Trade, Democracy, and the Size of the Public Sector

Home , Free Trade Party, Open economy, Public expenditure

Trade,Democracy, and the Sizeof the PublicSector: The Political Underpinnings of Openness Al´õ cia Adsera`and Carles Boix

Toaccount for thestrong and positive correlation found between trade openness andthe size of the public sector, scholars have developed theoretical explanations inwhichpolitics have remained conspicuously absent in two ways. 1 First,why some economiesare more open than others has been (implicitly) attributed to parameters exogenousto the political decisions of domestic actors. Second, the presence of a sizablepublic sector has been merely thought of asan automatic,and in somecases afunctional,requirement of a free traderegime. This mechanical link has come in twoforms. InRodrik shows that higher levels of trade integration (coupled with highsectoral concentration in the economy) increase the risks associated with the internationalbusiness cycle and call for publicly-nanced compensatory programs infavorof theexposed sectors. 2 Publicexpenditure, set by astatepurely conceived asa socialplanner, stabilizes aggregate income and delivers social peace and politicalstability. In Aukrust argues that the tradeable sector, modeled as an international-price-taker,employs public spending to buy the acquiescence of the non-tradeablesector to low wage increases, therefore ensuring the overall compet- itiveness(and survival) of the national economy. 3

Earlier versionsof this paper were presentedat theGerman-American Academic CouncilSummer Instituteat StanfordUniversity, the Department ofGovernment and Foreign Affairs oftheUniversity of Virginia,the Department ofPolitical Science at OhioState University,the Department ofEconomicsat OhioState University,the 1999 Annual Meeting of theMidwest Political Science Association,the 1999 MidwestEconomics Association Meeting, the Workshop on Politics,Nations, and States inComparative Perspectiveat theUniversity of Chicago, and the 2000 Annual Meeting of theAmerican PoliticalScience Association.We thankthe participants for their comments, particularlyGreg Caldeira,Dan Levin,Peter Gourevitch,Lloyd Gruber, Michael Hiscox, Dean Lacy,Peter Lange,Jim Peck,Ron Rogowski, and Herman Schwartz. 1.See Cameron 1978;Katzenstein 1985;and Rodrik 1998. 2.Rodrik 1998. 3.Aukrust 1977. Katzenstein includesboth arguments. Katzenstein 1985.In Cameron thelevel of publicspending is a functionof thepolitical power of unions, which in turnderives from the openness oftheeconomy. Cameron 1978.

Inthis article we modelthe choice of the level of trade openness and the determinationof the size of the public sector as two political decisions that are simultaneouslytaken by political agents. In this sense, we simplybuild upon the insightsof the literature on both the domestic determinants of tradepolicy and the originsof the welfare state,which respectively insist that the extent of trade opennessand the tax level derive from theparticular distribution of domestic interests.4 Oncethe choice of tariffs andtaxes is modeled as a simultaneouspolitical game, anymechanical correlation between trade openness and domestic compensation breaksdown. Domestic compensation emerges as just one possible strategy to build afree tradecoalition. Its useto securelow tariffs isneithernecessary (free trademay beachieved without increasing public spending) nor suf cient(a protectionist solutioncan still be imposedin spite of a compensatorypromise from thefree trade gainersto thelosers from free trade).More precisely, the following set of alternative political-economicequilibria constitute equally possible outcomes across the uni- verseof nations: 1.To insulate domestic actors from internationally-inducedchanges in relative prices,national policymakers may choose to close the domestic economy. Oncedomestic actors are relatively isolated from externalcompetition and the variabilityof the world business cycle, there are no incentives to resort to higherlevels of public expenditure to compensate voters for (temporaryor permanent)employment losses. 5 As aresult,economic insulation should be expected,ceteris paribus, to depress the level of public expenditure. 6 2.A secondpolicy outcome takes place whenever the sectorsthat gain from trade setup public programs to compensate trade losers to get the latter ’s political supportfor anopen economy. Accordingly, as we stressin this article, the expansionof thepublic sector is nota mere derivationof tradeopenness —as assumedin most of thecurrent literature — buta trulypolitical pre-condition neededto secure the liberalization of theeconomy. 3.Since the construction of apoliticalcoalition to securelow tariffs mayrequire hightaxes, policymakers may decide to keep the economy open while minimizingpublic expenditure. Since this solution may have substantial redistributiveconsequences, such as lower levels of welfare protectionand increasesin income inequality during economic downturns, this laissez-faire strategycan only be implementedafter excluding, in a systematicmanner, all tradelosers. In other words, as we shallshow, the combination of an open

4.On trade policy,see Altet al.1996; Gourevitch 1986; and Frieden and Rogowski 1996. On the welfare state, see Meltzer andRichards 1981; and Esping-Andersen 1990. 5.As amatter offact, oncethe economy is closed,aggregate demand management becomes an effective strategyto minimize theoccurrence ofrecessions. Policymakerscan thenengage in counter- cyclical policiesto smooth the business cycle. Forevidence showing that closed economies and in ation are stronglyand positively correlated, see Romer 1993;and Campillo and Miron 1997. 6. For a rst attempt topoint to autarkyas asubstitutefor domestic compensation under a free trade regime, see Bates, Brock,and Tiefenthaler 1991. TheSize of thePublic Sector 231

economyand no public compensation takes place, in most cases, under an authoritarianpolitical regime. We explorethese alternative combinations of trade policy, scalstrategy, and politicalregimes through a formalmodel, a teston a sampleof approximately sixty- venations from 1950to 1990, and historical evidence spanning a longer periodof time.The rst partof thearticle examines the choices policymakers make onboththe degree of openness and the size of thepublic sector. The model is based ona setof actorswith heterogeneous preferences in both policy dimensions (trade andpublic sector) due to theirexposure to tradeand the variability of theirpotential tradegains. Given a rathergeneric distribution of interests, we show, rst, that politiciansengaged in democratic competition may develop two alternative economicstrategies (to win elections): they either support a protectionistregime or defenda programthat combines free tradeand domestic compensation. The model thengoes on to allow politicians to choose, given the available set of policy alternatives,the political regime they would impose —eithera democracyor an authoritarianregime. This section shows under what conditions those sectors interestedin free tradehave an incentiveto impose an authoritarianregime to secure aliberaltrade regime without domestic compensation. In the second section, we test themodel on a samplethat includes all the countries for whichcomparable data on publicrevenue (current receipts) of the general government were availablefrom 1950to 1990.The statistical analysis is followedby theexamination of asetof key historicalcases from thenineteenth and twentieth centuries to bring out the dynamicsthat underlie the choice of tariffs andtaxes under different political regimes.In the last section, we draw ourconclusions.

Theory:Trade, Taxes, and Political Regime TheDistribution of Interests As shownin the literature on the political determinants of trade policy, different economicagents have different material interests, and therefore distinct policy preferences,depending on theirposition in the international economy. The attitudes ofeconomicactors toward trade and trade policy may be determinedby twofactors: rst,the potential bene tsto be derivedfrom trade,which go from netgains to net losses;and second, the variability of the returns derived from trade,that is, the probabilitythat trade will lead to gains or losses given the nature of the actor ’s assets.This second component simply captures the degree of uncertaintythat actors faceover the future realization of some gains (or losses)in trade and their incapacity,once they have embraced a free traderegime, to use standard counter- cyclicalpolicies to combat the volatility of the world business cycle. Inits simplest form, consideran economy in which there are three types of individualswith distinct trade interests, C, P, and O.Eachindividual can be thought ofasa representativeagent of aneconomic factor, a productivesector, or asetof homogeneous rms. Thismodel, therefore, takes the de nitionof each actor ’s 232 InternationalOrganization positionin the international economy and its interests as given —it does not adjudicatebetween factor-based or sector-based models of international political economyand should be applicable to any of them. 7 C,orthepro-closed-economy agent,only stands to lose from openingthe economy — hisreturns decline, in a linearfashion, as tariffs arereduced. In a Heckscher-Ohlinmodel, for example, C wouldbe an owner of capitalin a capital-scarceeconomy. Formally, the utility of C canbe represented as:

Uc 5 ~1 2 l !C c 1 l Co (1)

where l isthe extent to which the economy is open in an interval [0,1], Cc is the return to C whenthe economy is closed, Co isthe return to C whenthe economy is open, and Cc . Co. O,instead,consists of the agent who always gains from free trade — his returns alwaysrise as theeconomy becomes more open. Again, within the framework ofa Heckscher-Ohlinmodel, O wouldbe alandownerin a land-abundantcountry. The utility of O is de ned by:

Uo 5 ~1 2 l !Oc 1 l O o (2)

where Oc isthe return to O whenthe economy is closed, Oo isthereturn to O when theeconomy is open, and Oc , Oo. Finally, P representsan agent whose gains from tradeare always subject to considerablevariation over time. If theinternational business cycle (or thespeci c demand for P’sproducts)is booming, P bene tsfrom afree traderegime. If the internationalbusiness cycle is in recession, free tradeharms P substantially. Accordingly, P ranksher preferences in the following fashion: an open economy undera goodbusiness cycle is preferredto a closedeconomy regime, yet this latter regimeis certainly more attractive than a laissez-faireregime when the economic cyclegoes through a recession. 8 Theexpected utility of P canbe represented as:

7.For a discussionof thede nitionof each actor ’sinterests,see Friedenand Rogowski; and Alt et al. 1996. 8.The presence of P captures inpart the empirical fact thattrade is correlated tovariations in the businesscycle across countries.Using the Summers-Heston (1988) database (thatincludes 147 nations forthe period 1950 –90and 4,546 observations), the volatility of the business cycle (calculated as the standarddeviation of changesin thegrowth rate in5-year periods) can beshownto increase withtrade openness.More speci cally,for each loggedunit of trade openness,the volatility of thebusiness cycle goesup by 0.60 (which is statistically signi cant at the1 percentlevel). The result is robust to the introductionof controlvariables such as percapita income,economic structure, and weight of fuel and primaryexports. TheSize of thePublic Sector 233

EU p 5 ~1 2 l !Pc 1 l @p Pg 1 ~1 2 l !P b# (3) where Pc isthe return to P ifthe economy is closed; Pg isthe return to P if the economyis openand the international business cycle is booming; Pb isthe return to P iftheeconomy is openand the world business cycle is inrecession;the ranking ofreturnsis Pg . Pc . Pb; and p istheprobability that the world business cycle willbe expansive. For any P theexpected return under an open economy is smaller than under a closedeconomy:

p Pg 1 ~1 2 l !P b , Pc (39)

