the measurement and impact of corruption victimization ......are measures of corruption perception...

World Development Vol. 34, No. 2, pp. 381–404, 2006� 2005 Elsevier Ltd. All rights reserved

Printed in Great Britain
0305-750X/$ - see front matter
doi:10.1016/j.worlddev.2005.03.012www.elsevier.com/locate/worlddev

The Measurement and Impact of Corruption

Victimization: Survey Evidence from Latin America

MITCHELL A. SELIGSON *

Vanderbilt University, Nashville, TN, USA

Summary. — Does democracy suffer when corruption is high? Unfortunately, most studies of cor-ruption have been descriptive, making it difficult to develop a well-supported answer to this ques-tion. There are, of course, quantitative corruption measures, but the most popularly used of themare measures of corruption perception aggregated at the national level, rather than corruption expe-rience at the individual level. It is difficult to draw firm conclusions about the impact of corruptionfrom such data. This paper proposes a corruption–victimization measure based on a survey re-search approach to the measurement problem, and discusses (in a non-technical fashion) itsstrengths and limitations. The paper then presents data from studies conducted by the VanderbiltUniversity Latin American Public Opinion Project (LAPOP), and demonstrates the impact of cor-ruption victimization on belief in the legitimacy of the political system, a critical element in dem-ocratic political stability.
� 2005 Elsevier Ltd. All rights reserved.
Key words — corruption, Latin America, survey research, democracy

* This is a revised version of a paper presented at the

conference ‘‘Re/Constructing Corruption,’’ University

of East Anglia, England, April 30–May 2, 2003. I would

like to thank Nick Duncan for inviting me to that con-

ference and his helpful remarks on an earlier version of

this paper, as well as the many helpful suggestions of

Indranil Dutta, Maria Barron, and Margaret Sarles.

Final revision accepted: March 9, 2005.

1. CORRUPTION AS A THREAT TODEMOCRACY: IS THERE A LINK?

Widespread corruption is increasingly seen asone of the most significant threats to deepeningdemocratization in Latin America (and indeedmuch of the democratizing Third World). In arecent study, Weyland (1998) views that democ-racies in Latin America are threatened by anexponential growth in corruption that has oc-curred since the dictators of the past have leftthe scene. He argues forcefully that corruptionhas increased a great deal under democracy inLatin America, and points to several factorsthat are responsible for the increase. First, heargues that the dispersion of power in thehands of many (as has occurred as dictator-ships have been replaced by democracies) haswidened the opportunity for bribery. In effect,there are many more ‘‘veto players’’ today thanthose under the military, and therefore, therehas been an increase in the number of palmsthat need to be greased. Second, neoliberal re-forms have involved opening many areas ofthe economy to bribery, especially those involv-ing sales of public corporations. Third, theincreasing numbers of neopopulist leaders,who win elections based on personal appeals

381

through television, are driving aspiring politi-cians to corrupt practices in order to collectthe funds needed to pay for TV time.How might corruption threaten democratiza-

tion? This is the link that has not yet been spec-ified, but this paper attempts to do so. Manyclaims have been made that corruption is badfor democratization, but most of those areimpressionistic. Case studies, especially that ofVenezuela, are suggestive. In that country, thetwo mainstream political parties that hadshared political power for decades were sweptaside by a populist candidate (and former mili-tary man). It has been argued that the funda-mental cause was public outrage over thewidely perceived corruption of the traditionalparty system (Canache, 2002; Canache & Kulis-heck, 1998). The upshot has been severe social

382 WORLD DEVELOPMENT

turmoil, an ephemeral coup d’etat, and, mostimportantly, a steady weakening of democraticguarantees (especially the separation of powers)in that country. At the same time, it is also truethat over a decade of abysmal economic perfor-mance also had weakened support for themainstream parties, so one cannot be certainif corruption played a central role in the declineof democracy in Venezuela. Moving beyondcase studies, there is limited cross-national re-search that links corruption and democracy,but the causal arrows have been reversed. Onekey study (Treisman, 2000), for example, hasshown that over the long run, democratic coun-tries become less corrupt, but the reverse hasnot been shown, namely, that highly corruptcountries experience weakened democratizationor even democratic breakdown.Perhaps the most extensive empirical study to

date on the relationship between corruptionand democratization is the work of della Porta.In a series of papers focused on Europeancountries, corruption has been shown to de-crease satisfaction with democracy (della Porta,2000; della Porta & Meny, 1996). More recentwork, looking at Latin America, has alsopointed to weakening of democracy as a resultof corruption (Gingerich, 2004). In Japan,Pharr (2000) has shown that corruption bypublic officials lowers public confidence indemocracy. And in Italy, research has shownthat voters can punish corrupt politicians,although they do not always do so (Golden &Chang, 2001, 2004).These studies, while certainly suggestive of a

link between corruption and democratic weak-ening, suffer from a number of weaknesses, asthis paper discusses below. In particular, thereare two main difficulties: First, the studies focuson the perception of corruption rather than actsof corruption themselves, thus leaving open thepossibility that even if corruption is at low lev-els, if perception is that it is high, democracycould be weakened. Second, and more impor-tant, the studies do not demonstrate the mech-anism by which corruption could impactdemocracy. This paper, by developing a cor-ruption–victimization measure at the level ofthe individual and by showing how corruptionvictimization erodes belief in the legitimacy ofthe political system, hopes to draw the linkmore tightly.The specific argument made in this paper is

that corruption weakens democracy by under-mining citizen trust in their regimes, in effect,delegitimizing them. Once that happens, citi-

zens can seek alternatives to democracy,through support for the return of military rule,or through support for populist but anti-demo-cratic figures such as Hugo Chavez of Venezu-ela. Unlike dictatorships, that can employalmost unlimited coercion to stay in power,democracies rely on popular legitimacy to stayin power, as Lipset and others have argued(Easton, 1975; Kornberg & Clarke, 1992; Lip-set, 1961). If Weyland is right, and corruptionis on the rise, one can expect that the nascentdemocracies in Latin America, and by exten-sion the democratizing world, will have an evengreater difficulty in establishing and retainingtheir right to govern. One of the major limita-tions that authoritarian regimes have in estab-lishing their own legitimacy is that more oftenthan not they operate as cleptocracies, in whichthe state is corrupt to its core, and citizensknow it. In a similar fashion, democracies,based as they are on expectations of account-ability, may weaken their own legitimacy whencorruption is rampant.Is there evidence for this linkage between cor-

ruption and the erosion of legitimacy? In fact,the classic research in the field, widely respectedfor many years, argued persuasively for pre-cisely the reverse relationship. From that pointof view, corruption serves a very constructivepurpose in developing countries by establishingfunctional clientelistic ties that bind constitu-ents to public officials, allowing relatively low-cost circumvention of dysfunctional bureaucra-cies. In that way, corruption was seen as havinga positive impact on political systems. As Hun-tington (1968, p. 69) put it most succinctly inthis classic statement: ‘‘the only thing worsethan a society with a rigid, over-centralized, dis-honest bureaucracy is one with a rigid, over-centralized honest bureaucracy.’’ The view thatcorruption helps ‘‘grease the wheels’’ of dys-functional bureaucracies is prominent not onlyin the work of Huntington and classic workssuch as V.O. Key’s Southern Politics publishedin 1949, but also in many more recent studies(Heidenheimer, Johnston, & LeVine, 1989). Inshort, according to this view, corruption is justwhat the doctor ordered for the Third World.Is this argument accurate and does corrup-

tion help ‘‘grease the wheels’’ of the publicbureaucracy, thus making it more agile, func-tional, and most importantly more responsiveto public demands? If it does, then one wouldexpect that more corruption would lead togreater accountability and a public that is moresatisfied with the political system.

