Corruption and Ideology in Autocracies

Transcription

1 Journal of Law, Economics, and Organization Advance Access ublished October, 014 JLEO 1 Corrution and Ideology in Autocracies James R. Hollyer* University of Minnesota Leonard Wantchekon Princeton University Corrution is usually deicted in one of two ways: as stemming from a lack of government accountability, or from a lack of caacity. Neither deiction redicts that the structure of institutions meant to control corrution should vary across autocratic regimes. If corrution results from moral hazard between oliticians and citizens, then all unaccountable governments should eschew anticorrution bodies. If rent-seeking stems from moral hazard between oliticians and bureaucrats, all governments should create anticorrution bodies. We offer an exlanation for why unaccountable governments vary in their willingness to create anticorrution institutions. Autocrats create such bodies to deter ideologically disaffected members of the oulace from entering the bureaucracy. Anticorrution institutions act as a commitment by the elite to restrict the monetary benefits from bureaucratic office, thus ensuring that only zealous suorters of the elite will ursue bureaucratic osts. We illustrate these arguments with case studies of South Korea and Rwanda. (JEL D73, P48) The literature on corrution, broadly seaking, oints to two agency roblems as the source of corrut behavior. One such roblem exists between the oulace and its olitical leaders. A lack of government accountability gives rise to oliticians redatory behavior (e.g., Adersa` et al. 003). Political institutions, such as democracy, may serve as a check on these redatory tendencies. Alternative arguments oint to an agency roblem between governments and their bureaucrats as a source of corrution. The state s lack of olicing caabilities may allow bureaucrats to redate on the oulation, regardless of the intentions of olitical leaders (e.g., Shleifer and Vishny 1993). *University of Minnesota. jhollyer@umn.edu. We would like to thank Jose Antonio Cheibub, Sunkyoung Park, Peter Rosendorff, Kongjoo Shin, Alberto Simser, articiants in the 011 APSA Conference Panel on Audits and Government Accountability in the Develoing World, articiants in the 011 EPSA Conference Panel on the Political Economy of Authoritarian Regimes, three anonymous referees, and Alvin Klevorick in his role as editor for their comments and suggestions. We would also like to thank Vanessa Hofman for excellent research assistance. All remaining errors are our own. The Journal of Law, Economics, and Organization, Vol. 0, No. 0 doi: /jleo/ewu015 ß The Author 014. Published by Oxford University Press on behalf of Yale University. All rights reserved. For Permissions, lease journals.ermissions@ou.com

2 The Journal of Law, Economics, & Organization Neither of these accounts, however, exlains the substantial variation in the institutions designed to olice corrution under autocratic rule. 1 If corrution is the result of the redatory behavior of oliticians, we would not exect these same oliticians to erect checks on their redation. If corrution stems from agency roblems between olitical elites and bureaucrats, we would exect all governments to take stes to redress these roblems including the creation of effective bodies designed to olice corrution. In fact, nondemocratic regimes vary greatly in the extent to which they emower indeendent anticorrution bodies. For instance, in the Peole s Reublic of China, the agencies nominally charged with corrution investigations are tyically subservient to the regional-level arty aaratuses they are meant to monitor. These bodies exercise little indeendence and thus, likely do little to curb corrution (Sun 004). But, in other instances, anticorrution institutions are given real indeendent ower desite the fact that those most liable to rosecution are servants of the regime. For instance, the Park regime founded the Bureau of Audit and Insection in Korea and the Peole s Action Party gave teeth to the Corrut Practices Investigation Bureau in Singaore (Quah 1999). In this article, we exlore an alternative concetion of corrution that, we argue, hels to exlain variation in the creation of anticorrution bodies under autocratic rule. We contend that authoritarian rulers may tolerate, or even suborn, corrut behavior as art of an incentive scheme designed to reward their suorters. However, the use of corrution in this manner gives rise to an adverse selection roblem: oortunistic citizens may seek osts within the regime simly to gain access to corrution rents. Indeendent anticorrution agencies may act as a credible commitment to curtail corrution rents, thus ensuring the regime may staff ositions with ideologically zealous suorters. 1. Argument We contend that authoritarian states have an incentive to systematically rely on corrution as a means to motivate bureaucratic agents and ruling arty members. While corrution may imose large economic costs, it may also rove an oortune means of roviding ecuniary benefits to those carrying out the rulers will. In this sense, systematic corrution acts as a tool by which governments resolve moral hazard roblems involving their subordinates. 3 Corrut behavior and the efforts exerted 1. We focus on instances in which the ublic is unable to unish olitical leaders for corrut behavior. We classify regimes where the threat of electoral sanction is absent as autocracies.. Wintrobe (1998), for instance, claims that the distribution of olitical rents can be used to urchase loyalty under autocratic rule. 3. Leff (1964) argues that bribery may induce efficient erformance by corrut officials as bribes constitute an equivalent to a iece wage. Here we are less concerned with the incentives

