Corruption and development

Journal of Economic LiteratureVol. XXXV (September 1997), pp. 1320–1346

Bardhan: Corruption and Development

Corruption and Development: A Review of Issues

PRANAB BARDHANUniversity of California at Berkeley

I am grateful for useful comments on earlier drafts from Jean-Claude Berthelemy, AndrewGoudie, Mancur Olson, Dani Rodrik, Susan Rose-Ackerman, and Andrei Shleifer. The firstdraft of the paper was written for and presented at a meeting of the OECD DevelopmentCenter, Paris.

I. Introduction

CORRUPTION IS an ancient problem. In atreatise on public administration dating

back to the fourth century B.C. in India, Kau-tiliya writes in his Arthasastra:

Just as it is impossible not to taste the honey(or the poison) that finds itself at the tip ofthe tongue, so it is impossible for a govern-ment servant not to eat up, at least, a bit ofthe king’s revenue. Just as fish moving underwater cannot possibly be found out either asdrinking or not drinking water, so govern-ment servants employed in the governmentwork cannot be found out (while) takingmoney (for themselves). (R. P. Kangle 1972,p. 91)

In a passage of characteristically re-markable precision Kautiliya states thatthere are “forty ways of embezzlement”and then goes on to enumerate theseways.

While corruption in one form or an-other has always been with us, it hashad variegated incidence in differenttimes at different places, with varyingdegrees of damaging consequences.While the tenacity with which it tendsto persist in some cases easily leads todespair and resignation on the part of

those who are concerned about it, therecan be and have been ways in which awhole range of policy measures make asignificant dent. In this paper, we startwith a discussion of some of the alterna-tive denotations of the problem of cor-ruption; we then consider the ways inwhich the damaging consequences ofcorruption operate in the economy,while not ignoring its possible redeem-ing features in some cases; we pursuethe question of why corruption is per-ceptibly so different in different socie-ties; and finally, we examine the feasi-ble policy issues that arise. Ourapproach in this paper is primarily ana-lytical and speculative, given the inher-ent difficulties of collecting (and hencethe nonexistence of) good empiricaldata on the subject of corruption; werefer in the Appendix to some subjec-tive ordinal rankings of countries interms of corruption that are available inthe international media, for whateverthey are worth.

In common usage the word “corrup-tion” is used to mean different things indifferent contexts. Even if we choose toconfine ourselves only to the economic

1320

context, staying away, for example, fromrelated issues of political corruption(i.e., where the ill-gotten gains are pri-marily in terms of political power),there are alternative denotations of eco-nomic corruption. In a majority of casessuch corruption ordinarily refers to theuse of public office for private gains,where an official (the agent) entrustedwith carrying out a task by the public(the principal) engages in some sort ofmalfeasance for private enrichmentwhich is difficult to monitor for theprincipal. There are, of course, manyeveryday cases of other kinds of corrup-tion some of which may take place en-tirely in the private sector. For exam-ple, a private seller sometimes rationsthe supply of a scarce good (instead ofusing the price mechanism to clear themarket), and we use various ways ofbribing him or an agent to jump thequeue (paying a higher price to a“scalper” for a sold-out theater show ora game, tipping a “bouncer” for entryinto a crowded nightclub, using “con-nections,” i.e., some form of long-run gift exchange, to get a job, and soon).

Sometimes one invokes legality andalmost interchangeably uses the word“corrupt” and “illicit” in describing atransaction. But just as clearly not allillegal transactions are corrupt, nor areall instances of corruption or bribery il-legal1 (as when you tip the maitre d’ toget a better table at a restaurant thanother customers, or in the much moreimportant cases of gift-giving by lobby-ists to politicians, campaign contribu-tions to Political Action Committees, orpost-retirement jobs in private firms to

bureaucrats of agencies meant to regu-late them). Similarly, one should keep adistinction between “immoral” and“corrupt” transactions. When you pay ablackmailer, you may consider him im-moral, but you are paying to stop himfrom revealing some information whichmay be unpleasant for you but whichmay be neither illegal nor corrupt. Onthe other hand, one can think of in-stances of corruption and bribery whichsome people may not regard as immoral(particularly those for whom end justi-fies means), as when you bribe a police-man not to torture a suspect. Havingreferred to these alternative meaningsof even economic corruption, let mestate that in this paper I shall mostlyconfine myself to the application ofthis term to imply the use of publicoffice for private gain or the agencyproblem referred to in the precedingparagraph.

Even with this common use of theterm among economists, there are manyambiguities. Does striving for privategain include policies that are primarilyoriented to increasing the chances forremaining in office? The distinction be-tween political and economic corrup-tion can get blurred here. Then thereare problems in common comparativeuse of the term in the obvious absenceof any publicly available objective mea-sures. A particular African country maybe in some sense more corrupt than aparticular East Asian country, eventhough the actual amount of bribemoney exchanging hands may be muchlarger in the latter; this may be simplybecause rampant corruption may havechoked off large parts of economictransactions in the former. Then thereare cases where the bribe per unit oftransaction (and the consequent ineffi-ciency) may be higher (in the case ofdecentralized corruption, as we shallnote later) than in situations of central-

1 As Gordon Adams (1981, p. 177) notes, theU.S. Department of Defense directive 55007 al-lows gratuities when they are a part of a “custom-ary exchange of social amenities between personalfriends and relatives when motivated by such rela-tionships and extended on a personal basis.”

Bardhan: Corruption and Development 1321

ized (“one-stop shopping”) corruptionwhere the inefficiency may be less,even though the total amount of bribepaid may be larger.

II. Effects on Efficiency

There is a strand in the corruption lit-erature, contributed both by economistsand noneconomists, suggesting that, inthe context of pervasive and cumber-some regulations in developing coun-tries, corruption may actually improveefficiency and help growth. Economistshave shown that, in the second-bestworld when there are pre-existing pol-icy induced distortions, additional dis-tortions in the form of black-marke-teering, smuggling, etc., may actuallyimprove welfare even when some re-sources have to be spent in such activi-ties. The argument for efficiency-im-proving corruption is a simple extensionof this idea. As Nathaniel H. Leff(1964, p. 11) puts it simply: “if the gov-ernment has erred in its decision, thecourse made possible by corruption maywell be the better one.” As nonecono-mists usually point out, corruption isthe much-needed grease for the squeak-ing wheels of a rigid administration.Samuel P. Huntington (1968, p. 386)states it bluntly: “In terms of economicgrowth, the only thing worse than asociety with a rigid, over-centralized,dishonest bureaucracy is one with arigid, over-centralized, honest bureauc-racy.”

Even without pre-existing distortions,one may look upon corruption as part ofa Coasean bargaining process in which abureaucrat (who is in the illicit businessof selling property rights to a public re-source in the form of issuing permitsand licences) and the private agent (theprospective buyer) may negotiate theirway to an efficient outcome. If in abribery game there is competitive bid-

ding by private firms for a governmentprocurement contract, and the corruptofficial awards the contract to the high-est bidder in bribes, then allocation ef-ficiency is maintained, as only the low-est-cost firm can afford the largestbribe. (This, of course, assumes thatother goals of the program are not vio-lated: this bidding procedure is clearlynot acceptable in the case of Universityadmissions, for example.) That the pro-ducer surplus lines the pocket of thebureaucrat and does not go to the pub-lic treasury (as would have happened inan open auction for the contract) doesnot seemingly affect the allocation effi-ciency.

This argument, however, is morecomplex when a briber does not havefull information about the cost levelsand therefore the bribing capacity of hiscompetitors, and when he has to takeinto account strategic considerations inmaking any particular offer of a bribe.But the situation can be modeled as ann-person symmetric game with incom-plete information on the part of eachplayer and one can draw upon the the-ory of sealed-bid auctions. In such acontext Paul J. Beck and Michael W.Maher (1986) and Donald H. D. Lien(1986) have shown that under the as-sumptions of the model, the lowest-costfirm is always the winner of the con-tract, and thus bribery can reproducethe efficiency consequences of competi-tive bidding procedures under imper-fect information. Inefficiency may, ofcourse, result if the official is influ-enced by considerations other than justthe size of the bribe (for example, fa-voritism for a particular client or nepo-tism); or when the briber can get awaywith supplying a low-quality good at ahigh-quality price, and the official letsin unqualified applicants with a highwillingness to pay; or when bribery isused to limit the competition (as in the

1322 Journal of Economic Literature, Vol. XXXV (September 1997)

case of bribing the police or tax inspec-tors to harass rival firms).2

Another efficiency argument in favorof corruption is to look upon it as“speed money” (for which there are dis-tinct terms in different countries, likelagay in the Philippines), which reducesdelay in moving files in administrativeoffices and in getting ahead in slow-moving queues for public services.Queuing models which have receivedsome attention in the theoretical litera-ture allow the possibility for the corruptbureaucrat to practice price discrimina-tion among clients with different timepreference. In an interesting equilib-rium queuing model with some specialassumptions Francis T. Lui (1985) de-rives bribing functions where the size ofthe bribe (decided by the briber, notthe server of the queue) is linked to theopportunity costs of time for the indi-vidual client and shows that the bribingstrategies will form a Nash equilibriumof this noncooperative game that willminimize the waiting costs associatedwith the queue, thereby reducing theinefficiency in public administration.(The model can also be useful in de-signing schedules of incentive paymentsin the pay structure of civil servants.)