MEASUREMENT AND IMPACT OF CORRUPTION VICTIMIZATION 383

More recent research, empirically based, takesthe opposite approach, one that is consistentwith international donor concerns that corrup-tion is not only unhelpful, but it can do seriousdamage to both economic growth and demo-cratic development (Rose-Ackerman, 1999).Unfortunately, there are few empirically basedstudies on the subject, and the ones that do exist,do not test corruption itself, but only percep-tions of corruption. One study on 10 Centraland Eastern European countries and Russiasupports the linkage, finding that greater per-ceived corruption relates to lower trust in demo-cratic institutions (Mishler & Rose, 2001). Amore recent study (Anderson & Tverdova,2003), based on 16 countries, including the Uni-ted States, Japan, New Zealand, and countriesin Western and Eastern Europe, concludes that‘‘corruption undermines citizens’ faith in theirgovernments.’’ In the present paper, these find-ings will be supported, but with a direct measureof corruption experienced by individuals wholive in the countries being studied rather thanwith an index of perception of the degree of cor-ruption constructed by external observers.In order to establish the linkage between cor-

ruption and the erosion of support for democ-racy, the paper dwells on the issue of themeasurement of corruption. Early on in theinternational efforts to increase transparencyto fight corruption, it became evident that inorder to know precisely where to expend re-sources and how to measure the degree of pro-gram success, it was going to be very importantto be able to measure the magnitude of corrup-tion. However, unlike programs in otherareas—for example, health, where there areuniversally accepted objective indicators like in-fant mortality rates and life expectancy—thereis not as yet any universally accepted standardfor measuring corruption. But, unless it isknown how much corruption there is, whetherit is increasing or declining, little can be doneto target efforts to reduce it.In the absence of reliable data on corruption,

it is difficult to target anti-corruption spending.More important, perhaps, it is that without reli-able and specific information on corruption,one cannot determine what impact, if any,anti-corruption efforts are having, which is,after all, ‘‘the bottom line.’’Although constrained by the limited space

available in the format dictated by these shortresearch notes, this paper will respond to thechallenge of developing better measures of cor-ruption in the following six sections:

(1) a brief examination of the widespreaduse of corruption perception measure;(2) a discussion of an alternative measure ofcorruption experience;(3) an exploration into the level of corrup-tion detected by the corruption–victimiza-tion series, using data from severalcountries drawn from the Vanderbilt Uni-versity Latin American Public Opinion Pro-ject (LAPOP);(4) an examination of the data to find outwho the victims of corruption are;(5) a discussion of the research that hypoth-esizes about the positive and negative effectsof corruption on governance, providing evi-dence that shows just how dangerous cor-ruption is for democratic consolidation,that is, that corruption matters;(6) and finally, the paper will show how cor-ruption levels vary substantially withincountries.

2. THE CHALLENGE: MEASURINGCORRUPTION 1

It is not surprising that, until recently, cor-ruption research has been largely descriptiverather than empirical. Given its sub rosa nature,corruption is inherently an extremely difficultphenomenon to measure—thus presenting alarge obstacle for those researchers attemptingto study it. In other areas of development, itis easy to have participants or outsiders listaccomplishments. For example, health workerscan report on the number of vaccines adminis-tered or healthy babies delivered. We cannotask, however, for police officers to report onthe number of bribes they take in a given weekor for customs officials to report on how muchduty goes uncollected. The need to get somehandle on the level of corruption has led, overthe years, to pursuing different approaches tosolving this problem, but each approach hassuffered from its own limitations.The early efforts were based on official police

and court records. One simply counted thenumber of arrests and convictions for corrup-tion in a given country. The main difficulty withthat approach, of course, was the spuriousnessof the measure; the more vigilant the authori-ties, the more arrests and convictions, produc-ing a corruption index that was almostcompletely independent of the corruption levelitself. For example, in highly corrupt countries,there may be virtually no enforcement, and


therefore, very little corruption is reported,while in ‘‘squeaky clean’’ countries, there maybe frequent arrests and convictions for evenminor infractions. For the most part, this ap-proach has been abandoned.In order to overcome the measurement prob-

lem inherent in using official records, two newerapproaches have been taken—each with its ownlimitations, however. The first is that carriedout by Berlin-based Transparency Interna-tional (TI) with its annual highly respectedand widely used corruption–perception index(CPI). TI is dedicated to generating publicsupport for anti-corruption programs andenhancing transparency and accountability ingovernment. In recent years, the index has beengreatly strengthened due to the use of multiplesources of data and multi-year averages, thusincreasing the reliability of the measure. 2 TheCPI remains the most widely used measure ofcorruption, akin to the Freedom House mea-sure of democracy. Most economists rely uponthe CPI when they examine the impact of cor-ruption on growth and investment.The results of the TI index for Latin America

in 2002 are shown in Figure 1. The index rangesfrom a high of 10, in which the country is con-sidered to be ‘‘highly clean’’ to a 0, for coun-tries that are considered to be ‘‘highlycorrupt.’’ The 2002 report ranks 102 countries,

0.00 2.00

Score 10 =

ParaguayBolivia

EcuadorHaiti

GuatemalaNicaraguaVenezuelaHondurasArgentina

PanamaEl Salvador

Dominican RepublicColombia

MexicoBrazilPeru

Costa RicaUruguay

Chile

coun

try

1.702.202.202.20

2.502.502.50

2.702.80

3.0

Figure 1. Transparency international corruptio

and finds 70 countries—including many of theworld’s most poverty stricken—that score lessthan 5 out of a possible 10. For comparison,it is important to note that the countries witha score of higher than 9 are Finland, Denmark,New Zealand, Iceland, Singapore, and Sweden.As seen in Figure 1, no Latin American countryscores that high, with Chile scoring the highest,7.5. Only one other country scores above 5,Uruguay (at 5.1), while all other Latin Ameri-can countries do worse, placing them amongthe bottom 70 countries out of the 102 on thelist. It is obvious from this information thatcorruption in Latin America, at least as mea-sured by the TI perception index, is a seriousproblem.As illuminating as these results may seem to

be, we need to consider their limitations. Thescores are based upon a series of surveys con-ducted by different organizations and includea measure from the Economist IntelligenceUnit, the Pricewaterhouse Cooper’s OpacityIndex, and data from Freedom House. Thesedata sources rely heavily upon the impressionsof international business people, who are rea-sonably good, albeit limited, sources of data.The great strength of these data sources is alsoa weakness—they tap into the perceptions ofindividuals engaged in international businessand thus are good at evaluating business trans-

4.00 6.00

highly clean; 0= highly corrupt

03.403.503.603.60

4.004.00

4.505.10

7.50

n–perception index for Latin America, 2002.


actions, but they are weak at tapping into thewhole range of activities pursued by the citizensof the countries being evaluated. These activi-ties are not business related and thus requirea different kind of data. Moreover, a problemwith data from such sources is that it is difficultto separate stereotyping from reality. Theseindexes do not attempt to report on how fre-quently bribery actually occurs in the countriesbeing studied but on how corrupt the respon-dents perceive the countries to be. Of course,this may be based on their own experience oron what they happened to have heard about aspecific country. Consider Paraguay—longconsidered by most experts on Latin Americato be an endemically corrupt country. It is alsoa country with very little foreign investmentand very few international business arrange-ments. According to the World Bank, forexample, net private capital inflows for Para-guay in 2000 was a negative $16 million andtotal foreign direct investment was only $82million in an economy with a GDP of $6.9 bil-lion. Foreign direct investment in Paraguayamounts to only about 1% of the GDP, a levelonly one-sixth of that of its neighbor Brazil.How many of the respondents in the surveysof business people used by the data sourceswhose perceptions were incorporated into theCPI had actually conducted business in Para-guay? The chances are, not many. Rather it islikely that the ratings of Paraguay are basedlargely upon a stereotype which may or maynot be closely linked to reality. Furthermore,while TI draws on 15 different sources for itsdata, Paraguay’s results emerge from consult-ing only three sources because so few datasources have a large enough sampling of busi-ness people to enable Paraguay to be given ascore. Finland, on the other hand, with a pop-ulation nearly identical to that of Paraguay,was rated by eight sources. Indeed, an inspec-tion of the number of sources used for LatinAmerican countries as a whole reveals that,with the exception of the largest and richest,most were rated by only a handful of sources,raising doubts about the reliability of the infor-mation.A further complication with the CPI is that it

may suffer from an ‘‘endogeneity problem.’’That is, the results of the CPI may be stronglyinfluenced by factors other than direct observa-tion by participants. Consider the case ofArgentina. In 1995, when Argentina was thedarling of the development community, havingprogressed rapidly in its privatization pro-

