ALFRED MARSHALL LECTURE KLEPTOCRACY ANDDIVIDE-AND-RULE: AMODEL OFPERSONAL RULE

DaronAcemoglu Thierry Verdier MassachusettsInstitute of Technology DELTA-ENS JamesA. Robinson Universityof California, Berkeley

Abstract Manydeveloping countries have suffered under the personal rule of kleptocrats, who implementhighly inefŽ cient economic policies, expropriate the wealth of their citizens, and usethe proceeds for their own gloriŽ cation or consumption. We argue that the success of kleptocratsrests, in part,on theirability to usea divide-and-rulestrategy, made possible by theweakness of institutionsin these societies. Members of society need to cooperate in order todepose a kleptocrat,yet such cooperation may be defused by imposing punitive rates of taxationon any citizen who proposes such a move,and redistributing the beneŽ ts to those whoneed to agreeto it. Thus the collective action problem can be intensiŽed bythreatswhich remainoff the equilibrium path. In equilibrium, all are exploited and no onechallenges the kleptocrat.Kleptocratic policies are more likely when foreign aid and rents from natural resourcesprovide rulers with substantial resources to buy off opponents; when opposition groupsare shortsighted; when the average productivity in the economy is low;and when there isgreater inequality between producer groups (because more productive groups are more difŽcult to buy off). (JEL: O12, H00)

1. Introduction

Many countries in Africaand the Caribbean sufferunder “kleptocratic”regimes, wherethe state is controlled and run forthe beneŽt of anindividual, or asmall group, who use their power to transfera large fraction of society’s resources to themselves. Examples of kleptocratic regimesinclude the DemocraticRepublic of the Congo (Zaire)under Mobutu Sese Seko, the Dominican Republic under RafaelTrujillo, Haiti under the Duvaliers, Nicaragua under the Somozas, Uganda under Idi Amin, Liberiaunder Charles Taylor, and the Philippines

Acknowledgments: Paper presented as the Marshall Lecture at the European Economic Associ- ation’s Annual meetings in Stockholm, August 24, 2003. We thank Silje Aslaksen for pointing out an algebraic mistake in the Žrst version and Alexander Debs for excellent research assistance. E-mail addresses: Acemoglu: [email protected]; Robinson: [email protected]; Verdier: [email protected]

Journal of the European Economic Association April–May 2004 2(2–3):162– 192 ©2004 by the European Economic Association Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 163 under Ferdinand Marcos. Inall these cases, kleptocratic regimesappear to have been disastrous foreconomic performanceand caused the impoverishment of the citizens. Astudy of the political economy of such regimesmust depart fromthe standard presumptions of most researchin economics and political science, which assume that rulers makechoices within strongly institutionalized poli- ties.1 In these polities, formalpolitical institutions, such as the constitution, the structure of the legislature, or electoralrules, place constraints on the behavior of politicians and political elites, and directly in  uence political outcomes. In contrast, kleptocracy emergesin weakly institutionalized polities, whereformal institutions neither place signi Ž cant restrictions on politicians ’ actions nor make them accountable to citizens. While the academicstudy of strongly institutionalized polities is well advanced (e.g., Shepsle and Weingast 1995; Cox 1997; Persson and Tabellini 2000, 2003), there arefew studies, and less of aconsensus, on the nature of weakly institutionalized polities. What determines corruption, rent extraction and bad policies when institutions areweak? Indeed, the qualitative nature of politics appears to differmarkedly between strongly and weakly institutional- ized polities: when institutions arestrong, citizens punish politicians by voting them out of power; when institutions areweak, politicians punish citizens who failto support them. When institutions arestrong, politicians vie forthe support and endorsement of interest groups; when institutions areweak, politicians createand control interest groups. When institutions arestrong, citizens demand rights; when institutions areweak, citizens beg forfavors. The researchprogram proposed in this paper is asystematic study of policymaking in weakly institutionalized societies, and ultimately, astudy of the process by which strongly institutionalized societies emerge(see North and Weingast (1989) fora classic account of such aprocess). Totake a Ž rst step in this program, weconstruct amodel to study kleptocratic politics. Perhaps the most puzzling featureof kleptocracies, illustrated by the examples fromthe Congo, the Dominican Republic, Nicaragua, or Haiti, is their longevity, despite the disastrous policies pursued by the rulers. 2 This longevity is madeeven more paradoxical by the factthat such regimesapparently lacked apolitical base (a core constituency) that supported them. Despite the absence of formalinstitu- tional mechanisms fordeposing unpopular rulers, constraints on the behavior of rulers exist even in weakly-institutionalized societies (e.g., the threat of revo-

1. We owe this terminology and the distinction between strongly and weakly institutionalized polities to Robert Powell. 2. Despite the appealing intuition, perhaps derived from many political analyses of strongly institutionalized societies, that rulers who reduce the utility of citizens ought to be replaced, this appears not to be the case in weakly-institutionalized societies. Many kleptocrats rule for long periods, Mobutu for 32 years, Trujillo for 31 years, and the Somozas for 42 years. 164 Journal of the European Economic Association lution, or competition fromother strongmen). Why do, then, the heavily taxed producers or the poverty-stricken citizens not replacethe kleptocrat? Why do they rarelyform an effectiveopposition constraining the kleptocrat? How can a regimethat apparently bene Ž ts nobody outside of the narrowest of cliques survive? Our basic answer is that this is because the absence of strong institu- tions allows rulers to adopt political strategies which arehighly effectiveat defusing any opposition to their regime. The seminal book by Robert Bates, Markets and States inTropical Africa, provides many clues towards an answer. Bates described the web of inef Ž cient transfersand policies in effectin many parts of Africa,but most notably in Ghana and Zambia,and suggested the following logic: many of these inef Ž cient policies arein place to transferresources fromthe population to the ruling groups, while atthe sametime ensuring their political survival. Inparticular, the nexus of inef Ž cient policies appeared to be useful forcreating anenvironment whereany group that becamepolitically mobilized against the rulers could be punished, while those that remained loyal wererewarded. With this logic, the Ghanaian government heavily taxed cocoa producers, while atthe sametime subsidizing their inputs of seeds and fertilizers.The subsidies could be allocated selectively as apotential rewardfor not attempting to change the status quo. Similarly, the exchange ratewas kept overvalued because then the government could allocate or withhold valuable rations of foreign exchange in order to guarantee support. In this paper, wesuggest ageneralization of this reasoning, which wedub the divide-and-rule strategy. Divide-and-rule is amethod used by kleptocrats to maintain power in weakly-institutionalized polities while simultaneously pur- suing policies costly to society. The logic of the divide-and-rule strategy is to enable arulerto bribe politically pivotal groups offthe equilibrium path, ensuring that he can remainin power against challenges. To remove arulerfrom power requires the cooperation of distinct social groups which is madedif Ž cult by the collective action problem (Olson 1965). By providing selective incentives and punishments, the divide-and-rule strategy exploits the fragility of social cooperation: when facedwith the threat of being ousted, the kleptocratic ruler intensiŽ es the collective action problem and destroys the coalition against him by bribing the pivotal groups. Moreexplicitly, weconsider adynamic gamebetween the rulerand two producer groups. The kleptocratic rulertaxes production and uses the resulting tax revenue, the rents fromnatural resources and potential foreign aid from outside donors forhis own consumption. The two producer groups, ifthey can cooperate, can remove the rulerfrom power and establish democracy (aregime morefavorable to their interests). We model this cooperation asfollows: one of the two groups (the proposer) makes aproposal to remove the rulerfrom power, and ifthe other group (the proposed) agrees, the ruleris removed and democracy Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 165 is established. The ruler-friendly political institutions, however, imply that before the proposed group responds to the proposal, the rulercan makea counteroffer. This counteroffer enables him to use adivide-and-rule strategy: following achallenge, the ruleruses all his resources and tax revenues to bribe the proposed group (and compensate them forfuture higher taxes ifthey turn down the proposal and keep the rulerin power). Ifhe can do so successfully, he can Ž ght offthe challenge, and anticipating this, no group will challenge the ruler. Therefore,the divide-and-rule strategy remainsoff the equilibrium path, and its anticipation implies that the rulercan follow highly distortionary (klep- tocratic)policies without being challenged. Not only is the kleptocrat able to stay in power, but the threat of divide-and-rule implies that there will be no challenges to remove him frompower along the equilibrium path. 3 In addition to providing an answer to the question of how kleptocratic regimessurvive and formalizing the idea of divide-and-rule, the model also implies anumber of interesting comparative static results. These results stemnot fromvarying the formalpolitical institutions, but fromhow the feasibility of divide-and-rule depends on the socioeconomic environment. Inparticular, they follow directly fromthe logic of divide-and-rule: Kleptocracy survives if, offthe equilibrium path, it can successfully bribe the pivotal group. Consequently, kleptocracy is morelikely: (a)when producer groups or their political repre- sentatives aremore shortsighted, because they put less weight on future returns under democracy; (b)when there is moreforeign aid tothe ruler, which he can use to bribe pivotal groups; (c)when there aregreater natural resource rents that can be used to bribe pivotal groups; 4 (d)when average productivity in the economy is low (loosely speaking, because this makes natural resource rents and funds fromforeign aid moreeffective instruments to bribe the pivotal group); 5 (e)when there is less inequality between producer groups. This is because the main challenge to the rulercomes fromthe moreproductive group, which has moreto gain fromdemocracy. Ifinequality increases, it becomes moredif Ž cult forthe rulerto sustain his kleptocratic policies. These comparative static results areuseful inthinking about the experiences of anumber of Africanand Caribbean countries. First, kleptocratic regimes seemto be morecommon in natural-resource-rich countries such as the Congo,