3 Corrution and Ideology in Autocracies 3 by subordinate officials, therefore, act as strategic comlements. The regime regulates the oortunities for corrution, and rewards its most loyal agents with the ability to artake in rents. Thus, the most corrut officials are also likely to be the most loyal. Corrution, in our treatment, is tightly controlled by the government it is not a manifestation of a lack of government control. 4 For instance, the government may assign high-erforming officials to osts made lucrative by access to corrution rents. 5 Loyal bureaucrats may be assigned to customs or rocurement offices where there are amle oortunities to solicit bribes. By maniulating assignments in this manner, the elite increases the career concerns of lower-level officials. Alternatively, the government may tolerate corrution by officials that toe the arty line, but unish it by those that do not. 6 This selective maniulation of corrution or of the unishments officials may exect to face for corrut activities may be seen as equivalent to the use of an efficiency wage. Corrution serves to increase the benefits from office and heightens the exected costs from removal. This maniulation of corrution serves as a substitute for the use of high owered wage incentives. Rewarding officials through access to corrution rents may be referable to wage incentives for a variety of reasons. First, such rewards can be rovided at relatively low cost to the elite. The rulers need only turn a blind eye to the corrut activities of roductive officials rather than raising and distributing the funds for their wages. Second, corrution incentives can be maniulated in a nontransarent manner. This oacity may be useful insofar as it rovides a shield from ublic scrutiny and an autocratic elite may wish to encourage officials to behave in a manner viewed as undesirable by the ublic. Oacity also inhibits collective bargaining by officials whether exlicitly through unions or imlicitly through informal networks. Access to corrution introduces for bureaucratic erformance with regards to the ublic, and more concerned with corrution s effect on the willingness of bureaucrats and arty members to serve the elite. The moral hazard roblem we focus on exists between the elite and its agents. 4. We abstract from corrut behavior that arises in defiance of the ruling elite in our account. We note, however, that the additional rents that would accrue to subordinate officials from such corrution would enhance the returns to office, worsening the adverse selection roblems we document below. 5. For the maniulation of assignment to osts as a means of controlling bureaucrats, see Iyer and Mani (01). Lazarev (007) documents how nondemocratic regimes may use the assignment of lum ositions those attracting large rents from office as a means of attracting large numbers of new recruits. 6. Examles of this behavior abound: Urban (1985) notes that the Soviet Union imlicitly encouraged officials to engage in rohibited corrut behaviors, and would selectively rosecute those that did not show sufficient zeal in serving their sueriors. During the early 1990s in the PRC, the Shanghai rosecutor s office announced that able individuals would be granted leniency in corrution cases if they reented for their acts (Sun, 001). Darden (008) documents similar behavior under the Kuchma regime in Ukraine. Regional officials that were insufficiently zealous in backing the regime were often threatened with investigations and rosecution.

4 4 The Journal of Law, Economics, & Organization corrution can thus be maniulated to target the behavior of individual bureaucrats in a manner that is difficult through wage contracts. Unlike the tyical official wage contract for bureaucrats, corrution incentives are likely to be high owered. 7 The use of corrution rents as an incentive mechanism shaes ublic exectations about the returns to office. Members of the ublic exect that loyal government servants will escae corrution rosecutions, regardless of the extent of their corrut acts. They believe that the government will assign lucrative osts as a reward for revious behavior and these beliefs will drive their exectations about the otential rewards of government service. As a result of these beliefs, such service will likely seem an attractive otion to a broad swath of the ublic, including those who have no articular love for their rulers. Since the ruling elite cannot directly observe the ideological redilections of those seeking government osts, and since those seeking such osts have every incentive to disguise their true beliefs, individuals disaffected with the ruling regime may enter government service and crowd out true ideological adherents. For instance, after assuming ower in 1933, the Nazi Party was deluged with new members, many of whom were current civil servants or were seeking bureaucratic aointment. (Calan 1988, ) claims that, as civil servants sensibly flocked to join a arty that ut such a remium on olitical affiliation, so they devalued the meaning of membershi as well as altering the character of the arty itself. In theory, the ruling elite could avoid such a roblem if it could commit to rein in corrution. Should the elite crack down on corrut behavior, the ecuniary benefits from bureaucratic office would decline and oortunists would be deterred from serving the regime. But we argue that, absent constraining institutions, the ruling elite cannot commit to such behavior, giving rise to an adverse selection roblem. This commitment roblem arises from two underlying causes. First, in an authoritarian system, the ruling elite faces few constraints in its ursuit of its own self-interest. Second, attemts to combat corrut behavior are likely to be costly for the elite. Policing corrut behavior is likely to require greater resources than are necessary to maniulate access to corrution rents. More subtly, the unishment of lower-level bureaucrats for corrut deeds may involve substantial costs. The rosecution and relacement of sitting officials is likely to sacrifice skill-secific human caital built u over time served in office (Gailmard and Patty 007). Given the imerfect nature of monitoring technologies, such rosecutions may result in the removal of loyal servants from office. Finally, the rosecution of sitting officials is likely to cast a negative light on those resonsible for their 7. On the revalence of low-owered wage contracts for bureaucratic officials see Dixit (00) and Tirole (1994). The incentives we describe are analogous to those emloyed by olitical machines in democracies (Banfield and Wilson, 1963; Robinson and Verdier, 013).