One does not have to take a moralis-tic position on corruption to see thatsome of these arguments above in favorof the efficiency effects of corruptionare fraught with general problems, eventhough in individual instances some re-deeming features of corruption may bepresent. For example, in the second-best case made above, it is usually pre-sumed that a given set of distortions aremitigated or circumvented by the ef-fects of corruption; but quite oftenthese distortions and corruption arecaused or at least preserved or aggra-

vated by the same common factors. Thedistortions are not exogenous to the sys-tem and are instead often part of thebuilt-in corrupt practices of a patron-client political system. As we have indi-cated above, bidding procedures in sucha system may still end up in allocationalinefficiency.

As for speed money, Gunnar Myrdal(1968), citing the 1964 SanthanamCommittee on the Prevention of Cor-ruption appointed by the Governmentof India, has argued that corrupt offi-cials may, instead of speeding up, actu-ally cause administrative delays in orderto attract more bribes.3 (I am told thatin Russia there is a clear terminologicaldistinction between mzdoimstvo, takinga remuneration to do what you are sup-posed to do anyway, and likhoimstvo,taking a remuneration for what you arenot supposed to do.) Lui’s equilibriumqueuing model is meant to question thevalidity of Myrdal’s hypothesis at thetheoretical level. But, as Jens C. Andvig(1991) points out, from the point ofview of imperfect information and stra-tegic considerations queues as alloca-tion mechanisms are more complex andmany-sided than has been recognized inthe literature, and different ways of or-ganizing the queue may give rise to dif-ferent outcomes on the average waitingtime. In Lui’s otherwise very interestingmodel, for example, both sides in thecorrupt transaction are honest in thesense that they stick to a deal, that nonew bribe offers are made by the wait-ing clients after the new entrants havearrived, that there is no moral hazardabout the reliability of the sale bythe server of a priority in the queue,and so on. The model’s results may notbe robust to these kinds of considera-tions.

2 For an account of many such harmful effectsof corruption, see Rose-Ackerman (1996).

3 Abhijit Banerjee (1994) examines situationswhere bureaucrats create red tape and use it toscreen clients of different types.


This also suggests the problem withlooking upon bribes simply as side pay-ments in a Coasean bargaining processbetween officials or politicians andfirms (even apart from the agency prob-lem that the bribee is not representingthe interests of the principal, the pub-lic). Of course, the briber and thebribee may fail to agree on the appro-priate size of the bribe on account ofbargaining in a situation of asymmetricinformation and also, there are collec-tive action problems when several firmshave to get together to bribe a singlepolitician or bureaucrat. But more im-portant than these is the fact, empha-sized by Maxim Boycko, Shleifer, andRobert Vishny (1995), that corruptioncontracts are not enforceable in courtsand there is many a slip between thebribing transaction and the actual deliv-ery of the good or the service involved.The control rights on the latter areoften arbitrary and uncertain, leaving alot of leeway for the bribee to renegeon his understanding with the briber, orto come back and demand anotherbribe. (It used to be said of GeneralNoriega of Panama in his heyday thathe could not be bought, he could onlybe rented.) Of course, the bribee mayhave to worry about his reputation inthe long run about keeping promises(but many corrupt politicians have tooshort a time horizon), or sometimes thebriber can hire hoodlums to disciplinethe bribee (but the transaction costs forsuch ways of enforcement can be high).

A. Centralization of Bribery

Sometimes the bribee cannot delivernot because he wants to cheat, but be-cause there is a multiple veto powersystem in operation, which makes cen-tralized collection of bribes in exchangeof guaranteed favors very difficult. Onehigh official in New Delhi is reported tohave told a friend : “if you want me to

move a file faster, I am not sure if I canhelp you; but if you want me to stop afile I can do it immediately.” This abil-ity to “stop a file” at multiple points (asystem often installed to keep corruptofficials in check) may result in increas-ing the inefficiency as well as the rateof bribes. In general centralized corrup-tion has less adverse consequences forefficiency than decentralized bribe-tak-ing, because in the former case thebribee will internalize some of the dis-tortionary effects of corruption (assum-ing similar powers at all levels to deter-mine the overall rents in the system).

Shleifer and Vishny (1993) illustratethis point with an elementary modelcomparing a case of independent mo-nopolists (where different public agen-cies provide complementary govern-ment goods or services independently)with that of a joint monopolist agencyproviding the same goods or services.Suppose a customer needs two permitsor two complementary inputs from twodifferent agencies in the former case.Each agency as an independent monop-olist will take the other agency’s sales asgiven and so the bureaucrat in charge ofit will set the bribe-inclusive price insuch a way that marginal revenue isequal to the marginal cost, the bribeper unit of sale being the difference be-tween the price and the monopolist’smarginal cost (i.e., the official price ofthe good supplied). The joint monopo-list, on the other hand, takes into ac-count the effect of an extra unit sold onthe sales of the complementary goodand thus on the revenue from bribesfrom the other source as well, so that inequilibrium the marginal revenue in thesupply of each good is less than themarginal cost. Thus the per unit bribe ishigher and the supply of each goodlower in the independent monopolistcase than in the case of collusion. Ofcourse, the aggregate revenue from


bribes is larger in the latter case, butthe customer gets a larger supply ofboth inputs. The problem is made muchworse when complementarity can be ar-tificially created (just when you thinkyou have bribed two agencies to getthe required two permits, another in-dependent monopolist comes along andtells you that you need a third permitfrom him to get your business inplace) and corruption opportunities stim-ulate the entry of permit-dispensersarmed with new regulations. Free entryin this game allows the officials to“overfish” in the “commons” or therental havens.

Shleifer and Vishny would explain theincrease in the inefficiency flowingfrom corruption in post-CommunistRussia in comparison with CommunistRussia in these terms. Formerly, theCommunist Party used to centralize thecollection of bribes and effectivelymonitored (sometimes with the help ofthe KGB) deviations from agreed-uponpatterns of corruption. Now differentministries, agencies, and levels of localgovernment all set their own bribes in-dependently in a decentralized attemptto maximize their own revenue. It isusually suggested that the regulatorystate is at the root of the inefficiencydue to corruption spawned by the regu-lations; the above analysis suggests thata weak central government with its in-ability to stop the setting up of indepen-dent corruption rackets (a kind of eco-nomic warlordism) makes the problemof inefficiency particularly acute. Thismay be relevant in a comparison of cor-ruption in, say, Indonesia with that inIndia. Table 2 in the Appendix suggeststhat in the perception of foreign busi-nessmen the two countries are aboutequally corrupt;4 and yet the economic

performance by most accounts has beenmuch better in Indonesia. Could it bethat Indonesian corruption is more cen-tralized (controlled largely by the firstfamily and the top military leadership incahoots with the ethnic Chinese-runconglomerates) and thus somewhatmore predictable, whereas in India it isa more fragmented, often anarchic, sys-tem of bribery?

Centralization of the political ma-chine also makes it possible to have asystem approximating “lump-sum” cor-ruption, without distorting too many de-cisions at the margin. It has been sug-gested, for example, that corruption incountries like South Korea5 may havebeen more in the form of lump-sumcontributions by the major businessleaders to the president’s campaignslush fund, without taxing economic ac-tivity at the margin. The importantquestion here is how the ruler cancredibly promise to keep the contribu-tions lump-sum, and not come backagain for individual quid pro quo dealsat the margin. This ability to crediblycommit is a feature of “strong” statesthat very few developing countrieshave.

The idea of the differential efficiencyeffects of centralized versus decentral-ized corruption is akin to Olson’s (1993)idea of smaller distortionary effects ofthe tax impositions of the state as a “sta-tionary bandit” (having thus an “encom-passing interest” in the domain overwhich its rent-exacting power is exer-cised) as opposed to those of the “rov-ing bandit.” One may, however, pointout that even centralized corruption ismore distortionary than taxation (not tospeak of the extra burden of taxes thatpublic revenue losses from corruption

4 According to Table 1 in the Appendix, corrup-tion in Indonesia was perceived to be much worsethan in India in the early 1980s.

5 It is interesting to note from Appendix Table 1that perceived corruption in Korea in the early1980s is not significantly different from that inBrazil or even India.


may necessitate). This is because of theneed to keep corruption secret, asShleifer and Vishny (1993) point out.Efforts to avoid detection and punish-ment cause corruption to be more dis-tortionary than taxation. Because differ-ent activities have different chances ofdetection for bribes, there will be somesubstitution effect following from cor-ruption by which corrupt officials willtry to induce investment and transac-tions in the direction of lower-detectionactivities (or contractors who are lesslikely to squeal, even though they maybe less efficient). Bureaucrats in poorcountries may, for example, opt for im-ports of complex technology or goods(where detecting improper valuation oroverinvoicing is more difficult) in pref-erence to more standardized, but possi-bly more appropriate, technology orgoods. For similar reasons, allocatinggovernment funds in a few large de-fense contracts may look more attrac-tive to the officials involved than spend-ing the money in building numeroussmall rural health clinics. To preservethe secrecy of deals, a small elite groupmay also try to raise entry barriersfor outsiders, which in many situationshas the effect of discouraging the flowof new ideas and innovations. Secretpayments, particularly by foreign com-panies, also tend to be accumulatedand spent not inside the country butabroad.