grams, it was scored at 5.2 on a 0-to-10 scaleby TI. By 2002, however, the Argentine econ-omy was in ruins, and its score had plummetedto 2.8. If Argentina had retained its 1995 cor-ruption score, it would have been ranked sec-ond in all of Latin America and near the topof the developing world. But by 2002, it hadfallen into the bottom half in Latin America.Was that precipitous fall in the index, reflectinga putative increase in corruption, a function ofa sudden real increase in corruption or didthose who responded to the CPI surveys beginto note a discrepancy between low economicperformance and the TI score of reasonablylow levels of corruption? If so, it may well bethat corruption levels were, in reality, not muchdifferent in 2002 than they had been in 1995 butthat respondents to the TI surveys felt the needto adjust their corruption scores to match thelevel of Argentina’s economic performance. Inother words, it is likely that the TI corruptionmeasure on Argentina declined not becausecorruption increased, but because the poor per-formance of the economy told observers thatcorruption must be higher than they thoughtit had been. On the other hand, if we assumethat the TI index for Argentina was actuallypicking up on a sudden increase in corruptionover the period 1995–2002, then this would im-ply that decelerated growth actually increasedcorruption. However, the logic of the researchfrom the World Bank and others has been justthe opposite; namely, that high levels of corrup-tion retard growth. If poor economic perfor-mance is the cause of corruption rather thanthe other way around, then anti-corruptionprograms would be irrelevant since the ‘‘fix’’for corruption would be to spur economicgrowth.In recent years, a very widely used alternative

to the TI measure has emerged from the WorldBank Institute, which has been releasing a ser-ies of ‘‘governance indicators’’ that include ameasure of ‘‘control of corruption.’’ 3 Theseindicators are widely cited as the best availablemeasures of national governance. The resultsfor Latin America are shown in Figure 2. Dowe learn something new from these data thatwe did not know from the TI measure? It isobvious from comparison with the TI resultsshown in Figure 1 that the two scores corre-spond very closely. Indeed, the rank–order cor-relation (Spearman’s rho) between the twomeasures is an almost perfect 0.96. Rarely dothe social sciences produce such consistent re-sults in indexes based on perception, and we

coun

try

2.00 4.00 6.00 8.00

EcuadorParaguay

HaitiNicaragua

BoliviaGuatemala

HondurasVenezuela

PanamaColombiaArgentina

El SalvadorMexico

Dominican RepublicPeru

BrazilUruguay

Costa RicaChile

1.261.32

2.082.392.582.77

3.083.33

4.154.474.534.65

4.915.03

5.535.66

7.748.05

8.99

Figure 2. World Bank Institute Corruption-Control Index, 2001/2002.


are left with the impression of great precision ofmeasurement in these two apparently indepen-dent sources. Careful examination of the twodata sources reveals, however, that they relyon most of the very same sources from whichthey derive their indexes, explaining why theyarrive at similar conclusions. The problemsnoted with the CPI, then, are also to be foundin the World Bank Institute ‘‘control of corrup-tion’’ measure.The lesson from this examination of TI’s CPI

and the World Bank Institute’s Corruption-Control Index is that one ought to interprettheir results with care, while recognizing thatthey are probably the best measures currentlyavailable for a worldwide ranking. If a generalnotion of the level of corruption in a country isneeded, these measures are, at least for the mo-ment, the best source. They allow internationalagencies to target the countries with the highestlevels of corruption and, of course, can providean important tool for investors who want toknow what they are likely to confront if they in-vest in a specific country. Unfortunately, how-ever, the measures are of limited utility formuch of what anti-corruption programs aretrying to accomplish in transparency and anti-corruption programs because they provide onlya national aggregate measure, with no break-downs by type or locus of corruption. These mea-sures do not reveal, for any of the countries,those who are most likely to be affected by cor-ruption—making it very difficult to design acost-effective anti-corruption strategy. Simi-

larly, an aggregate measure for a country doesnot allow development practitioners to measurethe impact of specific anti-corruption pro-grams. Equally frustrating, the TI and WBImeasures are limited in that they are presentedas relative scores (i.e., ranks and percentiles). 4

Consider the situation where the worldwidetrend is toward improvement, and an individ-ual country has done better from one year tothe next; if others have also improved, theimpression will be given that the country inquestion has stagnated. Moreover, the aggre-gate, nationwide measures make it impossibleto know if corruption does in fact erode confi-dence in the political system, as the WorldBank argues, or if it has the positive impactsthat Huntington claimed for it. Moreover, na-tional aggregate measures of corruption donot allow for an examination of regions orgroupings within a country. It is risky to as-sume that corruption is uniform throughout acountry, and the data presented in this paperwill show that it is in fact not. If anti-corrup-tion programs want to target certain geo-graphic hot spots of corruption, for example,then national aggregate measures like the CPIare not helpful. If such programs want to targetcertain groups demographically, socioeconomi-cally, or occupationally, the CPI cannot help.How to solve the problem of finding a mea-

sure of corruption that is both valid and sub-national? A new approach has been taken byHarvard’s Susan Pharr, who constructs anindex based on newspaper reports of corruption


(Pharr, 2000). One advantage of this method isthat the newspaper articles report where thecorruption occurred and who was involved.Unfortunately, this measure alone cannot over-come the problem of validity. The accusationsreported in the articles may be entirely theinvention of the newspaper staff, whose motiva-tions for making the accusation may vary froma desire to increase circulation to an effort toweaken one party or candidate and strengthenanother. Indeed, the newspaper may publishstories on alleged acts of corruption becausethe paper is ‘‘on the take’’ from some politicalor business group that wants to make its oppo-nents look bad. The classic example of thiscomes from Mexico, where Excelsior, for dec-ades regarded as the ‘‘newspaper of record’’in that country, was subsequently found tohave been on the payroll of the PRI (the rulingparty) for decades.

3. MEASURING CORRUPTIONVICTIMIZATION

Recently, a more promising method has beendeveloped that largely overcomes both of thelimitations of the previously mentioned meth-ods. The new approach has been inspired bycrime-victimization surveys, which have be-come the mainstay of sociological investigationinto crime. Criminologists have long recognizedthat official reports of crime are highly unreli-able because of the likelihood that the datahave been manipulated. Police chiefs who wantnew patrol cars from their local governmentshave a strong incentive to help justify the re-quest by providing ‘‘evidence’’ that a crimewave has hit the town. To develop this evi-dence, the police chief may tell his/her officersto become especially aggressive when enforcingthe law, or it may be that the figures themselveshave been cooked. Alternatively, politicianswho want to take credit for successes in fightingcrime have the incentive to make it appear thatcrime is on the wane, and salary raises for thepolice force might subtly be made contingentupon less aggressive policing.In order to overcome these intractable prob-

lems, criminologists have increasingly come torely upon victimization surveys, which arewidely regarded as providing a more accuratetally of crime rates. 5 That approach is theinspiration for the current movement to usesurveys of national populations to estimate cor-ruption levels. Internationally, this approach

has been spearheaded by the UN Center forInternational Crime Prevention. Implementedin 1987, the International Crime Victim Survey(ICVS) now includes 55 countries, with samplesof between 1,000 and 2,000 respondents percountry. In 1996, for the first time, a questionwas included in the surveys on bribe victim-ization.The UN approach represented a good start,

but a broader series of questions is to be pre-ferred since anti-corruption projects need muchmore detailed information about the natureand level of corruption than a single questioncould provide. That is precisely the reason be-hind the approach that the LAPOP beganapplying in 1996 and the World Bank beganusing in 1998. These efforts, which may alsoincidentally ask about perception of corruption,focus on actual citizen experience (i.e., victim-ization) with public sector corruption. A reviewof the approach used by the World Bank hasbeen presented as a conference paper (Kauf-mann, 1998). The World Bank studies havebeen carried out in Nicaragua, Honduras,and, more recently, Ecuador, and, by 2003,have been carried out in many other countriesaround the world. 6 The World Bank surveysgenerally include samples of the national pub-lic, but also include specialized samples of ‘‘ser-vice users’’ and of government officials. In thatway, they are able to get at corruption from avariety of angles. Moreover, the World Bankhas made efforts to calculate the cost of corrup-tion by asking respondents to indicate the valueof the bribe solicited or paid. The World Banksurveys, however, do not include questions onattitudes toward democracy, thus making itimpossible to see how corruption might affectdemocracy.The effort has been to measure corruption