5 Corrution and Ideology in Autocracies 5 aointment, otentially jeoardizing the standing of at least some members of the elite. Consequently, the ublic will not find autocratic threats to limit corrution credible unless these threats are backed by the creation of constraining institutions. Indeendent anticorrution institutions may credibly act to limit corrut behavior without interference by the ruling elite. Delegation to indeendent anticorrution institutions gives rise to yet another commitment roblem: How can an elite commit to ensure the continued indeendence (or indeed continued existence) of the anticorrution bodies it creates? For anticorrution agencies to address the autocratic elite s credibility roblem, the elite must suffer some cost from abolishing or subverting these agencies (Jensen 1997). We contend that an authoritarian elite seeking to undermine an indeendent anticorrution body would likely incur such costs, largely because attemts to subvert the anticorrution institution would be layed out in ublic. 8 Case studies of anticorrution commissions in the develoing world reveal that success or failure of these bodies deends largely on institutional features notably the scoe of their investigative and rosecutorial authority. Once such authority is granted, subtle attemts to reduce the effectiveness of anticorrution agencies through ersonnel changes or cuts are unlikely to be effective. 9 For the elite to eviscerate an established anticorrution body, it is likely necessary to alter the mandate given to that institution. Changes to the scoe of the authority granted to anticorrution institutions likely require statutory changes which may attract ublic notice. This danger is articularly great given that many anticorrution commissions are charged with ublic outreach. 10 The sectacle involved in undermining anticorrution bodies is likely to be articularly rominent and damaging if these entities have already launched effective investigations or rosecutions. 11 Credible anticorrution may, therefore, enable the elite to screen jobseekers by restricting corrution rents. Credible anticorrution bodies 8. This claim mirrors Svolik (01) and Boix and Svolik (013) in ointing to the informational role of institutions in autocracies. 9. Bolongaita (010), for instance, notes that the Corrution Eradication Commission (KPK) of Indonesia was able to secure a large number of high-level convictions with a relatively small number of ersonnel (580 staffers). Lawson (009) finds that the Economic and Financial Crimes Commission (EFCC) in Nigeria was aggressive in ursuing olitical corrution desite the aointment of a chief that was handicked by the resident. 10. On the duties of anticorrution commissions in 18 countries, see Charron (008). On the secific imortance of ublic outreach to Botswana s anticorrution commission, see Heilbrunn (004). 11. This is not to contend that all anticorrution institutions will rove effective (Heilbrunn, 004). We merely contend that conditional on having established an effective anticorrution institution in time t the ruling elite will find it costly to abolish that institution in time t þ 1.

6 6 The Journal of Law, Economics, & Organization offer a ublic setting for the elite to demonstrate its commitment to refrain from corrut behavior and to discourage corrut activities on the art of its subordinates. As members of the ublic and arty adjust their exectations regarding the rewards from office following such a commitment, those less ideologically aligned with the leadershi s ositions will leave or refrain from entering the bureaucracy.. Related Literature This article builds on an emerging literature on the functioning of nondemocratic states and the role of institutions therein. Traditionally, most analysts have viewed nondemocratic regimes as constrained only by the effects of their resent actions on future consumtion (see, for instance, McGuire and Olson 1996), or by the threat of mass revolution (Acemoglu and Robinson 000). However, a more recent literature suggests that nondemocracies may rationally seek to build institutions to constrain their ower usually to overcome some commitment roblem. Myerson (008) argues that an autocrat may romote the creation of institutions that allow subordinates to coordinate his ouster as this allows the autocrat to credibly commit to reward these subordinates for their suort (for a similar argument, see Gehlbach and Malesky forthcoming). Gandhi and Przeworski (006) argue that dictatorshis may commit to share rents with oosition grous by including the oosition in a legislature. We build on these arguments by suggesting that autocratic governments may seek to limit their ability to reward officials to shift the ideological comosition of the ool of recruits to these ositions. We also build on the literature on agency roblems in nondemocratic governments. Egorov et al. (009) and Egorov and Sonin (011) examine roblems of moral hazard in dictatorshis. These articles argue, resectively, that some dictatorshis may have an incentive to encourage freedom of the ress to increase monitoring of bureaucratic agents and that dictatorshis have an incentive to romote less cometent agents than democracies, given the danger that cometent bureaucrats will stage a cou. Dixit (010) argues that autocratic governments are less willing to share olicy rents with bureaucrats than democratic governments and hence derive less effort from officials. We build on this literature by examining roblems of adverse selection and articularly adverse selection with resect to ideology in nondemocratic governments. We also borrow from the literature on agency roblems with motivated agents. Besley and Ghatak (005) examine the henomenon in which bureaucrats develo a formal sense of mission, and note that the existence of such ideological interests allows the rincial to relax high owered incentives. Prendergast (007) examines the ideological motivations of regulators and characterizes the situations in which biased bureaucrats are referred over nonideological alternatives. Prendergast also notes

7 Corrution and Ideology in Autocracies 7 that ideological motivations may lead to adverse selection roblems in recruiting agents. We characterize just such a roblem here. Finally, we build on an extensive literature on the institutional determinants of corrution. One strand of this literature focuses on how olitical accountability may limit government redation (Barro 1973; Ferejohn 1986). More recent work has emhasized how the effects of olitical accountability may be moderated by the amount of information available to the electorate (Adsera` et al. 003; Brunetti and Weder 003; Ferraz and Finan 008), by the resence or absence of checks and balances (Persson et al. 1997; Persson and Tabellini, 000), by the clarity of lines of resonsibility (Kunicova and Rose-Ackerman 005), or by the role of oosition arties (Kunicova 00). Here we offer an exlanation for how ideology may influence corrut behavior in the absence of electoral accountability. An alternative institutional argument stresses the imortance of state caacity for corrution. States that lack administrative caacity may rove unable to olice the bureaucracy for corrut behavior. Weak governments may be unable to coordinate bureaucratic corrution, resulting in articularly damaging cometing demands for bribes (Shleifer and Vishny 1993). Countries with inadequate fiscal caacity may rely on corrution to ensure that civil servants receive wages sufficient to meet their articiation constraint (Besley and McLaren 1993). Here we offer an argument as to how ideology may influence the level of corrution even within the set of caable states. 3. Model 3.1 Model Primitives Consider an interaction between two classes of layers: a government leadershi (L) and a ool of otential bureaucratic recruits, where an individual recruit is denoted i and i is indexed over the unit interval i ½0; 1Š. L will staff a single bureaucratic ost with a recruit selected from the ool of otential recruits. Potential recruits are characterized by their level of ideological affinity with the arty in ower i f;g, where 0 < <<1. 1 We label individuals with values of i ¼ as zealots and those with values of i ¼ as oortunists. We further denote a term ; ð1; 1 Þ, which reflects the extent of ideological olarization within the recruitment ool. As increases, the difference in the fealty of zealots and oortunists rises. The fraction of zealots in the recruitment ool is given by the arameter, while the fraction of oortunists is given by Of course, in reality otential recruits may be characterized by factors other than their ideology notably by their level of cometence. We see no reason to exect that cometence is systematically correlated with ideological suort, and abstract from any consideration of otential recruits cometence in the model.