B. Bribes Relative to Rents

Before we leave the subject of costsof corruption, it may be useful to com-ment on the magnitude of bribes in re-lation to that of the rent they are sup-posed to procure for the briber. Theearly literature on rent-seeking, as inAnne O. Krueger (1974), assumed aprocess of competitive bidding by therent-seekers which resulted in a com-plete dissipation of the rent. Since then

there have been models of barriers toentry in the rent-seeking sector (includ-ing models of dynamic games of movesand counter-moves of the contendingrent-seekers) and of the various transac-tion costs and risks that the rent-seek-ers have to face. But what is stillastonishing is the extremely small sizeof the usual bribe compared to the rentcollected (Gordon Tullock, 1980, hadpointed this out quite early, and thephenomenon is sometimes referred toin the public choice literature as the“Tullock paradox”). The anecdotes areendless. Tullock (1990) cites the case ofthe New York Congressman MarioBiaggi, who manipulated the federalgovernment to save from bankruptcy anenormous Brooklyn dockyard, for whichhe received three Florida vacationsworth $3,000. Spiro Agnew had to re-sign from the Vice Presidency of theNixon Administration for continuing totake bribes of an incredibly triflingamount from an arrangement made ear-lier in his political career. Most suchanecdotes are from democratic polities.On the other hand, there are anecdotesof corrupt income running to billions ofdollars for authoritarian rulers in muchpoorer countries, like Mobutu seseSeko in Zaire or Ferdinand Marcos inthe Philippines. This may point to a par-ticular coordination problem in bribecollection in democratic polities thatEric Rasmusen and Mark Ramseyer(1994) have tried to model.

They use a coordination game amongwealth-maximizing legislators to showthat, if the latter cannot coordinatetheir actions, they may supply private-interest statutes for bribes even lessthan the costs they incur. Only whenthey can enforce agreements with oneanother, solving a prisoner’s dilemmaproblem, will they come close to col-lecting the full benefits of the statutesthey pass. Rasmusen and Ramsayer


(1994) have a simple example to illus-trate the difference between a demo-cratic and an autocratic government inthis context. Suppose that private-inter-est statute S14 would provide a benefitof 14 for a lobbyist and would cost anautocratic government 50 because of,say, an increased probability of publicdiscontent or even rebellion. The auto-crat will supply this statute only if of-fered at least 50, which the lobbyist willbe unwilling to offer, so S14 will notpass. Suppose that a second statute,S80, would cost the autocrat 50 butbenefit the lobbyist by 80; the autocratwill supply this statute for a bribe any-where between 50 and 80.

Now take a democracy where fivelegislators vote on statutes S14 and S80.For each statute, each legislator loses 5by voting “yes” when the others vote“no,” but 10 if the statute passes. Thegovernment thus loses (again in termsof public discontent) a total of 50 if astatute passes, exactly the same cost asin the case of the autocratic govern-ment. Take first the statute S14. If eachlegislator thinks that the others willvote “no,” then all voting “no” will bethe equilibrium. The lobbyist couldovercome these expectations by offeringa bribe of 5 to three legislators, but thatis too costly for him for a statute worth14. But if each legislator thinks the oth-ers will vote “yes,” then each may aswell vote “yes” for an infinitesimallysmall bribe, because he will lose 10, nomatter how he votes (so that his mar-ginal cost of voting “yes” is 0). Thus ademocratic government may sell a pri-vate-interest statute at below cost whenthe autocratic government would not.Consider now the statute S80. Here toothere is an equilibrium in which thestatute passes in the democratic legisla-ture with an infinitesimally small bribe,when the autocrat would do it only for alarge bribe.

It is often said that autocratic rulersare more corrupt than democratic onesbecause the former do not have toworry about re-election. (This is notquite true as elections have becomevery expensive, and to dispense favorsin exchange for campaign contributionsis a major source fof corruption indemocratic regimes.) In the exampleabove, the cost of corruption is deliber-ately kept the same for both autocraticand democratic governments, and yetthe equilibrium bribe amount is largerunder the former. The essential prob-lem is due to an externality that eachdemocratic legislator’s vote potentiallyimposes on every other legislator, whenthey cannot coordinate their votes todemand a bribe which compensatesthem for that externality. In some ac-tual democratic polities, of course, suchcoordination problems are reduced bycommittee systems, disciplined factionsand party political machines.6 It is re-ported that in the past few decades Ja-pan’s Liberal Democratic Party (par-ticularly its so-called Policy AffairsResearch Council, where importantpolicies were made and payoffs werecoordinated behind closed doors) hasbeen quite successful in centralizingbribery and raking off billions of dol-lars’ worth in the process.

III. The Growth Process

Corruption has its adverse effects notjust on static efficiency but also on in-vestment and growth. A payment ofbribes to get an investment licenseclearly reduces the incentive to invest(even apart from affecting the composi-tion of investment, in view of the con-siderations of secrecy and uncertaintyalluded to in the previous section). One

6 Rose-Ackerman (1978) has noted that well-or-ganized legislators may be able to extort largeramounts than disorganized legislators.


might add that in the taxation system ofmany countries, negative profits (losses)can be deducted from taxable invest-ment income, but there is no corre-sponding loss offset in the case ofbribes, so that the latter are particularlyharmful for risk-taking in the context ofinnovation.

Similarly, when public resourcesmeant for building productivity-enhanc-ing infrastructure are diverted for poli-ticians’ private consumption (cementfor public roads or dams used for luxuryhomes) growth rates obviously will beaffected adversely. Another growth ef-fect follows from the fact that higherbribes imply declining profitability onproductive investments relative to rent-seeking investments, thus tending tocrowd out the former. As Kevin Mur-phy, Shleifer, and Vishny (1993) pointout, there are many reasons why thereare increasing returns to rent-seeking,so that an increase in rent-seeking low-ers the cost of further rent-seeking rela-tive to that of productive investment. Ingeneral when there is slow growth thereturns to entrepreneurship (particu-larly in production of new goods) fallrelative to those to rent-seeking, andthe ensuing increase in the pace ofrent-seeking activities further slowsdown growth. Besides, innovators areparticularly at the mercy of corruptpublic officials, because new producersneed government-supplied goods likepermits and licenses more than estab-lished producers. In any case, as Romer(1994) has suggested, corruption as atax on ex post profits may in general sti-fle entry of new goods or technologywhich require an initial fixed cost in-vestment.

Some of these growth effects havebeen statistically corroborated fromcross-country data. On the basis of cor-ruption rankings data assembled fromthe Business International correspon-

dents7 in 70 countries in the early1980s, as reported in Table 1 in the Ap-pendix, Paolo Mauro (1995) finds a sig-nificant negative association betweenthe corruption index and the investmentrate or the rate of growth (even aftercontrolling for some other determinantsof the latter, and correcting for a possi-ble endogeneity bias in the data). Aone-standard-deviation improvement inthe corruption index is estimated to beassociated with an increase in the in-vestment rate by about 3 percent ofGDP.8 The negative relation seems tohold even in subsamples of countrieswhere bureaucratic regulations are re-ported to be cumbersome, indicatingthat corruption as a way of by-passingthese regulations may not have beenvery beneficial.

Historians, of course, point to manycases when a great deal of corruption indispensing licenses, or loans, or miningand land concessions has been associ-ated with (and may have even helpedin) the emergence of an entrepreneurialclass. In European history the latterclass grew out of the sales of monopolyrights, tax farms, and other forms ofprivileged access to public resources. Inthe U.S. “gilded age” of 1860s and1870s widespread corruption of statelegislatures and city governments bybusiness interests and those seekingfranchises for public utilities is reportedto have helped rather than hindered

7 One problem with this data set is that it isbased on the perception of foreign businessmenwhose experience of corruption may be differentfrom what domestic businessmen face in a coun-try. The former may have less insider knowledgeabout the intricacies of the indigenous bureauc-racy and even less patience with its slow pro-cesses. So they may end up paying much largerbribes than what the latter settle for at the end oflong negotiations and endless cups of coffee in fa-miliar terrain. This discrepancy may vary fromcountry to country and thus bias the results of sta-tistical analysis on the basis of this data set.

8 These results are confirmed in Mauro (forthcoming)with a larger and more up to date data set.


economic growth.9 More generally, cor-ruption may have historically playedsome role in undermining the sway ofcollective passions that used to fuel in-ternecine group warfare. As RonaldWraith and Edgar Simpkins (1963, p.60) say of English history: “For twohundred and fifty years before 1688,Englishmen had been killing each otherto obtain power. . . . The settlements of1660 and 1688 inaugurated the Age ofReason, and substituted a system of pa-tronage, bribery, and corruption for theprevious method of bloodletting.” Inthis century, the highly corrupt systeminstitutionalized in the PRI enabledMexico to transcend the decade ofbloodletting that followed the Revolu-tion. Without denying the positive rolethat corruption may have played in his-tory in some situations, in many devel-oping countries today, however, corrup-tion is perceived to be so pervasive andendemic that it is unlikely to have goodnet effects, on grounds that we havediscussed earlier in this section and be-cause corruption tends to feed on itself(as we shall discuss in the next section)and it is impossible to confine corrup-tion to areas, if any, of relative benefi-cial effects.