and its impact. Corruption is measured by di-rect personal experience with it. In the courseof a public opinion survey of democratic valuesand behaviors, which is generally applied to aprobability sample (i.e., ‘‘random sample’’) ofthe country being studied, a module of ques-tions on corruption experience has been devel-oped and refined since its first application in1996. 7 In this approach, respondents are askeda series of questions recording their experiencewith corruption over the year immediatelyprior to the survey. A study of longer timeframes can be attempted, but not only is thereliability of memory over long periods prob-lematic, it is also especially important tolimit the time frame so that the impact of


anti-corruption projects and programs can bepicked up in subsequent measurements. Thesesurveys, then, include both the key ‘‘indepen-dent variable’’ (i.e., corruption victimization)and the key ‘‘dependent variable’’ (i.e., trustin the system of government) so that the impactof the former on the latter can be measured,controlling for other factors (e.g., respondentincome, education, gender, region, party affilia-tion, etc.).The forms of corruption measured were

decided upon based on focus groups and wereones that were found to be most commonlyexperienced in Latin America. The items havevaried from questionnaire to questionnaire,but the following questions are common to anumber of them:

—being stopped by a police officer for atrumped-up infraction of the law;—being asked to pay a bribe 8 to a policeofficer;—observing a bribe being paid to a policeofficer;—observing a bribe being paid to a publicofficial;—being asked to pay a bribe to a publicofficial;—being asked to pay an illegal fee to expeditea transaction at the municipal government;—being asked to pay a bribe at work;—being asked to pay a bribe in the courtsystem;—paying bribes to a public utility to avoidpaying the full bill;—being asked to pay a bribe in the schoolsystem; and—being asked to pay a bribe for services inthe public health/hospital system.The reader will note that most of the items in

the series focus on corruption directly experi-enced by the respondents but that some of themfocus on the respondents’ observations whenbribes were being paid by others. We have la-beled this as ‘‘indirect’’ or ‘‘vicarious’’ corrup-tion. In each country where the survey hasbeen applied, it was found that direct, personalexperience with corruption is less frequent thanindirect or vicarious experiences. The reasonfor this could be that individuals may be und-erreporting their own experience with corrup-tion out of fear that reporting it could getthem into some sort of legal trouble. Focus-group research has shown that an underreport-ing problem exists but is surprisingly limited.On the other hand, saying that one observedsomeone else paying a bribe might be a way

of actually reporting one’s own bribery experi-ence with the advantage of deniability. Yet, itmay be possible that these reports of observedacts of corruption could be inflated—reportingmore on perception than reality. In analyzingthe data in this paper, the focus is almost en-tirely on the direct forms of corruption.A second observation about the series of

questions relates to corruption in schools andin public health. Focus-group research demon-strated that users of these services tend to bevery well aware of the rules of the game. If,for example, primary education is supposed tobe free, many citizens consider it corruptbehavior if fees are demanded by the teachers,even if the purpose of the fee is to purchasebooks or supplies that ought to come fromthe state. For that reason, the original question-naire series referred to ‘‘improper payments’’ inthe schools or health clinics. However, somecritics of these questions suggested that thesepayments were not really signs of corruptionbut merely a reflection of the poor state of pub-lic financing of human services. In order to fo-cus directly on corruption itself, these itemshave been altered in the most recent adminis-trations of the questionnaire, and the term ‘‘im-proper payments’’ was dropped and ‘‘bribes’’substituted. This change resulted in a small de-cline in reported acts of corruption in the healthservice and a somewhat larger decline of re-ported acts of corruption in the school system.Yet, even with these adjustments the levels re-main distressingly high, as will be shown below.The analysis of the corruption–victimization

data has been carried out in two ways. First,when the particular form of corruption is thefocus (in order, perhaps, to develop an anti-cor-ruption campaign targeted to each specific formof corruption), individual items are chosen tobe analyzed so that the particular form of cor-ruption can be highlighted. The second ap-proach is to form an overall scale ofcorruption so that the total experiences can beexamined and linked to their impact on individ-ual variables, such as political legitimacy. Whenformed as a scale, the items have been found tobe reliable (Cronbach Alphas of around 0.75,depending on the country).In approach to corruption victimization pre-

sented here, national samples have been used.The World Bank, however, which has also car-ried out national samples of corruption victim-ization, has included specialized samples ofindividuals in various sectors, such as the busi-ness world or various sectors of the state. The


advantage of the World Bank approach is thatit allows an even finer-grained targeting of anti-corruption efforts; the disadvantage, of course,is cost.

(a) Limitations of the corruption–victimizationmeasure

No system of measuring attitudes or behav-iors is without limitations, and the corrup-tion–victimization measure suffers from twoof them, which will be pointed out here.

—Are all survey respondents who reporthaving paid a bribe really ‘‘victims?’’ Per-haps some bribe payers are not victims atall but in fact are willing participants inthe transaction, deliberately seeking to cir-cumvent rules and regulations in order toadvance their own objectives. Such individu-als, then, may not be ‘‘victims’’ but willing‘‘clients’’ in a patron/client relationship withthe bribe taker. On the other hand, if thesystem were both fair and efficient, therewould be little reason for citizens to becomeclients. If the police did not regularly stopmotorists in Latin America who have donenothing wrong, then there would be no rea-son to offer to pay a bribe. Similarly, ininstances when the motorist actually com-mits a traffic infraction and seeks to avoida ticket and the corresponding fine, an hon-est police officer would not demand oraccept a bribe. If the public sector were bothfair and efficient, individuals could easilyobtain the service they are seeking (e.g., apermit from the local government) and therewould be no need to pay a bribe. Addition-ally, if one sought to avoid paying a large,legally established fee for a service and thepublic officials were honest, no bribe wouldbe accepted. In short, while it is always pos-sible to view corruption ‘‘victims’’ as corrup-tion ‘‘partners,’’ bribery would be largelyfutile if not for the inefficiency and complic-ity of a corrupt public sector.—The survey approach is flawed because itmeasures only low-level corruption, but isincapable of measuring much high-level cor-ruption. This critique is valid, but only to alimited degree. It is true that surveys of thepublic do not measure—nor do they attemptto measure—high-level corruption (e.g.,bribes paid to ministers or legislators). But,it would be a serious error to assume thatlow-level corruption is not strongly corre-lated with the degree of corruption at the

top. It is simply not reasonable to imaginethat a political system that is very corruptat the level of day-to-day transactions(police, local government, schools, banks,etc.) would be squeaky clean at the top. Sim-ilarly, it is difficult to imagine the reverse sit-uation with a highly corrupt senioradministration combined with an absenceof corruption down on the street. Since thereseems to be no reasonable way to directlymeasure high-level corruption (rather, onlythe perception of that corruption), the vic-timization approach used by the LatinAmerican Public Opinion may be the bestinstrument currently available.

(b) Perception and corruption victimization:Are they linked?

