does democracy help? - world bankdemocracy and development,finding that compet-itive elections have...

ASTRIKING PHENOMENON OF THE

1990s was the rise in the numberof countries selecting their lead-

ers through competitive elections.1 The numberrose from 60 countries in 1989 to 100 in 2000.Among poorer countries (those with less than themedian country’s per capita income), the numbernearly tripled, from 11 in 1989 to 32 in 2000; 15percent of the poorer countries elected their gov-ernments in 1989 and 42 percent in 2000.

Unfortunately, democratization does not ensureeconomic development.The simple fact of com-petitive elections did not enable Haiti’s governmentto contain predation by the powerful or to establishminimal law and order. Nor did it prevent Kenya’sgovernment from exerting its authority to benefit asmall, privileged elite. Certainly, most poor demo-cratic countries control predation better and treatcitizens more generously than in these examples.But typically contractual and property rights,widely recognized as fundamental for investmentand economic growth, are less well enforced inpoorer democracies than in richer ones. Similarly,while rent seeking and corruption are higher inpoorer democracies, public services such as educa-tion, critical for both growth and poverty allevia-tion, are less well provided.

Accelerating economic development in devel-oping countries with elected leaders stands as oneof the important challenges of the 2000s.Why aredemocratic institutions less accountable—morevulnerable to narrow interests, rent seeking, and

venality—in some countries than in others? Whyare commitments by some governments more cred-ible than others?

To answer these questions, this chapter focuseson two propositions. First, elected governments aremost likely to make policies favoring narrow seg-ments of the population at the expense of the major-ity when citizens are ill informed, or cannot trustpromises made prior to elections, or are deeplypolarized. Second, elected governments are mostcredible and most likely to respect private propertyrights when they confront checks and balances ontheir decision making.2 These are not the onlyexplanations for democratic performance. Forexample, outside forces direct the policies of somecountries, and the consequences of a country’s his-tory and culture are surely important.3 But the argu-ments in this chapter suggest that it is through theireffects on political credibility (party development),clientelism, citizen information, and social polariza-tion that these other forces probably operate.

Section 1 looks at the relationship betweendemocracy and development, finding that compet-itive elections have only a modest effect on thequality of government; elected governments do notexhibit a systematic advantage in achieving eco-nomic development. Section 2 examines reasonswhy political decision makers do not always adoptpolicies in the broad public interest. Imperfectionsin electoral markets—lack of voter information, theinability of political competitors to make crediblepromises, and social polarization—are important to

Does Democracy Help?

313

Chapter 10

tions and the quality of government.The modestimprovements that took place in the policies ofnewly democratized countries are better explainedby increases in income per capita.Among all coun-tries that held competitive elections in the 1990s,purchasing power parity (PPP)–adjusted incomesrose by a third during the decade.4 Using a six-pointscale to compare the quality of government in thesecountries, three points separated the lowest-scoring25 percent of countries from the highest-scoring.And in half of these countries, the rule of law was nobetter than it was in the median country lackingcompetitive elections. Among countries that hadcompetitively elected governments in 1995, grosssecondary school enrollment varied more than 140percentage points from the minimum to the maxi-mum, and 60 percentage points separated the topand bottom quartiles.After accounting for the effectof income per capita, 40 percent of the countrieslacking competitive elections exhibited higher grosssecondary school enrollment than 40 percent of thecountries that held them.

Consistent with these findings, a large literaturefinds no consistent, significant effect of elections oneconomic growth. For example, Przeworski et al.(2000) find no difference in growth rates betweencountries that have competitive elections and thosethat do not.

Another factor that blurs the distinctionbetween democracies and nondemocracies is theheterogeneity of the latter group. Some autocratsfind that they can extract more rents and stay inoffice longer if they encourage investment and pro-mote long-run economic growth; indeed, in coun-tries with nonelected leaders, property rightsbecome more secure the longer the leaders havebeen in office (Clague et al.1996).5 But many auto-crats are unable to trigger this virtuous circle:investors are deterred by the fear that profits will beexpropriated, the rents obtained by the autocrat fall,economic performance drags, and threats to theautocrat’s tenure grow. It appears that in non-democracies in which the likely rates of return toinvestment are low (for example countries withuneducated workforces and no easy access to for-

E C O N O M I C G ROW T H I N T H E 1 9 9 0 s314

understanding policy formulation and explainingdifferences in economic performance between richand poor democracies. Section 3 looks at reasonsfor the lack of credibility of government commit-ments. It finds evidence that imperfections in polit-ical markets have a significant impact on economicgrowth and hence need to be taken into account indesigning strategies to speed growth and develop-ment. Section 4 discusses reform strategies for rem-edying some of the fundamental distortions thatcan plague democratic decision making. Adjust-ments in the way that governments and their devel-opment partners approach the more traditionaldevelopment agenda, from service deliveryimprovements to broader public sector reforms, cango a long way toward mitigating the shortcomingsin information and credibility that otherwiseundermine government accountability and per-formance. Section 5 concludes the chapter.

1. Elections Have an UnevenImpact on Development

Intuitively, one might expect that in countrieswhere most of the public cannot hold the govern-ment accountable, government decision makingwill tend to disregard the public interest. Moreover,the richest countries in the world (the countrieswith the longest record of sustained growth) haveexperienced relatively long periods of uninter-rupted elections.Thus the political upheaval anddemocratization in the 1990s offered reasons foroptimism regarding economic development.

Some policy progress could be seen among thedemocratizing countries. For example, amongcountries that lacked competitively elected gov-ernments in 1988 but had them by 1998, second-ary school enrollment rose by about 14 percentagepoints. Similarly, a measure of the rule of law—capturing the extent to which government actsarbitrarily—improved by three-quarters of a pointon a six-point scale.

Cross-country analysis shows, however, thatthere is little association between competitive elec-

eign markets), or in which the rates of return tonatural resource exploitation are high, leaders areless likely to curb their own authority to attractgreater investment.

Beyond geographic explanations, nondemocra-cies that emerge from broad social movementsappear to place more controls on their leaders. InMexico, for example, during the period that itdominated politics, the Partido Revolucionario Institu-cional (PRI) provided checks on the behavior ofpresidents.Nondemocracies also differ in the extentof their institutionalized sources of authority (evenwithin a single party) that might counterbalancethe authority of the top leader. Even nondemocra-cies with an unelected legislature have significantlyless corruption and greater rule of law than coun-tries without an unelected legislature.6 Gandhi(2003) finds that nondemocracies with unelectedlegislatures grow faster than those without suchlegislatures.

2. Characteristics of DemocraciesThat Influence Policy Successand Failure

In all settings where people come together to actcollectively, complaints of high-handed behaviorby leaders and of its converse, endless consensus-building, are endemic.Whether in town councils,sports clubs, or Musikvereinen, issues of fairness andequity, efficiency, and consistency regularly arise.Special interests curry favor in every country in theworld, and individuals everywhere succumb to thetemptations of venality and rent seeking.

Why, though, are rent seeking, special interestinfluence, and venality—the effects of governmentinefficiency—worse in poor countries than in rich?As shown in the following discussion, the activitiesof special interest groups explain policy outcomesgenerally, but they do not explain why policy out-comes in developing countries differ from those indeveloped ones. Similarly, differences in politicaland electoral institutions explain variations in pol-icy outcomes across countries, but they do not

D O E S D E M O C R AC Y H E L P ? 315

explain the divergence between policies in devel-oping and developed countries.

Instead, there are three other explanations, allrelated to imperfections in electoral markets, forwhy policies are more likely to neglect the publicinterest in poor democracies but not rich ones:lack of voter information, the inability of politicalcompetitors to make credible promises and betrusted, and social polarization. Each of these isimportant to understanding policy formulationand reform.

Special InterestsBefore the 1990s, attempts to explain governmentpolicy failures centered on the role of special inter-ests.The logic—“the logic of collective action,” asMancur Olson coined it in 1965—is clear. Small,homogeneous groups with much at stake confrontrelatively low costs to acting collectively in theircommon interest. In competing for benefits fromgovernment, this gives them advantages over largegroups whose interests are heterogeneous. Unfor-tunately, narrow interests rarely benefit from publicgoods, such as the provision of universal educationor an improved court system, as much as they dofrom diverting some fraction of societal resources tothemselves. Hence, to the extent that governmentincentives encourage targeting of benefits to specialinterest groups, policy failure—the underprovisionof public goods and the overprovision of regula-tions and laws that benefit special interests at theexpense of the whole society—is more likely.