3. We conjecture that these results extend to an environment with uncertainty, where, along the lines of the analysis in Green and Porter (1984) and Abreu, Pearce, and Stacchetti (1990), punishment will sometimes occur along the equilibrium path. Obviously, most of the examples in the case study literature feature situations where divide-and-rule is actually observed in practice. 4. However, this result requires producer groups not to be too patient. Otherwise, the prospect of obtaining some of those natural resource rents in the future, once the ruler is deposed, may make it sufŽ ciently attractive to oust the ruler, making kleptocratic policies harder to sustain. 5. However, as with the comparative static with respect to natural resource rents, the effect of average productivity on policies depends on the discount rate of the citizens: if citizens are very impatient, greater average productivity may make kleptocracy more likely. 166 Journal of the European Economic Association

Liberia, SierraLeone, and Nigeria, which is consistent with our result that, aslong asproducer groups arenot too patient, greaternatural resource rents facilitatekleptocrac y. For example, Ross (2001) shows that, other things equal, the presence of oil increases the probability that acountry will be a dictatorship. 6 Second, many accounts of the worst kleptocracies also illustrate how the rulerwas able to use foreign aid (aswell asrents fromnatural resources) to stay in power and Ž ght offchallenges. The case of Mobutu, which wediscuss in the next section, is especially pertinent here. This negative role of foreign aid on equilibrium policies might also explain the common Ž nding in the empirical growth literaturethat foreign aid seemsto have little positive effecton the rate of economic growth in less-developed countries, and in fact,it might even have anegative effecton non-democracies (e.g., Burnside and Dollar 2000; Easterly, Levine, and Roodman 2003). In addition there is alarge amount of anecdotal evidence (e.g., Dollar and Pritchett 1998; Brautigam 2000; Easterly 2001; van der Walle 2001), and some tentative econometric evidence (Knack 2000) that foreign aid leads to worse governance. The comparative static result with respect to inequality isalso interesting. One interpretation of this result is that kleptocracy is less likely to arisewhen there is abalance of power between the rulerand apowerful producer group in society. This is consistent with the experiences of anumber of moresuccessful economies insub-Saharan Africa,such asBotswana and Mauritius, which have both managed to become stable democracies and refrainfrom the most distor- tionary policies. Apossible explanation is the political power of majorproducer groups, such as the cattleowners in Botswana and the sugar planters in Mauritius, which has placed realconstraints on the behavior of political elites (see, forexample, Leith (2000) and Acemoglu, Johnson, and Robinson (2003) on Botswana and Bowman (1991) on Mauritius). Our result is also consistent with Bates’ emphasis that economic policies wereless distortionary in Kenya than in Ghana, because large landowners in Kenya werebetter able tosolve the collective action problem and actas an important forcein politics. The inter- pretation suggested by our model is that, in Kenya, the presence of asocial class with sufŽ cient wealth and power (thus greaterinequality among producers)

6. Moreover, the case study evidence also suggests that whether or not resources are acurse depends crucially on institutions (see Ross 1999 for asurvey). Similarly, Mehlum, Moene, and Torvik (2001) show that the negative effects of resources pointed out by Sachs and Warner (1995) are conditional on institutions —only countries with poor institutions, speci Ž cally weak constraints on the executive, suffer aresource curse. In our model this follows from the fact that once political institutions do not constrain rulers, natural resources help to facilitate divide-and-rule (in this context, see also Robinson, Torvik, and Verdier (2002) for arelated argument that when institu- tions are weak, resource booms are acurse because they intensify incentives for regimes to use inefŽ cient forms of redistribution to stay in power). Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 167 counterbalanced the power of the ruling party KANU(Kenya AfricanNational Union) under Jomo Kenyatta. 7 In addition to these comparative static results, weshould emphasize that kleptocratic rule can be sustained because of two featuresof our political game: (1)the ability of the rulerto makea counteroffer afterthe proposal of one of the groups; (2)the ability of the rulerto charge different tax ratesand make differential transfers to the two groups. The Ž rst featureis plausible in a weakly-institutionalized society wherepoliticians can exploit their power to consolidate their rule. The second featureis also plausible in the context of many less developed countries whereethnic, geographic and economic divisions enable policies with very differentdistributional consequences forvarious groups to be implemented. In the next section, weprovide abriefdiscussion of how these divisions wereimportant in the survival of some kleptocratic regimes. This paper is related to anumber of differentliteratures. As wenoted, weakly-institutionalized polities have received little attention by political econ- omists. For instance, the political economy literatureon dictatorships is very underdeveloped. Most models aresimilar to ours in that they formalizedicta- torship as the maximization of anindividual ’s or group’sutility function subject to the constraint of staying in power (e.g., North 1981; Levi 1988; Grossman 1991; Grossman and Noh 1994; McGuireand Olson 1996; Robinson 1998; Wintrobe 1998; Acemoglu and Robinson 2000b; LaFerrara and Bates 2001; Bueno de Mesquita etal. 2003; Acemoglu 2003). 8 However, these models do not capture the notion that the nature of politics under weak institutions is different than in strongly institutionalized polities. Our model develops this idea and shows how this can lead to the divide-and-rule strategy in weakly institu- tionalized societies, and how divide-and-rule leads to very differentcomparative static results. Political scientists often distinguish between different types of autocracies or dictatorships. For example, the seminal study of Linz and Stepan (1996) posits the existence of four types: totalitarian, posttotalitarian, authoritarian, and sultanistic. Sultanism, which is close towhat wemean by kleptocracy, is also equivalent to other concepts used by political scientists. Scholars working in Africanpolitics, forinstance, referto this asneopatrimonialism, personal rule, or prebendalism. In the words of Bratton and van der Walle (1997, p. 62):

7. In this context, it is also interesting that in the most extreme examples of kleptocracy in the Americas, the Dominican Republic, Haiti and Nicaragua, revolution and U.S.intervention de- stroyed traditional elites that might have blocked the power of kleptocrats. 8. There are interesting comparative static results from the literature on dictatorships. Existing ideas suggest that better economic policies are chosen by dictators with long-time horizons (Levi 1988, Grossman and Noh 1994), who are encompassing in the sense that alarge fraction of income accrues to them (McGuire and Olson 1996), who do not fear losing their political power (Robinson 1998; Acemoglu and Robinson 2000b), or who have to gain the support of alarge “winning coalition” (Bueno de Mesquita et al. 2003). See also Acemoglu and Robinson (2000b, 2004) for aframework to analyze transitions from dictatorship to democracy. 168 Journal of the European Economic Association

the right to rule in neopatrimonial regimesis ascribed toaperson ratherthan to an ofŽ ce, despite the of Ž cialexistence of awritten constitution. One individual ...often apresident forlife, dominates the state apparatus and stands above its laws. Relationships of loyalty and dependence pervade a formalpolitical and administrative system, and of Ž cials occupy bureaucratic positions less to performpublic service ...than to acquire personal wealth and status. Although state functionaries receivean of Ž cialsalary, they also enjoy access to various formsof illicit rents, prebends, and petty corruption, which constitute ...an entitlement of of Ž ce. The chief executive and his inner circleundermine the effectiveness of the nominally modern state administration by using it forsystematic patronage and clientelist practices in order tomaintain political order.

That such neopatrimonialism is atthe heart of Africa ’seconomic crisis is the conventional wisdom in political science (seeSandbrook 1985; Herbst 2000; Bates 2001). Jackson and Rosberg (1982, pp. 17 –19) note that personal rule is “asystem of relations linking rulers ...with patrons, clients, supporters, and rivals, who constitute the system. Ifpersonal rulers arerestrained, it is by the limits of their personal authority and power and by the authority and power of patrons, associates, clients, supporters, and — of course—rivals. The system is ‘structured’ ...not by institutions, but by the politicians themselves. ” Our contribution to this literatureis to provide aformalmodel of kleptoc- racy (personal rule), to systematically investigate how adivide-and-rule strategy supports kleptocracies, and to develop anumber of comparative static results on the likelihood of kleptocratic regimesand policies. These areall prerequisites to afull understanding of politics in weakly institutionalized societies. The restof the paper isorganized as follows. In the next section, webrie  y discuss the experiences of two countries under kleptocratic regimes, with special emphasis on how these regimeswere sustained. Section 3presents amodel that formalizesthe main ideas discussed above and derives the main comparative static results. Section 4concludes.

2. The KleptocraticExperience: Two Case Studies

In this section webrie  ydiscuss two classic kleptocracies. The rule of Mobutu Sese Seko in the DemocraticRepublic of the Congo from1965 to 1997 and the rule of RafaelTrujillo in the Dominican Republic from1930 to 1961. Our discussion emphasizes several characteristicfeatures of these regimes. First, both rulers weretrue kleptocrats in that their primeaim was to accumulate personal wealth forthemselves and their families.Second, neither regime, despite their longevity, had anextensive base of support in society. Third, each Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 169 rulermaintained control of power by manipulating people and groups along the lines of divide-and-rule. Finally, the grip on power of both rulers was greatly facilitated by the nature of society, natural resource rents, and foreign aid.