The corruption–victimization surveys alwaysalso ask at least one question regarding therespondent’s perception of corruption andtherefore provide for a test of the putative link-age between corruption perception (the basis ofthe TI CPI) and the fact of corruption as expe-rienced by residents of the countries included inthe TI series. Are the two linked? Yes and no.By examining Figure 3, one can see that as cor-ruption victimization increases, there is a signif-icant increase in the perception of corruptionon the part of individuals in the countries in-cluded in this series. Yet, it is important notto overstate this finding because on a 0–100scale, for all four countries in this sample, allaverages are in the 60–70 range for respondentswho have not had any direct personal corrup-tion experiences within the 12 months prior totheir being surveyed. In other words, eventhough respondents to these surveys have notbeen personally affected by corruption, at leastin the year prior to the survey, they nonethelessgenerally view corruption as being extensive intheir countries. In contrast (with the exceptionof Paraguay), even among those who had beenvictimized two or more times in the past year,the increase in the level of perception of corrup-tion is quite small, generally fewer than 10points on a 100-point scale. More troubling stillfor those who wish to rely on corruption per-ception is the case of El Salvador, where thosewho had experienced the highest frequency ofcorruption had a somewhat lower perceptionof the degree of corruption than those whohad been victimized only one time.These findings caution us to be very careful

when estimating corruption on the basis of

All sig. < .05

Number of direct personal experiences with corruption

Two or moreOneNone

Mea

n pe

rcep

tion

of c

orru

ptio

n

100

90

80

70

60

50

40

30

20

10

0

Paraguay

El Salvador

Nicaragua

Bolivia

Figure 3. Impact of corruption victimization on perception of corruption.


perception rather than experience since the twomay not have a close fit. Moreover, in countriesin which a major effort to reduce corruption hasbeen successful, it may well be that the cam-paign to do so heightens awareness of corruptpractices and therefore might produce an in-crease in the perception of corruption preciselyat the time when actual corruption is declining.The point becomes even more obvious in Fig-ure 4, where a direct effort was made to seehow a perception-based measure tracks withactual corruption experience. As is evidentfrom the results, although corruption variessubstantially from country to country in thesix-country sample, perception has only a weakrelationship to victimization levels. If one wereto use perception as the measure, for example,one would conclude that there is more corrup-tion in El Salvador than in Bolivia, when in factit is very clearly the other way around. The ac-tual TI ranking for these countries does cor-rectly show that Bolivia is worse than ElSalvador, but while Bolivia scores 2.2 on the1–10 scale (see Figure 4), El Salvador scoresonly slightly better at 3.4. The corruption–vic-timization scale in Figure 4 shows that Boliviahas a level of corruption more than four timesthat of El Salvador.However, these comparisons do not make the

indexes generated by TI and the World Bankobsolete, as they are the only ones currently

available for many countries. When using thesebroad measures for specific developmental pur-poses, however, considerable care must be ta-ken or these measures of perception couldbecome reified.

(c) Corruption victimization in sixLatin American countries

(i) How frequent is it?The surveys of Latin America reveal that cor-

ruption is widespread and emerging in manydifferent forms, some more frequently experi-enced than others. This paper, which can onlyhighlight a small portion of the data becauseof space limitations, focuses on direct corrup-tion victimization rather than observed corrup-tion.One very common way in which corruption

affects Latin Americans is through the very offi-cials who should be involved in controlling it—namely, the police force. The results shown inFigure 5 reveal that over one in four Bolivianshave been stopped annually by a police officerand asked to pay a bribe. In contrast, this hap-pens to only 3% of the population of El Salva-dor. Respondents in these countries were alsoasked about being asked to pay a bribe by apublic official, and the comparative results areshown in Figure 6. Once again, consider the ex-tremes—El Salvador has a 4% corruption–vic-

Spearman's rho = .098, sig. < .001

BoliviaParaguay

Ecuador

Nicaragua

Honduras

El Salvador

Inde

x va

lue

100

80

60

40

20

0

How commonare bribes?

Index of corruptionvictimization

27

2219

15

9

64

7775

696971

5

Figure 4. Do national level perceptions of corruptioin track with corruption victimization?

Sig. < .001

BoliviaEcuador

NicaraguaHonduras

El Salvador

% s

olici

ted

brib

e

30%

20%

10%

0%

27

14

76

3

Figure 5. Police officer solicited a bribe.


timization rate and Bolivia has a 28% rate. Butalso consider the consistency with which thecountries are ordered; the only difference be-tween bribes solicited by police and bribes solic-ited by the more generic ‘‘public employees’’ isthat in the later instance, data are availablefrom Paraguay, which is found to have corrup-tion levels somewhat lower than Bolivia buthigher than Ecuador. The overall pattern that

emerges here is the large gap in corruption vic-timization between Central America and SouthAmerica, with the former being far lower thanthe latter.The survey also collected information on in-

stances of corruption in specific settings. Forexample, respondents were asked questionsabout being asked to pay a bribe at work (Fig-ure 7). What is striking about these results is

BoliviaParaguay

EcuadorNicaragua

HondurasEl Salvador

% s

olic

ited

30%

20%

10%

0%

26

21

19

8

64

Figure 6. Public employee solicited a bribe.

Bribe solicited at work

BoliviaEcuador

NicaraguaEl Salvador

Honduras

% v

ictim

s

14%

12%

10%

8%

6%

4%

2%

0%

13

11

6

43

Figure 7. Bribe solicited at work.


how much lower they are than those of public-sector corruption (compare with figures above).It might be assumed that these low levels are afunction of the fact that significant percentagesof respondents are not employed (i.e., students,

retired, unemployed) and thus did not have theopportunity to be subjected to work-placebribery. 9 If, however, the analysis is limitedto employed respondents, corruption levels doincrease, but not by much, leading to the con-

Sig. < .001

Paraguay

Nicaragua

Bolivia

Honduras

El Salvador

% v

ictim

s

30%

20%

10%

0%

22

18

13

7

Figure 8. Bribe solicited in the courts.


clusion that corruption in the workplace is sim-ply not as common as it is in interactions withthe public sector (as shown in Figures 5 and 6).Bribery in the courts varies widely fromcountry to country (as shown in Figure 8). Inplaces like El Salvador, it seems to be veryuncommon, whereas in Paraguay, it is veryfrequent. 10

Recently, the World Bank Institute has beenfocusing on bribery at the level of users of ser-vices. In the questions analyzed thus far, all cit-izens are, at least in some fashion, users ofpolice services. But not all (or even most) citi-zens make use of the court system. In orderto focus on users of public services, the LAPOPhas recently been adding questions to itssurvey.The respondents are first asked if they use the

service, and then they are asked if they havehad to pay a bribe to use that service. Questionshave been asked regarding bribery when carry-ing out a transaction at the local municipalgovernment, using the court system, sendingchildren to public schools, and accessing a pub-lic health service (e.g., clinics and hospitals).The results for Ecuador are shown in Figure 9and are very disturbing. It should be kept inmind that Ecuador has already shown itself tobe a high-corruption country, at least whencompared to Central America, based on theLAPOP data and also when considered from

the World Bank Institute perspective (Ecuadorranks last in Latin America on corruption con-trol in the World Bank data). The survey re-sults for Ecuador show that one-fifth of itsadult citizens who carry out a transaction withthe local government pay bribes when doing so,over one-quarter of the users of courts havebeen asked to pay bribes, and over one-thirdwho use public health services and two-fifthsof users of public education facilities pay bribesthere. 11 These results are disheartening as theyindicate that those who use a public service inLatin America need to expect to be asked topay a bribe.

(d) Who are the victims? Variation withincountries

Victimization by corruption is not uniformacross the populations of the countries in whichit occurs. On the whole, males are much morelikely to be victimized by corruption than fe-males. As shown in Figure 10, for the generalitem on having a public employee solicit abribe, in every country except Honduras, malesare significantly more likely to be corruptionvictims than females. In El Salvador, althoughthe overall level of corruption is relativelylow, the gap is proportionately very wide. How-ever, this is largely because males make greateruse of many public services than females. As

All sig. except Honduras < .05

BoliviaParaguay

EcuadorNicaragua

HondurasEl Salvador

Mea

n Pu

blic

em

ploy

ee b

ribe

40%

30%

20%

10%

0%

Gender

Male

Female

21

1615

65

2

30

27

23

10

66

Figure 10. Impact of corruption on gender: six countries, public employee solicited bribe.

Sig. < .001

School

Health service

Courts

Municipality

% v

ictim

ized

50%

40%

30%

20%

10%

41

35

27

20

Figure 9. Bribes solicited from users of services: Ecuador.


shown in Figure 11, in Ecuador, where the sur-vey data make it possible to focus on the subsetof service users, the gender gap is much nar-rower and males are not significantly morelikely to be victimized than females in Ecuador(among users of the municipality, courts, andhealth services) but only somewhat more likelyto be victimized by the school system.Those seeking bribes know where to find the

‘‘deep pockets.’’ Figure 12 shows that as in-

come increases, the probability of being a cor-ruption victim increases sharply. Those withhigher incomes are two or three times morelikely to be victimized by corruption than thosewith low incomes. 12 Of course, those withhigher incomes invariably have more businessto transact, more permits and permissions torequest, and as a result, come into more fre-quent contact with the public bureaucracy. In-deed, they may be more likely to be seeking

Only school corruption sig.SchoolHealthCourtsMunicipal

% v

ictim

s

50%

40%

30%

20%

10%

Gender

Male

Female

39

32

24

18

44

37

30

21

Figure 11. Impact of corruption on gender: service users in Ecuador.