Ample evidence points to the importance of col-lective action considerations in the making of pub-lic policy in all countries. Bates (1981, 1983) andFrieden (1991) make compelling cases for the roleof special interests, indigenous or foreign, in shapingpolicies in Africa and Latin America, for example.

There is no strong evidence, however, that thelogic of collective action can explain differences indevelopment outcomes. For example, there is noindication that the intrinsic characteristics of specialinterests differ in developing countries, betterenabling special interests in these countries to

Though these associations might suggest an insti-tutional explanation for the differences in the policyexperiences of the two sets of countries, the evidenceis weak that these institutions are responsible for thedifferences.As researchers have predicted, spendingon education is about 2 percentage points of grossdomestic product (GDP) greater in parliamentarydemocracies with proportional electoral systems thanin presidential democracies with majoritarian sys-tems.11 However, the key public good is educationitself, not education spending, which turns out tohave little effect on gross secondary school enroll-ment.12 Political and electoral institutions are insignif-icant determinants of secondary school enrollmentprecisely because they have less of an influence inpoorer countries.13 Among the richer democracies in1997, school enrollment was about 38 percentagepoints greater in parliamentary than in presidentialsystems. In the poorer democracies, it was essentiallythe same regardless of political systems.14

Corruption is another indicator of the extent towhich government decisions on spending or policyare likely to translate into improved social welfare.


extract rents from the unorganized majority. Nor isthere evidence that special interests in developingcountries are more unified than in developed coun-tries, and therefore less likely to “cancel out” theirrespective influences.

Formal RulesGreat attention has focused on the role of institu-tions in development.The formal rules determin-ing how politicians attain office and make decisionsdecisively influence policy outcomes. However, it isless clear that institutional differences can explainthe differences in development performance amongdemocracies.

Substantial research in the 1990s focused pre-cisely on the effects of political and electoral insti-tutions on the magnitude of government spending,broad public goods, and rent seeking.7 Researchersfound that under some conditions majoritarianrules (first-past-the-post electoral systems withsmall electoral districts) lead politicians to focus onpivotal narrow constituencies, biasing spendingdownwards and away from broadly based publicgoods.8 Comparisons of presidential and parlia-mentary forms of government yielded similar pre-dictions: parliamentary systems, under someconditions, promote greater allocations to broadpublic goods than do presidential systems.9 Thesefindings are potentially important for our under-standing of development, to the extent that publicgoods such as universal education or law and orderare essential to economic growth.

Do poorer democracies, with less robust provi-sion of public goods and greater rent seeking,exhibit the electoral and political institutions thatare thought to promote these outcomes?

In 2000, of the countries with competitivelyelected governments, plurality or first-past-the-postrules dominated among the electoral systems of thepoorer countries but not among those of the richer(figure 10.1).10 Similarly with respect to politicalsystems,presidential systems were much more com-mon among the poorer democracies than amongthe richer (figure 10.2).

0

25

50

75

100

1990 2000 1990 2000

Proportional Plurality

Richer Poorer

%

FIGURE 10.1

Electoral Rules in Richer and Poorer Democracies

Sources: Database of Political Institutions (Beck et al. 2001) and WorldBank, World Development Indicators.

Note: Countries represented are those that held competitive electionsfor executive and legislative elections (the Legislative and ExecutiveIndices Indexes of Competitive Elections from the Database of Politi-cal Institutions were both equal to seven).

There is little evidence that political and electoralinstitutions can explain the greater prevalence ofcorruption in developing countries than in devel-oped ones.Arguments formulated in the 1990s (forexample by Persson and Tabellini 2000) suggestedthat under some conditions, presidential systemswould reduce corruption. However, there is littleevidence of this either in rich or poor countries(Adserà, Boix, and Payne 2003).

There is more evidence that electoral systemsaffect corruption,but the effect is subtle.Adserà,Boix,and Payne (2003) find no effect. But Persson,Tabellini, and Trebbi (2003) break down electoralinstitutions into their component parts—districtmagnitude (measured by the number of seats up forelection in the district) and voting rule—and arguethat the larger the district, the lower are the barriersto entry faced by competing parties and the morelikely it is that voters can drive out corrupt parties.When voters can express a preference for individualcandidates, as in plurality systems, they are better able


to remove corrupt legislators. Persson,Tabellini, andTrebbi (2003) find evidence that both effects are atwork.In practice,however,the two effects cancel eachother out, since countries with proportional electoralrules generally require voters to choose parties ratherthan candidates.Hence,electoral rules cannot explainthe greater levels of corruption in poor countries.

There is, then,no strong evidence that either spe-cial interest group organization or formal differencesin political and electoral institutions account for thedifferent policy choices of developed- and develop-ing-country democracies. Still, the arguments thatthese elements should matter are persuasive and seemto have great validity in richer countries.Their rela-tive weakness in explaining outcomes in poorercountries suggests that the underlying conditions ofpolitical competition in these countries differ fromthose in richer countries.We explore these condi-tions and their effects next.

Imperfections in Electoral MarketsDifferences in economic performance across democ-racies can be explained with respect to imperfectionsin electoral markets. Numerous imperfections inelectoral markets make it difficult for citizens to holdpoliticians accountable for policies.The discussionbelow focuses on three imperfections—uninformedvoters, noncredible political competitors, and socialpolarization—that offer powerful insights into theunderperformance of many democracies.

Uninformed VotersIn political markets, the information that votershave about the characteristics of political competi-tors and government performance is crucial.With-out information about the attributes of politicalcompetitors, about what politicians are doing, andhow their doings affect citizens’ well-being, citizenscannot easily reward high-performing politicians.This encourages poor performance.Politicians con-fronting uninformed voters can invest resources topersuade them of their accomplishments, throughadvertising or meetings, for example. But financingthese efforts, whether from their own pockets or

0

25

50

75

100

1990 2000 1990 2000

Presidential Parliamentary

Richer Poorer

%

FIGURE 10.2

Political Systems in Richer and Poorer Democracies

Source: Database of Political Institutions (Beck et al. 2001) and WorldDevelopment Indicators.

Note: Countries represented are those that held competitive electionsfor executive and legislative elections (the Legislative and ExecutiveIndexes of Competitive Elections from the Database of Political Insti-tutions were both equal to seventhe legislative and executive indicesof competitive elections from the Database of Political Institutionswere both equal to seven).

bility of receiving government transfers (Besley andBurgess 2002 and Strömberg 2002, respectively).

Figure 10.4 illustrates these effects, showing hownewspaper circulation, controlling for other influ-ences, suppresses corruption.15

Credibility of PoliticiansWhen challengers cannot make credible policycommitments to citizens, citizens have no reason toprefer them over incumbents. Even if incumbentsdo badly, citizens have no reason to believe thatchallengers will do better.16 This insulates incum-bents from pressure to perform.


those of special interests,or from government funds,carries a social cost: special interests demand poli-cies that diverge from the social interest in exchangefor campaign financing, while government fundingdiverts resources away from the provision of goodsand services to the electorate.

No data directly measure how well informed citi-zens are about the contributions that their representa-tives make to their welfare. One commonly usedproxy for citizen information is newspaper circulationper 1,000 of a country’s inhabitants. In 1995, amongcountries that hold competitive elections, newspapercirculation was,not unexpectedly, considerably higherin richer than in poorer countries (figure 10.3).

Controlling for income and other factors,highernewspaper circulation is associated with lower cor-ruption (Adserà, Boix, and Payne 2003), and withgreater rule of law, better bureaucratic quality, andgreater secondary school enrollment (Keefer2003a).As discussed later in this chapter, newspapercirculation and access to radios increase the proba-

45 42.8

35.4

22 21.7

11.8

0

25

50

Continuous years ofcompetitive elections (yrs)

Party age(yrs)

Newspapercirculation/1,000 people

Richer Poorer

FIGURE 10.3

Indicators of Political Market Imperfections in CountriesHolding Competitive Elections, 1995

Source: Newspaper circulation from World Development Indicators; party age (the averageage of parties under their current name); and continuous years of competitive elections, arefrom the Database of Political Institutions.

Note: Countries are all those with Legislative and Executive Indices Indexes of ElectoralCompetition (LIEC and EIEC) equal to the highest score of seven (see Figure 10.1). Theincome per capita threshold between richer and poorer democratic countries, dividing theminto roughly equally -sized groups, is US$6,193.