2.1. Mobutu Sese Seko and LeMal Zairois

Joseph Mobutu seized power in the Congo, which becameindependent from Belgium in 1960, in amilitarycoup in 1965 and quickly established himself as adictator atthe head of aone-party state, the MPR(Mouvement Populaire de la Re´volution). Thereis no doubt that the aimof Mobutu was touse the state for the enrichment of himself and his family. Hewas atrue kleptocrat. In the 1970s, 15–20% of the operating budget of the state went directly to Mobutu. In 1977 Mobutu’sfamilytook $71 million fromthe National Bank forpersonal use and by the early 1980s his personal fortune was estimated at$5 billion (Leslie1987, p. 72). Turner and Young (1985, pp. 178 –183) devote six pages to listing Mobutu’sassets and wealth. The social base of the Mobutu regimewas very narrow. Mobutu himself camefrom a smallethnic group, the Ngbandi, in the E´ quateur province and there is some evidence of systematic favoritism towards fellow E´ quateurians especially towards the end of his regime.For example, in 1990, 46 percent of the ofŽ cercorp, 34% of diplomats and 19% of the MPRCentral Committee were from E´ quateur. Nevertheless, it was relatively fewwho constituted the base of support. These people, called by Zairians les gros le´gumes or les barons du regime constituted “aninner core of persons with especially close kinship, ethnoregional, or personal ties to the president; variously known asthe ‘presi- dential brotherhood ’ or the ‘untouchables, ’ they wereable to conduct their mercantileaffairs in blatant disregard fornormal legal regulations. ” (Turner and Young 1985, p. 398) Mobutu used various strategies to maintain power, including the military and nationalist ideology, and most notably the philosophy of “Mobutuism” after 1974. However, his most important strategy was similarto what werefer to as divide-and-rule, creating an environment in which any person or group could be rewarded or punished selectively. For example, Leslie(1987, p. 70) describes this as “Individuals in public of Ž ceare totally dependent on him forselection and maintenance of power. By frequently rotating government posts, Mobutu manages to maintain uncertainty and vulnerability. ...Heplays the role of big chief ...bestowing favors on his subjects based on personal discretion. ” Turner and Young, in their seminal 1985 book on the Mobutu state, describe the samesituation as follows (pp. 165 –166): “Client ofŽ ceholders have been constantly reminded of the precariousness of tenure by the frequency of ofŽ cerotation, which simultaneously fuels the hopes of those Zairians anxiously waiting just outside the portals of power. The MPRPolitical Bureau, for 170 Journal of the European Economic Association example, was revamped adozen timesin the Ž rst decade of party life.No one (except of course Mobutu) has been continuously amember,and only six persons have Ž gured on as many as half of the membership lists. ...Insecurity has been sustained among top state and party personnel by the frequent appli- cation of sanctions. During the Ž rst decade of the Mobutu regime,29 of the 212 leaders went directly fromtheir posts to prison on either political or corruption charges. An additional 26 wereremoved on grounds of disloyalty or dishonesty, with penal sanctions. ...Frequently in presidential speeches referenceis made to unnamed disloyal or corrupt persons in the leadership ranks. ...Cumula- tively, these devices constitute apowerful mechanism of informalintimidation, and suggest why systematic opposition has never arisen within the top organs of the state.” Laterthey continue (pp. 397 –398): “Asaco-optative resource, apool of vacant high of Ž ces was sustained, ...the sanction fornot cooperating ...was imprisonment on corruption, nepotism, or subversion charges, or exile. Access to high rank in all state agencies depended upon presidential favor. ” They sum up the essence of Mobutu ’sstate as (p. 193): “the shifting patterns of membership [ofthe Political Bureau] ...constituted the very essence of patri- monial politics. ” Gould’s(1980, p. 83) analysis concurs when he notes that “the frequent cabinet shuf  es and transfers of of Ž cials fromregion to region ...maybe explained aslargely re  ections of the president ’sskill atusing people while they can provide assistance to him and atthe sametime keeping factions separated fromeach other, thus preventing autonomous power centers fromdeveloping. ” The careerof Nguza Karl-i-Bond, who started as anopponent of Mobutu in the 1960s, illustrates the workings of divide-and-rule and the nature of shifting personal alliances. Inthe 1970s he was brought into the Mobutu campand atone point becamethe foreign minister. In 1977 he was accused of treason and sentenced to death. In1979 he was released and madeprime minister. By 1981 he was inexile and gained fameby his book on the Mobutu regime(Karl-i-Bond 1982). By 1985 he was back in favor, and becamethe ambassador to the United States (Leslie1987, pp. 70 –71). Thereare many other cases like this, for example Kamitatu-Massambaand Mbujimayi Belshika (seeGould 1980, p. 83). Turner and Young (1985, p. 166) note in this context “though sanctions could be severe, afallfrom grace was not necessarily permanent. Those jailed seldom remained in prison forvery long. Repentance and renewed cultivation of the favor of the sovereign could makepossible areturn to full grace. ” The impact of divide-and-rule was catastrophic forthe ef Ž ciency of the bureaucracy and the state. Leslie(1987, p. 6) notes that in Mobutu ’s regime: “what is considered to be simply bureaucratic disorganization and economic mismanagement by external actors such asthe [World] Bank and the IMF,is to Zaire’sruling elitea rational policy of ‘organized disorganization ’ designed to maintain the status quo. ” Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 171

Added to this, the personal wealth accumulation of Mobutu and his clique destroyed much of the economy. The nationalization of foreign owned Ž rms, most of which ended up in the hands of Mobutu and the regime,under the Zairianization program initiated in November 1973 was particularly disastrous, destroying what was leftof the economy. According to Maddison (1995) GDP per-capita in 1992 was less than half of what it had been atindependenc e in 1960. What factors facilitated Mobutu ’suse of divide-and-rule? During the rule of Mobutu, the Congo was clearly very poor and characterizedby low productiv- ity, making it easy forhim to buy offpotential challengers. Moreover, the Congo isendowed with enormous natural resource wealth including 15 percent of the world’scopper deposits, vast amounts of diamonds, zinc, gold, silver, oil, and many other resources (Leslie1993, p. 3). This vast natural wealth gave Mobutu aconstant  ow of income to help sustain his power. The Congo is also ahuge and geographically diverse country (seeHerbst (2000) foran analysis of the “difŽ cult geography ” of the Congo) and many scholars have seen it as having auniquely complex cultural heritage. The common estimateis that there areabout 200 distinct ethnic groups inthe Congo (Leslie1993, p. 68) and Young (1965, p. 271) argued that “The particular colonial experience has worked to makeethnic loyalties in the Congo stronger and supra-tribal identi Ž cation probably weakerthan in most other African states.” According to numbers reported in Easterly and Levine (1997), the Congo is the second most ethnically diverse country inthe world afterTanzania. The fragmented nature of the country, combined with the impactof the Belgian colonial state, meantthat there was no large group or socioeconomic class that could offset the power of Mobutu and help obstruct his personal rule. As we noted in the Introduction, this is in stark contrast to Botswana and Mauritius wherethe existence of such classes placed checks on state power and limited the potential forkleptocracy. Moreover, the regional and ethnic identi Ž cation of different groups madeit straightforward forMobutu toengage in the types of targeted redistribution and punishment emphasized in our model. Finally, there is general consensus among scholars that foreign aid was a key factorenabling Mobutu to stay in power forso long. Leslie ’s (1987) analysis of the relationship between the Mobutu regimeand foreign aid agencies suggests that aid and foreign Ž nancial support was crucial in propping up the regime.She notes “the regime. ..views aid agencies such asthe [World] Bank and bilateral donors as sources of foreign exchange, not partners in develop- ment” (p. 135) and also emphasizes that Mobutu and his clique “sees develop- ment assistance as yet another means to accumulate personal wealth. ” Mobutu madeefforts to reformthe regimeand its economic policies only when “[the abysmal situation] undermined the con Ž dence of Western governments and 172 Journal of the European Economic Association investors, and thus  ows of foreign exchange, resources vital forMobutu and his clique” (p. 142).9 Thus the Mobutu regimeis aclassic example of how divide-and-rule can be used to sustain aregimewith little, ifany, popular support. One could ask why Mobutu, facedwith ruling such apolity, could not simply sustain himself through repression, ratherthan resort to the relatively complex strategy wehave outlined. One reason is that Mobutu always kept the militaryrelatively weak because he could never be sure of its support. Divide-and-rule was used with respect to militaryof Ž cersas well as civilian political elites. When Mobutu needed effectivemilitary forces he relied instead on the Belgians and the French, who could not be expected to repress domestic opponents. The expe- rienceof the Congo during this awful eraalso illustrates the importance of the factors highlighted by the comparative statics of our model.

2.2. TheDominican Republic under Rafael Trujillo

The Dominican Republic, the eastern half of the island of Hispaniola, became independent fromSpain in1821. However, in 1822 it was invaded by Haiti, and regained its freedomonly in 1844 aftera warof independence. The Haitian occupation destroyed the traditional ruling eliteof large landowners who  ed the country never to return. Inthe subsequent decades, the country was plagued by serious political instability and aseries of dictatorships. In1916 the United States took over the country (previously, they had controlled the customs since 1906 to recover delinquent loans), an occupation that lasted until 1924. During this period the country was run by amilitarygovernor. Toestablish control of the country, the United States trained adomestic constabulary, which became the Nation Armyafter 1924. Trujillo becamethe head of this constabulary in 1928 and in1930 used it as apower base to help acoup against the government, afterwhich he had himself elected President in afraudulent election. Through- out his rule Trujillo carriedout the facadeof elections, regularly being reelected by 100% of the vote and even on two instances, from1938 to 1942 and from 1952 to 1961, stepping down fromthe Presidency and controlling power from behind the scenes. As with Mobutu, there is no doubt about the kleptocratic tendencies of Trujillo. Indeed, relative to the size of the economy, Trujillo ’s regime was probably the most successful case of kleptocratic accumulation in history. By the end of his regime,the fortune of the Trujillo familywas equal toabout 100% of GDPatcurrent prices and the family, “controlled almost 80% of the country ’s

9. The reforms were never properly implemented, however, because “[such] reforms strike at the heart of Mobutu ’spatrimonial system and will only serve to undermine patron –client networks, efforts to buy loyalty, and wealth seeking strategies ” (Leslie 1993, p. 123). Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 173 industrial production. About 60% of the country ’slabor forcedepended on him directly or indirectly, with 45% employed in his Ž rmsand another 15% working forthe state ” (Moya Pons 1995, p. 398). Trujillo used many tools to maintain power. Heused coercion and force against potential rivals, and murdered and tortured many opponents. But as in the Congo, divide-and-rule was akey strategy. As aresult, Hartlyn (1998) explains that “Trujillo eventually becamethe single dominant forcein the country by combining abuse of state power, threats and co-optation. Although certain of the country ’seconomic elitemaintained some individual autonomy there was no possibility forindependent organization ” (pp. 99 –100).10 How the Dominican version of divide-and-rule worked is explained in Turits (2003, p. 7): “Trujillo continually shuf  ed cabinet membersand other important of Ž cials in and out of government posts to prevent their developing an independent power base. Heexerted still tighter control over the national legislature. Upon assumption of of Ž ce, membersof the National Congress were obliged tosign their own resignations, adocument that on Trujillo ’sorder could be handed to them atany moment. ...In certain cases, these resignations were reportedly delivered to legislators in the middle of their speeches before the assembly. Inthe 1942 to1947 period, forexample, there were32 resignations in aSenate of 19 members,and 139 resignations in aHouse composed of 42 deputies.” Wiarda’s(1968) analysis is similar: “Trujillo’swas ahighly personalistic dictatorship in which power was not shared, even among asmallclique, but concentrated in the hands of one man ” (p. 26). Hefurther demonstrates that “Trujillo’sprincipal method of controlling the governmental machinery was the constant shuf  ing and reshuf  ing of political of Ž ceholders. Hehad the consti- tutional authority ...to appoint and remove almost all government personnel and he used his power extensively. In addition, he kept a Ž leof signed but updated resignations forall government employees, and of Ž cials frequently arrived atwork only to learn that Trujillo had Ž lled in the date and that they had resigned. The technique of shuf  ing government personnel was employed for the samereason as Trujillo ’sfrequent changing of the armedforces and police commands. ...Anyone who gained apowerful position could expect to be replaced ...no potential opposition power centers wereallowed to develop ” (p. 62). Wiarda continues: “Trujillo’stechnique was to fragmentthe power of