8 8 The Journal of Law, Economics, & Organization Potential recruits must determine whether to seek office and conditional on entering office the level of effort they wish to devote to serve the leadershi e i 50. Should she not enter the bureaucracy, each otential recruit earns a rivate sector wage y i ¼ y50. If she enters the bureaucracy, she will enjoy a fixed official wage rate w in addition to rents from office r, both of which are constrained to be non-negative. 13 We assume that rents are awarded by the leadershi in a manner that is linear in roduction. Effort imoses a constant marginal cost equal to one on bureaucrats. When in office, the bureaucrat roduces goods for the leadershi according to the function gðe i Þ¼e 1 i. Potential recruits value both income and the ideological returns to government service. If she enters office, each recruit i will, therefore, enjoy ideological returns from service equal to i gðe i Þ and rents equal to rgðe i Þ. Her utility is thus: ( u i ð i ; e i Þ¼ w þðrþ iþgðe i Þ e i given entry ð1þ y i otherwise: The leadershi determines both the wage rate aid to the official w, and the level of rents she may enjoy from service r. The ool of funds used to rovide wages is derived from taxes which, for simlicity, we treat as exogenous and fixed at T > 0. In keeing with the literature on bureaucratic olitics (Dixit 00; Tirole 1994), we assume that official wages are unresonsive to effort levels that is, they are contractually fixed by the leadershi at the value w. By contrast, we contend that autocratic governments may maniulate access to rents to reward the erformance of bureaucrats, roviding relatively high-owered incentives for erformance. A second crucial difference between these two forms of remuneration lies in the leadershi s ability to commit to each ayment scheme. Official wages are fixed through contracts and law. Public sector unions (or informal grous of workers) may unish violations of official wage commitments through collective rotest and strikes. Remuneration 13. The assumtion that official wages are constrained to be non-negative is not innocuous. As we demonstrate below, reductions in official wages act as an alternative means for addressing adverse selection roblems in bureaucratic recruitment. Were these wages infinitely flexible, no need for anticorrution commitments would arise. However, ractical constraints limit the ability of governments to imose negative official wages that is, to charge recruits fees for osts. While some governments have exerimented with requirements that bureaucrats ost bonds for office, these schemes have not been widely adoted largely due to difficulties identified by Dickens et al. (1989). Requirements that bureaucrats ay for osts are likely to give rise to a variety of additional selection issues, largely due to the resence of credit/liquidity constraints for a discussion of related issues see Hollyer (011). Beyond the selection issues osed by the sale of offices, Guardado R. (013) finds evidence that the sale of office directly undermines bureaucratic functions through mechanisms exogenous to those discussed in this article.

9 Corrution and Ideology in Autocracies 9 schemes centered on corrution rents, however, are surretitious by nature. For obvious reasons, they are never secified in contracts and the exact terms and levels of this remuneration are unlikely to be common knowledge even within the bureaucracy. Given the lack of an institutional framework for corrut remuneration schemes, we assume that the leadershi cannot commit ex ante to a given rate of corrut remuneration for otential recruits. Finally, the leadershi may choose its level of anticorrution commitments. This may be thought of as reresenting a government s decision of whether to establish an indeendent anticorrution office, or to enter into an internationally sonsored anticorrution rogram. We treat these commitments as a continuous choice arameter 51 where ¼ 1 denotes an absence of anticorrution institutions and the extent of these institutions mandates or resources is monotonically increasing in. The resence of anticorrution institutions raises the costs to the leadershi of engaging in or suborning rent-seeking. The costs to the leadershi of rent-based remuneration is thus given by R, where R ¼ rgðe i Þ denotes the ool of rents raised for such schemes. Anticorrution bodies imose an additional direct cost on the elite. To the extent that such bodies are effective, they constrain corrut behavior at both the elite and the etty level. We, therefore, assume that increasing the level of imoses a direct enalty on the leadershi, where the extent of this enalty is exressed by the arameter >0. 14 The leadershi s utility is a function of the roduction of the lower-level official gðe i Þ which may include, but is not limited to, her official duties; the budget surlus T w; the costs of rent-seeking R; and the extent of constraints on elite corrution ð1 Þ. We denote the leadershi s utility as follows: u L ðr; w;þ¼gðe i ÞþT w R þð1 Þ; which it must maximize subject to the constraints that R ¼ rgðe i Þ, and T5w, in addition to the non-negativity constraints on r and w. In the event that the leadershi is unable to staff the oen ost, we assume that it receives an arbitrarily large negative utility as a enalty. The order of lay of the game is as follows: 1. The arty leadershi sets an official wage rate w ½0; TŠ and chooses its level of anticorrution commitments 51.. All otential recruits declare their willingness to enter office. Of those who exress a willingness to enter, one is randomly chosen as an entrant. ðþ 14. might also reresent the direct costs of creating an effective anticorrution institution that is, wages, equiment or mechanisms to ensure their indeendence. These costs may be thought to rise in the extent of the anticorrution body s authority. We would like to thank an anonymous reviewer for raising this oint.