–206.656 343.875

–3.32613

2.16705

Newspaper circulation per 1,000 people (adjusted)

Corruption(adjusted)

FIGURE 10.4

Newspaper Circulation and Corruption


Note: The figure depicts the effect of the component of newspapercirculation that is uncorrelated with the other explanatory variableson the component of corruption that is uncorrelated with the otherexplanatory variables (the orthogonal component of each), based onthe regression below. The sample is of countries that exhibit compet-itive elections (LIEC=EIEC=7 from the Database of Political Institu-tions), 1990–2000; economic variables are from World DevelopmentIndicators; t-statistics are in parentheses; ordinary least squaresregression controls for clusters of observations from the same coun-try that artificially inflate statistical significance.

Corruption = 3.85 + 0.00002PPP-adjusted income/capita + (2.25e–08)land area–5.33) (0.66)(0.65) (9.3e–10) population – .44 electoralsystem – 3.87percent population young + 1.2 percent rural ( –1.14)(–2.49) (–1.98) – 0.01 political system + 0 .02 continuous years com-petitive elections + 0.002 newspaper circulation per 1,000 (2.62)(–0.07) (3.94) (2.88).

<Q? Figure okay? Artwork did not seem toimport perfectly.Also, figs 10.5 & 10.6>

In practice, politicians never entirely lack credi-bility. Some political competitors are credible ononly one or a few issues unrelated to economicdevelopment, such as a country’s struggle for inde-pendence, or issues of religious importance, but insuch cases, the votes they attract do not provide amotivation for better economic policy performance.

Credibility may also be partial in the sense thatpoliticians can make credible promises to some vot-ers only. Credibility resides in individual politiciansor in “patrons” rather than in political parties.The


problem of credibility is therefore closely related tothe phenomenon of clientelism, which is widelyargued to characterize political relationships inpoorer countries, and involves patrons and clientswho are bound together by reciprocal, long-lastingpatterns of exchange.These exchanges form thefoundation of reputations that allow patrons todeliver votes at election time. Unfortunately, nar-rowly based credibility gives politicians incentivesto underprovide public goods and to extract largerents (box 10.1).

BOX 10.1

Clientelism, Credibility, and Politics

Only since the late 1990s have scholars begun tounderstand why clientelism is a more dominantcharacteristic of public policy in some countries

than in others. One explanation derives from the struggleto make credible promises to citizens. Clientelism in publicpolicy prevails when average citizens cannot believe thepromises of political competitors with whom they have nopersonal connection. Such a connection emerges moststrongly in the context of patron-client relations. Scholarshave long noted that these relations have two importantcharacteristics: patrons and clients interact over a longperiod and they exchange goods and favors. Bista (1991,91–92) describes the key role of reciprocity in the opera-tion of clientelism in Nepal (where it is called chakari):“The gift donor in chakari has certain rights. There is anobligation on the part of the recipient to respond to thechakariwal when the chakariwal so determines . . . . Ulti-mately, there has to be a balance in exchange relations.”

Scholars of clientelism from Africa to SouthernEurope to East Asia confirm this pattern (Lemarchand1972; Powell 1970; and Scott 1972). Extended compli-ance with reciprocal obligations forms a basis for cred-ible commitment, which patrons can use if they decideto become politically active.

In fact, Scott (1972) quotes Wurfel as pointing outthat “the Filipino politician . . . . does favors individually

rather than collectively because he wishes to create a per-sonal obligation of clientship.” He cites the work of Nashon the 1960 elections in Burma: “When a local patron wasapproached to join U Nu’s faction of the AFPFL on thepromise of later patronage, he was able to get thirty-nineothers—his relatives and those who owed him money orfor whom he had done favors, i.e., his clients—to join aswell.” The rents to patrons were potentially high, sinceparties often had to give a local patron significant author-ity over local administrative and development decisions inexchange for vote delivery (Scott 1972, 110).

Patron-client relations drive politicians to focus ontargeted favors and goods over broad public goods andpublic policy: to the extent that only clients believepatron promises (given the absence of well-developedpolitical parties, for example), political competition con-cerns primarily targeted transfers to clients rather thanpublic policy issues more generally. Wilder (1999) quotesformer members of the Pakistani National Assembly fromthe state of Punjab as saying, “People now think that thejob of an MNA and MPA is to fix their gutters, get theirchildren enrolled in school, arrange for job transfers . . . .[These tasks] consume your whole day . . . .” (p. 196).“Look, we get elected because we are ba asr log [effectivepeople] in our area. People vote for me because they per-ceive me as someone who can help them”(p. 204).

Source: Keefer 2002.

Figure 10.6 also indicates that education spend-ing,which has little effect on gross secondary schoolenrollment in general, has a strong conditionaleffect: once one controls for continuous years ofcompetitive elections, education spending has a sig-nificant positive effect on enrollments.This is a clearindication of an increasingly well-identified phe-nomenon: that without appropriate political incen-tives, financial resources do little to improvegovernment performance.18

Taken together, then, the evidence suggests thatthe divergent performances of rich and poordemocracies can be traced to differences in theirexposure to electoral market imperfections.

Social PolarizationSocial polarization undermines the accountabilityof government to citizens.One type of social polar-ization emerges when substantial groups of citizens


A dysfunctional public sector limits the abilityof politicians to make credible promises.This is theproblem of capability that was discussed in chapter9. If an education ministry is deeply dysfunctionaland is likely to take years to reform, and if citizenscannot observe changes in the ministry until theseare reflected in schools, even favorably inclinedpoliticians are unlikely to make promises abouteducation. For example, when Alberto Fujimoribecame President of Peru in 1990, he privatizedenterprises, revamped the tax administration,removed price controls, reformed the customs sys-tem, and built up a large and successful social fund,but he explicitly ignored education, which was atleast as troubled as the sectors that he did address.

The credibility of pre-electoral promises is difficultto measure empirically.However, it is likely to be asso-ciated to some extent both with the number of yearsthat countries have experienced continuous electionsand with the age of their political parties.The passageof time allows (though it does not require) politicalcompetitors and parties to build up a reputation fortheir stances on policy issues.Among countries thathold competitive elections,as figure 10.3 illustrates for1995, both of these factors are considerably higher inricher countries than in poorer ones.

Similarly, where political reputations are sturdier,the effects of clientelism should be reduced, publicgood provision should be greater, and rent seekinglower, since political competition encourages politi-cians to provide high-quality public services. Keefer(2003a) finds that this is the case in practice: thelonger a country’s unbroken series of elections, thegreater are secondary school enrollment, the rule oflaw, and bureaucratic quality, and the less are corrup-tion and public investment as a fraction of GDP(public investment having the greatest political pay-offs to targeted constituencies).These effects are oftenlarge.The number of continuous years of electionshas a greater impact on corruption than do any ofthe other usual determinants, from newspaper circu-lation to formal constitutional rules to demographics(figure 10.5). It has a greater impact on secondaryschool enrollment than do education spending andprimary school enrollment (figure 10.6).17

–40.7955 43.446–3.87134

2.13072

Continuous years of competitive elections (adjusted)

Corruption(adjusted)

FIGURE 10.5

Continuous Years of Competitive Electionsand Corruption


Note: See note to Figure 10.4 for specification, data sources, andestimation methodology. Figure drawn from regression below:

Corruption = 3.85 + 0.00002PPP income/capita + (2.25e-08)land area– (9.3e-10)total population (5.33) (0.66) ( 0.65) (–1.14) – .44 elec-toral rule – 3.87percent pop. young + 1.2 percent pop. rural – 0.01political system (–2.49) (–1.98) (2.62) (-0.07) + .02 continuousyears competitive elections + .002 newspaper circulation per 1,000(3.94) (2.88).

have deeply opposing interests on most salientpolitical issues.These divisions can run so deep thatone group of citizens cannot contemplate electinga representative from the other. Elected representa-tives from one group then have no incentive to sat-isfy the concerns of citizens in the other.Moreover,they may have little incentive even to satisfy theconcerns of citizens from their own group; this canhappen if groups choose their candidates in a dis-torted manner (for example if backroom dealsdetermine who will be the candidate from eachgroup for the general election).

Majority disdain for the interests of identifiableminorities is another manifestation of social polar-ization. The more pronounced the disdain, thegreater the distortion in the provision of publicgoods, and the more likely that minorities will be


excluded from government services.There is sub-stantial evidence for these effects, and not only indeveloping countries. In the United States,Alesina,Baqir, and Easterly (1999) find that the more ethni-cally fragmented is a community (the smaller is thewhite majority), the more limited is the provisionof public goods. In such cases, the failure of politi-cal accountability does not show up as an excessivewillingness to serve special interests at the expenseof average citizens, but as the opposite: an excep-tional unwillingness to protect the rights of minori-ties. Subjective measures of ethnic tensions aredramatically higher in poorer democracies than inricher ones.19 Outright abuse of minorities hasbeen documented in a wide range of countries.

