ELECTORAL RULES AS CONSTRAINTS ON CORRUPTION Jana Kunicova and Susan Rose-Ackerman * Department of Political Science Yale University First Draft: September 5, 2001 This Version: June 24, 2002 Abstract This paper investigates how electoral rules influence political corruption. We show that closed-list and open-list proportional representation (PR) systems are more susceptible to corruption relative to plurality systems. We argue that this effect is due to differences in the locus of rents between PR and plurality systems, and to monitoring difficulties under PR. Monitoring is limited under PR because (1) such systems produce districts with large numbers of voters producing severe collective action problems for voters, and (2) under PR opposition parties face incentive problems in monitoring corrupt incumbents. In addition, in CLPR systems the link between re-election and performance in office is weaker than in plurality or OLPR systems, making them especially susceptible to corruption. We also examine interaction effects between electoral rules and presidentialism. We empirically test our main predictions, interaction effects, and the proposed causal mechanism on a crosssection of up to 99 countries, controlling for economic, political, and social background factors. The empirical findings strongly support our theoretical hypothesis that closed-list PR systems, especially together with presidentialism, are associated with higher levels of corruption. This result is robust to different model specifications and to the deletion of influential observations. * We would like to thank José Antonio Cheibub, Rafael DiTella, Aaron Edlin, Eduardo Engel, Philip Levy, Fiona McGillivray, Jonathan Rodden, Frances Rosenbluth, Matthew Shugart, Peter Siavelis, Alastair Smith, and Georg Vanberg for helpful discussions and comments on earlier drafts, and Phil Keefer and Jessica Seddon for sharing their data. We also benefited from comments received in seminar presentations at Yale, Berkeley Law School, the Public Choice Society, the Midwest Political Science Association, and the Society for Comparative Research. All remaining errors are ours. Rose-Ackerman s contribution was partially supported by the Center for Advanced Study in the Behavioral Sciences. Contact Information: jana.kunicova@yale.edu; susan.roseackerman@yale.edu.
ELECTORAL RULES AS CONSTRAINTS ON CORRUPTION 1. Introduction Elections serve two functions in representative democracies. First, they select political actors who enact public policies in light of constituents preferences. Second, they permit citizens to hold their representatives accountable and to punish them if they are corrupt or self-serving. In other words, elections provide both incentives for politicians to enact certain kinds of policies and constraints on politicians malfeasance. In this paper, we focus on the second of these two functions and investigate how different electoral systems constrain political corruption, holding constant other political, economic, and social background factors. We study three stylized categories of electoral rules: plurality/majoritarian systems with single-member districts, and two kinds of proportional representation (PR) systems: closed- and open-list. Under a closed list system, party leaders rank candidates, and voters only cast votes for parties. Under an open list, voters both select a party and rank candidates given the party s selection of candidates. The control of political corruption depends both on the locus of corrupt rents and on whether any actors have both the incentive and the ability to monitor those politicians with access to rents. We argue that PLURALITY and PR systems differ in that under PLURALITY rule more rentseeking opportunities are available to individual legislators, compared to the leadership, and that under PLURALITY, the monitoring of these rent seekers is likely to be more stringent than the monitoring of rent seekers under PR. There are three reasons for this. First, PLURALITY rule produces districts with relatively small populations where voters face less severe collective action problems in monitoring the incumbents than in the large, often national, districts in PR systems. Second, the opposition has greater incentives to uncover corruption in PLURALITY systems. If the conditions of Duverger s law hold, PLURALITY rule leads to only one major party in opposition in each district. This eliminates the free-rider problems associated with exposing corrupt incumbents in systems with multiple opposition parties. Furthermore, opposition parties in PR systems may be reluctant to expose incumbents if they think that they may be asked to form a government with them in the next political round. Third, in PLURALITY systems the locus of monitoring effort at the district level is more effectively targeted toward the potentially corrupt than in PR systems where the leadership controls the rents and is difficult to monitor. As far as the difference between CLPR and OLPR is concerned, we expect CLPR systems to be more corrupt because they are more opaque. If voters cast votes for party lists, the link between re-election and performing well in office is weakened in comparison to OLPR and PLURALITY systems, where voters select individual candidates. 2