For sakeof simplicity,we haveassumed that P isrisk-neutralso thather expected utilityequals her expected return. If we hadcharacterized P asrisk-averse, P’s demandsfor compensationfrom O toacquiesce to free tradepolicies would have becomemore intense. 9 Therewould have been a setof pointsfor which,although P’sexpectedreturns would have been higher under an open economy (and thus condition[3 9]wouldnot hold), P wouldhave still preferred a closedeconomy to minimizerisks unless a compensationpolicy would have been promised by O. Theposition of theseeconomic actors affects, in turn, their attitudes toward trade policy. C and O espousecompletely different policy positions. Whereas C favors autarky, O isin favor of full free trade.In turn, P’spositionis conditional on the evolutionof theworld business cycle and, more interestingly, on theintroduction of somemechanism that could smooth the variability of the business cycle. If P can securea futurecompensation package (via public transfers) tomake up for any lossesin recession periods, its willingness to support free tradepolicies will increase.Conversely, if the public sector has devised no compensatorymechanisms, P’ssupportfor free tradewill decline. The position of eachagent is represented in Figure 1. C and O havetheir bliss points at theextreme opposites of the trade policy dimension(horizontal axis). Their support for eitherprotectionism or free tradeis independentof thelevel of compensation they may get. 10 Theirindifference curves (witha lossof utilityhigher along the horizontalthan on thevertical axis for O) have beendrawn to stressthat trade policy matters more than the level of compensation. P’ssupportfor free tradecannot be separatedfrom hisdemand for acompensatory mechanism.Thus two different equilibria may ensue. A coalitionbetween C and P wouldhinge upon high tariffs andno compensation mechanisms. Instead, in

9.With risk aversion, P’sexpectedutility function is:

EUp 5 p U@~1 2 l ! Pc 1 l Pg# 1 ~1 2 p !U@~1 2 l ! Pc 1 l Pb# where U9 . 0 and U99 , 0. 10.It could be argued, however, that, other things equal, O may havea higherinterest in getting some compensationgiven its higher exposure to trade —thatis, that O may alsobe affected bythe variability oftheterms oftrade. This possibility is notmodeled here forthe sake ofsimplicity. 234 InternationalOrganization

FIGURE 1. Supportfor free trade in a two-dimensionalpolicy space

exchangefor publiccompensation, P couldbe equally willing to support O’s free tradepolicy preferences.

TheChoice of Trade and Fiscal Policies under a Democratic Arrangement Considernow how both trade and scalpolicies are jointly determined in that economy.Political decisions take place through two procedures that may occur sequentiallyor simultaneously.First, policies are or maybe chosenthrough a voting ordemocratic mechanism. Second, (some) actors,unsatis edwith the actual or anticipatedoutcome in a democraticsetting, may contest the voting procedure and establishan authoritarian regime that, by excluding any opposition, secures their mostpreferred policy outcome. (See theappendix for theformal derivation of the equilibria.) Letus consider rst whathappens in a democraticregime. (In thissubsection we xtheconstitutional regime. In the next subsection, we explorethe policy outcomes whenwe allowactors to contest the regime itself.) At nationalelections held periodically, C, O, and P maychoose among two parties. A protectionistparty, party S,whichdraws its nancialand electoral support from C (thepro-closed economy sector),favors high tariffs. The pro-free trade party, party V,whichreceives its supportfrom pro-free tradevoters, O,defendsa free traderegime. Neither C nor O TheSize of thePublic Sector 235 controlsa majorityof the vote. 11 Accordingly, P,whois the median voter (in the trade-policydimension), holds the balance between the two forces. 12 Thesequence of the game between the economic actors ( C, O, P) and the two contendingparties ( S, V)isasfollows. Both parties, S and V,choosetheir policy programsbefore elections. Once they have committed to a certainpolicy strategy, theycannot credibly alter their initial promises. 13 Thechoice of their electoral promisesis made under conditions of uncertainty about how the international businesscycle will be after the election takes place. Voters then vote for parties — dependingon theirprograms and their forecast about the future. After theelection, theworld business cycle expands with a probability p andis in recession with a probability1 – p .Arecessionaffects all the P votersat the same time. 14 If neither C nor O hasa majorityof thevotes, party 3 hastwo possible choices. Onthe one hand, it mayjust promise low (or no)tariffs. On the other hand, it may promiselow (or no)tariffs jointlywith the introduction of apublicly-fundedscheme tocompensate the P sectorin case a worlddepression occurs. In turn, S always promisesan autarkic regime. S doesnot develop any compensatory mechanism relatedto tradesince, under a closedeconomy, there is novolatilityinvolved. Still, theprotectionist party may promise a transferto match the offer ofthe free trade party. Todetermine the strategy of parties and the nalresult of elections, let us examine rst howvoters vote. For thefree tradeparty V to win, O and P will have tovote for it.Consider rst thevote of the pivotal voters, P.At electionday, P votersdo notknow the evolution of thebusiness cycle. Their expected returns will bewhat they would gain in an open economy (and these gains are affected by the variationin the economic cycle) or what they would gain in a closedeconomy. P voterswill vote for thefree tradeparty, V,ifthe returns under an open economy plusthe compensation package this party promises are larger than the returns to be obtainedin a closedeconomy jointly with any transfers established by the protectionistparty S. Whichstrategy will V pick?Given that the ex ante expectedgains from openness arelower for P thanthe certain gains from autarky(as statedin condition[3 9]), V mustalways commit to a certaincompensation package, a o, to attract P. Such a

11.The lack ofamajorityof Cs or Os(andthe corresponding pivotal role of P)is aplausibleway ofcapturing the possibility of volatility, which we havefound to be inherently related totrade across countries.Again, C(O)always bene ts fromprotectionism (free trade). P’sreturnis conditionalon the businesscycle. 12.A similar situationarises if,instead of an electoral arena, we thinkof aparliamentin whichthe three agentshave attained representation in a proportionequal to their electoral strength.The party representing P wouldbe the median parliamentarian. 13.That is, since voterstake intoaccount the reputation of theparty, parties cannotmodify their party platformsonce in government to tailor them tocyclical economicconditions. 14.Altering the sequence ofthe game (withnature —thebusiness cycle —occurringafter parties commit butbefore voters vote) and the impact oftherecession on P (with every P havingan independent probability1 – p ofsufferinga recession insteadof all Psbeingaffected byitat thesame time), does notsubstantially change our results. 236 InternationalOrganization

packagehas to meet two criteria. First, the compensation program a o will have to outmatchany spending package a c promisedby the protectionist party, S. (Notice that C’scostof this compensation package a c cannever be higher than the differencebetween what each C gainsunder a protectionistregime and what he gainsunder free trade.If itwere, C voterswould not vote for theprotectionistparty.)

Second,the compensation package a o promisedby the free tradersshould not alienatethe O voters— otherwisethey would not vote for thefree tradeparty. (In a waysymmetrical to what happens to the protectionist parties and their core supporters, O voterswill only vote for V upuntilthe point in which the gains of an openeconomy are larger than the gains of aclosedeconomy even allowing for the compensationthey may have to pay to each P voter.)Finally, it is important to stress thatthe compensation package promised by thefree tradeparty will always be larger thanthe spending package offered by theprotectionist party, since the former hasto overcomethe lower expected return of thepivotal voter in anopeneconomy relative toa closedeconomy. In short, free trade(resulting from democraticelections) booststhe public sector more than protectionism. Severalimplications about what determines the winner of the electoral stage can bederivedfrom ourdiscussion. First, the size of eachgroup matters. As thenumber of O votersincreases, the free tradeparty is more likely to win an election. 15 Second,the returns to eachfactor or sectoralso matter. As thegains from free trade increase, O willhave more resources to pay off P voters,and the victory of V becomesalmost certain. 16 Finally,as thevolatility of the business cycle diminishes, thecost to lure P voterswill decline, and a free traderegime will become easier to establish.

Trade,Fiscal Policy, and the Choice of Political Regime

Ina democraticsetting, unless O commandsa majorityof votes, securing a free traderegime has a politicaland nancialcost. To havean open economy, O has to compensate P.If thecost involved in ensuring that P votesfor apro-free tradeparty is too high, O willchoose not to promiseany compensation. But, again, this decision will cost V theelection and will lead to the passage of protectionist policies.

15.In the presentation of themodel we haveassumed fullparticipation of voters.Naturally, if turnout declines(asymmetrically) across sectors, thepolicy regime may vary.Similarly, differential rates of participationcan beequated to different levels oflobbying effort across sectors (resultingin different proportionsof votes). 16.Notice thatthese twoimplications can becombined to predict the likelihood and traits ofthe policyoutcome in the following way. Assume afactor-basedmodel of trade preferences. Inlabor- abundanteconomies, the majority of thepopulation (at least inasettingas describedby Mayer)will lean towardfree trade and,as aresult,compensation policies may besmall insize (Mayer1984). By contrast, inlabor-scarce economies,where themajority loses fromtrade, compensation packages willtend to be substantial.The goal of this article is totest howtrade andpolitical regime affect scal policy.Still, we tentativelyconsider the impact oftheabundance/ scarcity oflabor in aset ofregressionsreported below innote 30. We are gratefulto Michael Hiscox for having drawn our attention to thispossibility. TheSize of thePublic Sector 237

Adifferentpolitical alternative is, however, still available to O. Instead of acceptingan electoral defeat, O maydecide to impose an authoritarian regime to securefree tradepolicies. Naturally, a similaroption is available to C. If S cannot winthe election, it maystill resort to adictatorshipto secure a protectionistregime. (Acaveatis inorder.In what follows, we donotclaim that the level of con ict over tradepolicy solely determines the choice of political regime. There is already an abundantliterature on those factors —mainlyeconomic modernization —that affect thelikelihood of ademocraticoutcome. 17 Accordingly,we simplyoutline a theory asto why, in the margin, with all the other relevant factors constant, trade con ict maylead to either a democracyor an authoritarian regime, and with what policy consequences.) Underwhat conditions will the two parties resort to an authoritarian solution? Supposethat there is a costto impose a dictatorship —w orthe cost of repressing eachmember of those economic sectors that are permanently excluded from government. O willimpose a dictatorshipif three conditions are met: 1. The nalburden to each O ofimposing a dictatorshipdoes not exceed the differencebetween the gains under an open economy and the gains under a closedeconomy. If theydid, a dictatorshipwould absorb all the economic advantagesof establishingit, and naturally, O wouldhave no logicalincentive topursue the authoritarian strategy. 2.The resources of O arelarger than the resources C hasto establish a dictatorship —thatis, when a pro-free tradedictatorship cannot be contestedby theprotectionist sector. Otherwise, C isable to impose an uncontested dictatorship. 3.Finally, whenever O couldcompensate P, O willimpose a tyrannyif this solutionis cheaper than paying for acompensationpackage. If compensating P were notfeasible for O, O wouldestablish a dictatorshipwhenever conditions(1) and(2) occurred. Solvinganalytically these three conditions, it is apparent that the outcome will hingeon thenumber of individuals in eachsector, the difference in thegains of each sectorunder each policy regime (open or closed),and the variability of the cycle in thefollowing way. 18 Consider,to start with, a scenariowhere O and C experiencesymmetrical gains (losses)as aresultof trade policy. That is, the losses that one sector (for example, O)experienceswhen policy is shifted (for example,going from free tradeto protectionism)are similar to thegains the other sector ( C inthiscase) enjoys. In this case,a democraticoutcome is unlikely:a dictatorshipis generally cheaper than the compensatorypolicies that must be enactedafter an election. The capacity of each sectorto impose a dictatorshipis balanced. Who prevails simply depends on their respectivenumbers. As C becomeslarger in numerical terms (relative to O), a

17.See Lipset1959; Rueschemeyer, Stephens,and Stephens 1992; Przeworski and Limongi 1997; andBoix and Garicano 2001. 18.The solutions are formallyshown at theend of theappendix. 238 InternationalOrganization protectionistoutcome will be established.Conversely, as O becomeslarger, a free tradedictatorship will be imposed (with the public sector remaining small). The variabilityof P’sgainshas no appreciable impact on O’sstrategy:even if the variabilityof P’sgainsis low (and the compensation package to lure P is small), democracyis notgenerally pursued by thefree tradesector. The protectionist sector caneasily outmatch any compensation since it has similar resources at stake. Resultschange when O standsto gain relatively more from opennessthan C from autarky. O’scapacityto imposean uncontesteddictatorship increases steadily (since itcanamass resources rapidly). But, interestingly, the gains of trademake democracy(with compensation) cheaper at a fasterrate. Thus democracy and free trade startbecoming the dominant strategy for O,evenin those circumstances in which adictatorshipalso could be imposed by O.Thiseffect, though weaker, holds even when P’s compensationpackage must be relatively high.