Both the nature and consequences of social polar-ization depend on the political environment. In thefirst case above,of “classic”polarization, there may bethird groups that are indifferent to the ideologicaldivide between the other two groups and whosesupport is needed to win elections. In the second,political institutions and circumstances can mitigateor exacerbate the effects of discrimination againstminorities. For example, scheduled tribes and castesin India received greater benefits once they wereguaranteed seats on local legislative bodies.Wilkin-son (2000) finds that Hindu-Muslim violence wasless common and elicited a more aggressive govern-ment response in those Indian states in which Mus-lims were pivotal voters. Rodrik (1999b) argues thatethnically fragmented countries had the greatest dif-ficulty reaching the agreement necessary to emergefrom crisis, though countries with a better gover-nance environment were able to offset this effect.

The consequences of social polarization can beworsened by all the factors that undermine voters’ability to hold politicians accountable. If voters arebetter informed about actions taken by members oftheir own social groups,or if they are more likely totrust promises by members of their own socialgroups, relative to members of other groups, theeffects of social polarization are likely to be wors-ened, and the rewards from controlling the govern-ment are more likely to flow to the groups whoserepresentatives control government. By contrast,

–40.3976 41.4514–42.2341

41.9462

Continuous years of competitive elections (adjusted)

Secondary school

enrollment(adjusted)

FIGURE 10.6

Gross Secondary School Enrollment and Continuous Years of Competitive Elections


Note: See Figure 10.4 for interpretation. Specification and estimationas in Figure 10.4 and Figure 10.5 , with the addition of primary grossschool enrollment and education spending.

Gross secondary school enrollment = 82.23 + 0.0009PPP income/capita– (3.68e–07) land area (6.96) (2.21) (–0.59) – (1.87e–08) totalpopulation – 176.7percent pop. young – 16.78 percent pop. rural(–1.23) (–8.53) (–1.79) + 0 .99 electoral rule +0.89 political system+0.16 continuous years of competitive elections (0.39) (0.52) (1.56) +.001 newspaper circulation per 1,000 + 0.34 gross primary enrollment+ 77.7 total education. spending/GDP (0.11) (3.7) (2.29).

promises to respect investors’ rights to their assets.When those promises are not credible, investmentsslow down or take inefficient forms:power plants areset up on barges rather than on land; older machin-ery is used at the expense of greater efficiency; inno-vation falls, in part because production techniquesare not at the cutting edge and in part because thefruits of innovation are themselves vulnerable toexpropriation.The growth effects are immediate:annual growth in income per capita in poor coun-tries with the most secure property rights is between2 and 4 percentage points faster than in poor coun-tries with the least secure property rights (Keefer andKnack 1997). Earlier chapters in this report attributethe weak effects of policy reform on growth partly toinstitutional weaknesses in countries.The inability ofcountries to secure property and contractual rights isa core element of these weaknesses.

Controversy emerged in the 1990s regardingtwo issues surrounding the growth–property rightsdebate: whether trade or other factors matter moreto growth than does the security of property rights,and whether the security of property rights isrooted in countries’ more fundamental geographi-cal features.The first debate is not resolved and maynot be, bound up as it is in intractable problems(Rodrik, Subramanian, and Trebbi 2002;Dollar andKraay 2003).21 The implications of the seconddebate loom larger. If geography determines thesecurity of property rights—that is, if geography isfate—the range of options for accelerating develop-ment is more limited.The role of geography is dis-cussed below, in the context of institutional andother determinants of government credibility.

. . . and Undermines PolicyA vast array of government policies need to be cred-ible to be effective.A key problem in monetary pol-icy, for example, is the threat that the government willenact a surprise increase in the money supply at theexpense of economic agents that have signed long-term contracts. Anticipating this, economic agentsfactor extra inflation into their contracts, raising thelong-term rate of inflation (Barro and Gordon 1983).


where the information and credibility gap betweenown-group and other-group representatives issmaller, social polarization is less likely to have dam-aging effects.

Controlling for numerous other factors, includ-ing income per capita, analysis shows that a com-mon subjective measure of ethnic tensions issignificantly worse, the lower are the average age ofpolitical parties (one measure of the credibility ofpolitical competitors) and newspaper circulation.20

The extent to which there are multiple large ethnicor linguistic groups in a country, though, has noeffect on ethnic tensions.

Another key element of social polarization is theability of competing groups to make credible com-mitments to each other. Previously quiescent inter-group relationships can suddenly explode intoconflict when the foundation for credible commit-ment crumbles. Bates, de Figuereido, and Weingast(1998) point to key events—such as the election ofSlobodan Milosevic in the former Yugoslavia—thatdisrupt arrangements that all groups believe have pro-tected them from aggression by other groups.Theinteraction of political and ethnic effects explains whysocial identity (ethnic, tribal, religious, geographic) isoften not politically salient and has no discernibleeffect on policy. It also explains why, as Posner (forth-coming) shows for Africa, the “identities” that matterfor politics often shift within the same population.

3. Government Credibility as aPrerequisite for Development

All of the foregoing relates to the reluctance orinability of political decision makers to adopt poli-cies in the broad public interest.A related problemfor development emerges when policies, onceenacted, are not credible.

Lack of Government Credibility Under-mines Growth . . .The most notable effect of credibility is on invest-ment and growth. Investors rely on government

Lack of government credibility also dampensincentives to invest in public infrastructure or tomake other social investments.The payoffs to theseinvestments depend on the willingness of economicactors to make complementary investments thattake advantage of them. Where expropriation ismore likely, private investors are slower to respondto improved public infrastructure, and governmentscorrespondingly reduce their allocations to theseinvestments. Keefer and Knack (2002) show that incountries with insecure property rights, measuredpublic investment is largely rent seeking. Whenproperty rights are insecure, an additional percent-age point of public investment as a fraction of GDPsignificantly reduces the rate of growth of incomeper capita; when property rights are secure, it adds0.3 percentage points to the growth rate.

The security of property rights has a similar effecton other long-term investments. In the 1990s,improved security of property rights increased grosssecondary school enrollment by an amount as large asdid a similar increase in government expenditure.22

Confirming the link between rent seeking and publicsector performance, Rajkumar and Swaroop (2002)find that child mortality rates and primary schoolattainment improve in response to increased publichealth and education spending only in countries withlow corruption and high bureaucratic quality.

Sources of Low Government Credibility:Lack of Reputation and Short Time HorizonsWhat makes government policies credible? Cer-tainly the elements of political competition thatallow political competitors to make credible pre-electoral promises help to ensure the credibility ofthe policies they implement after they take office.There is some evidence of this, if one accepts thatthe number of years that countries with competi-tively elected governments have continuously hadsuch governments corresponds to the opportuni-ties that political competitors have had to build rep-utations. The greater the number of years ofuninterrupted competitive elections, the moreestablished is the rule of law.23


However, many policies are not the subject ofpre-electoral debate. Even when they are, the gainsfrom reneging on policies, once implemented, areoften greater than the gains from reneging on pre-electoral promises to implement them in the firstplace.24 Finally, investors are always concernedabout a change in political control, from politicianswho have promised to support a particular policy tothose who have not.

The horizons of political actors—how long theyexpect to be in power or to be competing forpower—can mitigate these additional threats tocredibility. Governments that expect to be in officemany years have more to lose from current policiesthat upset future growth, such as investment-deter-ring expropriation, than do governments with shorthorizons.As noted earlier,Clague et al. (1996) showthat among nonelected leaders, the longer is theirhorizon, the more likely they are to respect prop-erty rights.These results do not imply that govern-ments should be immune to threats of removal.They do imply that in countries where accounta-bility mechanisms are flawed, extending the hori-zons of governments by making them more securein office may be the only means to create sufficientincentives to maintain secure property rights.