10. Those who bene Ž tted from the regime were few outside Trujillo ’sextended family. A1953 NewYork Times article (quoted in Wiarda 1968, p. 74) found that 153 relatives were employed by the government. Trujillo himself was minister of foreign relations, minister of social security, ambassador at large, and special ambassador to the United States. Family members held the presidency (his brother), two senatorial posts, six major diplomatic assignments, the positions of commander in chief of the armed forces, undersecretary of defense, chief of staff of the air force, inspector general of the army, inspector of embassies, plus agreat number of other posts. 174 Journal of the European Economic Association the many and, correspondingly, to concentrate all authority in his own hands. ” (p. 62) As inZaire,people who weredemoted, jailed or exiled wereoften pardoned and returned to positions of power (seeWiarda (1968, p. 63) and Crassweller (1966) formany examples). Hartlyn (1998, p. 100) adds: “Politically, Trujillo combined guile, cynicism, ruthlessness, and cooptation. Therewas ...incred- ible manipulation of individuals, who found themselves moved and removed frompublic of Ž cein complex and disconcerting fashion even as personal rivalries werepromoted and tested. ” Trujillo also employed the strategy of divide-and-rule to control the rural sector. Turits (2003, pp. 82 –83, 95) describes this asfollows: “in exchange for land access and state assistance, peasants werecompelled to farmin amore sedentary, intensive and land ef Ž cient manner. ...The dual nature of the Trujillo state ’s proteccio´n, or support, of the peasantry was evident in the severe methods it used to augment peasant production. The state distributed rewardsto those it deemed ‘man of work’ by variously providing them with land, irriga- tion, tools, seeds, credit, and technical assistance, while it harshly punished with vagrant penalties (jailand forced labor) and withdrawal of usufruct rights those it considered to be idle. ...The distribution of lands under Trujillo was repre- sented as agift morethan aright and thus served to dramatizethe dictator ’s personal power. ” Most notably, Turits (2003, p. 113) explains: 11 “the state’s mediation of peasants ’ access to land fostered ahigh level of political control. All individuals who wished to obtain land had to supply acerti Ž cate of ‘good conduct’ fromthe neighborhood authority. ...Certainly anyone suspected of political disloyalty ...would be excluded. ” The divide-and-rule strategy was facilitated by the extraordinary control that Trujillo accumulated over the economy. According toWiarda (1968, pp. 87, 90 –91): “Trujillo could hire and Ž rewhom he pleased when he pleased. Since the great majority of the population was absolutely dependent on him for day-to-day existence, his control over it was assured ...there was so little economic independence that even abare minimumof political independence was impossible. ...Trujillo frequently employed his economic power to de- stroy his political opponents. Banks could and did ...refuseloans and foreclose mortgages; government agencies refused export or import permits ...electricity or phone services wereinterrupted; and streets and sidewalks infront of selected business establishments weretorn up. ” As with divide-and-rule in the Congo, one effectwas to destroy the efŽ ciency of much of the state machinery. Turits (2003, p. 140) notes: “confu- sion within the state was an almost inevitable by-product of Trujillo ’s system of

11. Turits (2003) interestingly claims that despite the manipulative nature of the relationship between Trujillo and the peasantry, the fact that Trujillo did actually give peasants access to land created quite alarge degree of support for his regime. Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 175 continually circulating high-level functionaries into differentpositions almost every year forthe purposes of maintaining his control. ” Nevertheless, economic performanceunder Trujillo was not as bad as under Mobutu. Though there appears to be no reliable national account data forthis period, the scholarly consensus appears to be that the Dominican Republic experienced some growth during Trujillo ’s rule. What facilitated divide-and-rule in the Dominican Republic? Unlike the Congo, not all the factors suggested by our model Ž tthe Dominican experience. For example, though the country has very rich agricultural lands, it does not have large quantities of minerals, oil, or other natural resources. Also the Trujillo regimetook place in aperiod before the existence of the IMF,World Bank, and much international development assistance, so this cannot have played animportant role insustaining the regime.Although the United States was generally friendly to the regime,it did not provide great quantities of Ž nancial assistance. However, two factors clearly areimportant. When Trujillo took power in1930 the Dominican Republic was avery poor country and this madeit relatively feasible forTrujillo to buy offsupport. Moreover, although the country is not ethnically heterogeneous in the way the Congo is, the impact of the nineteenth-century Haitian occupation and the United States military intervention, meantthat there was no organized traditional elite to counterbal- ance Trujillo ’sriseto power and kleptocratic tendencies.

3. A Model

We now outline asimple model to formalizethe ideas discussed above, and to illustrate how, in weakly institutionalized societies, akleptocratic regimesus- tained by adivide-and-ruled strategy can emerge.Our purpose is not generality, but to construct asimple framework,which can be used to derive new com- parative static results on the incidence and sustainability of kleptocratic regimes.

3.1. TheEnvironment

We consider asmallopen economy (alternatively, an economy with linear technology) producing three goods: anatural resource, Z,and two goods, q1 and q2.We normalize the prices of all goods to 1, which is without loss of any generality, since wewill allow differences in the technology of production of the two goods. To start with, wefocus on the case wherethe production of the natural resource good Zt is constant in all periods,

Zt 5 Z. Natural resources createrents in this economy, which, in turn, affectpolitical 176 Journal of the European Economic Association equilibria. We assume that the natural resource rents accrueto the government, and can then be distributed to the producers or consumed by the ruler.

Thereare two (large)groups of agents, n1 that produce q1 and n2 that produce q2.We normalize n1 5 n2 5 1. Both groups have utility attime t given by:

` ` h bsu ~ y , l ! 5 bs y 2 l~11h!/h (1) O is is is O S is 1 1 h is D s5t s5t where b , 1is the discount factor, yit denotes their after-taxincome, and lit is labor supply attime t. This speciŽ cation implies that labor is supplied with elasticity h . 0. For each producer of group i,the production technology is:

qit 5 vilit, (2) where vi is the productivity of group i 5 1, 2. Without loss of generality, we assume that group 1is moreproductive, that is, v1 $ v2.To parametrizethe degree of inequality between the two groups, wedenote:

v1 5 v~1 1 x! and v2 5 v~1 2 x!, (3) where, by construction, v is the average productivity of the economy, and x [ [0, 1]. Agreater x corresponds to greaterinequality between the two groups. The only redistributive tools in the economy area linear income tax that is potentially speci Ž cto each group, and group-speci Ž clump-sum transfers. The option to use group-speci Ž ctaxes and transfers areimportant forthe results, and plausible in the context of Africansocieties, wherethere areclear geographic and ethnic distinctions between producer groups. 12 The after-taxincome of the two groups are

yit 5 ~1 2 tit!vilit 1 Tit, (4) where tit [ [0, 1] is the income tax imposed on group i at time t and Tit [ [0, `)is a(nonnegative) lump-sum transferto group i. In each period, each producer maximizeshis utility function (1)taking the tax rate tit asgiven. This implies alabor supply function:

h lit~tit! 5 @~1 2 tit!vi# . (5) This equation relateslabor supply, and thereforeoutput, to taxes, and illustrates

12. In practice, differential taxes also include indirect taxes levied on the producer and consumer groups, as well as the use of marketing boards to manipulate agricultural prices, afrequently used policy in many African countries (see Bates 1981). Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 177 the distortionary effectsof taxation: greatertaxes reduce labor supply and output. Using (5), the instantaneous indirect utility of arepresentative agent in group i is found to be: 1 U ~t , T ! 5 @v ~1 2 t !#11h 1 T , (6) i it it 1 1 h i it it and tax revenues are:

h 11h h 11h R~t1t, t2t! 5 t1t q1t 1 t2t q2t 5 t1t~1 2 t1t! v1 1 t2t~1 2 t2t! v2 . (7) The government budget constraint is

T1t 1 T2t 1 CKt # R~t1t, t2t! 1 Z 1 Ft, (8) where CKt [ [0, `)is the consumption of the (kleptocratic) ruler, R(t1t, t2t) is tax revenue given by (7), and Ft [ [0, `)is foreign aid, ifany. The ruleris assumed to have the utility function attime t:

` s O bK CKs, s5t where bK , 1is the discount factorof the ruler, which could differfrom those of the citizens. The political system is either “dictatorship ” (controlled by the ruler), K, or democracy, D.Our focus is whether dictatorship can survive and to what extent it will be kleptocratic (i.e., to what extent the rulerwill be able to tax producers forhis own consumption, while ensuring the survival of the dictatorship). To focus on this question, wemodel democracy in the simplest possible way, and assume that in democracy the two producer groups arein power jointly, thus they set zero taxes, and share the natural resource rents and foreign aid equally 13 (and therefore, set CKt 5 0). In what follows, weassume that only the rulerreceives foreign aid (i.e., there is no foreign aid in democracy). Moreformally, denoting the political state by St21, we have Ft 5 0 if St21 5 D and Ft 5 F if St21 5 K.This assumption is not as extremeas it appears, since any part of foreign aid that is perpetual can be included in the natural resource rents, Z,and therefore, Ft can be interpreted as the additional portion of foreign aid that the rulerreceives because under his rule there is morepoverty or famine,or because the rulerpursues aforeign