What about the effects of the growthprocess on the extent of corruption? Al-though the requisite time-series evi-dence in terms of hard data is absent,circumstantial evidence suggests thatover the last 100 years or so corruptionhas generally declined with economicgrowth in most rich countries (and insome developing countries, like Singa-pore, it is reported to have declinedquite fast in recent decades). While thehistorical relationship between eco-nomic growth and corruption is thuslikely to have been negative in general,it is possible to envisage some non-

linearities in this relationship: in par-ticular, in some countries with the pro-cess of modernization and growth cor-ruption may have got worse for sometime before getting better. What kindof forces work toward possibly increas-ing corruption at the earlier stages ofeconomic growth? As the economy ex-pands and becomes more complex, pub-lic officials see more opportunities formaking money from their decisions,which now go beyond simple functionslike maintaining law and order and col-lecting land revenue. As the markets inmany new products are “thin” for quitesome time, this gives scope for those of-ficials to milk the process of grantingmonopoly rights and franchises. In theprocess of transition from controlled tomarket economy in Eastern Europe,China, and Vietnam it has often beenobserved that there are some specialfactors increasing corruption even as in-come grows. For a considerable periodof time the transition economy is on adual-track system: a part of output isstill under obligatory delivery at con-trolled prices, while the rest is allowedto be sold at market prices. This createsall kinds of new opportunities for cor-ruption. The process of privatization ofstate-owned enterprises in many coun-tries has also given rise to opportunitiesfor public officials to get kickbacksfrom “crony capitalist” buyers of thoseenterprises and contractors.

Yet, it is probably correct to say thatthe process of economic growth ulti-mately generates enough forces to re-duce corruption. Rewards to entrepre-neurship and productive investmentrelative to rent-seeking investment risewhen there is sustained growth. A pros-pering economy can also afford to payits civil servants well, reducing theirmotivation for corruption. And to theextent prosperity in the long run bringsmore demand, at least on the part of9 See Robin Theobald (1990) for a discussion.


the middle classes, for democratic re-forms, the latter may install institutionsthat check corruption. Not merely is thecoordination problem in bribe-collec-tion among legislators rendered moredifficult under democracy, as we havediscussed at the end of the precedingsection, but, more important, demo-cratic institutions build mechanisms ofaccountability and transparency at dif-ferent levels which make it difficult forthe networks of corruption to be sus-tained for long. A qualifier to this argu-ment relates, as we have noted before,to campaign finance in democratic elec-tions which leads to influence peddlingon the part of politicians. Thus whilerich democracies have been quite suc-cessful in better enforcement of laws,they have been in some cases less suc-cessful in reducing the influence ofmoney on the process of enactment ofthose laws.

IV. Factors Behind DifferentialIncidence and Persistence

We now turn to the question of whythe incidence of corruption is so palpa-bly different in different countries andthe related question of why in somecases corruption is so persistent. Lib-eral economists, of course, have an easyanswer to this: it is the regulatory statewith its elaborate system of permits andlicences that spawns corruption, anddifferent countries with different de-grees of insertion of the regulatory statein the economy give rise to varyingamounts of corruption. This explanationis no doubt valid to a large extent, butinadequate. It cannot, for example, ex-plain why corruption, in the judgmentof many perceptive observers, may haveincreased in post-Communist Russia orin China after the onset of the marketreforms in recent years. Comparingacross countries in Appendix Table 1

(based on the Business Internationalsurvey data for the early 1980s), it can-not explain why corruption is supposedto be so much more in Mexico than in,say, South Korea or Taiwan in the early1980s (when in the latter countries thestate was not much less interventionistthan in Mexico).

Another common explanation of dif-ferential corruption, popular among so-ciologists, is that social norms are verydifferent in different countries. What isregarded in one culture as corrupt maybe considered a part of routine transac-tion in another. (Visiting Westernersare often aghast that an Asian or an Af-rican will sometimes not carry out hisordinary service without baksheesh ortips; the latter, on the other hand, findsthe high degree of monetization even inpersonal transactions in advanced capi-talist countries somehow “corrupt.”)But a more important issue is involved.It is widely recognized that in develop-ing countries gift-exchange is a majorsocial norm in business transactions,and allegiance to kinship-based or clan-based loyalties often takes precedenceover public duties even for salariedpublic officials. Under such circum-stances use of public resources to caterto particularistic loyalties become quitecommon and routinely expected. At thesame time, it will be wrong to suggestthat concern about public corruption ispeculiarly Western. In most of the samedeveloping countries, public opinionpolls indicate that corruption is usuallyat the top of the list of problems citedby respondents. But there is a certainschizophrenia in this voicing of con-cern: the same people who are most vo-cal and genuinely worried about wide-spread corruption and fraud in thepublic arena do not hesitate at all inabusing public resources when it comesto helping out people belonging to theirown kinship network. (It is a bit like the


U.S. Congressmen who are usually lividabout the rampant pork-barrel politicsthey see all around them but they willfiercely protect the “pork” they bring totheir own constituency.) Edward C.Banfield (1958) comments on the preva-lence of what he calls “amoral familism”in the Mezzogiorno in Italy, but RobertPutnam (1993) observes in his study ofcomparative civicness in the regions ofItaly that the amoral individuals in theless civic regions clamor most forsterner law enforcement. Mayfair Yang(1989) notes how people in China gen-erally condemn the widespread use ofguanxi (connections) in securing publicresources, but at the same time admirethe ingenuity of individual exploitsamong their acquaintances in its use.

A major problem with norm-based ex-planations is that they can very easily benear-tautological (“a country has morecorruption because its norms are more

favorable to corruption”). A more satis-factory explanation on these lines has togo into how otherwise similar countries(or regions in the same country likeNorth and South in Italy) may settlewith different social norms in equilib-rium in, say, a repeated game frame-work, and how a country may some-times shift from one equilibrium intoanother (as has happened in the case oftoday’s developed countries in recenthistory with respect to corruption).

The idea of multiple equilibria in theincidence of corruption is salient insome of the recent economic theorists’explanations. The basic idea is that cor-ruption represents an example of whatare called frequency-dependent equilib-ria, and our expected gain from corrup-tion depends crucially on the number ofother people we expect to be corrupt.At a very simple level the idea may beillustrated, as in Andvig (1991), with a

Figure 1.

A

C

BM Curve

N Curve

0


so-called Schelling diagram shown inFigure 1. The distance between the ori-gin and any point on the horizontal axisrepresents the proportion of a given to-tal number of officials (or transactions)that is known to be corrupt, so that thepoint of origin is when no one is cor-rupt, and the end-point n is when every-one is corrupt. The curves M and N rep-resent the marginal benefit for acorrupt and an honest official respec-tively for all different allocations of theremaining officials in the two catego-ries. The way the curve N is drawn, thebenefit of an honest official is higherthan that of a corrupt official when veryfew officials are corrupt, but it declinesas the proportion of corrupt officialsincreases and ultimately becomes evennegative when almost all others arecorrupt. The M curve goes up at thebeginning when more and more offi-cials are corrupt (for the marginalcorrupt official lower reputation losswhen detected, lower chance of detec-tion, lower search cost in finding abriber, etc.), but ultimately declines(when the size of bribe is bid down bytoo many competing bribers, for exam-ple), even though at the end-pont thepay-off for a corrupt official remainspositive.

In Figure 1 there are three equilib-rium points, A, B, and C. A and C arestable, but B is not. At point A all arehonest and it does not pay to be cor-rupt. At C all are corrupt, and it doesnot pay to be honest. At B, any givenofficial is indifferent (between beingcorrupt and honest) but if only onemore official is corrupt it pays to be-come corrupt; on the other hand, if onefewer is corrupt, the marginal officialwill choose to be honest. So initial con-ditions are important: if the economystarts with (or gets jolted into) a highaverage level of corruption it will movetoward the high-corruption stable equi-

librium C; if the initial average corrup-tion is low, the economy gravitates to-ward the honest equilibrium A. The dia-gram illustrates in an elementary wayhow two otherwise similar countries(both in socio-economic structures andin moral attitudes) may end up with twovery different equilibrium levels of cor-ruption; also, how small changes mayhave a large impact on corruption if onestarts out at points close to B.

The problem with such simple dia-grams is that the mechanisms throughwhich the economy reaches one or theother equilibrium are not fully spelledout. There are now several theoreticalmodels in the literature which try to dothat rigorously, and also get away fromthe naive informational presumptionsimplicit in the diagram. We shall brieflytouch upon the main ideas in a few ofthem. Olivier Cadot (1987) has a modelof corruption as a gamble, where everytime an official asks for a bribe in a bi-lateral situation, there is a risk of beingreported to and sacked by a superior of-ficer. The optimal Nash strategy of acorrupt official is derived under alterna-tive assumptions about the informationstructure. The comparative-static re-sults show that a higher time discountrate, a lower degree of risk-aversion,and a lower wage rate will induce him,under certain conditions, to be morecorrupt. Then Cadot goes on to intro-duce corruption also at the level of thesuperior officer who can be bribed (be-yond a certain threshold) to cover uplower-level corruption. The interactionof corruption at different hierarchicallevels of administration leads to multi-ple equilibria (one with only petty cor-ruption and the other with more perva-sive corruption), as the probability ofbeing sacked diminishes with the gen-eral level of corruption in the civil ser-vice, and corruption at each level feedson the other. In the rent-seeking litera-


ture also it has been pointed out byArye L. Hillman and Eliakim Katz(1987) that there are extra social costswhen there is a hierarchical structuresuch that a lowly customs official isobliged to pay a part of his take ofbribes to a superior. The usual pre-sumption of that literature—which is, aswe have seen, in any case question-able—that bribes used in contesting arent do not entail a social cost becausethey are only transfers, is seriously viti-ated when one takes into account multi-tiered rent-seeking, with the official po-sitions to which the bribes accrue arethemselves contested with real re-sources.

Andvig and Karl O. Moene (1990) intheir model assume, as in Cadot (1987),that the expected punishment for cor-ruption when detected declines as moreofficials become corrupt, because it ischeaper to be discovered by a corruptrather than a noncorrupt superior.There is a bell-shaped frequency distri-bution of officials with respect to theircosts of supplying corrupt services. Onthe demand side the potential bribers’demand for corrupt services decreasesas the bribe size increases and as thefraction of officials who are corrupt de-creases (raising the search cost for a po-tential bribee). This model generatestwo stable stationary equilibria of theNash type and highlights how the prof-itability of corruption is positively re-lated to its frequency and how tempo-rary shifts may lead to permanentchanges in corruption.