Empirical Evidence

Toovercome the absence of a trulypolitical explanation for thesimultaneous occurrenceof openness and a sizablepublic sector in a substantialnumber of countries,in theprevious section we modeledthe joint selection of tradepolicy, the sizeof the state, and political regime for anycountry. We showedthat three distinct political-economicequilibria may occur. First, in nations that hardly trade, the publicsector should be smaller, other things being equal. Again, the reason is straightforward:in aclosedeconomy policymakers have no incentiveto expandthe publicsector to compensate the losing sectors of the economy. Second, in fully democraticpolities, a free traderegime leads to the expansion of thepublic sector tocompensate losers from internationalrecessions or shocks. Finally, since open economiescan only shirk from committingresources to losing sectors of the economyby excluding them from thepolitical decision-making process, free trade willnot imply a biggergovernment under authoritarian regimes.

Sample Toexamine the strength of our theory, we usea samplethat includes all the countriesfor whichcomparable data on public revenue (current receipts) of the generalgovernment are available from 1950to 1990. We usedtwo sources for our sample:the United Nations (UN) NationalAccounts andthe Organization for EconomicCooperation and Development (OECD) NationalAccounts .19 The data startin 1950 and cover approximately 65 countries(22 are OECD members),with somevariation in the years covered, providing about 2,000 data points.

19.United Nations, various years. TheSize of thePublic Sector 239

DependentVariable

Thedependent variable “Sizeof Government ” ismeasuredthrough current receipts ofthegeneral government as a percentof grossdomestic product (GDP). 20 Current receiptshave been chosen over public expenditure to maximize the sample under analysis.Current disbursements of general government as a percentageof GDP (takenfrom theUN NationalAccounts )includemuch fewer countries(and, by excludingcapital spending, underreport the size of government). Although two otheralternative databases offer largersamples for partsof publicexpenditure, they arealso unsuitable for ouranalysis. Although the Penn World Tables report the shareof government consumption of more than a hundredcountries, government consumption,which is themeasure recently employed to testthe effect of tradeon publicspending, represents a fractionof allgovernment spending and is biased in favorof developing countries. 21 TheWorld Bank reports levels of government spendingfor, at most, around eighty countries. Still, these data (as wellas the InternationalMonetary Fund data) report spending only at the central government leveland lead to extremelybiased values for federalcountries (which tend to beless openon average).

Modeland Results

Todetermine the impact of tradeand the interaction of trade and political regime on thesize of government, we rst estimatethe following model on a time-series cross-sectionalpanel of nations in Table 1: Sizeof Government t 5 a 1 a 1Economyt 1 a 2Tradet 1 a 3PoliticalInstitu- tionst 1 a 4Tradet*PoliticalInstitutions t 1 e t We estimatethe pooled cross-sectional time-series model through ordinary least squares,adjusting the standard errors for unequalvariation within panels and correctingfor autocorrelation. 22 Table2 checksfor therobustness of theresults of Table1 inseveralways. It rst introducesthe lagged value of thedependent variable andthen re-estimates the model using both xed-effectsand random-effects procedures. InTable 3 we estimatethe same model using change rather than level of the parameters,that is: Changein Size of Government 5 a 1 a 1Sizeof Government t-1 1 a 2Change in Economy 1 a 3Changein Trade 1a 4Changein Political Institutions 1 a 5Change in (Trade*Political Institutions) 1 Gt

20.According to theUN/ SNA,the current receipts includeoperating surplus, property and entrepre- neurialincome, taxes, fees andcontributions, and other current transfers. 21.Rodrik 1998. 22.Beck andKatz 1995. 240 InternationalOrganization

TABLE 1. Thesize of governments across the world: trade, development, and politicalregimes

Publicrevenue as percent of GDP, 1950 –1990

Independentvariables (1) (2) (3) (4) (5) (6)

Constant 240.26*** 237.39*** 238.56*** 237.47*** 221.62** 239.58*** (5.54)(4.99) (6.40) (6.46) (9.31) (6.83) Per capitaincome 6.20***6.15*** 6.72*** 6.68*** 6.90*** 7.35*** (log)a (0.51)(0.51) (0.71) (0.71) (0.86) (0.73) Trade openness(log of 3.95***3.30*** 2.18*** 2.13*** 21.24 2.37* sumof exports and (0.54)(0.79) (0.86) (0.86) (1.42) (1.28) importsover GDP) b Democratic 2.69*** 22.35^ 24.87 24.69 28.87* 2.43 institutions c (0.68)(3.46) (3.41) (3.40) (4.81) (4.94) Democratic institutions 1.33^ 2.19**2.18** 3.22*** 20.14 3 Trade openness (0.96)(0.94) (0.94) (1.31) (1.37) Logarea (thousand 0.96***1.05*** 0.91** 0.00 km2) (0.30)(0.32) (0.39) (0.38) Distance 20.71** 20.70** 20.84** 20.68** (0.28)(0.28) (0.40) (0.32) SubsaharanAfrica 2.44 2.30 21.17 4.21 (2.48)(2.45) (3.11) (3.20) East Asia 21.52 21.33 26.12** 21.05 (2.13)(2.10) (2.81) (2.54) LatinAmerica 24.67** 24.92** 29.31*** 22.92 (2.21)(2.18) (3.02) (3.21) OECD countries 21.91 22.90 24.82* 20.19 (2.15)(2.25) (2.59) (2.50) Weightedlevel of 23.92* tariffs on (2.08) intermediate inputs andcapital goods Exportconcentration 230.08** (12.79) Exportconcentration 7.97** 3 Trade openness (3.14) Exportconcentration 0.55^ 3 Trade openness (0.87) 3 Democratic institutions Volatilityin terms of 4.98^^ trade (14.12) Volatility 3 Trade 23.15^^ openness (7.32) Volatility 3 Trade 20.82^^ openness 3 (2.86) Democratic institutions ModelChi-square 374.76 220.94 644.02 685.05 553.94 418.56 P . Chi-square 0.00000.0000 0.0000 0.0000 0.0000 0.0000 Number of 2006 2006 1602 1602 958 943 observations

Notes: aPer capita incomeis log of per capita GDPin $ in1985 constant prices. Source: PennWorld Tables. bTrade opennessis logof thesum ofexportsand imports over GDP. Source: PennWorld Tables.

(continued) TheSize of thePublic Sector 241

TABLE 1. continued cDemocratic institutions.Five-year average ofdemocratic institutions.Variable goesfrom 1 (democracy inthe previous 5 years) to0 (non-democracyin the previous veyears). Average calculated fromdata inAlvarez et al.1996. Estimation:Ordinary least squares estimation,with panel-corrected standard errors, and correction forautocorrelation and for heteroskedastic disturbances between panels. Standarderrors inparentheses. ***p , .01 **p , .05 *p , .10 ^Jointtest signi cant withseparate componentsand interactive terms signi cant at .05level. ^^Jointtest signi cant withseparate components,interactive term andopenness, signi cant at .05 level.

Theestimates in Table 3 canbe read as making direct predictions about how changesin openness and regime lead to changes in the size of the public sector — ratherthan giving equilibria values as in Tables 1 and2. Column1 inTable 1 reportsthe results of theestimation for thebaseline model inpreviousstudies on trade and the size of thepublic sector. It includes economic development,trade, and political regime. Economic development is measured as the logvalue of realper capita income (in constant dollars, Chain Index, expressed in internationalprices, base 1985), taken from thePenn World Tables. Trade corre- spondsto ameasureof theimpact of opennesson governments,calculated as thelog valueof theratio of trade(sum ofimportsand exports) to GDP, andis taken from thePenn World Tables. 23 PoliticalInstitutions is a variablethat indicates whether eachcountry was acompetitivedemocracy in the previous ve years—and thus rangesfrom 0(nodemocracy ever) to 1(democracyin the veyears).To measure thepresence of a democraticregime, we followthe index developed by Alvarez, Cheibub,Limongi, and Przeworski and the classi cationreported in appendix 1 of theirarticle. 24 Democraticregimes are de nedas those regimes “inwhich some governmentalof ces are lledas a consequenceof contested elections. ”25

23.Openness rather thantariff (andnontariff) barriers are employedfor two reasons. On the one hand, we are interestedin measuring the potential effects thattrade opennessmay haveon the size ofthepublic sector.Although tariffs are naturallyemployed to alter theextent of openness, there are cases where, althoughtariffs are low,the degree ofexposure to the world business cycle is relativelylow (for example, theUnited States dueto the size ofthe country) and hence politicalpressure fordomestic compensation is low.On theother hand, given the increasingly marginal role of tariffs, the direct measurement oftrade policyinstruments is substantiallyproblematic. Still, as shownbelow, we havealso attempted to separate thenatural from the policy-induced levels oftrade openness. 24.Alvarez et al.1996. 25.Ibid., 4. We havealso regressed thedependent variable on avariablethat indicates whether each countrywas a ‘bureaucracy ’ each year; avariablethat indicates whether each countrywas an ‘autocracy’ each year; anda variablethat indicates whether each countrywas independenteach year.Bureaucracies are thosedictatorships that have legislatures. Autocracies are thosedictatorships that do not and that thereforecan bethought of as nothaving any sort of institutionalizedrule for operating the government. Thepresence ofautocracies andbureaucracies is alsobased on the index developed by Alvarez et al. 1996. 242 InternationalOrganization

TABLE 2. Therobustness of the effects of trade and political regime

Publicrevenue aspercent ofGDP,1950 –1990

Fixed Fixed GLS GLS Beck-Katz effects no effects random no random lagged lag lagged lag lagged Independentvariables revenue revenue revenue revenue revenue