10 10 The Journal of Law, Economics, & Organization 3. The leadershi sets a contract for corrut incentives r50. i chooses her level of effort e i and roduction takes lace. 4. Corrution revenues R are raised, wages and rent-based incentives are aid, and all ayoffs are realized. The game is solved using backward induction, alying the erfect Bayesian equilibrium solution concet (Fudenberg and Tirole 1991) Equilibrium Once in office, a bureaucrat will devote effort in a manner that maximizes equation (1), conditional on the level of corrution incentives r and on her affinity with the ruling elite i. This maximum will be given by the following exression: e i ¼ðr þ i Þ : The leadershi will set values of w and r in light of the effort decision of the bureaucrat. The leadershi, thus, seeks to maximize the following: max w;r; E½gðe i ÞþT w Rþð1 ÞŠ s:t: R ¼ rgðe i Þ r50 04w4T where E½ : Š is the exectations oerator. The exectations oerator is necessary in this instance as the leadershi may be uncertain of the tye of recruit it faces. If equilibrium rents and wages are sufficiently high or rivate sector oortunities sufficiently oor (y sufficiently low) both oortunists and zealots may seek bureaucratic osts. If this is the case, then the recruit is a zealot with robability and an oortunist with robability 1. If, however, the returns to ublic office are relatively low, then only zealots will be willing to serve. In this instance, the leadershi can be certain of the tye of recruit with whom it is dealing and may set rents accordingly. We denote L s osterior beliefs over candidate ideology, given entry, as q Prð i ¼ j entryþ; q f; 1g. The leadershi s maximization roblem imlies that equilibrium rents are given by: r ¼ max 0; 1 q þð1 qþ : ð4þ ð3þ 15. We assume, throughout, that the bureaucracy is staffed by a single recruit. This is equivalent to assuming that there are enough zealots in the oulation of otential recruits to staff the bureaucracy.

11 Corrution and Ideology in Autocracies 11 Since the equilibrium rent r deends on q, which reresents L s osterior beliefs over the bureaucrat s tye given entry, the leadershi is unable to commit to a rate of rent-based remuneration. Ceteris aribus, rents decline in q because L must offer higher rates of remuneration to the bureaucrat if it believes that this recruit may be an oortunist. The leadershi cannot, absent constraining institutions, set these rates at a level sufficiently low to deter oortunist entry. Leaders must deal with the ool of bureaucratic recruits they have, not the ool they may wish to have. The critical question is, therefore, whether oortunists are willing to serve, given this equilibrium exression for the value of rents. Recall that otential recruits derive a rivate sector wage of y > 0. A given otential recruit i, therefore, will be willing to enter office if and only if: w þð i þ r Þgðe i Þ e i > y: An adverse selection roblem arises when inequality (5) is satisfied for oortunistic otential recruits. The leadershi ossesses two tools through which it may redress such an adverse selection roblem. First, it may lower the contractual wage rate w. This serves to reduce the exression on the left-hand side of inequality (5), directly dissuading oortunists from entering office. Second, the leadershi may erect anticorrution institutions that is, raise the value of. This serves to reduce the level of rents r candidates may exect from office, as is given in exression (4). Lowering the level of rents serves to reduce the left-hand side of inequality (5), again dissuading oortunists from seeking osts. Of the two tools at its disosal, the leadershi always refers to solve its adverse selection roblem via wage reductions. Reductions in the value of w are doubly beneficial to the leadershi: they both directly imrove its utility by increasing the budget surlus, and indirectly increase its utility by diminishing the incentive for oortunists to enter the bureaucracy. However, for certain configurations of arameter values, the leadershi cannot solve its adverse selection roblem through reductions in the official wage alone. Under these circumstances, it may rely on the erection of anticorrution barriers to dissuade oortunists from entering osts. The need to rely on anticorrution barriers is costly for the leadershi they both raise the costs it faces from using rents to motivate bureaucrats and they directly diminish the elite s ability to consume rents for itself. Consequently, the leadershi will rely on anticorrution bodies only when adverse selection issues ersist even after official wages are reduced to the lowest ossible level that is, w ¼ 0. Lemma 1 characterizes the circumstances under which such severe adverse selection roblems exist. Lemma 1. When w ¼ 0 and ¼ 1, oortunists will seek to enter the bureaucracy iff > ffiffi y þ 1. When > ffiffi y þ 1, the leadershi may deter oortunists from entry by increasing iff 4 ffiffiffi y. ð5þ