Sources of Low Government Credibility:Political InstitutionsMultiple institutional arrangements have been pro-posed to solve the problem of government credibil-ity, but in the end, only political institutions—particularly institutional checks and balances—havedemonstrated a consistent effect on the credibilityof government decision making. For example,Keefer and Stasavage (2003) find that only in coun-tries that exhibit political checks and balances doesthe legal independence of central banks suppressinflation. Moreover, even in countries withoutlegally independent central banks, political checksand balances can inhibit governments from reneg-ing on monetary commitments. Inflation is lowerin countries with political checks and balances thanin countries without them.

tries. Our earlier analysis suggests that the reasonmay be rooted in the underlying conditions ofpolitical competition. Improvement in these condi-tions, therefore, is likely to be an important com-plement to institutional reform.27

Growth and Accountability Boiled down to its essence, the foregoing argues thatelections alone are insufficient to ensure accounta-bility of governments to citizens. If this is true, onemight expect to find a stronger relationship betweengrowth and democracy if one takes account of thedifferent accountability mechanisms used in differ-ent democracies. In fact, there is evidence that theextent of imperfections in political markets has a sig-nificant impact on economic growth.

Easterly and Levine (1997) look at the effects ongrowth of one measure of social polarization: eth-nolinguistic fractionalization.Although they do notexplicitly consider the role of elections, they findthat their polarization variable has a significantimpact on economic growth. Keefer (2003b) pro-vides evidence that newspaper circulation, checksand balances, and the number of continuous yearsof elections have a significant impact on economicgrowth.Gerring,Brandt, and Bond (2003) similarlyfind that—controlling for whether a country isdemocratic or not (which has no impact ongrowth)—the total number of years that a countryhad elected governments through most of the 20thcentury had a significant impact on growth.

Taken together, these results make a compellingcase for reformers and development activists to takepolitical market imperfections into account in design-ing strategies to speed growth and development.

4. Lessons: Making Politics Workfor Policy When GovernmentsAre Not Credible and ElectoralMarkets Are Imperfect

How should we formulate strategies of policyreform,given imperfections in the market for polit-


This finding is consistent with a large body ofresearch on institutions and government credibility.North and Weingast (1989) argued that the intro-duction of checks on the English monarchy after theGlorious Revolution reduced the monarchy’s abilityto renege on sovereign debt obligations to foreigncreditors, and eventually brought down interestrates.25 Acemoglu, Johnson, and Robinson (2002)point, though less explicitly, to the role of the right tovote and political checks and balances as the key linkbetween the economic drivers of political power(relative prices and natural resource endowments)and the ultimate security of property rights.26

Engerman and Sokoloff (2002) make similararguments.These authors highlight the role of natu-ral endowments and other exogenous factors indetermining the types of institutions that countriesexhibit.The essential logic is that in countries whereownership of economic activity is concentrated inthe hands of a few,and assuming that political powerfollows control of the country’s most valuable assets,there is no incentive for the politically and econom-ically powerful to enfranchise the powerless.Thegains that the elite could reap—in the form ofincreased productivity of the masses, whose greaterpolitical rights would protect them from expropria-tion—would not offset the losses that the elitewould face in being forced to share the high returnsfrom their plantations, mines, or oil wells. Hence,these countries grow slowly, and the price of slowgrowth is borne by the disenfranchised.

If geography were the main determinant ofinstitutional development, there would be littlepurpose in institutional reform, and developmentpossibilities would be limited. But in practice manycountries have made institutional changes that seemto represent an escape from geographically deter-mined destinies. One example is the wave ofdemocratization in the 1990s.Another is the spreadof democracy in Latin America in the 1980s and1990s, precisely where conditions for democracyare supposed to have been the least propitious.

In all of these cases, the question remains whythe introduction of formal institutions is not suffi-cient to ensure sustained development across coun-

ical office and limitations on the credibility of gov-ernment commitments? And what reforms mightmitigate these political and institutional problemsdirectly?

The traditional answer to the first question is tobuy off the opposition to reform. This formularequires political leadership: buying off the loserswho are in a position to block reform, and exploit-ing windows of opportunity such as crisis or achange in government. However, nothing in thetraditional formula hints at the fact that reform maybe systematically more difficult in some countriesand policy areas than in others.

Political Market Imperfections Explain WhyBuying off Reform Losers Usually FailsThat compensation strategies have rarely succeededis not surprising. First, the compensation needed topersuade reform losers to support reform can beprohibitively high; for example, the benefits to thefertilizer industry of fertilizer subsidies in Indiaamount to 0.7 percent of GDP each year (Pana-gariya 2003). Even large payouts may be feasible ifthe gains are correspondingly large. But when theimperfections in the market for political officeloom as large as they do in many countries,or whenpolitical institutions provide few checks on oppor-tunistic behavior by politicians, adequate compen-sation may be impossible. If politicians cannot makecredible promises to voters, they cannot make cred-ible promises of compensation to losers fromreform. And if citizens are poorly informed aboutwhat politicians do in office, losers may be unableto observe whether governments have actuallydelivered the promised compensation. Hence rem-edying the underlying imperfections in electoralmarkets is a prerequisite for successful reform.

Second, institutional deficiencies can alsoundermine compensation strategies. If reform los-ers control government decision making, they can-not credibly promise to refrain from introducinginefficient policies that benefit them at the expenseof others: once they have received compensation,nothing prevents them from reverting to policies


that run counter to the public interest. In this case,institutional reform is an essential prerequisite ofpolicy reform.

Third, a reform itself can undermine the basesfor making credible agreements. Consider an effortto downsize a ministry or to close a money-losingstate-owned enterprise. If they are well organized,threatened workers can oppose these efforts bydemonstrating, targeting contributions to politi-cians who will help them,and purchasing advertise-ments to sway public opinion. To offset thisopposition, reform proponents could offer theworkers a large pension. Should this proposal beaccepted, the workers will be sent home.Once scat-tered, they cannot easily block subsequent efforts toreduce the generous pension. Reform has under-mined the political power that would allow them toenforce the agreement.28 Realizing that they wouldthen be vulnerable to government efforts to recap-ture the pension from the now-disorganized work-ers, workers therefore reject the proposal.

Dramatic increases in prices are relatively easy toattribute to political failure (Keefer and Khemani,forthcoming). But some other types of policyreforms—banking and social service delivery, forexample—are more vulnerable to political marketimperfections and institutions that fail to solidifygovernment credibility.

Banking crises emerge after years of regulatoryneglect and imprudent lending practices. It is diffi-cult to assign political responsibility for thembecause these practices may occur under multiplegovernments, and because politicians can easilyblame regulators for shirking and bankers for crim-inal behavior. Such claims are difficult for voters toevaluate in every country, so it is not surprising thatthe fiscal costs of banking crises are almost exactlythe same in poorer and richer countries.29

Social services are also vulnerable to electoralmarket failure.The goal of universal education isexactly contrary to clientelist political motiva-tions.30 It is quite difficult for citizens to assignblame to politicians for health and education fail-ures, which could be due to idiosyncrasies of indi-vidual health status, or to shirking by service

policies and programs.The spread of the radio par-ticularly improved information access for rural vot-ers, who had previously been disadvantaged relativeto urban voters, with the latter’s ready access toother information sources such as newspapers. Itaccounted for as much as 20 percent greater alloca-tion of social assistance funds to a rural county ascompared to an identical urban county (Strömberg2001). Besley and Burgess (2002) find that stategovernments in India are significantly more respon-sive to declines in food production and crop flooddamage via public food distribution and calamityrelief expenditure where newspaper circulation,particularly in local languages, is greater.

These findings raise some unresolved issues.Thestudies suggest that information (newspaper circu-lation) seems to have only a limited effect on theprovision of public goods (gross secondary schoolenrollment).This result, consistent with those ofother studies, suggests that the availability of infor-mation has its greatest impact on the provision oftransfers to voters, who can easily monitor suchtransfers with the assistance of a thriving mediaindustry. For example, even in societies with edu-cated societies and unrestricted media, voters tendto be relatively uninformed about the specifics ofgovernment performance.While government-con-trolled media are more likely to limit citizens toinformation favorable to the government, privatemedia can be controlled by special interests thathave their own biases. This is less problematic ifthere are low barriers to entry into the news busi-ness, but even if barriers to entry are low, it mightbe the case that other types of news are more prof-itable to report than information on governmentperformance (Strömberg 2002). For example, themedia might prefer to report extreme outcomesthat are not typical of the government’s perform-ance and that bias voter perceptions.

Indirect evidence of this emerges from work oncampaign finance reporting in the United States.Theresearch suggests that newspapers systematically biasthe information that citizens receive about campaignfinance.Specifically,newspaper reporting conveys theimpression that politicians receive more contribu-


providers, or to the fact that the country lacksresources.