Constant 22.20* 265.78*** 27.49*** 258.90*** 25.06*** (1.15) (2.88)(1.71) (4.51)(1.44) Publicrevenue t 2 1 0.94*** 0.89*** 0.92*** (0.01) (0.01) (0.01) Per capita income(log) a 0.33** 11.77***1.07*** 11.32*** 0.70*** (0.16) (0.41)(0.25) (0.39)(0.18) Trade openness(log of sum 0.18^ 21.49**0.43* 20.91 0.27^ ofexportsand imports (0.26) (0.60)(0.29) (0.57)(0.22) over GDP)b c Democratic institutions 20.55^ 221.48*** 21.34^ 220.70*** 21.03^ (1.01) (2.48)(1.25) (2.41)(1.00) Democratic institutions 3 0.27^ 6.35***0.49^ 6.16***0.40^ trade openness (0.26) (0.65)(0.33) (0.63)(0.26) Logarea (thousandkm 2) 0.08* 1.07***0.14** (0.05) (0.35)(0.06) Distance 20.06 20.66** 20.06 (0.04) (0.33)(0.05) SubsaharanAfrica 0.49 2.10 0.67 (0.34) (3.45)(0.61) East Asia 0.07 22.32 0.13 (0.27) (3.28)(0.57) LatinAmerica 20.09 29.69*** 20.25 (0.30) (3.44)(0.61) OECD countries 0.22 210.68***0.07 (0.31) (3.20)(0.59) ModelChi-square 63562.2 P . Chi-square 0.0000 R-squared 0.587 0.968 0.665 0.971 Numberof observations 1493 1605 1494 1605 1494

Notes: aPer capita incomeis log of per capita GDPin $ in1985 constant prices. Source: PennWorld Tables. bTrade opennessis logof thesum ofexportsand imports over GDP. Source: PennWorld Tables. cDemocratic institutions.Five-year average ofdemocratic institutions.Variable goesfrom 1 (democracy inprevious 5 years) to0 (non-democracyin theprevious veyears). Average calculated fromdata inAlvarez et al.1996. Estimation:Ordinary least squares estimation,with panel-corrected standard errors, and correction forautocorrelation and for heteroskedastic disturbances between panels. Standarderrors inparentheses. ***p , .01 **p , .05 *p , .10 ^Jointtest signi cant withseparate componentsand interactive terms signi cant at .05level. TheSize of thePublic Sector 243

TABLE 3. Theimpact of changes in political regime and trade on rates of changeof public revenue

Publicrevenue aspercent of GDP, 1950–1990

Independentvariables (1) (2)

Constant 0.04*** 0.06*** (0.01) (0.02) Lagof levelof public revenue 20.001*** 20.001*** (0.000) (0.000) a Changein log of per capita income 20.02 20.49 (0.51) (0.62) b Changein log of exports and imports over GDP 20.64 20.54` (0.73) (0.76) c Changein political regime 21.16 21.00` (0.72) (0.92) Changein interaction of trade opennessand political regime 1.14* 0.98` (0.71) (0.74) 2 Logarea (thousandkm ) 20.000 (0.000) Distance 20.001 (0.002) SubsaharanAfrica 0.001 (0.033) East Asia 20.005 (0.013) LatinAmerica 20.001 (0.018) OECD countries 0.009 (0.013) ModelChi-Square 43.90 49.84 Prob . Chi-Square 0.0000 0.0000 Numberof observations 1267 1050

Notes: aPer capita incomeis logof percapita GDP in$ in1985 constant prices. Source: PennWorld Ta- bles. bTrade opennessis logof thesum ofexportsand imports over GDP. Source: PennWorld Tables. cDemocratic institutions.Five-year average ofdemocratic institutions.Variable goesfrom 1 (democracy inprevious 5 years) to0 (non-democracyin theprevious veyears). Average calculated fromdata inAlvarez et al.1996. Estimation:Ordinary least squares estimation,with panel-corrected standard errors, and correction forautocorrelation and for heteroskedastic disturbances between panels. Standarderrors inparentheses. ***p , .01 **p , .05 ^Injoint test oftheinteractive term andits separate components,statistically signi cant with 2 Prob . x 5 0.00. 244 InternationalOrganization

Economicdevelopment and trade openness affect positively the size of government.Other things being equal, public revenue goes up from 20percentof GDPfor arealper capita income of $1,000 to about 30 percent of GDP for apercapita incomeof $5,000 and more than 40 percent of GDP for apercapita income of $25,000.26 Theresults con rm aconsiderableliterature relating the size of the publicsector to economicdevelopment. 27 For trade,in turn,public revenue goes up from around23 percent of GDP ina closedeconomy (where exportsand imports equal10 percentof GDP) toalmost30 percent of GDP(if tradeopenness equals 50 percentof GDP) andup to almost 35 percent when openness is very high (around 150percent of GDP). 28 Theresults con rm theexisting literature on the public sectoras a compensatorymechanism in open economies. 29 Thepresence of democraticinstitutions has a slightlypositive effect on the size of revenue —it is 2.69percent of GDP higherif the country is democratic. Totest for themodel developed in sections 1 and2, which showed that trade shapesthe size of thepublic sector conditional on thepolitical institutions in place, we add,to thebasic model of Table1, theinteractive term “DemocraticInstitutions *TradeOpenness ” (Table1, column 2). Per capitaincome and trade openness continueto boost public revenue. The presence of democratic institutions slightly reducespublic expenditure. But this result has to beset against the positive sign of theinteractive term. Given how strongly correlated it isto openness and democracy, itisnotstrange that the interactive term is only signi cantat the 16 percent level. Still,a jointtest with openness and democracy shows that it is statistically 2 signi cant (Prob . x 5 0.0000).In short, as tradegrows, the public sector grows indemocratic regimes. 30

26.This is estimated bysetting trade opennessequal to 50percent and assuming there isademocratic regime. 27.See Cutright1965; Wilensky 1975; Hicks andSwank 1992; and Huber, Ragin, and Stephens 1993. Exceptfor the weight of the primary sector andthe dependency ratio, all othereconomic and demographicvariables related todevelopment —theurbanization rate, thepercentage ofthe labor force inthe primary sector andin the manufacturing sector, the percentage ofagricultural land, and the proportionof thepopulation below 15 and25 —are notstatistically signi cant.Since trade is thevariable underanalysis in thisarticle, we haveonly included per capita incomein our regressions here. For a full analysisof the impact ofeconomic modernization in all itsdimensions on thesize ofthepublic sector, see Boix2001. 28.This is estimated bysettingper capita incomeequal to $4,000 and assuming there isademocratic regime. 29.See Katzenstein 1985;and Rodrik 1998. 30.As discussedin note 16 above,the size ofpubliccompensation to losers may berelated totheir size andtype. More speci cally,compensation may belarger inlabor-scarce economiesthan in labor-abundantones since inthe former alarger group(those deriving their income from wages) are hurt bytrade. Although this article isfocused on the impact oftrade across regimes, we havetentatively examinedthe impact oflaborabundance on the size ofthe public sector intwoways. First, we haveadded tothe model reported in column2 inTable1 aninteraction of per capita andtrade opennessunder the assumptionthat richer nationshave higher capital/ laborratios. Second, we haverun a modelwith the log ofcapital stockper worker (as reportedin the Penn World Tables), alone, in combination with trade opennessand in interaction with openness and democracy. As expected,the size ofthe public sector growswith capital abundance(that is, with scarce laborand then a more pressingneed to compensate). TheSize of thePublic Sector 245

Thebenchmark regression of column2 isrobustto the introduction of geograph- icaldummies, the log value of geographicalarea and the distance of each country from themajor world exporters —thisis done in column 3. The introduction of continentaldummies, a standardprocedure in economicstudies, controls for speci c effectsof particularworld regions. 31 Theinclusion of thelog value of geographical areaof each country and the distance to the twenty major exporters serves two purposes.First, the log of area controls for thepossibility that the provisionof public servicesis subject to economies of scale. Since area and trade openness are correlated,one could not exclude the possibility that the correlation between opennessand public sector could be aspuriousone due to other factors. However, asshown in column3, theeffect of tradeopenness is notaffected by theintroduction ofarea. As amatterof fact, the log of area positively enters the regression, disprovingthe economies of scale argument. Second, the log of area, jointly with distance,controls for theexogenous or structuralcomponent of tradeopenness. As shownby Frankel and Romer, distance, geographical area, and population explain asubstantialportion of the share of exports and imports in each country ’s economy.32 Thustheir inclusion in the regression as controlvariables allows us to measurethe impact that trade openness, as a policychoice, has on the size of the publicsector. 33 Threeresults emerge from theirintroduction. First, the coef cient of tradeopenness declines in size(an unsurprising result given the impact of areaand distanceon theopenness) but remains signi cant.More importantly, the interactive termof trade and political regime remains signi cantand has a slightlylarger impacton the size of government. Finally, notice that size has a positiveeffect (ratherthan the negative effect we wouldexpect from theidea that smaller, and hencemore likely to beopen,countries will spend more) anddistance has a negative butnot substantial effect (one standard deviation in distance changes public revenue byonly 1.4 percentage points of GDP). Inour opinion, these results mean that domesticcompensation is mainly driven by political decisions (as capturedby openness,net of the structural elements, and the interactive term) ratherthan by structuraldeterminants (estimated through area and distance). Figure2 simulatesthe results of Table 1, column 3. The simulation shows the evolutionof currentpublic revenue when trade openness goes up for acountryof a percapita income equal to $4,000. When trade openness is low, that is, when the sumof exports and imports amounts to 10 percent of GDP, publicrevenues are about23 percent of GDP inboth democratic and authoritarian regimes. As the economyopens, the size of the government increases at different rates. In author-

Ourmain resultson trade anddemocracy remain unaffectedby the inclusion of thatvariable. Results can beobtainedfrom the authors. 31.See Rodrik1998; and Barro 1997. 32.Frankel and Romer 1999. 33.The introduction of the variable “Population” inthe regression does not change the results shown inthe article. Sinceits coef cientis notstatistically signi cant andthe variable is tightlycorrelated to area, ithas notbeen included in column 3. 246 InternationalOrganization

FIGURE 2. Publicrevenue as a functionof trade openness and political regime

itarianregimes, it only goes up 5 percentagepoints to 28 percent (for exportsand importsequal to 100 percent of GDP). Indemocratic regimes, instead, it goesup by 10percentage points to reach 33 percent of GDP. Inshort, trade openness partly pushes public revenue upward. But the combina- tionof politicalregime and openness speeds up theformation of largergovernments. Inclosed economies, politicians have few incentivesto engage in substantialpublic spending.In open economies, a largepublic sector emerges as the price that the tradablepart of theeconomy has to payto ensurethe acquiescence and cooperation ofboth the sheltered economy and declining tradable industries. Finally, no compensatoryprograms develop in those countries in which, although openness is maintained(to the bene tofthe most powerful sectors in the economy —for example,plantation or mines ’ owners),political participation is restricted, and those sectorsmost hurt by economicopenness are systematically excluded from government.34 Columns4 through6 inTable1 probedeeper into the mechanisms through which tradeopenness leads, in a democraticregime, to a largerpublic sector. Column 4 includesLee ’sweightedlevel of tariffs onimportsof intermediateinputs and capital goods.35 Thismeasure, which ranges from 0.85in India to 0.36in HongKong, can