Raaj Sah (1988) has a model of cor-ruption with intertemporal behavioralexternalities in the context of overlap-ping generations and a Bayesian learn-ing process in belief formation. The bu-reaucrats and citizens both start offwith a subjective probability distribu-tion which tells them how likely it isthat the agent they will meet in a trans-

action is corrupt. Corrupt (noncorrupt)agents would prefer meeting agents onthe other side of the transaction whoare similarly corrupt (noncorrupt). Foreach corrupt agent they meet, they willrevise upwards their subjective prob-ability estimates of meeting corruptpeople, and are more likely to initiate acorrupt act in the next period. This ishow beliefs about the nature of an eco-nomic environment one faces formedon the basis of one’s past experience ofdealing with that environment feedsinto the perpetuation of a culture ofcorruption. Again, there are multipleequilibria and two economies with anidentical set of parameters can have sig-nificantly different levels of corruption;the particular steady state to which theeconomy settles is influenced by thehistory of the economy preceding thesteady state.

Sah’s model admits the possibilitythat sometimes there may be discrepan-cies between beliefs about corruptionfrequency and its actual incidence.Philip Oldenburg’s (1987) account ofthe land consolidation program in vil-lages in U.P. in Northern India providesan interesting case study in this context.A land consolidation program involves amajor reorganization of the mapping ofthe existing cultivation plots, their valu-ation and carving out of new plots in avillage, and thus provides a lot of scopefor corruption for the petty officials incharge. But Oldenburg’s field investiga-tions found very little evidence of ac-tual official corruption. Complaints ofcorruption usually came from farmerswho had not got precisely what theywanted, and did not understand theprocess fully, and so assumed that otherfarmers who in their perception didbetter must have bribed to get theirway. Bribes were often paid to a mid-dleman, who pocketed the money whiletelling the villagers that it was primarily


meant to bribe the Assistant Consolida-tion Officer. (He even made a show ofpaying a visit to the Officer.) There mayactually be more corruption in othercases, but Oldenburg makes a validpoint that the middlemen in generalhave a vested interest in spreading(dis)information that “nothing getsdone without bribing the Officials,” andwhen everybody believes that, it mayeven have the effect of inducing an offi-cial to indulge in corruption, as he isassumed to be corrupt anyway. Thisis a familiar self-fulfilling equilibriumof corruption.10 (The middleman’srole in corruption is similar to whatDiego Gambetta (1988, p. 173) ob-serves in his study of the Italian Mafia:“the mafioso himself has an interestin regulated injections of distrust intothe market to increase the demand forthe product he sells—that is, protec-tion.”)

In an overlapping generations frame-work with dynamic complementarity be-tween past and future reputation JeanTirole (1996) has argued that the per-sistence of corruption in a society maybe explained partly by the bad collec-tive reputation of previous generations:younger generations may inherit thereputation of their elders with the con-sequence that they may have no incen-tive to be honest themselves. Thismeans, if for some temporary reasons(say, due to a war or some other disrup-tion in the economic system) corruptionin an economy increases, it has lastingeffects: collective reputation once shat-tered is difficult to rebuild. Similarly, aone-shot reduction in corruption(through, say, an anti-corruption cam-

paign) may have no lasting effect: itmay take a minimum number of periodswithout corruption to return to a pathleading to the low-corruption steadystate.

We have discussed in this section thereasons for the persistence of corrup-tion that have to do with frequency-de-pendent equilibria or intertemporal ex-ternalities. Let us end it by referring toa simpler reason for persistence in thecase of some types of corruption. Thereare many cases where corruption is mu-tually beneficial between the officialand his client, so neither the briber northe bribee has an incentive to report orprotest, for example, when a customsofficer lets contraband through, or a taxauditor purposely overlooks a case oftax evasion, and so on. Shleifer andVishny (1993) call it corruption withtheft (a better name may be collusivecorruption), to distinguish it from caseswhere the official does not hide thetransaction in which the client pays therequisite price, fee, or fine to the gov-ernment, but only charges somethingextra for himself, what Shleifer andVishny call corruption without theft.The former type is more insidious, diffi-cult to detect and therefore more per-sistent. One should add that this typealso includes many cases of official re-laxation of quality control standards, ininspection of safety in construction ofbuildings and bridges or in supplies offood and drugs, in pollution control,etc.

V. Policy Issues

We now turn to policy issues arisingfrom our analysis above. We shall ingeneral avoid paying much attention tothe policy positions taken by the “mor-alists” and the “fatalists” on corruption,even though it is sometimes tempting totake their side. The “moralists” empha-

10 Myrdal (1968, pp. 408–09) quotes Prime Min-ister Nehru: “Merely shouting from the house-tops that everybody is corrupt creates an atmos-phere of corruption. People feel they live in aclimate of corruption and they get corruptedthemselves.”


size that without fundamental changesin values and norms of honesty in publiclife—a kind of ethic cleansing throughactive moral reform campaigns—no bigdent in the corrosive effects of corrup-tion is likely to be achieved. The “fatal-ists” are more cynical, that we havereached a point of no return in manydeveloping countries, the corruption isso pervasive and well entrenched thatfor all practical purposes nothing muchcan be done about it. Our discussion inthe last section on the history-depen-dence of the high-corruption equilib-rium and the forces that tend to per-petuate it does point to the difficultiesof getting out of the rut, but there existsome examples of success in controllingcorruption even in the recent history ofdeveloping countries: Robert Klitgaard(1988) cites several examples, of whichthe cases of the Hong Kong Police De-partment and the Singapore Customsand Excise Department are the mostsuccessful, but in some sense the valiantefforts by one tax commissioner to fightpervasive corruption in the Bureau ofInternal Revenue and the substantialimpact he made in the 1970s in a hope-lessly corrupt country like the Philip-pines under Marcos provide the moststriking case. Without minimizing theimportance of moral exhortations inanti-corruption campaigns, our focushere will be on incentive structures thatmay induce even opportunists to foregocorrupt practices and the general prob-lems and prospects of implementingthem.

The first point that is commonlymade, no doubt with a great deal of jus-tification, is that regulations and bu-reaucratic allocation of scarce public re-sources breed corruption, and so theimmediate task is to get rid of them. Insome sense the simplest and the mostradical way of eliminating corruption isto legalize the activity that was formerly

prohibited or controlled. As Klitgaard(1988) notes, when Hong Kong legal-ized off-track betting, police corruptionfell significantly, and as Singapore al-lowed more imported products dutyfree, corruption in customs went down.Sometimes, however, turning over agovernment agency’s functions to themarket implies essentially a shift from apublic monopoly to a private monopoly,with a corresponding transfer of therent,11 but without much of an improve-ment in allocational efficiency (exceptthat due to a removal of the distortioncaused by secrecy discussed in SectionII).

While regulations designed primarilyto serve the patronage-dispensingpower of politicians and bureaucrats arecommon, there are many regulationswhich serve some other valued socialobjectives, and there may be a tradeoffbetween these objectives and that of re-ducing corruption through deregula-tion. Suppose a scarce but essentialconsumer good (like food) in a poorcountry is currently rationed by the gov-ernment so that the poor people canhave some access to it. The rations areadministered by corrupt officials. Whatwill be the welfare consequences forthe poor of replacing this system by themarket?12 To simplify, let us assumethat the government is the only sourceof food under the rationing scheme,that food obtained under ration cannotbe resold, and that corruption takes theform of the official charging a pricehigher than the stipulated ration price.In Figure 2 the ration price p is givenby the slope of AB and the consumer’sincome by OA. The ration, x–, is binding

11 The history of privatization in the last fewyears in many developing countries is replete withinstances of corrupt transfers to cronies of politi-cians.

12 For a taxonomic analysis of different cases for thisquestion, see Peter J. Gordon (1994). The idea of Figure2 is due to T. N. Srinivasan.


in the sense that the consumer with hisincome OA and ration price p wouldlike to buy x which is more than x–. Ifthere were no corruption his rationedconsumption equilibrium will be atsome point G on the line AB to the leftof D. Suppose the alternative nonra-tioned market equilibrium is given bypoint E, where the market price line,which is the slope of AC, is tangent toindifference curve I’, and the consump-tion is given by x’. As long as G is to theright of F (where the indifference curvethrough E intersects AB), the consumerprefers the uncorrupt ration scheme tothe market system. Now suppose thecorrupt official charges a price higherthan the ration price while distributinga stipulated total amount of food. Thebroken curve in Figure 1 is the locus ofpoints of tangency on the indifferencemap as the price line is rotated with Aas its focus starting at AB and converg-ing to the vertical axis. It is easy to seethat this locus is also the locus of con-sumption points to which the consumer

is driven to by the corrupt official as theration is reduced from x– to zero, be-cause at each point on the locus theslope of the indifference curve repre-sents the maximum price the consumeris willing to pay for the associated ra-tion. As long as the ration exceeds x’,the consumer will prefer the corrupt ra-tion scheme to the market system. Thebasic point is simple, although it can bemade with more complicated models,and should be brought to the attentionof those who in their zeal for deregula-tion and the market system with a viewto reducing corruption lose sight ofthe social objective that the regulationwas supposed to serve. (It is not anaccident that getting rid of the corruptpublic distribution systems in foodunder structural adjustment programsin developing countries has beenpolitically so unpopular.) In generalthe literature on corruption oftenoverlooks the distributional implica-tions of corruption (apart from notingthat the poor do not have the resources

Figure 2.