The essential lesson is that for policies and coun-tries in which electoral market failures loom large,reform efforts should focus on mitigating these fail-ures rather than on paying off losers or encouragingleadership or awaiting the opening of windows ofopportunity.Where market failures are too large,the first may be too expensive, the second unrealis-tic, and the third may never occur.

Mitigating Electoral Market FailuresWhat measures might alleviate imperfections in themarkets for political office?31 Mitigating electoralmarket failures essentially means reducing politi-cians’ incentives to engage in clientelist behavior.How to shift political competition away from clien-telism is a key challenge of institutional reform thatis not yet well understood. Some steps are probablykey to reform, however: increasing public informa-tion, and increasing the credibility of politicalpromises.

Increasing Public InformationAn important step is to encourage, or removeimpediments to, nongovernmental sources of infor-mation on reform needs and direction. If their cred-ibility is established, such sources can validatereform strategies outlined by interested politicalactors.The report cards undertaken in India by thePublic Affairs Centre of Bangalore, mentioned inchapter 9, are one way in which nongovernmentalorganizations (NGOs) can credibly collect infor-mation about the performance of public officialsand use it to stimulate reform.

The media also appear to be key for increasinggovernment responsiveness. Research on India andthe United States during the Great Depressionhighlights how information can improve access togovernment assistance. Between 1933 and 1935 inthe United States, federal assistance to low-incomehouseholds was significantly greater in those coun-ties where more households had radios and werethus more likely to be informed about government

tions overall and a higher fraction from corporationsthan they in fact do.In turn,college-educated Amer-icans—those who are most likely to read newspa-pers—believe campaign financing flows areapproximately what newspapers report, while lesseducated Americans believe they are considerablyless (Ansolabehere, Snowberg, and Snyder 2004).

The media may solve the following coordinationproblem: voters unhappy with a government, forwhatever reason, may be reluctant to oppose thegovernment if they think their own experience isisolated. By conveying a general impression of gov-ernment performance to which all voters areexposed, individual voters who share that impressioncan be more confident that others share it as well.This reduces their reluctance to support or opposeperforming or nonperforming governments.

Information reforms must also grapple with theconditions under which politicians respond to therevelation of information about their performance.Scandalous information frequently has no politicalimpact: even public knowledge of criminal behav-ior by politicians is not a sufficient condition forpoliticians to leave office, in either developed ordeveloping countries. Newspaper circulation canreduce corruption, as seen earlier in this chapter,but appears to have no effect in countries that lackcompetitive elections.Among countries with com-petitive elections, the influence of newspaper circu-lation on corruption depends, though less robustly,on the existence of political checks and balances.32

Competing political forces inside government,eachwith the right to influence government decisions,have both the incentive and ability to use evidenceof each other’s mal- or misfeasance for their ownpolitical advantage.

The efficacy of other information reforms alsodepends on the political environment.A key char-acteristic of government in many developed coun-tries is the transparency with which newregulations emerge from the executive branch ofgovernment. These range from the issuance ofwhite and green papers in the United Kingdom toopen meeting requirements and freedom of infor-mation laws.These transparency requirements are


almost always imposed by politicians on themselves,and are potentially but not always enforceable bycourts.This means, however, that the requirementshave less effect to the extent that there is little polit-ical cost to politicians who decide not to abide bythem and to the extent that the courts are reluctantto require adherence to them. Unfortunately, thepolitical costs of ignoring transparency laws arelikely to be lowest precisely where government per-formance in general is likely to be poor: wherethere are few political checks and balances, andwhere political competition is organized aroundclientelist favors rather than overall governmentperformance.

Increasing the Credibility of Political PromisesThe other major electoral market imperfection—the lack of credibility of political competitors—ismore difficult to address. In principle,political cred-ibility should provide political competitors with acompetitive advantage. Clientelism (the defaultoption for political competition when politicianscannot make credible promises) is expensive.Theresources needed to give 50 voters jobs couldfinance broad public goods, such as improved edu-cation, offering equivalent benefits to hundreds ofvoters.That is, politicians who can offer crediblepolicy or public goods to a large number of voterscan defeat politicians who can only operate in apatronage mode.

However, moving out of clientelism is risky forpoliticians. Shifting resources to public goods mayleave clients sufficiently dissatisfied to desert theirpatron, while public good benefits may materializetoo slowly to attract new bases of political supportbefore the next round of political competition. Inany case, the beneficiaries of improved public ser-vices may not credit the incumbent politician forthe improvements.

How can politicians build the credibility of theirpromises to improve the quality of public goods?Leaders can build credibility by being vocal,emphatic, and specific about their reform goals.Specificity makes it easier for citizens to judge whenleaders have failed.Emphasis makes it clear that lead-

erate project implementation, but it reinforces pat-terns of political accountability in which votersexpect only targeted or clientelist benefits fromtheir leaders. Donors also often agree with govern-ments to set up enclaves of bureaucratic excellenceto carry out particular tasks.As discussed in chapter9, while enclaving can assure governments thatpromises will be carried out, enhancing the sustain-ability of reform, this potential benefit is rarely real-ized, since the end of a task often means thedissolution of the enclaved agency.

The development community is also doing con-siderable work to “institutionalize” political parties,improving their ability to communicate with votersor to organize at the grassroots level.This is poten-tially important for achieving the ultimate goal ofimproving policy credibility and voter informationabout the policy stances of political competitors.However, it has no guarantee of success, since partyleaders may prefer to construct patronage machinesor vehicles for personal advancement rather thanrely on the institutionalization of their party’s stanceon policy issues (box 10.2).

Mitigating Political Market Failures:Institutional and Legal Reforms Even though institutional factors do not systemati-cally explain the underperformance of somedemocracies relative to others, institutional reformscan promote policy reform. Such reforms includechanging electoral rules, reinforcing checks and bal-ances, introducing laws that regulate campaign con-tributions, and decentralization.

Electoral Reforms Can Spur Sustainable Policy ReformReform of electoral laws can both spur reform andserve as a vehicle for mitigating electoral marketimperfections. One indication of the policy effectsof such reform emerged in the 1990s in Japan.Priorto its 1994 reform, the electoral system in Japan wasa mix of plurality voting and multimember districtsthat essentially compelled candidates from the samepolitical party to compete with one another.


ers expect to be judged on their performance regard-ing these goals, rather than on other issues, and inde-pendent of shocks or difficulties that might emerge.Together these improve credibility. As is often thecase, there may be a tradeoff between reform successand building up credibility. Publicizing reform mayincite resistance that stifles reform, while successfulreform undertaken unpublicized has fewer politicalbenefits and may be less sustainable.

Public sector reform can help too. A politicalcompetitor is unlikely to promise improved provi-sion of public goods if the organization needed tosupply those goods is dysfunctional, since citizenscannot easily distinguish whether reform failure iscaused by bureaucrats or politicians.33 Unfortu-nately, public sector reform is itself an arduousprocess that requires political commitment. In sys-tems where politicians have a strong political inter-est in satisfying clientelist demands, their incentivesto improve the functioning of the civil service areweak.Cox (1987) demonstrates that the profession-alization of the justly acclaimed British Civil Ser-vice followed, rather than preceded, the shift in thebasis of political competition from clientelism topartisan or policy differences.

Donors can help developing-country govern-ments with this dilemma by coordinating theirassistance for public sector reforms with their assis-tance for improving the provision of public goods,while being sensitive to the political timetableaccording to which citizens express their judgmentsabout these reforms.A successful reform strategy isone that devises and links “on-the-ground” out-comes to intermediate stages of public sectorreform, such that politicians can get credit forreform in a timely fashion. (Again, governmentleaders need to be vocal in promising results, or thecredibility effects will be diminished and voters willhave little reason to change their judgments aboutincumbents based on the reform experience.)

Donor strategies for project implementation arerelevant here. A donor focus on specific projectstouching a fraction of the population, rather thanon broad policy goals and public good improve-ments that affect most of the population,may accel-

Because they could not use party labels to distin-guish themselves from competitors, candidatesspent considerable sums of money distinguishingthemselves in other ways, thereby building up per-sonal constituencies. These constituencies hadclientelist attributes. Politicians, for example, wouldappear at weddings and funerals, making cash con-tributions to the newly married or bereaved.Theirneed for financial resources led incumbent politi-cians to be especially generous toward special inter-ests, including the banking industry. The laxregulatory standards to which banks were held con-tributed to soaring nonperforming loans. Thesewere exposed when rapid economic growthground to a halt in 1990.

