correct for the ways in whichassumptions underlying ordinary least squares(OLS) estimations are violated by cross-nationalpanel data (cf. Hsiao 1986, Hicks 1994).

Appendix

To determine the variables that influence sizeof government, I have estimated the followingmodel on a cross-section of nations:

Public Revenues = α + α1(Economy) +α2(Trade) +α3(Political Institutions) + εt

1. ‘Economy’ includes the set of variables thatmeasure the effects of economic development(and the general effects of modernization) on thesize of government:

(a) The log value of real per capita income (inconstant dollars, Chain Index, expressed ininternational prices, base 1985), taken from thePenn World Tables;

(b) the average percentage of urbanpopulation in 1970-90, taken from the WorldBank;

(c) the average share of the agricultural sectorover GDP in 1970-90, taken from the World Bank;

(d) the average proportion of the labor forcein the manufacturing sector in 1970-90, reportedby the World Bank;

(e) the ‘dependency ratio’, that is the numberof years life expectancy goes beyond 60, in 1970-90; life expectancy is taken from the World Bank.

2.‘Trade’ includes:

(a) a measure of the impact of openness ongovernments, calculated as the log value of theratio of trade (sum of imports and exports) toGDP, and is taken from the Penn World Tables.

(b) the ratio of fuel exports over total exports,for 1970-90, taken from World Bank tables;

33

sectional studies, see Cutright (1965) and Wilensky (1975) onadvanced and developing nations, Jackman (1975) onAmerican states, and Korpi (1989) and Cameron (1978) onOECD nations alone.

(9) Both data sources overlap substantially and their data arestrongly correlated (r=.9556). To build the sample I haveprimarily used the data from UN National Accounts. Data fromthe Governmental Financial Statistics has been only taken forcountries not reported by the UN.

(10) Rodrik (1998) and Cheibub (1998) have recently builtbroader samples that encompass developed and developingnations. Rodrik (1998), however, employs public consumptionas a percentage of GDP. This is too limited a tool to measure thesize of the welfare state and provides highly biased results(given how important public consumption is among developingcountries). Cheibub (1998) employs data on centralgovernment, which also measures very imperfectly total publicexpenditure (especially for large, closed economies, that tend tobe decentralized), and focuses only on the tax capacityassociated to different political regimes.

(11) For a discussion of this point in the context of economicgrowth, see Olson (1993). Przeworski and Limongi (1997) offer,however, a less favorable vision of the monitoring capacity ofdemocracy.

(12) For an extended discussion, see Boix (1999).

(13) For the results of estimating the impact of the averagepercentage of urban population, the average share of theagricultural sector over GDP, the average proportion of thelabor force in the manufacturing sector, and the ‘dependencyratio’, that is the number of years life expectancy goes beyond60, in 1970-90, on the size of the public sector, see Boix (1999).

(14) In the sample under analysis, there are very fewauthoritarian cases (some oil exporters) with a per capitaincome over $8,000. The lack of dictatorships at high levels ofdevelopments is a well established fact in the literature. SeeLipset (1959), Limongi and Przeworski (1997).

(15) For a formal treatment of this question, see Adserà andBoix (1998).

(16) In fact, aggregate demand management may become aneffective strategy to minimize the occurrence of recessions. Thatis, imposing a closed economy actually allows policymakers toengage in countercyclical policies to smooth the business cycle.Romer (1993) and Campillo and Miron (1997) show thatclosed economies and inflation are strongly and positivelycorrelated. By contrast, demand management in openeconomies can only take place under particularly stringentconditions (Alt 1985; Lange and Garrett 1985), and it is onlysustainable in a temporary manner.

(17) For some evidence that the origins of the Scandinavianwelfare state at the turn of the century may lie on the demandof agricultural-based exporting sectors to minimize production-related risks, see Baldwin (1990).

32

Footnotes(1) Figures 1 to 5 are based on data taken from the UnitedNations National Accounts (UN, several years) and theGovernmental Financial Statistics Yearbook (IMF, 1971-90).

(2) Figure 5 also shows that, at least among OECD nations, the1960s brought a considerable growth of the public sector,particularly among several Scandinavian countries, theNetherlands and Belgium. After the mid-1970s, the publicsector went back to expand at an extremely uniform paceacross all nations. In the mid-1980s, public revenue as aproportion of GDP experienced a slight contraction in bothdeveloped and developing countries.

(3) For more extensive reviews see Lybeck (1988), Holsey andBorcherding (1997).

(4) For sociological accounts of the process of economicmodernization, see Wilensky (1975) and Flora and Alber(1981). Among economists, resorting to explanations based onprocesses of technological and structural change has taken twomain forms. On the one hand, the so-called Wagner's law statesthat public expenditure rises with social progress because thetypes of goods and services provided by the public sector have ahigh income elasticity of demand. On the other hand, Baumol'scost disease predicts that the combination of similar real wagesincreases at both the public and the private sector and a lowerproductivity growth rate in the public sector (which is a servicesector and hence a relatively labor-intensive industry)compared to the manufacturing sector leads to an increase ofthe costs of government services in real terms over time.