13. This formulation of policies in democracy is adopted to simplify the discussion, and the results are robust to more realistic models of decision making in democracy, as long as democracy is more attractive for both producer groups than the kleptocratic regime. For example, we could allow the two groups to be of different sizes, and the larger group to impose limited redistributive taxes on the smaller group. 178 Journal of the European Economic Association policy in line with the donors ’ interests. Moreover, ifa democracy will receive moreforeign aid than the ruler, wecan allow this by letting F , 0. Finally, we can interpret F as the fungible part of foreign aid. In this case, even ifforeign aid will continue afterdemocracy it will be allocated to some speci Ž c purposes such aspoverty reduction or education fordisadvantaged groups. Inthis case in democracy the two producer groups will not have access to aid, and this situation is equivalent tothe one here with Ft 5 0 when St21 5 D. Given this assumption, the instantaneous utilities of the two groups in democracy are:

11h vi Z UD 5 1 . (9) i 1 1 h 2 In contrast, in kleptocracy, the rulerwill maximizehis consumption, subject to the constraint that he keeps power (alternatively, he can be removed from power, and in this case, democracy will result, and CKt 5 0forall future periods). Before describing the constraints facing the rulerin detail, let us write the unconstrained solution. This is given by maximizing R(t1t, t2t), which is achieved atthe tax rates: 1 t* 5 t* 5 t* ; , (10) 1t 2t 1 1 h

* * * and paying 0transfers, i.e., T it 5 0, thus setting CKt 5 R(t , t ) 1 Z 1 F. The instantaneous utilities of the two groups under these tax ratesare given by:

11h 11h vi h U * 5 U ~t 5 t*, T 5 0! 5 . (11) i i it it 1 1 h S 1 1 hD

3.2. ThePolitical Gameand DeŽnition of Equilibrium

The timing of events in the political gameare as follows. In each period, t, society inherits apolitical state, either St21 5 D or St21 5 K. Note that St21 5 D is an absorbing state, so ifthe economy ever becomes ademocracy, it remains so forever. If St21 5 D,the two producer groups play the simple gamedescribed above, denoted by Gt(D)forconvenience, where they set the taxes and share the natural resources rents equally. Ifsociety is adictatorship, that is, St21 5 K, then the following game, Gt(K),isplayed:

1. The rulerannounces tax rates( t1t, t2t)and transfers ( T1t, T2t). 2. Each group i decides whether to makea proposal to remove the rulerfrom

power. pit 5 1denotes that group i has madea proposal and pit 5 0 denotes otherwise (ifboth groups simultaneously choose to makea proposal, one of

them is chosen randomly to have pit 5 1and the other one has pit 5 0). If Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 179

one of the two groups makes aproposal to replacethe ruler, wedenote this

by Pt 5 1, with Pt 5 0otherwise.

3. If Pt 5 0, then (t1t, t2t, T1t, T2t)is implemented and the political system remains at St 5 K. 4. If Pt 5 1, i.e., if pjt 5 1forone of the groups, then: the rulermakes anew r r r r offerof taxes and transfers, ( t 1t, t 2t, T 1t, T 2t)such that this policy vector satisŽ esthe government budget constraint, (8). Group i Þ j then responds to the proposal of “proposer” group j and the ruler ’snew policy vector. If

the proposed group i, chooses dit 5 1, the ruleris replaced and there is a switch to democracy, i.e., St 5 D. If dit 5 0, the political system remains r r r r at St 5 K, and (t 1t, t 2t, T 1t, T 2t)is implemented. r r r r 5. Given the policy vector, either ( t1t, t2t, T1t, T2t) or (t 1t, t 2t, T 1t, T 2t), individuals in both groups choose labor supply. 6. Output is produced, tax revenues arecollected and consumption takes place.

7. If Pt 5 1and the proposed group plays dit 5 1, then in the next period the

stage gameswitches to Gt11(D),and otherwise it is Gt11(K). Thereare a number of noteworthy features: Ž rst, weassume that all individuals within aproducer group actin cohesion in the political game. This is anatural assumption here, since there areno costs of political action, and all agents within agroup have the samepreferences, so there is no free-rider problem. Second, there is aspeci Ž c(political) structure built in the timing of the political game:the rulercan only be replaced ifthe two groups agreeto replace him. This assumption captures the factthat in weakly-institutionalized societies, those controlling the state mayhave considerable power, and cannot be easily removed fromof Ž ceby one of the social groups alone. 14 An alternative political game, wherethe party in power needs toreceivesupport fromall social groups or compete against potential rivals, would correspond to political institutions placing checks on politicians. However, such strong political institutions are absent in anumber of less developed countries. Inthese weakly institutionalized polities, the implied power of the ruler, combined with the factthat afterthe proposal to remove him frompower he can offera different policy vector, gives him the opportunity to use adivide-and-rule strategy, which will be the focus of our analysis. To simplify the analysis, wewill focus on the (pure strategy) Markov PerfectEquilibria (MPE)of the above game(though this restriction is not important forthe results). An MPEis amapping fromthe current state of the game(and fromthe actions taken previously in the samestage game)to

14. In addition, in some countries the ethnic and social divisions enabling different taxes and transfers on distinct groups may be absent, thus also undermining the logic of the divide-and-rule strategy. 180 Journal of the European Economic Association

strategies. Here,the only state variable is St21,which denotes whether the political state is either democracy or dictatorship.

3.3. Analysis

The MPEwill be characterizedby backward induction. When St21 5 D, there areno interesting actions, and the rulerreceives zero utility, while the two groups receivelifetime utilities of:

D Ui VD 5 , (12) i 1 2 b

D D with Ui given by (9). Note also that Vi given by (12) is what the proposed group will receiveif it chooses dit 5 1and removes the rulerfrom power. r r 15 On the other hand, if,in response to the reaction of the ruler( t it, T it), the proposed group chooses dit 5 0, its memberswill receive

e e bUi~t i , T i ! VC~t r , T r ut e, T e! 5 U ~t r , T r ! 1 , (13) i it it i i i it it 1 2 b

e e where Ui is given by (6)and ( t i , T i )isthe MPEtax transfercombination that applies to this group. The reasoning forthis expression is that inthis period, the r r proposed group receives ( t it, T it),and the kleptocrat remainsin power, so in the e e 16 future, the play goes back tothe equilibrium policy of ( t i , T i ). In addition, the response of the rulermust satisfy the government budget constraint:

r r r r T 1t 1 T 2t 2 CKt # R~t1t, t2t! 1 Z 1 F. (14) The divide-and-rule strategy will be successful and the rulerwill keep power only if

C r r e e D V i ~t it, T itut i , T i ! $ V i . (15) Itis useful to distinguish two cases: 1. The rulerwill be able to maintain power, with the equilibrium strategy of e * e tit 5 t [ 1/(1 1 h)as given by (10) and T it 5 0 for i 5 1, 2and forall t.Wedenote the set of parameterssuch that this happens by S*, i.e., if s9 5 * e e e e * * (h, b, Z, F, v, x) [ S , then (t 1t, t 2t, T 1t, T 2t) 5 (t , t , 0, 0). 2. The rulerwill not be able to maintain power ifhe set ( t*, t*, 0, 0), thus

r r r r r 15. The reaction of the ruler is the vector ( t 1t, t 2t, T 1t, T 2t),but since only the component ( t it, r T it)is relevant for group i,wewill use this lower-dimensional vector to simplify notation. 16. Since the deviation does not affect any payoff-relevant variable in the continuation game, the MPEinvolves return to the same strategy, and since weare focusing on pure strategy equilibria, e e this is simply ( t i , T i ). Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 181

s9 ¸ S*.As wewill see later,in this case, the rulercan reduce taxes and increase transfersso as to maintain power. We will also see that in this case e e * * (t1, t2) , (t , t ),that is, the rulerwill necessarily be forced to reduce taxes, and policy will be less distortionary. To characterize S*,let us startwith the subgame in which group j has proposed to replacethe ruler, and denote the policies initially chosen by the ruler e e e e r r by (t 1t, t 2t, T 1t, T 2t).Ifthe rulerresponds with ( tit, T it) for i Þ j such that C r r e e D V i (t it, T itut i, T i) , V i ,then he will be replaced. This shows that the rulermust ensure (15). To analyze how, and when, the rulercan do so, let us Ž rst deŽ ne

C e e C r r e e V i @ti , Ti # 5 max V i ~t it, T itut i , Ti ! (16) r r r r t1t , t2t ,T1t , T2t

C e e D subject to(14). If V i [ti , Ti ] , V i for i 5 1or 2, then group j Þ i,anticipating that its proposal will be accepted, will propose to replacethe ruler, and the ruler C e e D will be deposed. Therefore,the rulermust guarantee that V i [ti , Ti ] $ V i for i 5 1and 2to remainin power. C e e Consequently, we Ž rst need to Ž nd V i [t i , Ti ],the maximumutility that the rulercan give to the proposed group offthe equilibrium path. To do this, we need to maximize(16) subject (14) and set the consumption of the rulerto 0, that is, CKt 5 0. Straightforward differentiation establishes that 1 t r 5 0 and t r 5 . i j 1 1 h Therefore,in Ž ghting offa challenge fromgroup j,the rulerwill set the revenue-maximizing tax rateon this group, and set zero taxes on the proposed group i.Inaddition, the rulerwill clearly give the minimumpossible amount to r the proposer group, thus T j 5 0. Then the government budget constraint, (8), implies:

r r r Ti 5 R~ti , tj ! 1 Z 1 F. Using these expressions, wecan derive the maximumoff-the-equilibrium- path payoff of the proposed group, as afunction of the MPEpolicy vector e e e e (t1t, t2t, T1t, T2t). This is