34.As alreadyemphasized above,we donotclaim thatthe choice of regime isexclusivelydetermined bytrade factors.We rather arguethat, ceteris paribus,the alignment of trade interests contributesto the choiceof aparticularpolitical regime. 35.Lee 1993. TheSize of thePublic Sector 247 bethought of as a directmeasure of the extent to which domestic policy imposes barriersto trade. The results strongly support our theory. The coef cientsof both tradeopenness and its interaction with democracy remain stable and statistically signi cant.Moreover, as expected, an increase in tariffs iscorrelated with a reductionin the size of the public sector. Columns5 and6 considerin turntwo variables that may be thought of asmostly tappingthe degree to which the domestic economy is affected by a higherrisk of havingto bear a volatileworld business cycle as it opens its borders. 36 Column 5 introducesan index of theproduct concentration of exportsin each country, alone andin interaction with trade and democracy. This export concentration measure, whichis aGini-Hirschmanindex of concentrationbased on 239three-digit standard internationaltrade classi cationcategories of exports as estimated by the United NationsCouncil on Tradeand Development, varies from 0.06(a highlydiversi ed economy)to 1 (wheneveronly one product is exported).Openness and its interac- tionwith political regime continue to perform strongly. The level of export concentrationdepresses public revenue signi cantly.37 Theinteractive variables of tradeopenness with export concentration and of these two measures and democracy havepositive and statistically signi cant coef cients.As thetradable sector be- comesless diversi edand has a morecentral role in the domestic economy, the pressurefor domesticcompensation clearly goes up. Under democratic regimes, this pressureintensi eseven more. Column6 introducesa moredirect measure of risk using the variability of the termsof trade.Following Rodrik, we de ne p asthe natural logarithm of theprice ofexportsrelative to imports,that is, the terms of trade. 38 Assumingthat the log of theterms of trade follows a randomwalk, the unanticipated component of the incomeeffects of a changein the terms of trade can be expressed(as apercentof GDP) as1/ 2[( x 1 m)/y] [dp – a ] where x standsfor thevolume of exports, m for thevolume of imports, y for GDP, and a for thetrend growth rate in the terms of trade.To calculate the level of volatilityof theeconomy, we calculatethe standard deviationof the change in the terms of trade over the previous veyears to each observation.An increase in volatilityis positively correlated with an increase in the sizeof the public sector. The interaction of trade openness with the standard deviationof thelog of theterms of tradeand the interaction of thesetwo terms with democraticinstitutions, which are also included in column 6, have negative coef cients—thatis, they depress the role of government. 39 Thestatistical signi - canceof the three new variables is relatively weak: they are only statistically

36.Rodrik 1998. 37.Since export concentration is inpart negatively related todemocracy (thecorrelation between the twovariables is –0.38for the whole sample and –0.50for the late 1980s),the negative result is likely tobe proxyingthe lack ofredistributiveprograms in authoritarianregimes. See Boixand Garicano 2001 fora discussionof the relationship between economicstructure and democracy. 38.Rodrik 1998. 39.As aresultof considerable colinearity, the introduction of these volatilitymeasures causes the interactiveterm oftrade anddemocracy tobecome notsigni cant. 248 InternationalOrganization signi cantin a jointtest with the log of trade openness, and they do not become signi canteven after we dropopenness and the interactive term of openness and democracy.In part, these results may be dueto thedifference of gathering complete andreliable data series on theterms of tradein developing countries. In our opinion, however,the subdued role of the terms-of-trade measure mainly points out that the strategyof domesticcompensation emerges, above all, to compensatelosing sectors ratherthan to smooth,as somesort of insurancemechanism, the short-term volatility ofthe business cycle. InTable 2 wereportseveral tests to check the robustness of ourresults. First we introducethe lagged value of the dependent variable “Sizeof Government. ” Both tradeopenness and the interactive term have a positiveeffect on publicrevenue and arestatistically signi cantin ajointtest. The long-run coef cientof the interactive term,at around 4, isclose to the estimates of Table1. 40 Columns2 and3 thenreport theresults of estimatingthe benchmark model (with and without the lagged value ofthe dependent variable) using a xed-effectmodel, to account for potential idiosyncraticeffects for differentcountries. Finally, columns 4 and5 reportthe resultsof estimating a random-effectsmodel (again with and without the lagged valueof thedependent variable) with a standardprocedure of a “variance-component” generalizedleast squares (GLS) technique(to correct for theways in which assumptionsunderlying ordinary least squares estimations are violated by cross- nationalpanel data). 41 Inall cases, the model is extremely robust to different statisticalprocedures. InTable 3 we examinethe impact that a changein either the level of trade opennessor the type of politicalinstitution has in the rate of growthof thepublic sector.The lagged value of publicrevenue and the change in tradeopenness restrain theexpansion of the public sector. By contrast, and in line with our previous estimations,a positivechange toward more trade openness in democratic regimes hasa directimpact on the size of thestate.

Historical Evidence

Sofar we haveemployed both analytical methods and econometric evidence to stresshow tariffs andtaxes are jointly selected by politicians, conditional on the politicalmechanisms (democratic or authoritarian) employed to aggregate preferences.To bolster our analysis, we nowexamine a setof historicalcases taken from verydifferent key periods and geographical areas. First, we describehow scal and tradelaissez-faire policies imposed under a veryrestrictive suffrage system in the rst halfof thenineteenth century were graduallyreplaced by anationalframework basedon free tradeand compensation policies (in the United Kingdom and

40.The long-run coef cientis calculated as a 2/(1 – a 1), where a 1 is the coef cientof the lagged value ofthedependent variable and a 2 is the coef cientof thevariable of interest(the interactive term). 41.See Hsiao 1986;and Hicks 1994. TheSize of thePublic Sector 249

Scandinavia)and protectionism without compensatory policies (in Australasia) in thetwentieth century. The historical evidence for thesecases provides us withall the outcomesthat may take place, so to speak,in theupper threshold of Figure 2 —that is,under democracy. Second, we discusshow free trade,the development of the welfare state,and the choiceof politicalregime interacted in thosecountries affected byauthoritarianismand the so-called third wave of democratization,paying special attentionto Southern European countries. This discussion sheds light on the set of casesin thelower band of Figure2 andon theway in which shifts in the structure ofthe electorate are related to changes in the role of the state.

FromLaissez-Faire to Compensatory Policies in Democratizing Europe(1830 –1950) Theintroduction of a laissez-fairetrade regime in the rst halfof the nineteenth centuryin Britain and its gradual extension to continental Europe in the following decadeswas achievedwithout any simultaneous expansion of domesticmechanisms ofcompensation. Free tradewas introducedwith the support of commercial and urbaninterests in Britain and the backing of working class associations. 42 But the triumphof Manchesterian liberalism (the quasi-welfare system structured by the PoorLaws was dismantledaround that period) was equallyrelated to the extremely restrictivenature of thefranchise. Only one in eight men were entitledto voteafter 1833and about three out of ten after 1868. Moreover, the electoral system was extremelybiased against both rural areas, which had borne most of thelosses of the tariff reform of1846, and the urban poor. Thestability of the Cobdenite regime was putinto question, however, by two paralleldevelopments at the turn of thecentury. On theone hand, after the electoral reform of1884, which equalized the franchise conditions of the rural counties to thosealready in place for urbancounties, the British electorate doubled to encom- passbetween two-thirds and four- fthsof theadult male population. 43 On the other hand,a fallin agriculturalprices and, above all, the growth of Germancompetition, unnervedBritish public opinion. Several anti-free trade episodes, such as an early resolutionof theNational Conservative Associations in 1887in favorof ‘fair trade,’ the ‘Madein Germany ’ panicof 1896,and the reimposition of sugardues, the coal exportduty, and the corn duty in the late 1890s and early 1900s, were preludesto anewpolitical realignment on trade policy. Chamberlain ’sproposalsfor tariff reform becamehegemonic within the Conservative party by 1906. In turn, among Liberalsand Labour intellectuals, free tradeand state interventionism rapidly mixed inwhatHowe hastermed the ‘New Liberal’ synthesis.44 Followinga spateof works

42.See Rogowski1989; and Schonhardt-Bailey 1991. 43.The proportion of enfranchisedadult men varies across authorsdepending on theage chosento countmen andassumptions about the weight of pluralvote in rural counties. For relatively low estimates, see McKibbin,Matthew, and Kay 1990.For high estimates, see Blewett 1972. 44.Howe 1997. 250 InternationalOrganization byAdam, Haldane, and the Webbs calling for increasedpublic spending on education,the Liberal League pamphlets of 1902 already defended a muchmore aggressivestance of the state in themarket to defendBritain ’scommitmentto free trade.Although the Liberal party won in the 1906 landslide election under the bannerof free trade,the economic downturn of 1907 – 08and stagnant real wages resultedin a markedpopular shift to Tariff Reformcandidates in several by- elections.45 TheLiberal government, now headed by Asquith and with Lloyd Georgein the treasury, immediately responded by creating an old-age pension programin 1908; raising land taxes through the ‘People’s Budget’ andintroducing laborexchanges and trade boards the following year; establishing national insurance for sickness,invalidity, and unemployment in 1911; and passing the Miners ’ MinimumWage Act of 1912. The combination of free tradeand compensation embracedby the Liberal cabinet pushed Conservatives and moderate Liberals into thetariff reform camp.As theDuke of Northumberland,a former opponentof tariff reform, wroteto Stratchey in the autumn of 1909 in reaction to Lloyd George ’s scal plans, “protectioncannot be worse thanSocialism. . .Andas. . .Tariff Reform orSocialismare the only possible alternatives at this moment, I amquite prepared toswallow the former. ”46 Thepolitical debate that emerged at the turn of the twentiethcentury continued to structure the agenda of the interwar period. The Conservativeparty led the battle for imperialprotection in the 1923 elections and was able,with the growing support of manufacturers and the City, to impose its solutionin 1931. By contrast, Labour, which had succeeded the Liberals as the progressivealternative, almost unanimously defended free trade. 47 The asco of the 1930spolicies and the victory of Labourin 1945 eventually brought Britain to the campof open borders and sizable public intervention. Asimilarevolution, with a muchfaster and radical commitment to the compen- sationstrategy, took place in Scandinavia.As shownby Baldwinfor Denmarkand partlyfor Sweden,the basis of universalist compensatory policies were alreadyin placeat the turn of the century. 48 As soonas the Liberal party, sustained by the Danishfarming community, secured a strongmajority in parliament, all-inclusive, non-contributory,tax- nancedpensions were establishedin the 1890s. The type andsize of pensionsdirectly responded to the tradable character of farming sector. First,they were “oneof themore successful measures tried ” toattractlabor needed bythe farmers tokeep being competitive “justas competition and falling prices fetteredtheir ability to improve conditions and stem migration. ”49 Second,because Danishfarming producers by nature are takers of international-prices, their costs(and bene ts) were spreadacross the whole population. As hasbeen well