Sig. < .001Income

% v

ictim

s

50%

40%

30%

20%

10%

0%

COUNTRY

Paraguay

Ecuador

Figure 12. Respondent income and public employee solicitation of bribes.


ways to pay bribes in order to acquire a givenservice. At the highest income levels, victimiza-tion seems to taper off, at least in Ecuador andParaguay, perhaps because the individuals in-volved are wealthy enough to be able to usepolitical influence to achieve what they needout of the system and do not have to confrontoutright bribery. It may also be that at thehighest level of income in the survey, respon-dents have more to hide, and therefore they re-duce their reporting of corruption. Since the

surveys do not measure the dollar value ofbribes, it may be that while the rich pay bribesmore frequently than the poor, that the relativecost to them is lower than it is for the poor,making bribery regressive. This has been a find-ing of the World Bank.There is also a life-cycle effect that determines

who the corruption victims are likely to be.Typically, the youngest respondents in the sur-veys are somewhat less likely to be corruptionvictims. Once they move out of their parents’


homes and set up their own families, however,they have greater contact with the bureaucracy,and corruption victimization increases. On theother hand, once Latin Americans age and be-gin to withdraw from active engagement withthe larger society, corruption victimizationfalls. In El Salvador (Figure 13), corruption vic-timization peaks abruptly in the 21–30 year agegroup, and then falls off sharply. In Bolivia

Sig. < .001

3121-3018-20

Mea

n ex

perie

nce

with

cor

rupt

ion

7

6

5

4

3

2

Figure 13. Personal experience with

Sig. < .001

36-4526-3518-25

% v

ictim

ized

28%

26%

24%

22%

20%

18%

16%

14%

Figure 14. Corruption victim

(Figure 14), where there are much higher over-all levels of corruption than El Salvador, thepeak is also reached in the 20s but continueshigh until middle age, when it then fallssharply.Although there are many other demographic

and socioeconomic patterns that corruptionvictimization follows—a number of which arefound in the various country-focused reports

Age51-9541-50-40

National average

corruption and age: El Salvador.

Age66+56-6546-55

ization and age: Bolivia.


of the LAPOP—only one remaining key findingwill be covered in this brief overview. Urbani-zation greatly increases exposure to corruption.In Bolivia, corruption is much higher in urbanareas than it is anywhere else in the country(Figure 15). In Honduras, the same pattern isfound, but in this case, the results comparethree different measures of corruption: hearingor observing corruption, being a victim, and a

Sig. < .001Urban 2,000-20,000

Urban >20,000

Cor

rupt

ion

inde

x

30%

28%

26%

24%

22%

20%

18%

16%

Figure 15. Corruption victimizat

Sig. < .001Location of

San Pedro Sula

Tegucigalpa

Tegucigalpa

% v

ictim

ized

40%

30%

20%

10%

0%

Figure 16. Corruption victimizatio

combined index. All three measures show thesame general pattern (Figure 16).The findings from Honduras in the figure

provide clear guidance to those who wish toexamine sub-national variation in corruption.Even though both Tegucigalpa and San PedroSula are the country’s only large urban areas,the corruption victimization in the Tegucigalpaarea, the seat of the central government, is far

Dispersed rural

Rural 500-1,999

ion and urbanization: Bolivia.

interview

Rural area

Municipal capital

Department capital

Heard/saw

Total index

Personal victim

n and urbanization: Honduras.


higher than that found in San Pedro Sula. Theappropriate conclusion from this finding is thatit is not urbanization per se but the presence ofgovernment that has had the major impact oncorruption levels. But, it is important not tolose sight of other factors, including regionaltraditions that influence the level of corruption.It has already been shown, for example, thatcorruption victimization varies by age and in-come, and other reports in the series haveshown that it varies by wealth as well. Whathappens when all of these factors are controlledfor, and corruption levels are examined net ofthem? Figure 17, which shows graphical varia-tion in corruption in Nicaragua, reveals thatthere is a hot spot that includes the country’scapital, Managua. But it is also clear that regio-nal traditions play a role; otherwise, it would beimpossible to explain the very low level of cor-ruption victimization in the department ofLeon, which borders on Managua, and the highlevel of corruption in the department of Chi-nandega on its other border. It is also clear thatregional traditions matter in Bolivia, where thedepartment of Oruro has higher corruption lev-els than the national capital, La Paz. The re-sults shown in Figure 18 are presented alongwith their ‘‘confidence intervals,’’ which is

Madriz

Chinandega

León

Managua

Nueva Segovia

Boaco

Carazo

Chonta

Estelí

Granada

Jinotega

Masaya

Matagalpa

Rivas

Figure 17. Corruption experience in Nicaragua: residua

really the margin of error produced by the sam-ple. This means that the average corruptionscore shown for each department, representedby the square box, is surrounded by a bandabove and below the point that represents therange in which, 95% of the time, the actualaverage would be found if the survey had inter-viewed the entire population. Confidence inter-vals help distinguish differences amongdepartments that are merely a function of thesample design and those that are true differ-ences. Thus, as Figure 18 demonstrates, forexample, Oruro and La Paz have significantlymore corruption than El Beni, but Coc-habamba’s confidence interval overlaps thatof La Paz, and therefore, the two should be as-sumed to be insignificantly different from eachother.An overall pattern emerges in these findings;

those who use the public sector more frequentlyare more likely to be victimized by it. Thisseems to be a trivial truism, until one pausesto think about the impact this has on the polit-ical system—the more Latin Americans usetheir government the more likely they are tobe victimized by it. The implications for erod-ing public confidence are obvious and will beshown empirically below.

Rió San Juan

RAAS

RAAN

les

ls controlled for age, income, education, and wealth.

300298294289400397297402295N =El Beni

Chuquisaca

TarijaPando

Santa Cruz

Cochabamba

Potosí

La Paz

Oruro

% v

ictim

ized

40%

30%

20%

10%

0%

Figure 18. Public employee solicited bribe: confidence intervals by department in Bolivia.


4. CORRUPTION ERODES POLITICALLEGITIMACY

One of the most frequently heard commentsis that corruption erodes trust and confidencein the political system. Typical of those who de-cry the negative effects of corruption, the WorldBank frequently argues the thesis, ‘‘Corruptionviolates the public trust and corrodes socialcapital. . . Unchecked, the creeping accumula-tion of seemingly minor infractions can slowlyerode political legitimacy.’’ (World Bank,1997, pp. 102–104). Unfortunately, althoughthe Bank presented substantial evidence thatcorruption negatively affects the economy, itprovided no support for the claims that minorcorruption (or even major corruption) erodespolitical legitimacy. While the Bank has pre-sented a great deal of evidence on the level ofthe independent variable (corruption), it haspresented no corresponding evidence on thedependent variable (political legitimacy). Themore recent World Bank Institute studies, asreported above, use a multi-index measure ofgovernance, including perceptions of corrup-tion worldwide, and have found that per capitaincomes are lower and infant mortality andadult illiteracy are higher when governance ispoor (Kaufmann, Kraay, & Zoido-Lobaton,1999). Yet once again, there is no effort to linkcorruption on the one hand and political legit-imacy on the other.