12 1 The Journal of Law, Economics, & Organization Lemma 1 secifies two thresholds for. If the ideological affinity oortunists feel for the elite is sufficiently low, the leadershi may redress its adverse selection roblem by maniulating official wages. That is, the leadershi may set some wage rate w > 0 such that zealots are willing to enter the regime s service, while oortunists are not. If, by contrast, is sufficiently high, the leadershi s adverse selection roblem is insoluble. Recall that anticorrution commitments serve to redress this roblem only by reducing the value of rents r. If oortunists are sufficiently ideologically motivated, they will enter office even when both w and r are ushed to zero. The leadershi will, therefore, be faced with an adverse selection roblem, which it may choose to solve through anticorrution commitments if and only if ð ffiffi y þ 1 ffiffi ; y Š.If falls outside of this interval, the leadershi will never choose to erect anticorrution barriers that is, it will set ¼ 1 either because it can address its adverse selection roblem through other methods or because it realizes this roblem cannot be solved. In Aendix, we demonstrate that this interval is nonemty if rivate sector wages y are sufficiently low. 16 From Lemma 1, it follows that when is sufficiently high, the equilibrium value of official wages will be zero. L will set a ositive official wage rate w only to satisfy the articiation constraint of zealots. When 5 ffiffi y þ 1 and anticorrution commitments are absent, the articiation constraint for all otential recruits is satisfied when w ¼ 0. Hence, in equilibrium w ¼ 0. We state this result in the following remark: Remark 1. When ffiffi y þ 1 <, w ¼ 0. To examine the creation and extent of anticorrution commitments, we will assume that the leadershi can solve its adverse selection roblem through such commitments. This amounts to a restriction on the value of : Assumtion 1. < ffiffiffi y Assumtion 1 guarantees that arameter values are such that the elite can solve any adverse selection roblem by raising the value of. However, since increasing the value of is costly for the leadershi, we must characterize the conditions under which L refers to raise the value of to a level sufficient to deter oortunist entry. The leadershi must comare the costs and benefits of acceting adverse selection in bureaucratic recruitment instead of solving this roblem. ffiffiffi 16. More recisely, the interval is nonemty for y < 1 1. Thus, the interval is nonemty for any ; if y < 1 that is, the marginal cost of effort. As! 1, it is nonemty for any y R þ. The range of values ;y for which it is nonemty is rising in, and the range of values ; y for which it is nonemty is decreasing in.

13 Corrution and Ideology in Autocracies 13 To undertake this comarison, we first characterize the value of necessary to deter oortunist entry, which we denote : ¼ maxf1; g ð6þ where 1 ¼ ffiffiffi y þ ðþ ð7þ We derive by setting the equilibrium remuneration an oortunist can exect from attaining office equal to her articiation constraint when w ¼ 0. For > ffiffi y þ 1, such that oortunists have an incentive to seek government osts, takes on a value strictly greater than one secifically, ¼. For 4 ffiffi y þ 1, oortunists eschew osts, even absent anticorrution commitments. Thus, the value of is equal to 1. We can now consider L s decision with regard to equilibrium levels of. On the one hand, L may raise to and ensure that the bureaucracy is staffed only with the most zealous ossible recruits. (It will never raise > since, as noted above, this serves only to imose a cost by limiting the elite s ability to consume rents.) Alternatively, when ¼ > 1 (i.e., when > ffiffi y þ 1 ), L may set ¼ 1 and accet the otential resence of oortunists in the bureaucracy. Proosition 1 characterizes this decision for different ranges of the arameter. Proosition 1. If 4 ffiffi y þ 1, then L will set ¼ ¼ 1. If ð ffiffi y þ 1 ; ffiffi y ð1 Þ Š, L will set ¼ > 1 iff the following inequality holds: 1 þ 1 þ Þþð1 Þ5 ½1þ½1þqð 1ÞŠŠ. Otherwise, L sets ¼ 1. If > ffiffi y ð1 Þ, L will set ¼ > 1 iff the following inequality holds: þð1 Þ5 ½1þ½1þqð 1ÞŠŠ. Otherwise, L sets ¼ 1. Proosition 1 divides the arameter sace for into three ranges, defined by two inequalities. The first inequality, < ffiffi y þ 1, follows directly from Lemma 1. For sufficiently low levels of, no oortunist seeks entry into the bureaucracy, and thus the equilibrium value of is equal to ¼ 1. The second inequality, ffiffi y ð1 Þ >, dictates whether the equilibrium rent rate r is ositive when ¼ > 1. Thus, if ð ffiffi y þ 1 ; ffiffi y ð1 Þ Š, L may solve its adverse selection roblem by erecting anticorrution barriers that lower equilibrium rents r. If > ffiffi y ð1 Þ, then L can successfully deter oortunists from

14 14 The Journal of Law, Economics, & Organization entry only by raising barriers to corrution to such levels that rents are eliminated altogether (r ¼ 0). 3.3 Comarative Statics We now turn our attention to the comarative statics of the model, which will be exressed in terms of the equilibrium level of. The equilibrium value of will be dictated by two considerations. First, what level of commitment is necessary to deter oortunists from office? Second, given this value, is the leadershi willing to undertake this commitment? We first state the following lemma: Lemma. There exists a threshold ^ ¼ 1 ffiffi ½ð1 þ Þ þ 1 y Š such that for 5^, the equilibrium value of is given by ¼ 1. This threshold is defined by the condition under which oortunists seek to enter the bureaucracy when ¼ 1. When exceeds this threshold, no oortunists ever seek to enter the bureaucracy and ¼ 1. In other words, will always fall into the first range of the arameter sace defined by Proosition 1, < ffiffi y þ 1, when 5^. By contrast, when <^; ¼ > 1. L can solve its adverse selection roblem only by erecting anticorrution barriers. Having stated this lemma, we now consider the effects of varying the degree of ideological olarization within the ool of otential recruits. To do so, we fix the affinity of oortunists, which must be as secified by Assumtion 1, and vary the affinity of zealots. This is equivalent to varying the arameter, where higher values of denote higher levels of ideological olarization. Proosition. Consider two values of f 0 ; 00 g, where 0 < 00. Each value of has an associated value of f 0 ; 00 g where If L sets ¼ 0 when ¼ 0, then it also sets ¼ 00 when ¼ 00. Loosely seaking, as ideological olarization rises, L grows (weakly) more willing to establish anticorrution commitments. The intuition for this result is straightforward: An increase in the level of olarization imlies that the desirability of recruiting a zealous suorter (as oosed to an oortunistic one) is rising. The cost of any adverse selection roblem becomes more severe as olarization rises. Thus, the leadershi is more willing to bear the cost of anticorrution commitments when is high than when it is low. This relationshi is weak because, for some arameter values, L will never be willing to enact anticorrution commitments. Recall that these commitments limit the ability of the elite to consume rents for themselves. The extent of the resultant costs is arameterized by the term. When is sufficiently high, the elite will never erect anticorrution barriers,