11h 11h h e e vi vj h bUi~ti , Ti ! V C@t e, T e # 5 1 1 Z 1 F 1 . (17) i i i 1 1 h 1 1 h S1 1 hD 1 2 b This expression is the maximumutility that the rulercan give to group i, following aproposal by group j,as afunction of the equilibrium tax and transfer rateson group i. 182 Journal of the European Economic Association

Given this analysis, the problem of Ž nding the MPEis equivalent to Ž nding asolution to the following maximization problem of the ruler:

1 e e max @R~t1, t2! 1 Z 1 F# (18) e e e e 1 2 bK t1t,t2t,T1t,T2t subject to the constraint set:

C e e D V i @ti , Ti # $ V i for i 5 1, 2. (19) We now characterizethe solution to this constrained maximization problem. First, notice that combining (11), (12), and (17), the constraint set, (19), can be rewrittenas:

11h 11h h vi vj h 1 1 Z 1 F 1 1 h 1 1 h S 1 1 hD (20) 11h b 1 1 vi Z 1 @v ~1 2 t !#11h 1 T $ 1 , 1 2 b S 1 1 h i i iD 1 2 b S 1 1 h 2D for i 5 1, 2 and j Þ i.Then exploiting the factthat v1 5 v(1 1 x) and v2 5 v(1 2 x),wecan writethe constraint set as

1 C~t , T , x! $ Z b 2 2 ~1 2 b! F and (21) 1 1 S 2D 1 C~t , T , 2x! $ Z b 2 2 ~1 2 b! F, (22) 2 2 S 2D where

~1 2 b!v11h~1 2 x!11h h h C~t, T, x! ; 1 1 h S 1 1 hD (23) bv11h~1 1 x!11h 2 ~1 2 ~1 2 t!11h! 1 bT. 1 1 h Moreover, in the case wherethere isno inequality between the two groups, i.e., when x 5 0, the constraint set is simply:

1 C~t , T ! ; C~t , T , x 5 0! $ Z b 2 2 ~1 2 b! F. (24) i i i i S 2D Itis already possible to see why the divide-and-rule strategy can arisein equilibrium. The constraint set, characterizedby (21) and (22), will be satis Ž ed when, off-the-equilibrium path, the rulercan shift enough resources to the Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 183 proposed group. In other words, avery inef Ž cient set of policies can be supported when each group knows that ifit proposes to replacethe ruler, the rulerwill bribe the other group successfully and remainin power. Recognizing this off-the-equilibrium path threat, no group will challenge the ruler, who will then be able to pursue kleptocratic policies along the equilibrium path.

3.4. Equilibrium without Inequality

Letus startwith the case in which x 5 0and there isno inequality between the two groups. First, note that whenever he can, the rulerwould like toset the tax e e e e * * ratesthat maximize(18), that is, ( t1t, t2t, T1t, T2t) 5 (t , t ,0, 0). Therefore,the Ž rst step is to characterizethe set of parameters S* such that these best tax rates (fromthe point of view of the kleptocratic ruler)can be supported as equilibria. e e e e * * Using (24) and substituting ( t1t, t2t, T1t, T2t) 5 (t , t ,0, 0), weimmediately obtain the set S* asthe set of parameterssuch that C(1/11h, 0) $ Z(b 2 1/2) 2 (1 2 b) F,or moreexplicitly: v11h h h S* 5 s 5 ~h, b, Z, F, v! : H 1 1 h S 1 1 hD (25) bv11h 1 2 $ Z b 2 2 ~1 2 b! F . 1 1 h S 2D J * e e e e * * If s 5 (h, b, Z, F, v) [ S ,then the MPEinvolves ( t1t, t2t, T1t, T2t) 5 (t , t , * e e e e e e e e 0, 0). What happens if s ¸ S . Then, (t1t, t2t, T1t, T2t) 5 (t , t , T , T ) will be chosen such that C(t e, T e) 5 Z(b 2 1/2) 2 (1 2 b) F (given the symmetry between the two groups, the rulerwill choose the sametaxes and transfers for both groups). Moreover, inspection of (23) establishes that aslong as C(tˆ, T 5 0) 5 Z(b 2 1/2) 2 (1 2 b) F for some tˆ [ [0, t*],the rulerwill reduce taxes to tˆ and sets 0lump-sum transfers (this is intuitive, since taxes aredistortion- ary). The important point to note is that the rulercan always satisfy (25), and thereforeremain in power. 17 This highlights the importance of the underlying political institutions in this context: by allowing the rulerto use divide-and-rule, the current set of political institutions makesure that he always remainsin power. Nevertheless, the extent to which he can transferrents to himself and distort the allocation of resources will depend on parametervalues as wewill see next.

17. Tosee this, suppose that for some parameter values, the kleptocrat is removed from power. C e e D Then it must be the case that Vi [ti , Ti ] , V i for i 5 1or 2. But in this case, it is credible for e e C e e D the kleptocrat to reduce ti and increase Ti ,and by construction, Vi [ti 5 0, Ti 5 Z/2] . Vi , providing acontradiction and establishing that the kleptocrat will not be removed from power along the equilibrium path. This argument would not necessarily work when F , 0, however. In amore general setup with stochastic productivity or costs of removing the ruler, there would also be parameter values for which dictatorship cannot survive, similar to, and for the same reasons as, the models of transition to democracy in Acemoglu and Robinson (2000a, 2004). 184 Journal of the European Economic Association

This discussion establishes the following proposition.

* PROPOSITION 1. Let S be given by (25). Then we have: 1. If s [ S*, then the unique MPEis an unconstrained kleptocratic regime e e e e * * where (t1t, t2t, T 1t, T 2t) 5 (t , t , 0, 0) for all tand i 5 1, 2. 2. If s ¸ S*, then the unique MPEis aconstrained kleptocratic regime where e e e e the equilibrium policies are (t1t, t2t, T 1t, T 2t) 5 (tˆ, tˆ, 0, 0) if 1 C~tˆ, T 5 0! 5 Z b 2 2 ~1 2 b! F, (26) S 2D

* e e e e ˆ ˆ for some tˆ [ [0, t ], and (t1t, t2t, T1t, T2t) 5 (0, 0, T, T) where C(t 5 0, Tˆ ) 5 Z(b 2 1/2) 2 (1 2 b)F otherwise.

r r In both cases , achallenge from group j , that is, pjt 5 1, is met by (tjt, T jt) 5 * r r C r r e e D C r (t , 0) and (tit, Tit) for i Þ j such that V i (tit, Tituti , Ti ) 5 V i where Vi (tit, r e e D Tituti , Ti ) is given by (13) and Vi is given by (12). The discussion above establishes this proposition. The only part that may need morecomment is the uniqueness of equilibrium. Recall that ifgroup j makes aproposal to remove the rulerfrom power, the rulerwill respond with r * * e * tjt 5 t , and when s [ S ,wealso have tjt 5 t .Itmay therefore appear that wecan construct equilibria wherethere arechallenges along the equilibrium path when s [ S*,and thus the equilibrium described in part 1of Proposition 1is not unique. This is not the case, however. Any combination of strategies e e * where pjt 5 1cannot be anequilibrium. Ifitwere,a deviation to( tjt, T jt) 5 (t , «) for « . 0would be abest response forthe ruler, and the strategy of pjt 5 1 would then cost group j an amount « . 0. Since asmaller « is always preferred by the ruler, the only combination of best response strategies is when « ¡ 0, which is the one described in the proposition. This proposition thereforeformalizes how the rulerremains in power and is able to transferresources tohimself thanks to the divide-and-rule strategy. He achieves this asfollows: when threatened by the proposer group, he can always gain the allegiance of the other, proposed group, by shifting resources to them. Because the proposed group ispivotal, the rulercan remainin power ifhe can successfully buy offthe proposed group. Ifthis is the case, anticipating this outcome, neither group will attempt to remove the rulerfrom power, and he will be able to establish akleptocratic regimetransferring resources tohimself atthe expense of the productive groups insociety. The proposition also highlights the notion of constrained kleptocratic re- gime, where the ruleris able to pursue kleptocratic policies transferring re- sources tohimself, but in this endeavor he is constrained by the threat that the two groups will coordinate and remove him frompower. To avoid this possi- Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 185 bility, the rulerreduces the equilibrium taxes (orsometimes sets zero taxes and makes positive transfers)to the two groups. Notice that when s [ S*,the equilibrium does not featurethe notion of punishment, which was discussed inthe introduction. According to this notion, kleptocrats arein power because they can threaten to punish challengers and rewardloyal groups. In this case, both on and offthe equilibrium path, the group that challenges the ruleris taxed atthe rate t* and receives 0transfers. In * e * contrast, when s ¸ S , tj , t ,and ifgroup j challenges the ruler, not only will group i Þ j be bribed to cooperate with the ruler, but also group j will be punished with the tax increased to t*. Next weturn to adiscussion of anumber of natural comparative statics in this model. Most of those areimmediate from the inspection of (25) and (26): 1. Greater Fmakes s [ S* more likely, and when s ¸ S*, greater F increases taxes. This comparative static is intuitive: greater F,that is, greater foreign aid, relaxes the budget constraint of the rulerand provides him with moreresources to buy offthe pivotal group offthe equilibrium path. Therefore, greater F makes the kleptocratic regimeeasier to sustain. This comparative static result suggests that the foreign aid given tomany Africanregimes by the United States and the United Nations during the Cold War period may have had the unforeseen consequence of consolidating kleptocratic regimes. As wedis- cussed in the introduction, this comparative static result may help us understand why in the postwar period, foreign aid appears to have had no positive effecton economic growth on average, and in fact,it mayhave had anegative effecton economic outcomes incertain non-democratic countries. 2. Greater b makes s [ S* less likely, and when s ¸ S*, greater b reduces taxes. Greater b means that both groups aremore patient. Since the beneŽ tof replacing the ruler — greaterreturns in democracy —accrues in the future, greaterpatience makes itless likely that the rulerwill be able to maintain his kleptocratic regime.This comparative static suggests that kleptocratic re- gimes aremore likely to emergein societies wherecitizens or their political representatives value the future less. 3. If b , 1/2, then greater Zmakes s [ S* more likely, and when s ¸ S*, it increases taxes. If b . 1/2,the opposite comparative statics apply. Inspection of (25) shows that greaternatural resource rents createtwo opposing forces. First, like foreign aid, greater Z relaxes the budget constraint of the ruler, and enables him to sustain his kleptocratic regimeby buying offpivotal groups when challenged. Second, greater Z increases the value of democracy. When b , 1/2,the two groups aresuf Ž ciently short-sighted that the Ž rst effect dominates. When b . 1/2,the second effectdominates. The reason why the relevant threshold is 1/2is that in democracy natural resource rents will be divided between the two groups, whereas offthe equilibrium path, the rulercan pay all the rents to the proposed group. Casual empiricismsuggests that the case 186 Journal of the European Economic Association with b , 1/2appears morerelevant here, and suggests that natural-resource-rich countries may be moreprone to kleptocratic regimes. 18 Infact,the comparative static that greaterrents fromnatural resources makekleptocracy morelikely (i.e., the case with b , 1/2) mayhelp us explain why kleptocratic and neopatrimonial regimeshave often emergedin resource-rich countries, such as the DemocraticRepublic of the Congo (Zaire),Sierra Leone, Liberia, and Nigeria. 4. If b . (h/(1 1 h))h,then greater v makes s [ S* less likely, and when s ¸ S*,it reduces taxes. If b , (h/(1 1 h))h,the opposite comparative statics apply. This result can be obtained by differentiating (25). The intuition forthe result may be better understood by considering condition (20), which separately shows the effectof the productivity of the proposed and the proposer groups.