Corruption surveys, though embryonic, ap-pear to be the most promising of the effortsundertaken to date to test the corruption/ero-sion of the legitimacy thesis. As discussed inthis study, these surveys obtain corruptionexperience data at the level of the individualwhile simultaneously obtaining informationfrom those same individuals about their beliefin the legitimacy of their government. The ana-lytical task, then, becomes searching for theconnections between corruption experienceand legitimacy beliefs after appropriate controlvariables are introduced.Latin America, where currently there is

extensive attention focused on corruption, is agood place to test the hypothesized relationshipbetween corruption and legitimacy for two rea-sons (Tulchin & Espach, 2000). First, as shownat the outset of this paper, this is a region of theworld frequently found to have high levels ofcorruption. Second, Latin America has longhad problems of political stability, suffering aseemingly endless succession of coups throughmuch of its history (Seligson & Carrion,2002). If, as Easton (1975) and also Lipset(1994) and Lipset and Lenz (2000) have argued,legitimacy is a fundamental requisite for demo-cratic stability, then it is plausible that legiti-macy is questionable in many Latin Americancountries. From an empirical point of view,considerable evidence exists to show that legit-imacy levels remain low in many countries in


the region despite 10 or more years of demo-cratic rule (Seligson, 2000).Legitimacy is measured in the surveys by a

scale of diffuse support attempting to tap intoconfidence in the key institutions of govern-ment (Muller, Jukam, & Seligson, 1982; Selig-son, 1983). The scale is based on five items,each scored on a metric of 1–7. The items, ini-tially developed in studies of Germany and theUnited States and refined in several studies ofLatin America, sought to tap into generalizedsupport for the basic institutions of governmentrather than support for the incumbent govern-ment (Muller, Seligson, & Turan, 1987).Respondents were asked (1) ‘‘To what extentdo the courts guarantee a fair trial?’’ (2)‘‘How much respect do you have for the polit-ical institutions of the country?’’ (3) ‘‘Howmuch pride do you feel living under the politi-cal system of the country?’’ (4) ‘‘How muchsupport do you have for the political systemof the country?’’ and (5) ‘‘How much trust doyou have in the police?’’ In order to make themetric consistent with the range of the corrup-tion-experience measure, the items weresummed into an overall scale and transformedinto a 0–100 basis. The overall scale was reli-able for each country as well as for the pooleddata (Pooled standardized item Cronbach’sAlpha = 0.78; mean interitem r = 0.37).The first task is to determine if corruption has

a negative or positive impact on legitimacy.When a citizen pays a bribe to either receive apublic service or to avoid sanctions from an ac-cused violation of law, two reactions couldemerge depending upon how the bribery is per-ceived. On the one hand, the bribe could beviewed as a ‘‘user fee,’’ much like the fee onewould willingly pay to use a toll road or a camp-ground. Those who pay such fees might view theassessment and payment as an entirely legiti-mate transaction, implying no negative evalua-tion of the political system. Indeed, anindividual who pays a ‘‘processing fee’’ in orderto facilitate the granting of a driver’s license, forexample, might be pleased with a system that al-lows the granting of such licenses even when therequisite requirements (vision test, driving skillstest, etc.) have not been met. Or the individualmay feel that the salaries paid to public officialsare properly kept low so that overall taxes re-main low, but those who use the service oughtto legitimately pay these user fees so as to sup-plement the salaries of public officials.Those asked to pay bribes may have an en-

tirely different reaction to the experience, view-

ing the bribe not as an appropriate user fee, butas what economists call a ‘‘DUP,’’ a directlyunproductive profit-seeking activity, otherwiseknown as rent seeking. When a municipal clerkasks for a payment above and beyond the offi-cially established fee to process a birth certifi-cate, the payment represents a surplus valueabove that of the established price, and hencecan be considered rent-seeking behavior. Rentseeking is possible only because those demand-ing the rent (in the form of a bribe) have beengiven state license (unofficially) to do so. Wecan predict, therefore, that individuals whoview such fees as rent seeking are likely to formnegative views about the state.The empirical results to test these two possi-

ble reactions are presented in detail in a study(Seligson, 2002) in which several countries fromthe data set are examined. For the purposes ofthis brief summary, the results of Nicaragua,which parallel those from the other countries,can be examined. The overall legitimacy scaledeclines steadily as corruption victimization in-creases, declining from above the nationalmean for those who had not been victimizedby corruption to one standard deviation belowthe national mean (Figure 19). One especiallyrelevant item from the legitimacy scale, the be-lief that people can get a fair trial, clearly dem-onstrates that corruption does erode confidencein the system (Figure 20).These results show that those who experience

corruption are less likely to believe in thelegitimacy of their political system, thus con-firming the World Bank speculation and refut-ing the Huntington hypothesis. Research withthese data sets have found that corruption vic-tims are also less likely to exhibit high levels ofinterpersonal trust, a possible important social–psychological contributor to belief in legiti-macy. Interpersonal trust has been linked tosuccessful efforts to build civil society and, asPutnam (1993) terms it, to ‘‘make democracywork.’’These findings demonstrate that there are

real costs to corruption overlooked by the func-tionalists who argued for the positive benefitsof it and by those who believe the clientelisticties based on corruption can breed political sta-bility. They do not, of course, demonstrate thatgovernment operates less efficiently as a resultof widespread corruption, for to do so wouldinvolve a study beyond the scope of publicopinion and take us back to the economic liter-ature. Some of the survey data do, however,allow us to determine respondent beliefs about

Sig. < .001Corruption victimization scale

76543210

Legi

timac

y sc

ale

50

40

30

20

10

0

Mean legitimacy

1Std. Dev.

Figure 19. The impact of corruption victimization of legitimacy in Nicaragua.

Sig. < .001Total personal experience with corruption

100.0075.0050.0025.000.00

Mea

n fa

ir tri

al

3.4

3.2

3.0

2.8

2.6

2.4

2.2

2.0

Figure 20. Belief in ability to get a fair trial in Nicaragua and total personal experience with corruption.


the efficacy of corruption and then to determinethe impact of that belief on system support.In the Nicaragua survey, we asked a straight-

forward operationalization of the functionalistargument: ‘‘Do you believe that the paymentof bribes facilitates getting things done withthe bureaucracy?’’ Among the 85% of the sam-ple who responded to this question, 60% agreedwith it, indicating fairly strong support withinNicaragua for the belief that bribery works.Moreover, belief in the functionality of corrup-

tion is positively associated with our index ofvictimization by corruption (r = 0.21; sig. <0.001), suggesting that those who have directexperience with corruption are more likely tobelieve that it gets them what they need. Yet,belief in the efficacy of corruption does nottranslate into a positive view of the politicalsystem—quite the contrary, in fact. Those whoagree that corruption helps get things done withthe bureaucracy are significantly less likely tobelieve in the legitimacy of the political system.


These results show that corruption erodes sup-port for the system even among those who rec-ognize that it may have some utility inovercoming bureaucratic barriers.

5. CONCLUSIONS

In order for political systems to function rea-sonably well, actions taken by leaders need tobe viewed as legitimate. If not, the ‘‘degrees offreedom’’ with which decision makers have tooperate are reduced considerably. Immobilismis a potential outcome of political systems inwhich the public does not believe in the legiti-macy of the system. In extreme cases, whichhave frequently occurred in Latin America,governments are swept aside by popular pro-tests or military coups that are supported bymuch of the public.This paper has emphasized the importance of

measuring corruption experience in order to be

able to show who its victims are and to demon-strate that those who saw corruption as build-ing ties between rulers and the ruled werewrong. Corruption does indeed erode the legit-imacy of democracies. It follows that if corrup-tion increases, as Wayland argued, at the outsetof this paper, erosion of support should also beoccurring. It is also shown in this paper thatcorruption lowers interpersonal trust, presum-ably negatively affecting civil society relations.On the positive side, there is some recent evi-dence that over the very long run democraticcountries become less corrupt, but the observedeffect took place over half a century, more yearsthan many fragile democracies may have tospare (Treisman, 2000). It would be the ulti-mate irony that an artifact of the rise of democ-racy itself may help contribute to its ownweakening in the developing world. That is allthe more reason to be concerned about theproblem of corruption and to find ways ofreducing its prevalence.

NOTES

1. This section draws on Seligson (2002).

2. These efforts are explained in detail in the TIwebsite. The specific document that presents the meth-odological issues is www.transparency.de/documents/cpi/cpi_framework.html.

3. These data are easily available at http://www.world-bank.org/wbi/governance/govdata2001.htm. The origi-nal data are listed as a 0–100 percentile rank and havebeen converted here to a 0–10 score to correspond to theCPI 0–10 format.

4. The WB web site (http://www.worldbank.org/wbi/governance/govdata2001.html) states: ‘‘As discussed indetail in the accompanying papers, countries’ relative

positions on these indicators are subject to margins oferror that are clearly indicated in the data files andgraphs downloadable below.’’ One reviewer of this paperargued that the WB measures are ‘‘objective scales,’’when in fact, the WB states on its web site that theindicators ‘‘reflect the statistical compilation of percep-tions of the quality of governance of a large number ofenterprise, citizens, and expert survey respondents.’’These measures should not be confused with the WBstudies of corruption experience.