(5) The extent to which politicians would tax the rich, however,is constrained by the extent to which excessive taxes on the tophalf discourage work among high earners and therefore depressthe total amount of income available for redistribution.

(6) In addition to median-voter models, other redistributivemodels explain the size of government as a result of particularpolitical coalitions among groups or, in the context of moderndemocracies, parties (with different ideologies) in government.If, given certain conditions, politicians do not converge towardsthe median voter, that is, if parties or political groups diverge,public expenditure can be expected to be determined by theparty or group in power (Hibbs 1987), regardless of the positionof the median voter. For sociological models that related publicspending and welfare states to the strength of a particular classor to cross-class alliances, see Esping-Andersen (1990).

(7) For an analysis on the dependence of the state on capital, seeHirschman (1981) and Przeworski and Wallerstein (1988).Thedegree of capital mobility varies in fact depending on the type ofcapital and is inversely related to the latter’s specificity. The lessspecific capital is (that is, the more alternative uses it can be putto), the more mobile it is, and the more power or influence capitalhas over the state. See Alt (1987) and Frieden (1991: 19-22).

(8) For initial studies on a limited number of cases, see Titmuss(1958) and Peacock and Wiseman (1961). For initial cross-

3534

Hibbs, D. A. Jr., 1987. The Political Economy of IndustrialDemocracies. Cambridge, Mass.: Harvard University Press.

Hicks, A. M., 1994. “Introduction to Pooling”. In TheComparative Political Economy of the Welfare State, eds.

Thomas Janoski and Alexander M. Hicks. Cambridge:Cambridge University Press.

Hicks, A. M. and D. H. Swank, 1992. “Politics, Institutions andWelfare Spending in Industrialized Democracies, 1960-82”,American Political Science Review 86, 658-74.

Hirschman, A. O., 1981. “Exit, Voice, and the State”, In AlbertO. Hirschman, Essays in Trespassing: Economics to Politicsand Beyond. Cambridge: Cambridge University Press, 246-265.

Holsey, C. M. and T. E. Borcherding, 1997. “Why DoesGovernment’s Share of National Income Grow? AnAssessment of the Recent Literature on the U.S. Experience”.In Dennis C. Mueller, ed. Perspectives on Public Choice: AHandbook. New York: Cambridge. 562-589.

Hsiao, C., 1986. Analysis of Panel Data. Cambridge:Cambridge University Press.

Huber, E., C. Ragin, and J. Stephens, 1993. “SocialDemocracy, Christian Democracy, Constitutional Structure andthe Welfare State”, American Journal of Sociology 99, 711-49.

Jackman, R. W., 1975. Politics and Social Equality: AComparative Analysis. New York: Wiley.

Katzenstein, P., 1985. Small States in World Markets:Industrial Policy in Europe. Ithaca: Cornell University Press.

Keohane, R. O. and H. V. Milner, eds. 1996.Internationalization and Domestic Politics. Cambridge:Cambridge University Press.

Korpi, W., 1989. “Power, Politics and State Autonomy in theDevelopment of Social Citizenship: Social Rights duringSickness in Eighteen OECD Countries since 1930s”, AmericanSociological Review 54, 309-28.

Lange, P. and G. Garrett, 1985. “The Politics of Growth:Strategic Interaction and Economic Performance in theAdvanced Industrial Democracies, 1974-1980”, Journal ofPolitics 47, 792-827.

Lipset, S. M., 1959. “Some Social Requisites of Democracy:Economic Development and Political Legitimacy”, AmericanPolitical Science Review 53, 69-105.

Lucas, R. E. Jr., 1990. “Why Doesn’t Capital Flow from Rich toPoor Countries?”, American Economic Review, 82, 1-13.

References

Adserà, A. and C. Boix, 1998. “Trade, Democracy and the Sizeof the Public Sector: The Political Underpinnings ofOpenness”, Political Economy Working Papers Archive.

Alvarez, M., J. A. Cheibub, F. Limongi and A. Przeworski,1996. “Classifying Political Regimes”, Studies in ComparativeInternational Development 31, 3-36.

Alt, J. E., 1985. “Political Parties, World Demand, andUnemployment: Domestic and International Sources ofEconomic Activity”, American Political Science Review 79,1016-40.

Alt, J. E., 1987. “Crude Politics: Oil and the Political Economyof Unemployment in Britain and Norway”, British Journal ofPolitical Science 17, 149-199.

Alt, J. E. and K. A. Chrystal, 1983. Political Economics.Berkeley: University of California Press.

Baldwin, P., 1990. The Politics of Social Solidarity. New York:Cambridge University Press.

Boix, C., 1999. “Democracy, Development, and the PublicSector”, Manuscript, The University of Chicago.