The electoral reform of 1994 introduced single-member districts and changed rules in multimem-ber districts to proportional rather than pluralityelectoral rules.These changes raised the electoralvalue of partisan affiliation and reduced the need


for money in campaigns. Soon after, in 1996, theruling Liberal Democratic Party forced banks tobail out their mortgage-lending subsidiaries andabsorb huge losses rather than socialize the losseswith taxpayer-financed bailouts.34 Policy reformsthat had been urged on Japan for years finallyoccurred, but only after the adoption of institu-tional reforms that changed political incentives.Whether such reforms would have an equal effectin countries without well-established political par-ties and informed voters is less clear.

Checks and Balances: Difficult to Introduce,Easy to UnderminePolitical checks and balances have a significant effecton government credibility and, as a consequence,onthe effects of policies in areas ranging from taxationto public investment and monetary policy.35 It isdifficult to introduce political checks and balanceswhere none exist, however.They require both for-

BOX 10.2

Political Parties and Reform

In many countries political parties are suppressed orlimits are placed on the extent to which they canmake ethnic appeals. Candidates in some elections

in Pakistan and Uganda have been prohibited from run-ning under a party affiliation. In Bulgaria, ethnicallybased parties have been excluded.

While parties are far from a sufficient condition foreliminating electoral market imperfections, they maybe necessary. Mature political parties with well-definedpositions on economic and social issues help solveproblems of both information and credibility that oth-erwise plague competition for political office. Matureparties convey information to voters on the policystances of party members, particularly relative to mem-bers of other parties. Unlike individual candidates, theyare more likely to have policy reputations that allowthem to make credible promises to voters. When partiesare credible entities, voters can more easily assign

blame and credit to the parties in control, relievingthem of the need to identify specific individuals to holdresponsible.

Unfortunately, history is replete with partieshijacked by personal interests or dedicated to patron-age politics or serving as a locus for ethnic rivalry orreligious conflict. Parties often fail to offer voters acredible choice in terms of economic policies.

At the same time, policy-based political parties canemerge from or succeed in a clientelist milieu, as maybe indicated by the fall of the PRI in Mexico from dom-inance and the persistence of the Partido dos Trabal-hadores (Workers Party) in Brazil. Nor is policy reformimpossible in clientelist environments—the mosteffective means for politicians to capture the vastmajority of disaffected voters who do not benefit fromclientelist payoffs is to develop a reputation for policyperformance that benefits the majority.

solving electoral market failures. Among countriesthat exhibit political checks and balances, the ruleof law and corruption are still strongly affected byvariables that capture the effects of some of thesefailures. However, their absence underminesprospects for sustainable reform and their develop-ment is therefore important.

Campaign Finance Reform: Attacking theSymptom, If Not the Disease Other institutional reforms can reduce both elec-toral market failures and the lack of credibility—although they can potentially exacerbate them aswell. One is campaign finance reform. Popular inboth developed and developing countries, the gen-eral notion is that to prevent special interests fromusing money to distort political outcomes,one mustplace caps on campaign finance or increase publicfinancing of elections.The evidence is not in on theefficacy of either solution, though the latter is likelyto be more effective than the former.

Evasion and enforcement have everywhere beena serious problem with campaign finance reforms.In the United States, caps on one form of contribu-tion have led to dramatic increases in other forms.Even when caps are comprehensive, as in Franceand Germany, reports on campaign finance scandalssuggest that the flows continued nonetheless. Eva-sion and nonenforcement are more likely in coun-tries in which politics is clientelist and large policyissues are not germane to political competition.

Even where caps are binding, some observersargue that they actually increase the returns to lob-bying. Drazen, Limão, and Stratmann (2004) arguethat moderate caps on political contributions caninduce more lobbyists to enter the political market,offsetting the reduction in contributions by existinggroups.They find some evidence for this,moreover,across U.S. states, which exhibit sharply differentcampaign finance regulations.

Decentralization: Finding More Perfect Political MarketsDecentralization embraces a range of institutionalreforms that have the possibility both of upsetting


mal institutions that endow multiple branches orindividuals of government with authority over gov-ernment decision making; and they require thatthose individuals or branches enjoy independentsources of political authority.The first is relativelyeasy to accomplish, through statutory or constitu-tional amendment.The second is difficult.

When formal institutions of checks and balancesare present (formal legislative, judicial and executivebranches of government, for example), steps can betaken to reinforce checks and balances even whenthe branches do not enjoy independent sources ofpolitical authority, as is often the case.Their effectsare likely to be small,however,until political author-ity is more equally shared among the branches.

For example, public sector financial manage-ment reforms increase the information available tolegislators inside government. Often, though, legis-lators have no incentive to act on this information:their prospects for reelection depend on maintain-ing good relations with the executive branch, suchthat the executive branch will fund projects in theirconstituencies, and this weakens their incentives tosupervise the executive’s overall performance.Thisdependency is less important in countries (such asthe United Kingdom) where strong parties providean offsetting check on political excess. But withoutstrong political parties, budget rules that deny legis-lator influence over spending undermine politicalchecks and balances.

Moreover, formal institutions are often incom-plete in these circumstances: budget-making pro-cedures deny them the policy-making leveragethey need to act on the information. Legal andconstitutional changes that endow legislatureswith very limited authority over spending preventthem from imposing budgetary sanctions on gov-ernment ministries that diverge from agreed allo-cations and amounts.

Where political checks and balances are weak,implementation of reforms—or of donor-sup-ported projects—is more likely to be undermined.Closer donor supervision is the most effectiveshort-run response to avoid this.At the same time,political checks and balances are not a substitute for

clientelist political patterns and of reinforcing them.To the extent that political competition and decisionmaking are less subject to political market imperfec-tions—information, credibility, or social polariza-tion—in subnational than in national governments,policy outcomes are likely to be more conducive todevelopment. Similarly, by splitting up issuesbetween national and subnational governments,decentralization facilitates voter efforts to holdpoliticians accountable for specific policy areas, andalso assists political efforts to develop policy reputa-tions that go beyond clientelism.36 However, thesepreconditions for successful decentralization are fre-quently absent, and in their absence decentralizationcan exacerbate the policy distortions of clientelism.

5. Conclusion

The arguments in this chapter paint a broad pictureof the role of political economy in development andhighlight a few characteristics of political systemsthat help explain some development outcomes:

• Can voters observe the decisions of governmentofficials and the effects of these decisions? Caneven informed observers attribute politicalresponsibility for policy failure? They cannot ifpolitical parties are amorphous and individual par-ticipation in political decision making is opaque.

• Are policy differences at all relevant to politicalcompetition? Do party platforms exist and, if so,do they diverge? Can the average citizen recog-nize and rely on policy differences among theparties? If not, political competition is sure tofocus on the allocation of narrowly targetedbenefits—projects, jobs, exemptions from oner-ous regulations—and promises of broadly basedreform are unlikely to be credible.

• Are checks and balances present and operative?

These questions are important in seeking tounderstand societies’ collective decision-makingprocess.


Notes

1. Countries are regarded as having competitively electedgovernments if they are reported in Beck et al. (2001),Database of Political Institutions, as having the highest score(seven) on the Executive Index of Electoral Competi-tion (EIEC) and on the Legislative Index of ElectoralCompetition (LIEC), where seven implies that there aremultiple parties competing and no party gets more than75 percent of the vote.The rule of law measure is fromPolitical Risk Services’ International Country Risk Guide.

2. These same sources of heterogeneity, especially informa-tion and social polarization, may also matter in non-democracies. This possibility is not exploredbelow—there is little evidence bearing on the question—but future work needs to explore the overarching deter-minants of good government performance that might becommon to both democracies and nondemocracies.

3. Banerjee and Iyer (2002) show that British colonialpractices affect land tenure relationships and land pro-ductivity in parts of India to this day.Acemoglu, John-son, and Robinson (2002) argue that a complexinteraction of relative price changes, natural endow-ments, and institutional choice has consequences thatlast for generations.

4. These results persist even when countries in EasternEurope and Central Asia are excluded.

5. The effect is nearly as large as that of a standard andalways powerful control, income per capita: an increaseof one standard deviation in the years a nonelectedleader is in office reduces the risk of expropriationalmost as much as does a one standard deviation in acountry’s income.

6. Controlling for total population, population living inrural areas, land area, population under the age of 16,and purchasing power parity–adjusted income percapita, and looking only at countries that did not exhibitfully competitive elections in 1995, the absence of a leg-islature of any kind, elected or not, was associated witha one standard deviation worsening of the rule of lawand corruption measures.