Income

0X X’ C X X B Food

I’

F

E

G

D

I


or the “connections” to be able to bribetheir way through).13

One way of reducing bureaucraticcorruption is to reduce the monopolypower of the bureaucrat when a clientfaces her in trying to get a licence orsome subsidy or transfer. Rose-Acker-man (1978) has suggested that, insteadof giving each official a clearly definedsphere of influence over which she hasmonopoly control, officials should begiven competing jurisdictions so that aclient who is not well-served by one of-ficial can go to another. When collusionamong several officials is difficult, com-petition will tend to drive the level ofbribes to zero. Of course, without anappropriate incentive payment system,this can encourage laziness in some offi-cials, because clients who are tired ofwaiting can turn to another official, in-stead of complaining to the official’s su-perior. Also, in cases of what Shleiferand Vishny (1993) call corruption withtheft, competitive pressure might in-crease theft from the government (in-cluding relaxation of minimum qualitystandards) at the same time as it re-duces bribes. So in such cases, competi-tion in the provision of government ser-vices has to be accompanied by moreintensive monitoring and auditing toprevent theft. Rose-Ackerman (1994)has suggested that multiple officialswith overlapping jurisdictions may alsohelp in such cases, because the poten-tial briber has to face the prospect of“persuading” all the officials involved,which raises costs and uncertainty for thecorrupt project. (It has been reportedthat in the United States the overlap-ping involvement of local, state, andfederal agencies in controlling illegaldrugs has reduced police corruption.)

In case of legitimate business projects,however, this raises the multiple vetopower problem discussed in Section II.

In some cases, on account of largefixed costs, indivisibilities, and coordi-nation problems, bureaucratic competi-tion through overlapping jurisdictions isnot feasible (nor desirable, if bargainingadvantages are to be pressed), as in thecase of large defense contracts or whenthe government buys in bulk in worldcommodity markets (say, in petroleum)or expensive single items like aircrafts.Not surprisingly, some of the major cor-ruption scandals in developing coun-tries (with substantial kickbacks fromforeign contractors) involve politiciansand bureaucrats in charge of such largeprocurement cases. On the bribe-givers’side it should be noted that when com-petition among the foreign contractorsis intense, very few governments of in-dustrially advanced countries discouragethe bribing of officials in the purchasingcountries (in fact tax-deductibility ofbribes by the companies often makes thetax-payers complicit in the payment ofsuch bribes). Even in the exceptionalcase of the U.S. where there is the 1977Foreign Corrupt Practices Act forbid-ding American companies from makingpayments to foreign officials, what aredescribed as “grease payments” to speedup transactions are not ruled out. (In fact,the 1988 amendments to the Act expandthe range of such payments allowed.)

Many countries launch periodic“spring-cleaning” through anticorrup-tion campaigns. How effective are they?It varies from situation to situation. Tobe effective, they have to be credibleand sustained. As suggested by the fre-quency-dependent equilibrium models,a critical mass of opportunist individu-als have to be convinced over a longenough period that corruption is not costeffective. But as has happened manytimes in the recent history of Africa or

13 In some cases the poor may not be completelyleft out. They get the rationed good after waitingin line (unless the good is extremely scarce), whilethe rich bribe to jump the queue.


China, anticorruption campaigns areusually ad hoc, and targeted at politicalenemies or at best at small fry, exempt-ing the big fish, or the important cro-nies and accomplices of the politicalrulers. Short-lived campaigns and re-peated amnesties to offenders (de-signed to wipe the slate clean) only in-crease the cynicism about the nextround and give out the wrong signals.As we have discussed in connectionwith Tirole’s (1996) intertemporal col-lective reputation model, trust takesseveral periods to reestablish itself. Whatis important is to institutionalize variouskinds of accountability mechanisms (likean independent office of public audit-ing, an election commission to limit andenforce rules on campaign contributionsin democratic elections, independentinvestigating agencies (like the HongKong Independent Committee againstCorruption which directly reports to theGovernor General), an office of localombudsman with some control over thebureaucracy, citizens’ watchdog commit-tees providing information and monitor-ing services and pursuing public-inter-est litigation, a vigorous and independent,even muckraking, press, less stringentlibel laws or laws protecting official se-crecy, etc.).14 For the watchdog com-mittees it is important not merely to un-earth and publicize egregious cases ofpublic corruption, but also to highlightcredible cases where the automatic andcynical presumption of the local peoplethat the officials are corrupt turns outto be gross exaggerations, thus cuttingdown on the feedback effects of rumorsand designs of middlemen.

Many other measures of reformwithin public administration have beensuggested: cutting down on the prolifer-ating functions of government depart-

ments (using vouchers and competitionwith private suppliers to serve a publicneed when customers can “vote” withtheir feet) and concentrating thesefunctions largely in areas where, on ac-count of elements of natural monopolyor a public good or quality standardsnot easily discernible to the custom-ers, a voucher plan is not an efficientway of providing the service; making su-pervisors answerable for gross acts ofmalfeasance by their subordinates; well-established procedures of encourag-ing “whistle-blowers” and guaranteeingtheir anonymity; authorization of peri-odic probing of ostensible but “unex-plainable assets” of officials; working inteams (for example, in Singapore cus-toms agents were asked to work inpairs) when lower bureaucrats face acustomer instead of one-on-one so thatthere is some check in the bargainingprocess (this is a simpler form of theoverlapping jurisdictions case discussedabove); well-defined career paths incivil service that are not dependent onthe incumbent politicians’ favor; peri-odic job rotation so that a bureaucratdoes not become too cosy with a cus-tomer over a long period; a more elabo-rate codification of civil service rulesreducing the official’s discretion ingranting favors; and so on. Of course, inmany of these cases one can also argueon the opposite side. Too many rulesrather than discretion may have the per-verse effect of providing opportunitiesfor corruption simply to circumventmindless inflexibilities. The practice offrequent job rotation may provide an in-centive to officials for maximum loot15

in the shortest possible time, discouragelearn

14 For a discussion of some of the “countervail-ing actions” that victims of corruption can under-take, see M. Shahid Alam (1995).

15 In Robert Wade’s (1985) case study in SouthIndia, an Executive Engineer in charge of irriga-tion may pay as bribe up to 14 times his annualsalary in order to obtain a two-year tenure at aparticular location. This suggests the lower boundof how much he expects to earn in bribes in twoyears.


ing on the job, and in general providethe politician (or the senior officer) aweapon to transfer an honest officialbent on rocking the boat of existing pa-tronage distribution. The opportunity toprobe the private finances of an officialis sometimes abused against rivals andpolitical opponents. Working on teams infacing a customer may sometimes en-courage unnecessary delays or collusionin demands for larger bribes. And so on.

VI. Incentive Payments for CivilServants

Let us now turn to the important pol-icy issue of an incentive pay structure inpublic administration that is often citedas one of the most effective ways offighting corruption. In imperial Chinaunder the Ch’ing dynasty district magis-trates were paid an extra allowancecalled yang-lien yin (“money to nourishhonesty”). Klitgaard (1988, p. 81) citesa quote from the historian Macaulay’saccount of Robert Clive’s attempt to re-duce the corruption rampant in theBritish East India Company in 1765:“Clive saw clearly that it was absurd togive men power, and to require them tolive in penury. He justly concluded thatno reform could be effectual whichshould not be coupled with a plan forliberally remunerating the civil servantsof the Company.” In recent times bothSingapore and Hong Kong have fol-lowed an incentive wage policy for pub-lic officials with a great deal of success.Current reforms in tax enforcement inmany countries, which include a bonusto the tax officer based on the amountof taxes he or she collects, have oftenbeen associated with significant im-provements in tax compliance (see, forexample, Dilip Mookerjee 1995). Insome cases (like in Singapore) a wagepremium above private sector salarieshas been found useful, consistent with

the efficiency wage theory. The poten-tial cost of job loss (including the wagepremium and seniority benefits) on de-tection may stiffen official resistance totemptation for corruption. Internationalagencies pushing for structural adjust-ment policies sometimes ignore that,while deregulation reduces opportuni-ties for corruption, another part of thesame policy package aimed at drastic re-ductions of public spending may resultin lower real wages for civil servants in-creasing their motivation for corrup-tion. One should also keep in mind thatwhen today’s rich countries had beatenthe worst of corruption in their history,the average salary of an official wasmany times that of what obtains in mostpoor countries.

While the argument for incentivepayment is clear, the relationship be-tween public compensation policy andcorruption can sometimes be quitecomplex. This is because our objectiveis not merely to reduce corruption in anofficial agency but, at the same time,not to harm the objective for which theagency was deployed in the first place.Much of the theory of rent-seeking doesnot worry about this, because the pre-sumption often is that government isnothing but organized theft and the lessof it the better. But, as we have alreadyseen in the case of rationed distributionof food to the poor, if we have anothervalued social objective, there may becases where the corrupt administeredsystem is preferable to the market. Weshall now discuss the compensation pol-icy for corruptible enforcers of a regula-tion when the latter has a valued socialpurpose. Let us take, for example, thecase of public inspectors charged withmonitoring pollution from a factory. Weshall follow the theoretical model ofMookherjee and I. P. L. Png (1995) tounderstand the nature of the tradeoffamong corruption, pollution, and en-


forcement effort and consider the con-sequences of strategic interaction be-tween the polluting factory and the cor-ruptible inspector.