Brennan, G. and J. M. Buchanan, 1980. The Power to Tax:Analytical Foundations of a Fiscal Constitution. New York:Cambridge University Press.

Cameron, D. R., 1978. “The Expansion of the Public Economy:A Comparative Analysis”, American Political Science Review72, 1243-1261.

Campillo, M. and J. A. Miron, 1996. “Why Does Inflation DifferAcross Countries?”, Manuscript, Boston University.

Cheibub, J. A., 1998. “Political Regimes and the ExtractiveCapacity of Governments: Taxation in Democracies andDictatorships”, World Politics 50, 349-376.

Cutright, P., 1965. “Political Structure, Economic Developmentand National Social Security Programs”, American Journal ofSociology 70, 537-50.

Esping-Andersen, G., 1990. The Three Worlds of WelfareCapitalism. Cambridge: Polity Press.

Flora, P. and J. Alber, 1981. “Modernization, Democratizationand the Development of Welfare States in Europe”. In P. Floraand A. Heidenheimer, eds., The Development of Welfare Statesin Europe and America. London: Transaction Books.

Frieden, J. A., 1991. Debt, Development, and Democracy:Modern Political Economy and Latin America, 1965-1985.Princeton: Princeton University Press.

36

Lybeck, J. A., 1988. “Comparing Government Growth Rates:The Non-Institutional vs. the Institutional Approach”, inLybeck, Johan A. and Magnus Henkerson, eds. 1988.Explaining the Growth of Government. Amsterdam: North-Holland. 29-48.

Meltzer, A.H. and S. F. Richards, 1981. “A Rational Theory ofthe Size of Government”, Journal of Political Economy 89,914-927.

Mueller, D., 1988. Public Choice II. Cambridge: CambridgeUniversity Press. Chapter 17.

OECD, 1991. OECD National Accounts. Main Aggregates.1960-89. Vol. I.

Olson, M., 1993. “Dictatorship, Democracy, andDevelopment”, American Political Science Review 87, 567-576.

Pampel, F.C. and J.B Williamson, 1988. “Welfare Spending inAdvanced Industrial Democracies, 1950-80”, American Journalof Sociology 93, 1424-56.

Peacock, A. T. and J. Wiseman, 1961. The Growth of PublicExpenditure in the United Kingdom. Princeton: PrincetonUniversity Press.

Przeworski, A. and F. Limongi, 1997. “Modernization: Theoriesand Facts”, World Politics 49 (January).

Przeworski, A. and M. Wallerstein, 1988. “StructuralDependence of the State on Capital”, American PoliticalScience Review 82, 11-30.

Rodrik, D., 1998. “Why Do Open Economies Have BiggerGovernments?”, Journal of Political Economy 106, 997-1032.

Romer, D., 1993. “Openness and Inflation: Theory andEvidence”, Quarterly Journal of Economics 107, 869-904.

Titmuss, R., 1958. Essays on the Welfare State. London: Allenand Uwin.

Weingast, B., 1995. “The Economic Role of PoliticalInstitutions. Market-Preserving Federalism and EconomicDevelopment”, Journal of Law, Economics and Organization11.

Wilensky, H. L., 1975. The Welfare State and Equality.Berkeley: University of California Press.

5. Why does the Public Sector Grow?The Role of Economic Development, Trade and DemocracyCarles Boix (november 99)

Opuscles already published

1. Reconsidering Spanish UnemploymentRamon Marimon (june 97)

2. Reducing Unemployment. At Any Cost?Fabrizio Zilibotti (december 97)

3. Capital and Labor Taxes, MacroeconomicActivity, and RedistributionAlbert Marcet (november 98)

4. The Lender of Last Resort in Today'sFinancial EnvironmentXavier Freixas (november 99)

Ramon Trias Fargas, 25-27 - 08005 BarcelonaTel: 93 542 24 98 - Fax: 93 542 18 60E-mail: crei@upf.eshttp://www.econ.upf.es/crei

C E N T R E D E R E C E R C AE N E C O N O M I A I N T E R N A C I O N A L

Generalitat de CatalunyaDepartament de Presidència

Pts

. 1.

000

Carles Boix

Carles Boix holds a Master in Public Administrationfrom the John F. Kennedy School of Government(1990) and earned his Ph.D. in Political Science fromHarvard University (1995).

He teaches Political Science at the University ofChicago, where is co-director of the Nations, Stateand Politics Workshop. He has previously taught atthe Ohio State University and has been a visitingprofessor at Universitat Pompeu Fabra and the Centerfor Advanced Study in the Social Sciences in the JuanMarch Institute in Madrid.

His main areas of research are: political economy, thesocial and institutional basis of democratic performance,and the origin and development of political institutions.

Boix has published several articles in professionaljournals such as the American Political Science Review,the American Journal of Political Science, the BritishJournal of Political Science and Electoral Studies. Hisbook Political Parties, Growth and Inequality(Cambridge University Press, 1998) has won the 1999American Political Science Association Best BookAward for the best book on political economy.