7. Persson and Tabellini (2000).8. Kontopolous and Perotti (1999) and Persson, Roland,

and Tabellini (2003) argue that proportional representa-tion systems encourage small parties, which increasesthe prevalence of minority governments or multipartycoalition governments, which in turn increases taxesand spending.Majoritarian systems,as argued by Milesi-Ferretti,Perotti, and Rostagno (2002) and others, shouldlead to greater attention to pivotal voters, and thereforemore targeted spending, rather than spending on broad-based public goods or redistributive programs.

for which this is not the case, there is no obvious reasonthat would impel a government to allow a free press thatreports on how well the government is performing.

16. A common refrain of voters in many countries is, “Allpoliticians are the same, and none is interested in thepeople.”Uninformed voters would naturally express thisopinion. So too would voters confronting politicianswho cannot make credible promises.

17. From cross-sectional regressions using democraticepisodes (1975–2000) as the units of observation, con-trolling for land area, total population, percentage of thepopulation that is young, and political and electoral sys-tems.

18. See, for example, Rajkumar and Swaroop (2002). It ispossible that the relationships between continuous yearsof competitive elections and corruption or education,for example,are due to omitted effects that in turn influ-ence both of these. One can control for this possibilityby identifying instrumental variables that explain com-petitive elections but not education or corruption.Results are robust to instrumental variable estimation.using the share of nonmanufacturing activity in totalindustrial activity in a country in 1965 and/or 1975.These capture reliance on natural resources (especiallymining), which in turn is often thought to discouragepolitical development. It does not explain either cor-ruption or secondary school enrollment but is stronglyrelated to years of continuous competitive elections.

19. One and one-half standard deviations higher, using ameasure of ethnic tensions from the International Coun-try Risk Guide.

20. The regression controls for income per capita, percent-age of the population that is young or rural, land area,and total population of a country, for all years since1989.

21. Rodrik, Subramanian, and Trebbi (2002) argue that thesecurity of institutions (measured as the security ofproperty rights, the rule of law, and so on) matters morefor economic development than geography and trade.Dollar and Kraay (2003) show that the instrumentalvariables used to control for the endogeneity of bothtrade and measures of governance or the security ofproperty rights yield a high correlation between thetwo, making their independent effects difficult to assess.

22. Results from regressing yearly data on gross secondaryschool enrollment from 1990 to 1997 on gross primaryschool enrollment, PPP-adjusted income per capita,land area, the fraction of the population that is young,total population, education spending as a fraction ofGDP, and expropriation risk.

23. Controlling for PPP-adjusted income per capita, totalpopulation, the fraction of the population that is young,land area, whether the political system is presidential or


9. Persson and Tabellini (2000) argue that vote of confi-dence procedures in parliamentary democracies bindlegislative majorities together, allowing them to makecredible agreements that taxes raised will serve the inter-ests of the majority.This encourages them to establishhigher taxes and spending. In presidential systems, leg-islative minorities (for example chairpersons of legisla-tive committees) are more powerful, but they areassumed to be unable to make credible agreements witheach other. Spending is targeted to the constituencies ofthese legislative minorities, but because they cannotcredibly agree with one another that higher tax rev-enues will be targeted to their constituencies, overallspending is lower.

10. Elections in which there were multiple competing can-didates or parties, more than one party contesting, andno candidate or party winning more than 75 percent ofthe vote, taken from the Legislative and ExecutiveIndexes of Electoral Competition, in Beck et al. (2001).

11. This result uses institutional data from Beck et al. (2001)and economic and social data from World DevelopmentIndicators.

12. Education spending, controlling for primary schoolenrollment, has a small effect on secondary schoolenrollment. One estimate suggests that a full percentagepoint increase in education spending as a fraction ofGDP (where the average country spends approximately3.3 percent of GDP) increases gross secondary schoolenrollment by fewer than 5 percentage points (wheregross secondary school enrollment in the average coun-try is approximately 65 percent).

13. These results are from an ordinary least squares regres-sion of gross secondary school enrollment on PPP-adjusted income per capita; the percentage of thepopulation that is young; land area; gross primary schoolenrollment;whether a system is parliamentary,presiden-tial, or semi-presidential; the voting rule used to electthe majority of representatives in the lower chamber ofthe legislature; and the average district magnitude of thechamber.The data is yearly, from 1990 to 2000. Signifi-cance tests based on robust standard errors assumingcountry observations from different years are not inde-pendent.The economic variables are from World Devel-opment Indicators; the political variables from the Databaseof Political Institutions (Beck et al. 2001).

14. In neither group are electoral institutions a significantdeterminant of gross secondary school enrollment. Spec-ification is as described in footnote 15.Standard errors areWhite-corrected (robust).The economic variables arefrom World Development Indicators; the political variablesfrom the Database of Political Institutions (Beck et al. 2001).

15. The sample includes only country-years in whichcountries had competitive elections, since for countries

parliamentary, whether the electoral system is propor-tional or plurality, and the average district magnitude(yearly data only for countries that have leaders chosenby competitive elections).

24. For example, the gains from reversing tax cuts meant toencourage investment are potentially substantial afterfixed investments have been made in response to the taxcut.

25. Stasavage (2003) revisits this episode and concludes thatparliament only restrained opportunistic behavior bythe government when the minority of parliamentarymembers who favored honoring sovereign obligationswere able to make a deal involving religious freedomwith those who were less favorable.That is, he showsthat not only did institutions matter, but so did politics.

26. This point is explicit in the theoretical work of Ace-moglu and Robinson (2001), who argue that only bysharing power can the disenfranchised be persuadedthat the enfranchised will not expropriate them.

27. Acemoglu and Robinson (2001) explain the poor per-formance of some democracies by arguing that, incountries exhibiting high inequality, as in Latin Amer-ica,democratization would give rise to significant redis-tribution and lay the groundwork for democracy’scollapse as the elites aimed to take back power. Cer-tainly, Latin American democracy throughout the 20thcentury has been notably unstable. However, simpleaverages of government expenditure as a fraction ofGDP and of education spending specifically, as a frac-tion of GDP, show little difference between democraticand nondemocratic periods since 1975. If anything,government spending was slightly higher in the nonde-mocratic country–years than in the democratic; educa-tion spending was almost identical.

28. See, for example,Acemoglu and Robinson (2000).29. The total cost is 12 percent of GDP. Data from Hono-

han and Klingebiel (2000), data on real incomes fromAten, Heston, and Summers (2001).

30. Social funds, in contrast, which are intended to distrib-ute resources to particular groups, are potentially usefulto clientelist politicians. Schady (2000), for example,shows that the Peruvian social fund, FONCODES, waswell targeted to the poor,conditional on the poor resid-ing in areas where President Fujimori thought political


transfers would be most useful.The poor in oppositionstrongholds were particularly unlikely to receive funds.

31. In advocating efforts to organize the demand forreforms, chapter 9 points out that reform winners areoften disorganized and confront significant barriers tothe collective action that would make them effectivesupporters of reform. Lowering these barriers is onerecommendation of chapter 9. The political marketimperfections and institutional deficiencies discussedhere are additional obstacles to successful reform, andalso undermine the efforts of reform proponents to shiftgovernment policy.

32. Controlling for income per capita, population variables,and the years that elections have been continuouslyheld, the effect of an increase in newspaper circulationon corruption approximately doubles, moving from aparliamentary political system in which the party of theprime minister party controls the legislature to one inwhich a four-party coalition government is in power.

33. See Shepherd (2003) for a thorough review of the argu-ment that meritocratic and well-performing civil ser-vants improve government credibility.

34. Rosenbluth and Thies (2001).35. Development assistance can have the unfortunate side

effect of undermining political checks and balanceswhere they do exist. Chapter 9 described a number ofreforms, particularly medium-term expenditure frame-works, that are meant to ensure that all public spendingis subjected to the scrutiny of multiple actors in thepolitical system. It also identifies the hazards of funnel-ing outside resources directly to line ministries, outsidethe normal budget processes.

36. Besley and Coate (2001) argue in the context of citi-zens’ initiatives that a key problem in politics is that gov-ernments make decisions on numerous policydimensions, but voters can only cast votes for a singlepolitician or party.They are confronted, therefore, withan “all or nothing” offer: politicians can shirk on somemargins, but still be reelected if they are sufficientlyforthcoming on the “salient” dimensions of policy.Decentralization eases this problem by splitting issuesbetween multiple levels of government, allowing morepolicies to become politically salient than would other-wise be the case.

does democracy help? - world bankdemocracy and development,finding that compet-itive elections have...

Documents