Suppose the regulator can directlycontrol neither the inspector’s monitor-ing effort nor his underreporting of thefactory’s pollution for which he getsbribed, a double moral hazard problemin a principal-agent model. The regula-tor has three instruments: a rate of re-ward r for the inspector (a percentagecommission based on the fines for pol-lution collected from the factory), apenalty p (depending on the amount ofunderreporting of pollution) on the in-spector when corruption is discovered,and a penalty q (a mark-up over theusual fine for the evaded pollution ) onthe factory for bribing the inspector.The probability that the inspector willunearth the factory’s true pollutionlevel (assuming, of course, that he willnot overreport) depends on the moni-toring effort exerted by the inspector.There is also an exogenous probabilitythat the inspector’s underreporting andthe bribe paid are discovered by theregulator. Given the regulator’s policypackage (r, p, q), the factory and theinspector simultaneously choose thepollution level and the monitoring in-tensity respectively. The two parties(assumed risk-neutral) then jointly de-termine the bribe, if any, as part of aNash bargaining solution.

Suppose the factory has polluted andthe inspector has found out about it. Ifbribery is going on, then small increasesin r or p may merely raise the level ofthe bribe: a compensation policywhereby the larger reward for the in-spector or a higher penalty for taking abribe, raises the cost borne by the in-spector for underreporting pollution,and so the inspector demands and re-ceives a larger bribe, and corruption in-creases. Mookherjee and Png show that

it takes a sufficiently large, discrete, in-crease in the reward or the penalty toeliminate corruption (when the inspec-tor’s demand for bribe rises beyond thefactory’s willingness to pay). One way toreduce the bribe, however, is to raise q,the penalty on the bribe-giver (makingbribing more costly for him), while re-ducing the penalty p for the bribe-taker(so that the latter does not demand alarger bribe): this contrasts with thetypical practice of punishing bribe-giv-ers less severely than bribe-takers.

What effect does the compensationpolicy have from the point of view ofthe primary objective of regulating pol-lution? A small increase in the rewardrate r, by raising the bribe and hencethe price of pollution will lower the in-centive for the factory to pollute. Thelarger bribe will increase the inspector’sincentive to monitor, further deterringthe factory. The reduction in pollution,on the other hand, will discourage theinspector from monitoring. In equilib-rium the net effect is to reduce pollu-tion. By contrast, when the regulatorraises the penalty rate p on the inspec-tor, this will reduce his incentive tomonitor; the reduction in monitoringcan reduce the expected penalty forpollution for the factory, and hence theresult may be more pollution. Thus al-though the inspector is risk-neutral, thecarrot (reward for reporting pollution)and the stick (penalty for taking abribe) can have opposite effects on thelevel of pollution. All this is not to dis-courage a suitable incentive paymentsystem in the context of corruption butto point to the nature of complexitiesinvolved.16 The analysis also suggests

thatth

16 Timothy Besley and John McLaren (1993)show that in the case of tax collectors hetero-genous in their corruptibility, the revenue author-ity may sometimes prefer a regime of corruptionamong the tax collectors than paying them effi-ciency wages.


e reward system should be more gearedto the incidence of the primary harmthat the regulator is supposed to con-trol. (This indicates that in the case ofcontrolling corruption in the Customsdepartment the value of paying rewardsto customs officials should be assessedby their effect on the open-marketprice of the product subject to importcontrols.)

Finally, policy issues on corruptioncannot be discussed without involvingthe larger question of the nature of thestate that is supposed to carry out thepolicies. This is too large a topic to becovered here,17 but one may neverthe-less point out that to assume that allstates are predatory, as is customary inmuch of the public choice literature inthe context of developing countries,does not help in understanding why cor-ruption is more in some countries thanin others (even with similar extent ofstate intervention), and why countrieswith similar over-all levels of corruptiondiffer in its effect on productivity andgrowth. We have noted in Section IIIthat political competition can reducecorruption (unless the transaction costsin the political market, in the form, say,of campaign finances, are too large),but what is particularly important in de-ciding the economic consequences ofcorruption is the extent of centraliza-tion in the rent-collection machinery.Weak and fragmented governments(even under authoritarian rulers) withrampant economic warlordism can letloose a regime of decentralized lootingthat is particularly harmful for staticand dynamic efficiency.

Some African states in recent historybecame predatory in their rent-extrac-tion not because they were strong, butbecause they were weak: the state could

not enforce the laws and property rightsthat provide the minimum underpin-nings of a market economy and thus lostrespect; disrespect quickly led to disloy-alty and thievery among public officials.The strong states of East Asia with theircentralized rent-collection machineryand their dense “encompassing” net-work with business interests stand insharp contrast, even though by somemeasures corruption has been quitesubstantial. As we have emphasized inour discussion of “lump-sum corrup-tion” in Section II, the ability to pre-commit credibly may have been an im-portant feature of the “strength” ofsuch states. This is not to deny that get-ting rid of many of the dysfunctionalregulations remains a major first step inanti-corruption policy, whatever the na-ture of the state. In addition, it is im-perative to institutionalize the variouskinds of accountability mechanisms atdifferent levels of the government (thatwe have briefly discussed in Section V)as part of the agenda for any meaning-ful policy reform in this context.

REFERENCES

ADAMS, GORDON. The iron triangle: The politicsof defense contracting. New Brunswick: Trans-action Books, 1981.

ALAM, M. SHAHID. “A Theory of Limits on Cor-ruption and some Applications,” Kyklos, 1995,46(43), pp. 419–35.

ANDVIG, JENS CHRISTOPHER. “The Economics ofCorruption: A Survey,” Studi Economici, 1991,43, pp. 57–94.

ANDVIG, JENS CHR. AND MOENE, KARL, OVE.“How Corruption May Corrupt,” J. Econ. Be-hav. Organ., Jan. 1990, 13(1), pp. 63–76.

BANERJEE, ABHIJIT. “A Theory of Misgover-nance.” Working Paper. MIT Economics Dept.,1994.

BANFIELD, EDWARD C. The moral basis of abackward society. New York: Free Press, 1958.

BARDHAN, PRANAB K. “The Nature of Institu-tional Impediments to Economic Develop-ment.” Working Paper. Dept. of Economics, U.of California at Berkeley, Mar. 1996.

BECK, PAUL J. AND MAHER, MICHAEL W. “AComparison of Bribery and Bidding in ThinMarkets,” Econ.Letters, 1986, 20, pp. 1–5.

BESLEY, TIMOTHY AND MCLAREN, JOHN. “Taxes17 For a discussion of the role of the state in the

context of development, see Bardhan (1996).


and Bribery: The Role of Wage Incentives,”Econ. J., Jan. 1993, 103(416), pp. 119–41.

BOYCKO, MAXIM; SHLEIFER, ANDREI ANDVISHNY, ROBERT. Privatizing Russia. Cam-bridge MA: MIT Press, 1995.

CADOT, OLIVIER. “Corruption as a Gamble,” J.Public Econ., July 1987, 33(2), pp. 223–44.

GAMBETTA, DIEGO. “Mafia: The Price of Dis-trust,” in Trust making and breaking coopera-tive relations. Ed.: DIEGO GAMBETTA. Oxford:Blackwell, 1988, pp. 158–75.

GORDON, PETER J. “Welfare of the Poor Given aCorrupt Non-Market System.” Working Paper.U. of West Indies, Jamaica, 1994.

HILLMAN, ARYE L. AND KATZ, ELIAKIM, “Hierar-chical Structure and the Social Costs of Bribesand Transfers,” J. Public Econ., Nov. 1987,34(2), pp. 129–42.

HUNTINGTON, SAMUEL P. Political order inchanging societies. New Haven: Yale U. Press,1968.

KANGLE, R. P. The Kautiliya Arthasastra, Part II.Bombay: U. of Bombay, 1972.

KLITGAARD, ROBERT E. Controlling corruption.Berkeley: U. of California Press, 1988.

KRUEGER, ANNE O. “The Political Economy ofthe Rent-Seeking Society,” Amer. Econ. Rev.,June 1974, 64(3), pp. 291–303.

LEFF, NATHANIEL H. “Economic Developmentthrough Bureaucratic Corruption,” The Ameri-can Behavioral Scientist, Nov. 1964, 8(2), pp.8–14.

LIEN, DONALD H. D. “A Note on CompetitiveBribery Games,” Econ. Letters, 1986, 22, pp.337–41.

LUI, FRANCIS T. “An Equilibrium Queuing Modelof Bribery,” J. Polit. Econ., Aug. 1985, 93(4),pp. 760–81.

MAURO, PAOLO. “Corruption and Growth,” Quart.J. Econ., Aug. 1995, 11093), pp. 681–712.

———. “The Effects of Corruption on Growth,Investment, and Government Expenditure,” inCorruption in the world economy. Ed.: KIM-BERLY A. ELLIOTT. Washington DC: Institutefor International Economics, forthcoming.

MOOKHERJEE, DILIP. “Reforms in Income TaxEnforcement in Mexico.” IRIS-India WorkingPaper No.6. U. of Maryland, 1995.

MOOKHERJEE, DILIP AND PNG, I. P. L. “Corrupt-ible Law Enforcers: How Should They be Com-pensated?” Econ. J., Jan. 1995, 105(428), pp.145–59.

MURPHY, KEVIN M.; SHLEIFER, ANDREI ANDVISHNY, ROBERT W. “Why Is Rent-Seeking SoCostly to Growth?” Amer. Econ. Rev., May1993, 83(2), pp. 409–14.