Higher productivity of the proposed group, vi,makes the condition less likely to hold (and kleptocracy morelikely to survive), because the proposed group has moreto gain fromdemocracy. Onthe other hand, agreaterlevel of productivity of the proposer group, vj,makes kleptocracy morelikely to survive, because it implies greatertax revenues that the rulercan use to bribe the proposed group. Consequently, higher average productivity v createstwo opposing forces. When the discount factor b is sufŽ ciently large, that is, b . (h/(1 1 h))h, the own effectdominates and greaterproductivity makes kleptocracy less likely. 19 This comparative static suggests that, aslong as b isnot very low, societies that are otherwise less productive arealso morelikely to sufferfrom kleptocratic regimesand distortionary policies of rulers.

3.5. Equilibrium with Inequality

Letus now return to the case wherethe two groups do not necessarily have the sameproductivity, that is, x [ (0, 1]. First, recallthat despite the differences in productivity between the two groups, the analysis in Section 3.3 established that e e e the most preferred(unconstrained) policy forthe ruleris still ( t1t, t2t, T 1t, e * * T 2t) 5 (t , t ,0, 0). The question is when this policy vector will be possible for the ruler. To answer this question, recallthat s9 5 (v, h, b, Z, F, x)is the vector of * parameters,and let S1 the the set of parameterssuch that C(t , T 5 0, x) $ Z * (b 2 1/2) 2 (1 2 b) F and S2 the set of parameterssuch that C(t , T 5 0, 2x) $ Z (b 2 1/2) 2 (1 2 b)F.In other words, s9 [ S1 implies that ifGroup 2makes aproposal to remove the rulerfrom power, the rulercan makea

18. The relevant time period here is not ayear, so b , 1/2may not be unreasonable. 19. It can be veri Ž ed that (h /(1 1 h))h is monotonically decreasing in h, with (h/(1 1 h))h ¡ e2 1 . 0.37 as h ¡ `.Therefore, for the set of parameters where greater natural resources rents make kleptocratic policies more likely, while simultaneously greater average productivity make them less likely, weneed b [ (0.37, 0.5). Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 187 counteroffer that coopts Group 1even when along the MPEGroup 1 ’s members * aretaxed atthe rate t and receiveno transfers. S2 is the corresponding set for Group 2. Formally, these two sets arede Ž ned by

~1 2 b!v11h~1 2 x!11h h h S 5 s9 : 1 H 1 1 h S 1 1 hD (27) bv11h~1 1 x!11h h 11h 1 2 1 2 $ Z b 2 2 ~1 2 b! F 1 1 h F S 1 1 hD G S 2D J and

~1 2 b!v11h~1 1 x!11h h h S 5 s9 : 2 H 1 1 h S 1 1 hD (28) bv11h~1 2 x!11h h 11h 1 2 1 2 $ Z b 2 2 ~1 2 b! F . 1 1 h F S 1 1 hD G S 2D J

* Since C(t , T 5 0, x)is adecreasing function of x,wehave that S1 , S2. In other words, when the producers in one of the groups become moreproductive they also become morewilling to oust the ruler. Consequently, the tighter constraint facedby the ruleris to satisfy the moreproductive group offthe equilibrium path. This result re  ects the factthat the moreproductive group has moreto gain fromdemocracy, whereits memberswill not be taxed (ormore generally, wherethey will be taxed morelightly). The logic of the political game above thereforeimplies that, everything else equal, the constraints that the ruler has to worry about is Group 2making anofferand Group 1, the moreproductive group, accepting this proposal. This observation and asimilaranalysis to the one in the previous subsection lead to the following proposition:

PROPOSITION 2. Let S1 and S2 be given by (27) and (28). Then we have:

1. When s9 [ S1, then the unique MPEis an unconstrained kleptocratic e e e e * * regime where (t 1t, t 2t, T 1t, T 2t) 5 (t , t , 0, 0) for all tand i 5 1, 2. 2. When s9 ¸ S1 but s9 [ S2, then the unique MPEis apartially constrained e e kleptocratic regime where the equilibrium policy combination is (t 1t, t 2t, e e * ˆ T 1t, T 2t) 5 (tˆ1, t , T1, 0) with 1 C~tˆ , Tˆ , x! 5 Z b 2 2 ~1 2 b! F. (29) 1 1 S 2D

3. When s9 ¸ S2, then the unique MPEis afully constrained kleptocratic 188 Journal of the European Economic Association

e e e e ˆ ˆ regime where (t 1t, t 2t, T 1t, T 2t) 5 (tˆ1, tˆ2, T1, T2) with 1 C~tˆ , Tˆ , x! 5 C~tˆ , Tˆ ,2x! 5 Z b 2 2 ~1 2 b! F. (30) 1 1 2 2 S 2D

This proposition extends Proposition 1to asituation with potential heter- ogeneity in the productivities of the two groups. Italso introduces the notion of partially and fully constrained kleptocratic regimes:as before, anunconstrained kleptocratic regimepursues the policy most preferredby the ruler. When it is partially constrained, the rulerhas to reduce the tax rateon the moreproductive group, but can tax the less productive group as heavily ashe wishes. When the regimeis fully constrained, the tax rateson both groups areconstrained. The comparative static results discussed in the previous subsection continue to apply in this extended model. The new result here is with respect to x, the degree of inequality between the two producer groups. Agreater x— greater inequality between the producer groups —makes the unconstrained kleptocratic regimeless likely (i.e., it makes itless likely that s9 [ S1).Intuitively, the more binding constraint fromthe point of view of the ruleris to satisfy the more productive group: when this group becomes even moreproductive, democracy becomes moreattractive forthe producers in this group, and therefore, it becomes moredif Ž cult forthe rulerto buy them offwhen challenged. This comparative static captures the notion that when there is astrong producer group in society, the rulerhas less room to maneuver, and thereforethe unconstrained kleptocratic regimeis less likely to emerge.Loosely speaking, wecan say that ahighly productive producer group createsa balance of power, and this balance between one of the majorproducer groups and the rulerprevents the most egregious kleptocratic policies. This comparative static result might help us understand why kleptocracies arerare in Africancountries with powerful producer groups, such asthe cattleowners in Botswana or the sugar planters in Mauritius.20

4. Concluding Remarks

Most current political economy researchfocuses on the study of politics in environments whereformal political institutions place effectiveconstraints both on politicians and on political strategies. This approach does not seemsuf Ž cient

20. Note, however, that adifferent timing of events might lead to the opposite comparative static with respect to inequality. For example, if after the two groups agree, the ruler can make a counteroffer to one of the groups, then he will always choose the “weakest link, ” the group more likely to agree, which will be the less productive group. In that case, greater inequality may make kleptocracy more likely. Webelieve that the timing of moves in the political game in the text is more plausible. Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 189 to understand the nature of politics in weakly institutionalized societies such as Liberia, Haiti, the Dominican Republic, or the DemocraticRepublic of the Congo. Inthese polities, differences in formalpolitical institutions cannot be the main determinant of differences in policies. For example, even though the Congo had several constitutions during the rule of Mobutu, it appears that this had little impacton his behavior. For example, Gould (1980, p. 63) shows that the 1974 ZairianConstitution put Mobutu: “above the constitution, giving him the right ...to run forpresident as often as he wishes, to modify the conditions governing elections, to change the number of seats on the Political Bureau [the governing body of Mobutu ’spolitical party and effectively of the state], to dismiss as many political commissioners as he would like, and indeed to abrogate the constitution as he sees Ž t.” Indeed, Mpinga Kasenda, aprofessor of lawand subsequently Zairianprime minister noted in1975 before an audience atthe Institut Royal des Relacions Internacionales in Brussels that (quoted in Turner and Young 1985, p. 70): “Zairians have no need to referto Montesquieu to determine what formof government they need. ” Such asituation is perhaps extreme.Zimbabwe currently has amuch more conventional constitution, written during the LancasterHouse negotiations of 1979 with the British government before the end of the Rhodesian state. Nevertheless, it is clearthat the Mugabe regimeacts in  agrant disregard forthe formalstructure of the constitution. Motivated by these observations, and as a Ž rst step on the road to building aframeworkfor analysis of weakly institutionalized polities, wedeveloped a model of how kleptocratic rulers that expropriate the wealth and incomes of their citizens remainin power without asigni Ž cant base of support insociety. The success of kleptocrats rests on their ability to use aparticular type of political strategy — divide-and-rule. Membersof society (producer groups) need to cooperate in order to depose akleptocrat. The kleptocrat may undermine such cooperation, however, by using tax revenues, the rents fromnatural resources and the funds fromforeign aid to bribe other groups to maintain his position. Thus kleptocrats can intensify the collective action problem by threats which remainoff the equilibrium path. In equilibrium, all areexploited and no one challenges the kleptocrat because of the threat of divide-and-rule. We argue that such amodel provides agood stylistic description of the behavior of many kleptocratic regimes, including that of Mobutu Sese Seko in the Democratic Republic of the Congo (Zaire)and RafaelTrujillo inthe Dominican Republic. Our model not only formalizesdivide-and-rule, but yields anumber of comparative static results useful inthinking about the emergence, longevity, and implications of kleptocracy. We showed that foreign aid and rents fromnatural resources typically help kleptocratic rulers by providing them with greater resources to buy offopponents. Kleptocratic policies arealso morelikely to arisewhen producer groups or their political representatives areshortsighted 190 Journal of the European Economic Association and when the average productivity in the economy is low. Also interestingly, greaterinequality between producer groups mayconstrain kleptocratic policies because moreproductive groups aremore dif Ž cult to buy off. Wediscussed how these comparative static results areuseful in interpreting anumber of cases of kleptocratic rule. We view this paper as only a Ž rst tentative step towards asystematic analysis of equilibrium politics in weakly institutionalized polities. To under- stand the political economy of less developed societies, weneed atheory of behavior and change in weakly institutionalized societies. We developed a simple model to study aspeci Ž caspect of political behavior that seemsto be endemic in such polities, but did not attempt to analyze how weakly institu- tionalized societies maydevelop their institutions, and what makes asociety possess weak institutions in the Ž rst place. These arenot only important, but also exciting areasfor future research.