15 Corrution and Ideology in Autocracies 15 regardless of the extent of ideological olarization. We summarize these results with the following remark: Remark. Define a value of such that, for 5, ¼ 1 for all values of. For <, there exists a corresonding threshold value of ðþ such that L will set ¼ 51 for any 5ðÞ and will set ¼ 1 for any <ðþ. Proosition does not suffice to characterize the equilibrium value of because, the level of anticorrution commitments necessary to deter oortunists when > ffiffi y þ 1, deends on the value of. To characterize the relationshi between equilibrium levels of anticorrution commitments and olarization, it is necessary to consider the relationshi between and. The value of is given by exression (7). Holding oortunists affinity for the regime fixed, an increase in zealots zealotry or, equivalently, an increase in the value of decreases the value of given by this exression. Hence, is decreasing in values of, and thus the value of is weakly decreasing in. The intuition for this claim is as follows: holding all else constant, as the affinity zealots feel for the regime,, rises, the rate of rents, r, decreases. In exectation, zealous recruits are willing to devote greater effort for less ecuniary reward; thus L has an incentive to lower rents. However, as the rate of rents falls, oortunists are less willing to seek office. Thus, L can deter oortunists entry with a lower value of anticorrution commitments. This result indicates that the relationshi between ideological olarization and the equilibrium level of anticorrution commitments may be nonmonotonic. Remark allows us to state this claim more recisely: Proosition 3. When <, the equilibrium value of is nonmonotonic in. For values of <ðþ, the equilibrium value of ¼ 1. At ¼ ðþ, the equilibrium value of jums to ¼ > 1. For all values of >ðþ, the equilibrium value of ¼ 51 and is (weakly) decreasing in. When 5, the equilibrium value of ¼ 1 for all values of. We lot the equilibrium value of for varying values of in Figure 1. We now turn our attention to shifts in the roortion of zealots in the recruitment ool. We state the relationshi between and the equilibrium value of in the following roosition: Proosition 4. The equilibrium value of is weakly decreasing in. To be more recise, the equilibrium value of is strictly decreasing in for some configuration of arameter values and invariant for others. In the following remark, we clarify the conditions under which values of are decreasing in. Remark 3. If, when ¼ 1, 4^, we can define a threshold value of ~ such that, for all 4 ~, the equilibrium value of ¼ 51, and for all > ~,

16 16 The Journal of Law, Economics, & Organization Comarative Statics λ Figure 1. The Equilibrium Value of Anticorrution Commitments as a Function of Polarization. The equilibrium value of as a function of. Parameter values are fixed such that ¼ 0:01, ¼ 0.5, and ¼ 0:1. the equilibrium value of ¼ 1. is strictly decreasing in, and thus is weakly decreasing in. ~ is interior to the unit interval for sufficiently small values of. A decrease in the value of has two effects. First, and most directly, it worsens any adverse selection roblems faced by the elite. As the recruitment ool becomes increasingly dominated by oortunists, the odds that any given recruit is a zealot fall. Consequently, redressing this roblem by enacting anticorrution commitments becomes more attractive. Second, and less directly, the decline in the roortion of zealots imlies that the elite must rely more heavily on ecuniary incentives to motivate bureaucratic effort. Holding all else constant, the value of r rises as falls. To deter oortunists from seeking office, therefore, L must erect more severe anticorrution barriers. Thus, the value of increases. Of course, as ¼ maxf1; g increases, the costs the leadershi faces in solving its adverse selection roblem also rise. Thus, as falls, both γ

17 Corrution and Ideology in Autocracies 17 the costs and benefits of deterring oortunists increase. For sufficiently low values of, the rise in benefits outaces the rise in costs for high values of, this is no longer true. To summarize the results of our comarative statics examination: anticorrution commitments in autocratic regimes tend to fall as the breadth of the regime s suort () rises and are nonmonotonic (first rising then falling) in the level of ideological olarization. Adverse selection in bureaucratic recruitment is most costly when the regime enjoys the suort of only a small number of zealous backers, who are likely to be crowded out of the recruitment rocess by oortunists. The incentive to screen candidates is thus highest when is small. Similarly, as the olarization between zealots and oortunists () rises, the incentive to screen candidates increases. But since the rate of rents, and thus the ecuniary rewards from office, decline in so too does the level of commitment necessary to achieve this screening effect. Thus the equilibrium value of first rises, then falls, as olarization increases, as Figure 1 illustrates. 3.4 Extension: Observability of Ideology Until this oint, we have considered the ideological affinity of a otential recruit for bureaucratic office i as wholly unobservable. This assumtion seems reasonable, as the olitical elite is unable to observe the thoughts of those seeking osts. Moreover, those wishing to obtain a osition within the government have every incentive to disguise their true beliefs. But, in many instances, the ideological beliefs of otential recruits are correlated with observable factors. For instance, cues such as ethnicity, lace of birth, or arental occuation may correlate with ideological affinity for the regime. 17 Ethnicity may lay a articularly imortant role in this regard. Ethnic identities are observable and difficult to change or conceal (Chandra 006) and, in many olities, olitical leaders rely heavily on ethnically based aeals to maintain suort. To incororate observable tyes, we consider an interaction that is isomorhic to the one deicted above. Only now, otential recruits can be characterized by their observable tye i f;g. We denote the fraction of the oulation with characteristic as x ð0; 1Þ and the fraction of the oulation with characteristic as 1 x. Without loss of generality, let i ¼ if i4x and i ¼ if i > x. Tye may thus reresent a otential recruit s ethnicity or geograhic background. Further assume that the roortion of ideological zealots may differ across observable tyes i. Let z 1 denote the fraction of otential recruits of tye who are zealots, and let z denote the same with regards to otential recruits of tye. As before, will denote the fraction of zealots in the entire oulation imlying that xz 1 þð1 xþz ¼. Without loss 17. These factors were all identified as influencing the advancement of military officials under the Rhee regime in our case study of South Korea (Kim, 1971).