45.Searle 1992. 46.Quoted in Blewett 1972,79. 47.As late as 1931,93 percentof Labourcandidates supported free trade intheir manifestos. Howe 1997, 285. 48.Baldwin 1990. 49.Ibid., 75. TheSize of thePublic Sector 251 documentedin the literature, the strategy of opennessin conjunction with compensatorymechanisms deepened in the 1930s and intensi edagain in the 1960s and 1970s.50 Theformation of a free tradeplus compensation regime in Northern Europe contrastswith the combination of protectionistschemes and a smallerwelfare state adoptedby both Australia and New Zealand. 51 Inresponse to depressed economic conditionsin the late nineteenth century, Australian Labor agreed to support tariff reform inexchangefor thelegal recognition of aminimumwage for unskilledlabor. Legalwage regulation, which was systematicallysustained through a national systemof compulsory arbitration in industrial disputes enshrined in the federal constitution,had the objective to secure, in the terms of the 1907 Harvester Judgmentfrom theCourt of Conciliation and Arbitration, a “fair andreasonable wage” to meet “thenormal needs of an average employee regarded as a human beingliving in a civilizedcommunity. ”52 Sustaininga wagethreshold required uncoupling(parts of) thedomestic economy from internationalmarkets. A restric- tiveimmigration policy in favorof preservinga ‘whiteAustralia ’ toblockthe in ow oflow-wage, non-white workers became the masthead of the federal Labour platformin 1905.Similarly, both Australia and New Zealanderected a strongtariff systemto sustain prices in the domestic manufacturing industry in the 1920s and 1930swith clear success. Whereas export prices fell by 40 percent between 1920 and1935, real weekly wages of workers onlydecreased by 5 percentin the same periodin New Zealand.The use of methods to shape the wage structure signi cantly lessenedany social demands for alargewelfare state.In 1949 –50,Australia only spent4.7 percent of its GDP onsocialsecurity —comparedwith an average of 8.0 percentin 14 advanced industrial democracies. 53 By1975, tax revenue as a proportionof GDP was 7.5percentage points below the OECD averagein both Australiaand New Zealand.Still, equality of conditions was wellpreserved in both countries.The Gini index for Australiaand New Zealandhas been consistently lowerthan the average OECD Giniindex by abouta wholestandard deviation of the sample(or 5points). 54

50.See Cameron 1978;and Katzenstein 1985.A briefset ofdata willgive a sense ofthedifference inpublic interventionism by level of openness. In the early 1970sand among OECD nations,public spendingin education averaged 5.4 percent of GDP inopeneconomies (exports equal to 40 percent or more ofGDP) and3.7 percent in closed countries; income maintenance programswere 12.9percent of GDP and8.6 percent of GDP,respectively; public xedcapital formationwas 4.5percent and 3.7 percent ofGDP respectively;subsidies were 2.5and 1.2 percent, respectively; and labor market policies amounted(in 1985) to 1 and0.5 percent in each set ofcountries. OECD variousyears. 51.See Castles 1985and 1989; and Mabbett 1995. 52.Quoted in Castles 1989,34 –35.The introduction of a protectionistregime alsorequired buying offthe support of farmers througha system ofsubsidies. 53.Castles 1985. 54.Our own estimations based on Deiningerand Squire ’sdata set. Deiningerand Squire 1996. 252 InternationalOrganization

FreeTrade, Democracy, and Taxes in Southern Europe (1930 –1980) As shownin ourprevious discussion of historicalevidence for theadvanced world, themaintenance of ademocraticregime forces states to deal with the pressures that derivefrom theinternational economy — eitherto close borders or to compensate theirelectorates. We nowexamine how, under institutional conditions making possiblethe exclusion of keysectors of thepopulation, policymakers can evade that politicaland economic dilemma. With that goal in mind,we focuson thetrajectory ofSouthernEurope, and particularly Spain, since that area has shifted trade policy regimesand constitutional systems several times in the last century. TheGreat Depression and the establishment of authoritarianregimes triggered the introductionof strongautarkic economic policies in Southern Europe in the 1930s and1940s. They were onlyabandoned in Italyafter military defeat in 1945 and in Portugaland Spain in the 1960s after almost two decades of economicstagnation. Considerthe case of Spain. Following an economic stabilization plan in the late 1950s,the peseta was madeconvertible, import quotas were progressivelydisman- tled,and foreign capital was aggressivelycourted. Economic liberalization was followedby arapidgrowth of thetradable sector. The sum of exportsand imports asa percentageof GDP rosefrom about10 percent in 1958 to 22 percent in 1962 andthen to 34 percentin 1974.The in owofforeignprivate long-term capital went from 15million US dollarsin 1958 to 435 million ten years later. The share of employmentin theprimary sector was cutin half, and real per capita income more thandoubled in fteenyears. 55 Themaintenance of an authoritarian regime until 1975 ‘freed’ theSpanish state from activelyresponding to the rapid dislocation causedby the process of economic liberalization. Tax revenues as a proportionof GDP uctuatedaround 17 percent throughout the 1960s and then climbed slightly toabout23 percentin 1974 —alevelequal to abouthalf of thetax effort of anyother mid-sizeEuropean country. Expenditure on social policies was halfthe European level.56 Expenditurein education averaged less than 2 percentin the 1960s —about athirdof the German and French level. Very similarpolicies were pursuedin Portugal.Even with higher levels of tradeopenness (the sum of exportsand imports asa percentof GDP was around55 percent in the 1960s), public revenues stood below20 per cent of GDP underSalazar ’sauthoritarianrule. 57 Authoritarianism operatedin a verysimilar fashion in Latin America ’sSouthernCone in the 1970s andin East Asia: it reconciledtrade and nancialliberalization with subdued growth ofthe welfare state. 58 Thecombination of economicliberalization and minimal compensatory policies brokedown with the transition to democracyin the mid 1970s. The process of trade

55.Donges 1971. 56.Maravall 1995. 57.For an analysis of thePortuguese case, see Corkill1999. 58.Again, our main theoreticalclaim is neitherthat authoritarianism leads tofree trade northat the introductionof the latter requiresan authoritarian (insulated) state, butrather thatauthoritarianism enables states toimpose laissez-faire strategies inboth trade and scal policies. TheSize of thePublic Sector 253 liberalization,which culminated in the integration with the European Union, was overwhelminglysupported by voters. 59 Yet,in thecontext of ademocraticsystem, itwas nowaccompanied by generous compensatory policies. After 1975,Spanish publicexpenditure grew byover 1 percentagepoint of GDP peryear in real terms toreach 49.6 percent of GDP in1993. Although an important part of that growth was simplydue to theexplosion of political demands that followed the introduction offree elections,part of theexpenditure was relatedto the new conditions imposed bythe rapid internationalizati onof the Spanish economy. In response to adverse internationalconditions, the Spanish government rst spentheavily on unemployment bene tsand injected money into entire industrial sectors —subsidiesand capitaltransfers rose to 5.6 percent of GDP by1982.In the mid- and late 1980s, the publicsector then shifted the content of publicexpenditure to supportstrong capital formationpolicies that could increase Spain ’scompetitiveness.Whereas subsidies andcapital transfers were cutsubstantially by almost 2 percentof GDP between 1982and 1989, public xedcapital formation rose by 2.1 points of GDPupto 5.2 percentof GDP in1991, general education expenditure went up to 4.7 percent of GDP in1994, and active labor market policies reached over 1 percentof GDP. Some ofthesenew programs were supportedwith European structural funds, themselves aresultof an explicit deal in which the Spanish cabinet supported German and Frenchplans to forge the European Union in exchange for substantialtransfers to Spain’spoorerregions. 60 Theexpansion of the Spanish state was reinforcedby the decisionof theSpanish government to integrate the peseta in theEuropean exchange ratesystem in June 1989. With its hands tied in the realm of macroeconomic management,the Spanish government tried to buy off supportfrom votersand unionsfor newplans to make Spain competitive (mostly through wage moderation) withgenerous programs that raised public spending by 2 percentagepoints from 1988 to 1990—thatis, even before the economic downturn of 1991 –93pushed both thepublic de citand public spending upward by another 4 pointsof GDP.

Conclusion

Inexploring the consequences that the international economy has on the domestic politicalarena, a growingliterature shows that higher levels of tradesystematically leadto a largerpublic sector across both developed and developing nations. 61 As opennessincreases, the state, mainly acting as a benevolentdictator or a welfare maximizer,adopts a salientrole to minimize the risks of highereconomic integra- tionand to compensate declining economic sectors. Politics is, however, promi-

59.Over halfof the Spanish population judged membership in EU tobe a ‘good thing’ in the early 1980s.This level of support climbed to two-thirds by the late 1980s.By contrast, only 5 percent consideredmembership to bedetrimentalto Spanish interests. Alvarez-Miranda 1996. 60.Boix 1998, 105 –29. 61.See Cameron 1978;and Rodrik 1998. 254 InternationalOrganization nentlyabsent in this approach. Disregarding the most recent research on thepolitical sourcesof different tariff regimes,trade is taken as an exogenously-determi ned variable.62 Neglectingthe literature on the redistributive consequences of public spending,the growth of the public sector is then regarded as a merelyfunctional responseto the requirements of trade. 63 Amoresatisfactory understanding of the relationship between the international economyand domestic politics requires, by contrast, taking into account the set of economicand political trade-offs that simultaneously underlie the choice of trade and scalpolicies. As we haveshown, once this is done countries may pursue three (stylized)alternative strategies: 1.National policymakers may keep the domestic economy closed. Due to high tariffs orsimplyto natural barriers to trade (such as geographical distance or thesize of the country), closed economies escape the variability of world businesscycles and demand lower levels of compensation than open economies.In short, ceteris paribus, economic insulation depresses the level of publicexpenditure. According to our empirical results, in closed economies, suchas Argentina, India, Iran, Japan, Mexico, or the United States, public revenueas a proportionof GDP isatleast 10 points lower than it should be, giventheir respective level of development. 2.Once free tradepolicies are embraced, and given that Keynesian demand managementis hardly available to open economies, policymakers can only ensurehigh levels of social welfare (andtherefore the support required to govern)by expanding the public sector to shore up declining economic sectors.64 As smallEuropean nations democratized at theturn of thetwentieth century,pressure from tradablesectors to establish welfare systemsgrew steadily.But it was mostlyin responseto the economic shock of the1930s that politicalelites used welfare andinvestment spending to structure a pro-free tradecoalition in small European states. 65 Thatsolution contrasted with the decisionto setup protectionistpolicies as awayof steadyingrelative prices at homewithout having to raise taxes and redistribute income through the public sectorin LatinAmerica (and in New Zealandand Australia to some extent) by thatsame period of time. 3.Since the combination of openness and compensation requires higher taxes, policymakersmay consider favoring a thirdpolitical strategy. Excluding in a systematicmanner —thatis, through authoritarian rule —thosesectors that may losefrom increasingeconomic integration, governments avoid increasing publicspending. As shownin ourempiricalanalysis, in free tradeauthoritarian regimes,such as the East Asian economies, the public sector is 10 percentage