5. Homicide rates, however, are used as reliable indi-cators of one form of extreme crime. Only in places like

South Africa under apartheid, where the murder ofblacks was often not reported by the police, or incountries undergoing civil wars where record-keepingfunctions often break down entirely, are these ratesunreliable.

6. See Comite Nacional de Integridad and WorldBank-CIET International (1998). See also World Bank(2000).

7. The surveys reported on in this project are allnational probability samples carried out at varioustimes from 1998 through 2002 in El Salvador, Hondu-ras, Nicaragua, Ecuador, Paraguay, and Bolivia. A website is currently being constructed where this data canbe easily accessed. The samples each have between2,500 and 3,000 respondents, except for Paraguay,which included only 1,463. The first survey, conductedin Nicaragua, was developed with Casals and Associ-ates. I would like to thank Dr. Sergio Dias Briquets ofCasals and Dr. Andrew Stein, now of the US Depart-ment of State, for assistance in that early work. Dr.Orlando Perez of Central Michigan University hasassisted in more recent studies in Honduras andEcuador.

8. The word for ‘‘bribe’’ in Spanish differs among thecountries sampled. In Central America, the surveygenerally used ‘‘mordida’’ while in South America the

http://www.transparency.de/documents/cpi/cpi_framework.html

http://www.transparency.de/documents/cpi/cpi_framework.html

http://www.worldbank.org/wbi/governance/govdata2001.htm

http://www.worldbank.org/wbi/governance/govdata2001.htm

http://www.worldbank.org/wbi/governance/govdata2001.html

http://www.worldbank.org/wbi/governance/govdata2001.html


word ‘‘coima’’ was utilized. In both areas, however, theadditional term ‘‘soborno’’ was utilized.

9. Some respondents might have worked at some pointduring the year prior to the interview but were notworking when the survey took place. Those individuals,therefore, could have been subjected to corruption atsome point during the year prior to the survey.

10. Ecuador is excluded from this series because aslightly different question was used there.

11. The original wording of the school and healthquestions referred to ‘‘improper payments’’ rather than

bribes in particular, while the question dealing with localgovernment focused on ‘‘payments above what the lawrequires.’’ In a subsequent study in Ecuador, it wasfound that if the word ‘‘bribe’’ was substituted for thesephrases, the reporting of corruption decreased some-what. For example, in a special sample focusing on 15municipalities in Ecuador (N = 4,500), bribery in themunicipality was found to be 17.4% compared to the19% reported in Figure 11.

12. This pattern is also observed in the other coun-tries in the series, but the inclusion of all six countriesin the line chart of Figure 12 makes it very difficult toread.

REFERENCES

Anderson, C. J., & Tverdova, Y. V. (2003). Corruption,political allegiances, and attitudes toward govern-ment in contemporary democracies. American Jour-nal of Political Science, 47(1), 91–109.

Canache, D. (2002). Venezuela: Public opinion andprotest in a fragile democracy. Coral Gables, Fla:North-South Center Press at the University ofMiami.

Canache, D., & Kulisheck, M. R. (Eds.). (1998).Reinventing legitimacy: Democracy and politicalchange in Venezuela. Westport, Conn: GreenwoodPress.

Comite Nacional de Integridad and World Bank-CIETInternational (1998). Encuesta nacional sobre integr-idad y corrupcion en la administracion publica:Encuesta de hogares. Managua.

della Porta, D. (2000). Social capital, beliefs in govern-ment, and political corruption. In S. J. Pharr & R. D.Putnam (Eds.), Disaffected democracies: What’s trou-bling the trilateral countries? (pp. 202–228). Prince-ton: Princeton University Press.

della Porta, D., & Meny, Y. (1996). Democracy andcorruption in Europe. London: Pinter.

Easton, D. (1975). A re-assessment of the concept ofpolitical support. British Journal of Political Science,5, 435–457.

Gingerich, D. (2004). On unstable ground: Parties,patronage and political corruption in contempo-rary Bolivia. Paper presented at the Meeting ofthe American Political Science Association,Chicago.

Golden, M., & Chang, E. (2004). Does corruption pay?The survival of politicians charged with malfeasancein the Postwar Italian chamber of deputies. Paperpresented at the American Political Science Associa-tion, Chicago, IL.

Golden, M. A., & Chang, E. C. C. (2001). Competitivecorruption: Factional conflict and political malfea-sance in postwar Italian Christian democracy. WorldPolitics, 53(4), 588–622.

Heidenheimer, A. J., Johnston, M., LeVine, V. T.(1989). Political corruption: A user’s manual. NewBrunswick: Transaction.

Huntington, S. P. (1968). Political order in changingsocieties. New Haven: Yale University Press.

Kaufmann, D. (1998). Corruption diagnostics: A newtechnocratic framework for the analysis of corrup-tion and its implications for the design of actionprograms. Miami, Florida.

Kaufmann, D., Kraay, A., Zoido-Lobaton, P. (1999).Governance matters (Vol. 2196). Policy ResearchWorking Paper. Washington, DC: World Bank.

Kornberg, A., & Clarke, H. D. (1992). Citizens andcommunity: Political support in a representativedemocracy. Cambridge: Cambridge University Press.

Lipset, S. M. (1961). Political man: The social basis ofpolitics (1981 expanded ed.). Baltimore, MD: JohnsHopkins University Press.

Lipset, S. M. (1994). The social requisites of democracyrevisited. American Sociological Review, 59, 1–22.

Lipset, S. M., & Lenz, G. S. (2000). Corruption, cultureand markets. In L. J. Harrison & S. P. Huntington(Eds.), Culture matters: How values shape humanprogress (pp. 112–124). Basic Books: New York.

Mishler, W., & Rose, R. (2001). What are the origins ofpolitical trust? Testing institutional and culturaltheories in post-communist societies. ComparativePolitical Studies, 34(1), 30–62.

Muller, E. N., Jukam, T. O., & Seligson, M. A. (1982).Diffuse political support and antisystem politicalbehavior: A comparative analysis. American Journalof Political Science, 26, 240–264.

Muller, E. N., Seligson, M. A., & Turan, I. (1987).Education participation, and support for democraticnorms. Comparative Politics, 20, 19–33.

Pharr, S. J. (2000). Officials’ misconduct and publicdistrust: Japan and the trilateral democracies. In S. J.Pharr & R. D. Putnam (Eds.), Disaffected democra-cies: What’s troubling the trilateral countries?(pp. 174–201). Princeton: Princeton University Press.

Putnam, R. D. (1993). Making democracy work: Civictraditions in modern Italy. Princeton, NJ: PrincetonUniversity Press.

Rose-Ackerman, S. (1999). Corruption and government:Causes, consequences, and reform. Cambridge: Cam-bridge University Press.


Seligson, M. A. (1983). On the measurement of diffusesupport: Some evidence from Mexico. Social Indica-tors Research, 12, 1–24.

Seligson, M. A. (2000). Toward a model of democraticstability: Political culture in Central America. Estu-dios interdisciplinarios de America Latina y el Caribe,11(2), 5–29.

Seligson, M. A. (2002). The impact of corruption onregime legitimacy: A comparative study of four LatinAmerican countries. Journal of Politics, 64, 408–433.

Seligson, M. A., & Carrion, J. (2002). Political support,political skepticism and political stability in newdemocracies: An empirical examination of masssupport for Coups D’etat in Peru. ComparativePolitical Studies, 35(1), 58–82.

Treisman, D. (2000). The causes of corruption: A cross-national study. Journal of Public Economics, (3),399–458.

Tulchin, J. S., & Espach, R. H. (2000). In Combatingcorruption in Latin America. Washington, D.C.:Woodrow Wilson Center Press.

Weyland, K. (1998). The politics of corruption inLatin America. Journal of Democracy, (2), 108–121.

World Bank (1997). World development report, 1997.Washington, DC: Oxford University Press.

World Bank Institute (2000). Ecuador: Governance andanticorruption empirical diagnostic study; evidencefrom surveys of households, enterprises, and publicofficials. Washington, DC: World Bank.

the measurement and impact of corruption victimization ......are measures of corruption perception...

Documents