18 18 The Journal of Law, Economics, & Organization of generality, we assume that z > z 1. z j ; j f1; g is thus a reflection of the correlation between ideology and observable tye. The game roceeds exactly as before, excet we now assume that in the first eriod of lay the leadershi may restrict the set of otential recruits eligible for office to those of tye or, or it may choose not to imose restrictions. Under these circumstances, the leadershi always refers to restrict the ool of eligible candidates to those whose tye indicates that they are more likely to be zealots that is, to those for whom i ¼. In so doing, the leadershi increases the likelihood with which any given otential recruit is a zealous tye. One may trivially extend the conclusions of Proosition 4 to cover this case, only here the arameter z relaces the arameter in that roosition. Thus, as z the correlation between observable tye and ideological affinity for the regime increases, the equilibrium value of must weakly decline. Moreover, since z >, the leadershi is (weakly) less likely to enact anticorrution commitments when ideology is artially observable than when it is wholly unobservable. Ethnic or geograhic olarization serves to reduce anticorrution commitments. 4. Illustrative Cases Our theory thus advances several hyotheses. First, the equilibrium value of anticorrution commitments should be rising as the size of the ool of zealots falls. Second, the creation of credible and caable anticorrution bodies is less likely in states that are characterized by ethnic or geograhic cleavages. Finally, we can make redictions about the effects of shifts in the value of ideological arameters. Consider a country that currently lacks anticorrution institutions and where olarization is low. Proosition 3 holds that any increase in olarization will (weakly) increase anticorrution commitments. This is equivalent to a shift from the far left of Figure 1 toward the right. Unfortunately, these hyotheses do not readily lend themselves to large-n analysis. Measures of the distribution of ideological suort for autocratic leaders are few and of dubious quality. We therefore choose to demonstrate our argument through a discussion of several illustrative cases. Our emirics focus on changes within countries over time. We thus attemt to minimize confounding covariates to the extent ossible in a qualitative analysis by holding time-invariant country characteristics constant. While our cases are consistent with model redictions, we cannot claim that they offer definitive suort for the theory we advance. Our model is limited it abstracts from many significant olitical and economic factors. The countries we consider exerience substantial economic and olitical shocks over the eriod we cover, which may have influenced decisions to establish anticorrution agencies indeendently

19 Corrution and Ideology in Autocracies 19 of the ideological changes that are our central concern. One should see our case studies as evidence of the lausibility of the claims, we advance, rather than as definitive tests. 4.1 Korea under the Rhee and Park Regimes Syngman Rhee was elected resident of South Korea in 1948, and elections continued throughout his tenure ( ). But, his rule was marked by ballot-rigging and assassinations of olitical oonents (Haggard 1990; Kim 1971; Moran 1998). The rosect of electoral sanctioning was far-fetched electoral accountability was minimal. Rhee s rule was notable for its lack of ideological underinnings and the disconnect between the executive and any mass olitical movements (Kim 1971,. ). 18 A arty of government the Liberal Party was created in 1951, but, it never develoed a grass-roots following (Cole and Lyman 1971; Han 1974). Rhee thus lacked stalwart ideological suorters levels of both and were low. Corrution flourished under Rhee s rule. 19 At the etty level, corrution was articularly rife in the olice and military institutions whose activities were crucial to Rhee s gri on ower. Members of the military, articularly senior members, could rofit from their ositions by selling equiment on the black market. One general was said to have commented on corrution: Everyone is in it. Privates steal on foot. Officers steal in jees. Generals steal by trucks (Clifford 1998,. 91). Military and olice officials were laced in ositions to benefit from such graft by virtue of their reutation for loyal service to the Rhee regime (Kim 1971). 0 In keeing with model rimitives that treat rents r as a motivation for bureaucratic effort, it seems that Rhee tolerated corrution by senior officials to encourage their activities on behalf of his regime. These conditions ersisted until Rhee was ousted and, subsequently, Park Chung-Hee was installed in ower via a military cou. The leadershi of this cou rimarily consisted of lower-tier military officers ossessed of a nationalistic and develomentalist ideology (Kim 1971,. 100). The Park regime rallied oular suort based on a revolutionarynationalist ideology that emhasized the imortance of work, clean government, and develoment that aealed to the ublic at large, and articularly to the lower tiers of the military. These ideological aeals increased Park s oularity such that, following his assassination in 1979, an estimated 9.5 million Koreans turned out to ay resects at his funeral 18. In a notable examle, Rhee dissociated himself from the Korean Democratic Party (KDP) on whose ticket he had run immediately after assuming ower. He aointed a cabinet all but devoid of KDP members (Han, 1974). 19. At the elite level, this corrution largely consisted of the referential allocation of exort licenses and funds from US-sonsored aid to olitical backers (Haggard, 1990; Moran, 1998). 0. Promotion and hiring decisions were also influenced by observable characteristics for examle, geograhic background that were associated with suort for Rhee.