62.Keohane and Milner 1996. 63.See Esping-Andersen1990; and Holsey and Borcherding 1997. 64.On demand management inopen economies, see Alt1985. 65.This argument is parallel toand encompasses Garrett ’sanalysisof the decision of social democratic parties toestablish compensatory policies in open economies. Garrett 1998. TheSize of thePublic Sector 255

pointsof GDP smallerthan in a democraticsystem with similar levels of economicintegration. Theissues and results raised in thisarticle are relevant to contemporary debates onthepolitical and economic consequences of trade(and nancial)integration on, atleast, two counts: the sustainability of thewelfare state,and the changes that the internationaleconomy may induce on thenumber and system of states.Consider the rst question.Broadly speaking, the current literature on the effects of economic opennesscan be dividedinto two camps. On the one hand, the most extended (and popular)approach sees the process of economic globalization as simply imposing increasingconstraints on the ability of statesto govern the economy. On the other hand,a setof scholarspoints to the striking correlation between openness and the sizeof thepublic sector, calling into question the former ’sconclusions:the fact that mostopen economies consistently espouse larger governments shows, in an un- equivocalmanner, that more trade does not require more scaldiscipline —and that theopposite may be true.We ndthat,once all the relevant variables are taken into consideration,both approaches can be reconciled.Besides the size of eacheconomic sector,two factors are relevant to the choice of thepolitical-economic regime: the variabilityof risks derived from tradingwith other nations; and, above all, the productivityof each economic sector (especially the pro-free tradesector) —that is, howmuch gain (or loss)each sector derives from free traderelative to autarky. If varyingproductivity levels affect the policy mix that is chosen, two conclusions are inorder. First, more openness does not automatically constrain the spending capacityof states. This is particularly true whenever public expenditure is geared towardthe production of publicgoods, such as infrastructuresor humancapital, that raisegrowth rates. 66 Second,how sustainable a largepublic sector is over time dependson thecompetitive advantage of theexporting sectors that pay for it.If this competitiveadvantage erodes, the incentive to sustaina largegovernment declines – andcountries (the South American path of the 1930s and 1940s) start shifting towardeither a protectionistsystem or an authoritarian free traderegime. That publiccompensation may run into limits seems to beforgottenby theliterature on tradeand government growth. By modelinga conditionalrelationship between trade andthe public sector, we takeaccount of that fact without destroying the strong empiricalassociation between globalization and the welfare state. Theway in which scalpolicy and trade regimes are related suggests also that our modelmay be usefulto shed light upon the causes that explain the evolution of the statesystem and any historical variation in the number of nations. In its most stylizedterms, our model describes a policymakerinterested in maximizing the welfare functionof the median voter to win elections (or, more generally, stay in power).The policymaker ’s rst choiceconsists in either establishing a relatively

66.The article employsan aggregate measure ofthe size ofpublic sector andtherefore does not address thecauses (andconsequences) of differenttypes of publicspending (capital investmentversus publicconsumption). For a discussionof these questions,see Alesina andRodrik 1991; and Boix 1998. 256 InternationalOrganization closedeconomy —wheresmoothing the business cycle is possible — oropeningthe economy—wheredemand management is fraught with risks. But the many ways throughwhich autarky may come about have been left unexplored. By assumption, we haveequated autarky to raising domestic tariffs. Nonetheless, a closedeconomy canbe achieved by the integration of previously separated countries. In the framework ofourmodel, the process of Europeanuni cationcan be understoodas analternative (and a morecost-effective) response to globalizationthan expanding thewelfare statein each European nation one step further. In short, the model we presentmay be takenas a startingpoint to explorewhy and how politicians shift the locusof sovereigntyto govern the economy and maximize the welfare ofvoters.

Appendix

Thechoice of policy is the result of two steps. In the rst,voters vote in elections. In the second,they may consider restricting (or not) the vote given the outcomes of a fully democraticsystem.

TheElectoral Game

Consider rstthe electoral game. C, O, and P (where C # O # P , 50percent of voters) canchoose among two parties in national elections. Both parties, S and V,committo their policyprograms before voters vote. After the election, the world business cycle either expandsor recedes. V,thefree trade party, may just promise no tariffs l 5 1 or it may promisefree trade jointly with an scheme to compensate the P sector with a o inrecession times.In turn, S,theprotectionist party, favors high tariffs l 5 0.67 Evenif it does not have tocompensate for volatility, it may offer a transfer a c to Pstooutmatch V’s offer. For V to win, O and P willhave to vote for it. At electionday, P votersonly know their expectedreturns under free trade (with possible compensation promised by V) and under autarky(with possible transfer promised by S). If V wins, P’sexpectedutility is

EU p 5 p Pg 1 ~1 2 p !Pb 1 ~1 2 p !a o (1)

If S wins, P’sexpectedutility is

EU p 5 P c 1 a c (19)

Accordingly, P voterschoose V whenever(1) is bigger than (1 9):

p Pg 1 ~1 2 p !Pb 1 ~1 2 p !a o . Pc 1 a c (2)

67.A cornersolution in l is ensuredby thelinearity of P’sexpectedutility. If Pswere riskaverse, apartialreduction of tariffs,that is 0 . l .1, would be V’sbeststrategy under some parameters. TheSize of thePublic Sector 257

Givenboth that P choosesunder uncertainty and that the expected gains from openness are lowerthan the gains from autarky (condition 3 9 inthe main text), V mustalways commit to acompensation a o to attract P.Thiscompensation has to meet two criteria. First, it has to outmatchany spending package a c promised by S.Second,it should not alienate the O voters— otherwisethey would not vote for the free trade party.

Assumingthat the package a c is nancedfrom taxes on C voters,the cost for every C of thespending package a c promised by S toeach individual P cannotbe higher than the differencebetween what each C gainsunder a protectionistregime ( Cc)andwhat it gains underfree trade ( Co).Otherwise, C wouldnot vote for S. That is,

C a , C 2 C (3) c P ~ c o!

Allowing S to tax both C and O (and,similarly, letting V tax O and C) to nance the compensationpackage a c(a o)doesnot alter the solution of thegame because each side will tryto extract the maximum resources from their opponents and will be willingto accept a tax thatleaves them marginally better off. O votersvote for V ifthe gains of anopeneconomy are greater than the gains of a closed economyeven allowing for the expected cost of the compensation they have to pay to each Pvoterin a recession:

P O , O 2 1 2 p a (4) c o ~ ! o O

Solvingfor expression (4), for O to vote for V,thefollowing inequality has to hold:

O a o , ~Oo 2 Oc! (49) P~1 2 p !

Giventhe maximum a c that S canoffer (3) and the compensation V canpromise (4 9) and substitutingin (2), P willalways vote for V whenever:

O C O 2 O . P 2 p P 2 1 2 p P 1 C 2 C (5) P ~ o c! c g ~ ! b P ~ c o!

Inshort, the free trade party, V,willbe morelikely to win, other things being equal, the larger thenumber of Os,thelarger the probability of agoodcycle, the fewer the number of Cs, and the lower P’sreturnsare under autarky. If thenumber of Psincreases,party V will be more likelyto win the smaller the distance in P’sutilityunder trade with respect to autarky. 68

68. P votersdo not have an incentive to set uptheir own party. To set itup, the new partyshould promisemore than S or V.If,for example, the P partyoffered more than a o, the amount O voters are willingto accept, O voterswould be better off voting for S, and the P partywould never be able to commanda majorityof votes. 258 InternationalOrganization

Moreover,if P votersare risk averse, V willstand more chances to win the lower the degree ofrisk aversion.

PoliticalRegime

Supposethat w isthe minimum cost of repressing each member of those economic sectors thatare permanently excluded from government. For the O sector,the total cost of a dictatorshipwill be ( P1C)w (thatis, the cost of repressionmultiplied by the number of Ps and Cs)69 and O’sminimumper capita expenditure to impose a dictatorshipwill be ([P1C]/O)w. In turn, C’sminimumper capita cost to establish a dictatorshipwill be 70 ([P1O]/C)w. Giventhe minimum per capita cost of authoritarianism, O willimpose a dictatorshipif three conditions are met: 1.First, the costs of imposinga dictatorshipdo not eat away the gains from trade:

~P 1 C! O 2 O . w (6) o c O

Similarly,the maximum per capita cost Cs arewilling to pay to impose a dictatorshiphave tobe less than their net gains from protectionism:

~P 1 O! C 2 C . w (7) C o C

2. Second, O’sresourcesfor repression are larger than those of C sothat a pro-freetrade dictatorshipcannot be contestedby the protectionist sector. Solving expressions (6) and (7) for w,thisoccurs whenever:

O C ~O o 2 Oc! . ~C c 2 C o! (8) ~P 1 C! ~P 1 O!

3.Finally, the cost of maintaininga dictatorshipmust be lowerthan the cost of compensating P.Thiscondition includes two scenarios: 3a.On the one hand, the minimum compensation for P maybe toocostly for V to promise it(condition 5 doesnot hold)and S accordinglywins the election. Still, O may be able toimpose an uncontested dictatorship (8 holds). This happens whenever:

2 2 2 P ~P c 2 p Pg 2 @1 2 p #Pb! . C ~C c 2 C o! 2 O ~O o 2 Oc! (9)

69.If we consideredthat P votersonly have to berepressed inrecession times, theexpected cost of dictatorshipwould shrink to ([1 – p ]P1C)w. 70.We assume thatthe introduction of a dictatorshiprequires that the economic sector ispreviously organizedthrough a party.As aresult, P cannotconsider imposing a dictatorship. TheSize of thePublic Sector 259

Clearly,this is more likely to happen the larger the number of Ps and/or Os and the lowernumber of Cs,as well as for the larger relative welfare loss of Ps in an open economywithout compensation. 3b.On the other hand, although V cancommit itself to a compensationpackage (and thus winthe election), a dictatorshipis cheaperthan paying a compensationpackage. This occurswhen the per capita cost of repression is lower than the per capita cost of compensating:

~P 1 C! P C w , P 2 p P 2 1 2 p P 1 C 2 C (10) O OF c g ~ ! b P ~ c o!G

If the Cswereready to spendmore than w perrepressed individual, we should substitute that particularamount in the place of w in (10):

~P 1 C! C~Cc 2 Co! P C , P c 2 p Pg 2 ~1 2 p !Pb 1 ~Cc 2 Co! (11) O ~P 1 O! O F P G

Tobetter understand how changes in the parameters affect the choices of eachsector, let us lookmore closely at the solution. The following equation (rewritten from 5) shows the number of Csrequiredto make voters indifferent between the two parties, given the other parameters:

~Oo 2 Oc! @Pc 2 p P g 2 ~1 2 p !P b# C 5 O 2 P (59) ~Cc 2 Co! ~Cc 2 C o!