References

Abreu,Dilip, David G. Pearce,and Ennio Stacchetti (1990). “Towarda Theoryof Discounted RepeatedGames with Imperfect Monitoring. ” Econometrica, 58, 1041–1063. Acemoglu,Daron (2003). “WhyNot a PoliticalCoase Theorem? ” forthcomingin the Journal ofComparative Economics, 31, 620–652. Acemoglu,Daron, Simon Johnson, and James A. Robinson(2003). “AnAfrican Success Story:Botswana. ” In InSearchof Prosperity:Analytic Narratives on EconomicGrowth, editedby Dani Rodrik. Princeton University Press. Acemoglu,Daron and JamesA. Robinson(2000a). “Whydid the WestExtend the Franchise? Democracy,Inequality and Growth in Historical Perspective. ” QuarterlyJournal of Economics, 115, 1167–1199. Acemoglu,Daron and James A. Robinson(2000b). “PoliticalLosers as Barriers to Economic Development.” AmericanEconomic Review, 90, 126–130. Acemoglu,Daron and James A. Robinson(2004). EconomicOrigins of Dictatorship and Democracy, forthcoming. Bates,Robert H. (1981). Marketsand States in Tropical Africa. Universityof California Press. Bates,Robert E. (2001). Prosperityand Violence: The Political Economy of Development. W.W.Norton& Co. Bowman,Larry W. (1991). Mauritius:Democracy and Development in the Indian Ocean. Boulder,CO: WestviewPress. Bratton,Michael and Nicolas van der Walle (1997). DemocraticExperiments in Africa: RegimeTransitions in Comparative Perspective. CambridgeUniversity Press. Brautigam,Deborah (2000). AidDependence and Governance. Stockholm:Almqvist and Wicksell. Buenode Mesquita, Bruce D., James D. Morrow,Randolph M. Siverson,and Alastair Smith (2003). TheLogic of Political Survival. MIT Press. Burnside,Craig and David Dollar (2000). “Aid,Policies and Growth. ” AmericanEconomic Review, 90, 847– 868. Cox,Gary W. (1997). MakingVotes Count. CambridgeUniversity Press. Crassweller,Robert D. (1966). Trujillo:The Life and Times of a CaribbeanDictator. Macmillan. Acemoglu, Robinson &VerdierKleptocracy and Divide-and-Rule 191

Dollar,David and LantPritchett (1998). AssessingAid; What Works, What Doesn ’t, and Why. OxfordUniversity Press. Easterly,William (2001). TheElusive Quest for Growth: Economists ’ Adventuresand Misadventuresin the Tropics. MIT Press. Easterly,William and Ross Levine (1997). “Africa’sGrowthTragedy: Policies and Ethnic Divisions.” QuarterlyJournal of Economics, 112, 1203–1250. Easterly,William, Ross Levine, and David Roodman (2003). “New Data,New Doubts: Revisiting ‘Aid,Policies, and Growth ’.” WorkingPaper No. 26, Center for Global Development.Available at ^http://www.cgdev.org/wp/cgd_wp026.pdf &. Gould,David J. (1980). BureaucraticCorruption and Underdevelopment in the Third World: TheCase of Zaire. New York:Pergamon Press. Green,Edward J. andRobert H. Porter(1984). “NoncooperativeCollusion under Imperfect PriceInformation. ” Econometrica, 52, 87–100. Grossman,Herschel I. (1991). “AGeneralEquilibrium Model of Insurrections. ” American EconomicReview, 81, 912–921. Grossman,Herschel I. andSuk Jae Noh (1994). “ProprietaryPublic Finance and Economic Welfare.” Journalof Public Economics, 53, 187–204. Hartlyn,Jonathan (1998). “TheTrujillo Regime in the Dominican Republic. ” In Sultanistic Regimes, editedby H. E.Chehabiand Juan J. Linz.Johns Hopkins University Press. Herbst,Jeffrey I. (2000). Statesand Power in Africa: Comparative Lessons in Authority and Control. PrincetonUniversity Press. Jackson,Robert H. andCarl G. Rosberg(1982). PersonalRule in Black Africa. Universityof CaliforniaPress. Karl-i-Bond,Nguza (1982). Mobutu,ou, L ’incarnationdu MalZairois. London:Rex Collins. Knack,Stephen (2000). “AidDependence and the Quality of Governance.” PolicyResearch workingpaper no. 2396, World Bank. LaFerrara, Eliana and Robert H. Bates(2001). “PoliticalCompetition in Weak States. ” Economicsand Politics, 13, 159–184. Leith,J. Clark(2000). WhyBotswana Prospered. Workingpaper, Department of Economics, Universityof Western Ontario. Leslie,Winsome J. (1987). TheWorld Bank and Structural Transformation in Developing Countries:The Case of Zaire. Boulder,CO: LynneRienner Publishers. Leslie,Winsome J. (1993). Zaire:Continuity and Political Change in an Oppressive State. Boulder,CO: WestviewPress. Levi,Margaret (1988). Of Ruleand Revenue. Berkeley:University of CaliforniaPress. Linz,Juan J. andAlfred Stepan (1996). Problemsof DemocraticTransition and Consolida- tion:Southern Europe, South America, and Post-Communist Europe. JohnsHopkins UniversityPress. Maddison,Angus (1995). Monitoringthe World Economy 1820 –1992. Paris:OECD Devel- opmentCentre. McGuire,Martin C. andMancur C. Olson(1996). “TheEconomics of Autocracy and MajorityRule. ” Journalof EconomicLiterature, 34, 72–96. Mehlum,Halvar, Karl Ove Moene, and Ragnar Torvik (2001). “Institutionsand the Resource Curse.” Workingpaper, Department of Economics, University of Oslo. Available at ^http://econwpa.wustl.edu/eps/ge/papers/0210/0210004.pdf &. MoyaPons, Frank (1995). TheDominican Republic: A NationalHistory. Princeton,NJ: MarkusWeiner. North,Douglass C. (1981). Structureand Change in Economic History. Norton. North,Douglass C. andBarry R. Weingast(1989). “Constitutionsand Commitment: The Evolutionof Institutions Governing Public Choice in Seventeenth-Century England. ” Journalof Economic History, 49, 803–832. 192 Journal of the European Economic Association

Olson,Mancur C. (1965). TheLogic of Collective Action; Public Goods and the Theory of Groups. HarvardUniversity Press. Olson,Mancur C. (2000). Powerand Prosperity. BasicBooks. Persson,Torsten and Guido Tabellini (2000). PoliticalEconomics: Explaining Economic Policy. MIT Press. Persson,Torsten and Guido Tabellini (2003). TheEconomic Effects of Constitutions: What Dothe Data Say? MIT Press. Robinson,James A. (1998). “Theoriesof BadPolicy. ” Journalof PolicyReform, 3, pp. 1–46. Robinson,James A., Ragnar Torvik, and Thierry Verdier (2002). “PoliticalFoundations of theResource Curse. ” CEPR DiscussionPaper No. 3422. Ross,Michael (1999). “ThePolitical Economy of the Resource Curse. ” WorldPolitics, 51, 297–322. Ross,Michael (2001). “DoesOil Hinder Democracy? ” WorldPolitics, 53, 325–361. Sachs,Jeffery D. andAndrew Warner (1995). “NaturalResource Abundance and Economic Growth,” DevelopmentDiscussion Paper No. 517, Harvard Institute for International Development. Sandbrook,Richard (1985). ThePolitics of Africa ’sEconomicStagnation. CambridgeUni- versityPress. Shepsle,Kenneth A. andBarry R. Weingast(1995). PositiveTheories of Congressional Institutions. Universityof Michigan Press. Turits,Richard Lee (2003). Foundationsof Despotism: Peasants, The Trujillo Regime, and Modernityin Dominican History. StanfordUniversity Press. Turner,Thomas and Crawford Young (1985). TheRise and Decline of the Zairian State. Universityof Wisconsin Press. vander Walle, Nicolas (2001). AfricanEconomies and the Politics of Permanent Crisis, 1979–1999. CambridgeUniversity Press. Wiarda,Howard J. (1968). Dictatorshipand Development: The Method ’sofControl in Trujillo’sDominicanRepublic. Universityof Florida Press. Wintrobe,Ronald (1998). ThePolitical Economy of Dictatorship. CambridgeUniversity Press. Young,Crawford (1965). Politicsin the Congo. PrincetonUniversity Press.