MYRDAL, GUNNAR. Asian drama. Vol. II. NewYork: Random House, 1968.

OLDENBURG, PHILIP. “Middlemen in Third-World Corruption,” World Politics, 1987, 39,pp. 508–35.

OLSON, MANCUR. “Dictatorship, Democracy, andDevelopment,” Amer. Polit. Sci. Rev., Sept.1993, 87(3), pp. 567–75.

PUTNAM, ROBERT D. Making democracy work:Civic traditions in modern Italy. Princeton:Princeton U. Press, 1993.

RASMUSEN, ERIC AND RAMSEYER, J. MARK.“Cheap Bribes and the Corruption Ban: ACoordination Game among Rational Legisla-tors,” Public Choice, Mar. 1994, 78(3–4), pp.305–27.

ROMER, PAUL. “New Goods, Old Theory, and theWelfare Costs of Trade Restrictions,” J. De-velop. Econ., Feb. 1994, 43(1), pp. 5–38.

ROSE-ACKERMAN, SUSAN. Corruption: A study inpolitical economy. New York: Academic Press,1978.

———. “Reducing Bribery in the Public Sector,”in Corruption & democracy: Political institu-tions, processes and corruption in trasitionstates in East-Central Europe and in the formerSoviet Union. Ed.: DUC V. TRANG. Budapest:Institute for Constitutional & Legislative Policy,1994.

———. “When is Corruption Harmful?” WorkingPaper. The World Bank, 1996.

SAH, RAAJ KUMAR. “Persistence and Pervasivenessof Corruption: New Perspectives.” Yale Eco-nomic Growth Center Discussion Paper: 560,48, Aug. 1988.

SHLEIFER, ANDREI AND VISHNY, ROBERT W.“Corruption,” Quart. J. Econ., Aug. 1993,108(3), pp. 599–617.

THEOBALD, ROBIN. Corruption, development, andunderdevelopment. Durham, NC: Duke U.Press, 1990.

TIROLE, JEAN. “ A Theory of Collective Reputa-tions,” Rev. Econ. Stud., Jan. 1996, 63(1), pp.1–22.

TULLOCK, GORDON. “Rent-Seeking as a Negative-Sum Game,” in Toward a theory of the rent-seeking society. Eds.: JAMES M. BUCHANAN,ROBERT D. TOLLISON, AND GORDON TUL-LOCK. College Station: Texas A&M U. Press,1980, pp. 16–36.

–––––– “The Costs of Special Privilege,” in Per-spectives on positive political economy. Eds.:JAMES E. ALT AND KENNETH A. SHEPSLE.Cambridge and New York: Cambridge U. Press,1990, pp. 195–211.

WADE, ROBERT. “The Market for Public Office:Why the Indian State Is Not Better at Develop-ment,” World Devel., Apr. 1985, 13(4), pp. 467–97.

WRAITH, RONALD E. AND SIMKINS, EDGAR. Cor-ruption in developing countries. London: Allenand Unwin, 1963.

YANG, MAYFAIR MEI-HUI. “The Gift Economyand State Power in China,” Comparative Stud-ies in Society and History, Jan. 1989, 31(1), pp.25–54.


TABLE 1

Country in descending order of

GNP per capita

Index of lowcorruption

(1)

Index of lowred tape

(2)

Index of efficiencyof the legal system

(3)

Efficiency of thegovernance

structure, averageof (1)−(3)

Switzerland 10 10 10 10 Japan 8.75 8.5 10 9.08Sweden 9.25 8.5 10 9.25Denmark 9.25 9.5 10 9.58Norway 10 9 10 9.67United States 10 9.25 10 9.75Germany 9.5 7.5 9 8.67Austria 8 7.25 9.5 8.25France 10 6.75 8 8.25Finland 9.5 8.5 10 9.33Belgium 9.75 8 9.5 9.08Canada 10 9.5 9.25 9.58Netherlands 10 10 10 10 Italy 7.5 4.75 6.75 6.33United Kingdom 9.25 7.75 10 9 Australia 10 9.25 10 9.75Singapore 10 10 10 10 Hong Kong 8 9.75 10 9.25Kuwait 7.75 6.25 7.5 7.17Spain 7 6 6.25 6.42Israel 9.25 7.5 10 8.92New Zealand 10 10 10 10 Ireland 9.75 7.5 8.75 8.67Saudi Arabia 4.75 5.25 6 5.33Portugal 6.75 4.5 5.5 5.58Greece 6.25 4 7 5.75Korea 5.75 6.5 6 6.08Argentina 7.66 6.66 6 6.77Trinidad/Tobago 6.5 4 8 6.17Mexico 3.25 5.25 6 4.83Uruguay 8 6 6.5 6.83Venezuela 5.75 4 6.5 5.42Malaysia 6 6 9 7 Brazil 5.75 4 5.75 5.17South Africa 8 7 6 7 Chile 9.25 9.25 7.25 8.58Iraq 10 3 6 6.33Panama 5 7.25 6.75 6.33Iran 3.25 1.25 2 2.17Turkey 6 5.33 4 5.11

Appendix


TABLE 1 (Cont.)

Country in descending order of

GNP per capita

Index of lowcorruption

(1)

Index of lowred tape

(2)

Index of efficiencyof the legal system

(3)

Efficiency of thegovernance

structure, averageof (1)−(3)

Thailand 1.5 3.25 3.25 2.56Algeria 5 2.5 7.25 4.92Jamaica 5 4 7.33 5.44Colombia 4.5 4.5 7.25 5.42Jordan 8.33 6.33 8.66 7.77Ecuador 5.5 5 6.25 5.58Dominican Rep. 6.5 6 6.75 6.42Morocco 5.66 5.33 6.66 5.88Peru 7.25 5.75 6.75 6.58Cameroon 7 6 7 6.67Philippines 4.5 5 4.75 4.75Ivory Coast 6 7.75 6.5 6.75Indonesia 1.5 2.75 2.5 2.25Egypt 3.25 3 6.5 4.25Zimbabwe 8.75 7.75 7.5 8 Sri Lanka 7 6 7 6.67Ghana 3.66 2.33 4.66 3.55Pakistan 4 4 5 4.33Nicaragua 8.75 4 6 6.25Nigeria 3 2.75 7.25 4.33Kenya 4.5 5 5.75 5.08India 5.25 3.25 8 5.5 Haiti 2 2 2 2 Bangladesh 4 4 6 4.67Liberia 2.66 5 3.33 3.66Angola 8.66 5.33 4 6 Zaire 1 2.66 2 1.89

Source: Mauro (1995). The scale is 10 for no corruption to 0 for maximum corruption. The indices are based onstandard questionnaires filled in by Business International correspondents stationed in about 70 countries in1980−83. It is likely that the indices reflect what are faced by foreign businessmen in a country, not necessarily whatown citizens face.


TABLE 2PERCEIVED CORRUPTION RANKING, 1996

Rank Country Mean Score Variance No. of Surveys Used

1 New Zealand 9.4 0.39 6 2 Denmark 9.3 0.44 6 3 Sweden 9.1 0.30 6 4 Finland 9.1 0.23 6 5 Canada 9.0 0.15 6 6 Norway 8.9 0.20 6 7 Singapore 8.8 2.36 10 8 Switzerland 8.8 0.24 6 9 Netherlands 8.7 0.25 610 Australia 8.6 0.48 611 Ireland 8.5 0.44 612 U.K. 8.4 0.25 713 Germany 8.3 0.53 614 Israel 7.7 1.41 515 USA 7.7 0.19 716 Austria 7.6 0.41 617 Japan 7.1 2.61 918 Hong Kong 7.0 1.79 919 France 7.0 1.58 620 Belgium 6.8 1.41 621 Chile 6.8 2.53 722 Portugal 6.5 1.17 623 South Africa 5.7 3.30 624 Poland 5.6 3.63 425 Czech Rep. 5.4 2.11 426 Malaysia 5.3 0.13 927 South Korea 5.0 2.30 928 Greece 5.0 3.37 629 Taiwan 5.0 0.87 930 Jordan 4.9 0.17 431 Hungary 4.9 2.19 632 Spain 4.3 2.48 633 Turkey 3.5 0.30 634 Italy 3.4 4.78 635 Argentina 3.4 0.54 636 Bolivia 3.4 0.64 437 Thailand 3.3 1.24 1038 Mexico 3.3 0.22 739 Ecuador 3.2 0.42 440 Brazil 3.0 1.07 741 Egypt 2.8 6.64 442 Colombia 2.7 2.41 643 Uganda 2.7 8.72 444 Philippines 2.7 0.49 8


TABLE 2 (Cont.)PERCEIVED CORRUPTION RANKING, 1996

Rank Country Mean Score Variance No. of Surveys Used

45 Indonesia 2.7 0.95 1046 India 2.6 0.12 947 Russia 2.6 0.94 548 Venezuela 2.5 0.40 749 Cameroon 2.5 2.98 450 China 2.4 0.52 951 Bangladesh 2.3 1.57 452 Kenya 2.2 3.69 453 Pakistan 1.0 2.52 554 Nigeria 0.7 6.37 4

Source: Transparency International. The perceived corruption score is an average of data from several surveys, therespondents of which are mostly people working for multinational firms and institutions: it is a “poll of polls” (withthe variance in the score from different surveys given in column 4). The scale is 10 for no corruption to 0 formaximum corruption.


Corruption and development

Documents

Transcript of Corruption and development