Similarly,from (8), the expression below shows the number Csrequiredto counterthe threat of an O’sdictatorship:

~Oo 2 O c! P 6 P 2 1 4O ~P 1 O! Î ~Cc 2 Co! C 5 (8 ) 2 9

Both (59) and (89),jointlywith (11) can be plottedin the O-C axisto study their interaction. When Cs and Osexperiencesimilar relative gains with their preferred trade policy, the Os are morelikely to (if, at all) open the economy through dictatorship than through a feasible democraticgovernment by V.Inthose cases S caneasily outmatch any compensation packageoffered by V andthe free-traders can only resort to violence. From (5 9),itis evident that Oshaveto more than outnumber the Cstowin democratically. This equilibrium is reinforcedby any increase in P’s ex ante predispositiontoward autarky. In that case, while (89)doesnot shift in the O-C axis,the space under which V canwin democratically shrinks. However, as O’srelativegains from trade ( Oo 2 Oc)increase,the space for democracy increasesat a fasterspeed than that for dictatorship. Notice that the slope of (5 9) increases linearlyin ( Oo 2 Oc)whereasit does so only in squared terms in (8 9).Therelative cost of bothregimes, given by (11), is notaffected by thischange. Eventually (5 9) and (89) cross at increasinglylower values of O meaningthat, a largerspace for a freetrade democracy is both 260 InternationalOrganization

possibleand nanciallysound. Since the transfers promised by protectionists are now relativelysmaller, it is both feasible to outmatchthem and cheaper only to compensate the Ps ratherthan repress a largernumber of people.Again, this result hinges on the strength of P’s predispositiontoward autarky.

References

Alesina,Alberto, and Dani Rodrik. 1991. Distributive Politics and Economic Growth. NBER Working PaperSeries 3668.Cambridge, Mass: NationalBureau of EconomicResearch. Alt,James E.1985.Political Parties, World Demand, and Unemployment: Domestic andInternational Sourcesof EconomicActivity. AmericanPolitical Science Review 79 (4):1016–40. Alt,James, Jeffry Frieden,Michael J. Gilligan,Dani Rodrik, and Ronald Rogowski. 1996. The Political Economyof InternationalTrade: Enduring Puzzles andan Agenda for Inquiry. ComparativePolitical Studies 29 (6):689–717. Alvarez, Mike,Jose ´ AntonioCheibub, Fernando Limongi, and Adam Przeworski.1996. Classifying PoliticalRegimes. Studiesin Comparative International Development 31 (2):3–36. Alvarez-Miranda,Berta. 1996. Elsur de Europay laadhesio ´nalaComunidad: los debates pol õ´ticos. Madrid:Centro de Investigaciones Sociolo ´gicas—SigloVeintiuno de Espan˜a Editores. Aukrust,Odd. 1977. In ationin the Open Economy: A NorwegianModel. In WorldwideIn ation: Theoryand Recent Practice ,editedby Lawrence B.Krause andWalter S.Salant,109 –26.Washing- ton,D.C.: Brookings Institution. Baldwin,Peter. 1990. ThePolitics of SocialSolidarity: Class Bases oftheEuropean Welfare State . New York:Cambridge University Press. Barro,Robert. 1997. Determinantsof EconomicGrowth: A Cross-CountryEmpirical Study. Cambridge, Mass.:MIT Press. Bates, RobertH., PhilipBrock, and Jill Tiefenthaler. 1991. Risk and Trade Regimes: AnotherExplo- ration. InternationalOrganization 45 (1):1–18. Beck,Nathaniel, and Jonathan N. Katz. 1995.What to Do (and Not to Do) with Time-Series Cross-SectionData. AmericanPolitical Science Review 89 (3):634– 47. Blewett, Neal. 1972. ThePeers, theParties, and the People: The British General Elections of 1910. Toronto:University of TorontoPress. Boix,Carles. 1998. PoliticalParties, Growth, and Equality. Conservative and Social Democratic EconomicStrategies in the World Economy .New York:Cambridge University Press. ———.2001.Democracy, Development,and the Public Sector. AmericanJournal of PoliticalScience 45 (1):1–17. Boix,Carles, andLuis Garicano. 2001. Democracy, Inequality,and Country-Speci cWealth.Unpub- lishedmanuscript, University of Chicago. Cameron,David R. 1978.The Expansion of the Public Economy: A ComparativeAnalysis. American PoliticalScience Review 72 (4):1243–61. Campillo,Marta, and Jeffrey A.Miron.1997. Why Does In ationDiffer Across Countries?In Reducing Ination: Motivation and Strategy, editedby Christina D. Romer andDavid H. Romer,chap. 9, 335–57.Chicago: University of ChicagoPress. Castles, FrancisG. 1985. TheWorking Class and Welfare: Re ections on thePolitical Development of theWelfare State in Australia and New Zealand,1890 – 1980. London:Allen and Unwin. ———.1989.Social Protection by Other Means: Australia ’sStrategyof Coping with External Vulnerability.In TheComparative History ofPublic Policy ,editedby Francis G. Castles, chap.2, 6 –52.New York:Oxford University Press. Corkill,David. 1999. TheDevelopment of the Portuguese Economy: A Case ofEuropeanization . New York:Routledge. TheSize of thePublic Sector 261

Cutright,Phillips. 1965. Political Structure, Economic Development, and National Social Security Programs. AmericanJournal of Sociology 70 (5):537–50. Deininger,Klaus, and Lyn Squire. 1996. A New Data SetMeasuring Income Inequality. World Bank EconomicReview 10 (3):565–91. Donges,Juergon B. 1971.From an Autarchic Toward a CautiouslyOutward-Looking Industrialization Policy:The Case ofSpain. In TheEconomic Development of Spain since 1870 ,editedby Pablo Mart´õ n-Acen˜aandJames Simpson,350 – 89. Aldershot,U.K.: E. Elgar. Esping-Andersen,Go ¨sta. 1990. TheThree Worldsof WelfareCapitalism .Cambridge:Polity Press. Frankel,Jeffrey A.,andDavid Romer. 1999. Does Trade Cause Growth? AmericanEconomic Review 89 (3):379–99. Frieden,Jeffry A.,andRonald Rogowski. 1996. The Impact oftheInternational Economy on National Policies:An Analytical Overview. In Internationalizationand Domestic Politics ,editedby Robert O. Keohaneand Helen V.Milner,chap. 2. New York:Cambridge University Press. Garrett, Geoffrey.1998. PartisanPolitics in the Global Economy .New York:Cambridge University Press. Gourevitch,Peter A.1986. Politicsin Hard Times: ComparativeResponses to InternationalEconomic Crises.Ithaca, N.Y.:Cornell University Press. Hicks,Alexander M. 1994.Introduction to Pooling. In TheComparative Political Economy of the WelfareState, editedby Thomas Janoski and A. M.Hicks,169 – 88.Cambridge: Cambridge University Press. Hicks,Alexander M., andDwane Swank.1992. Politics, Institutions and Welfare Spendingin Industri- alized Democracies, 1960 – 82. AmericanPolitical Science Review 86 (3):658–74. Holsey,Cheryl M., andThomas E. Borcherding.1997. Why Does Government ’sShareof National Income Grow? AnAssessment ofthe Recent Literatureon the U.S. Experience. In Perspectives on PublicChoice: A Handbook ,editedby Dennis C. Mueller,562 – 89.New York:Cambridge University Press. Howe,Anthony. 1997. Free Tradeand Liberal England, 1846 –1946. New York:Oxford University Press. Hsiao,Cheng. 1986. Analysisof Panel Data .Cambridge:Cambridge University Press. Huber,Evelyn, Charles Ragin,and John D. Stephens.1993. Social Democracy, ChristianDemocracy, ConstitutionalStructure, and the Welfare State. AmericanJournal of Sociology 99 (3):711– 49. Katzenstein,Peter J.1985. SmallStates in World Markets: Industrial Policy in Europe .Ithaca, N.Y.: CornellUniversity Press. Keohane,Robert O. andHelen V.Milner,eds. 1996. Internationalizationand Domestic Politics . Cambridge:Cambridge University Press. Lee, Jong-Wha.1993. International Trade, Distortions, and Long-Run Economic Growth. IMF Staff Papers. 40:299–328. Lipset,Seymour M. 1959.Some Social Requisites of Democracy: EconomicDevelopment and Political Legitimacy. AmericanPolitical Science Review 53 (1):69–105. Mabbett,Deborah. 1995. Trade,Employment, and Welfare: A ComparativeStudy of Tradeand Labour MarketPolicies in Sweden and New Zealand,1880 –1980. Oxford:Clarendon Press. Maravall,Jose ´ Mar´õ a. 1995. Losresultados de la democracia: un estudio del sur y el este deEuropa . Madrid:Alianza Editorial. Mayer,Wolfgang. 1984. Endogenous Tariff Formation. AmericanEconomic Review 74 (5):970– 85. McKibbin,Ross, Colin Matthew, and John Kay. 1990. The Franchise Factor in the Rise ofthe Labour Party. In TheIdeologies of Class:Social Relations in Britain ,editedby Ross McKibbin, 66 –100. New York:Oxford University Press. Meltzer, AllanH., and Scott F. Richard.1981. A RationalTheory of theSize ofGovernment. Journal ofPoliticalEconomy 89 (5):914–27. OECD.Various years. OECD NationalAccounts. Main Aggregates. 1960 – 89.Washington,D.C.: OECD. 262 InternationalOrganization

Przeworski,Adam, andFernando Limongi. 1997. Modernization: Theories and Facts. WorldPolitics 49 (2):155– 83. Rodrik,Dani. 1998. Why Do More Open Economies Have BiggerGovernments? Journalof Political Economy 106 (5):997–1032. Rogowski,Ronald. 1989. Commerce andCoalitions: How Trade Affects Domestic PoliticalAlignments . Princeton,N.J.: Princeton University Press. Romer,David. 1993. Openness and In ation:Theory and Evidence. QuarterlyJournal of Economics 108 (4):869–904. Rueschemeyer, Dietrich,Evelyne Huber Stephens, and John D. Stephens.1992. CapitalistDevelopment andDemocracy .Chicago:University of Chicago Press. Schonhardt-Bailey,Cheryl. 1991. Lessons in Lobbying for Free Trade in19th-Century Britain: To Concentrateor Not. AmericanPolitical Science Review 85 (1):37–58. Searle, GeoffreyR. 1992. TheLiberal Party: Triumph and Disintegration, 1886 –1929. New York:St. Martin’s Press. Summers, Robert,and Alan Heston. 1988. A New Setof International Comparisons of Real Productand Price Levels:Estimates for130 Countries, 1950 –1985. Review ofIncome and Wealth 34 (1):1–26. UnitedNations. Various years. NationalAccounts .New York:United Nations. Wilensky,Harold L. 1975. TheWelfare State and Equality .Berkeley:University of CaliforniaPress. WorldBank. Various years. WorldDevelopment Report .New York:Oxford University Press.