Although our primary focus is on the methods by which the legislature is chosen, we also recognize that a complete model should include other institutional features of a political system such as presidentialism/parliamentarism, federalism, bi-cameralism, and the strength of parties. We examine the interaction between presidentialism and electoral rules, and include other institutional variables, most notably federalism, as controls in our empirical work. Our paper stands at the intersection of two broad literatures: one examining electoral rules and their effects, and the other attempting to explain political corruption. Electoral rules have been shown to affect the incentives of political actors to organize and hence the number of political parties (Duverger 1954, Rae 1971, Riker 1982, Powell 1982, Taagepera & Shugart 1989, Lijphart 1990, 1994; ), as well as the way in which parties and politicians compete for votes, producing personalistic versus party-centered systems (Carey & Shugart 1995 Shugart 1999, Myerson 1993a, Gaviria et al. 1999, Seddon et al. 2001, Panizza 2001). Theoretical arguments conclude that, under plausible conditions, PLURALITY rule results in two major parties and PR produces several competing parties (for qualified arguments, see Taagepera & Grofman 1985, Ordeshook & Shvetsova 1994, and Neto & Cox 1997). In addition, electoral rules are believed to affect party discipline: where politicians have incentives to cultivate a personal vote, party discipline will be low ( Carey & Shugart 1995, Ames 1995, Reed 1994, Bowler et al. 1999). Our paper builds directly on these stepping stones. We argue that electoral rules affect the incentives and ability of voters, opposition politicians, and intra-party actors to organize and monitor the corruption of incumbents. The result critically depends on the process of nominating candidates, the number of parties in the opposition, and on party discipline-- factors that are, to some extent, the by-products of electoral rules. In addition to the large and developed literature on the effects of electoral rules, there exists a new and growing literature that addresses the relationship between political institutions and corruption. Our paper makes several contributions to this developing field. First, it explicitly distinguishes between criminal corruption (embezzlement of funds, bribery) and pork-barrel spending (perfectly legal activity), which tends to be conflated elsewhere (Geddes 1994, Rasmussen and Ramseyer 1994, Bicchieri 1995, Bueno de Mesquita et al. 1999). This distinction is crucial because electoral systems that constrain corruption may encourage pork-barrel spending, and vice versa. Second, on the theoretical side, our causal mechanism linking electoral rules and corruption is different from existing models (Myerson 1993, Holmstrom 1982, Persson and Tabellini 2000, ch. 9). Those models focus on district magnitude or party lists as the driving forces determining the impact of electoral rules on corruption; 1 we emphasize the effect of district size (that is, voters per district, as opposed to the number of representatives elected per district), and the number of parties in opposition. Third, there is only one other study that has attempted to assess the link between electoral rules and corruption empirically (Persson, Tabellini, and Trebbi 2001). Although we 3
confirm their basic finding that proportional elections are associated with higher corruption levels, we test a different causal mechanism. Most importantly, we are more attentive to other institutional details that were assumed away by Persson, Tabellini, and Trebbi (2001). In addition to differentiating between closed- and open-list PR, we control for the effects of federalism and presidentialism because they have been shown to influence corruption in two recent papers (Treisman 2000 and Kunicova 2001, respectively). We also explore the interaction effects of electoral rules and presidentialism. Finally, we use a more comprehensive and up-to-date dataset covering up to 99 democracies and test the robustness of our results to alternative measures of corruption, different specifications, and to deleting influential observations. The remainder of this paper is organized as follows. Section 2 discusses the importance of distinguishing between corruption and pork-barrel spending. Section 3 presents our basic theoretical argument about the link between electoral rules and corruption. Section 4 adds presidentialism to this framework. Section 5 states our hypotheses, and Section 6 describes the data and methods used to test them. Section 7 presents the results of the regression analysis. Section 8 concludes. 2. Corruption versus Pork-Barrel Spending A standard definition of corruption is misuse of public office for private gain (cf. Bardhan 1997; Rose-Ackerman 1999; Treisman 2000; Sandholtz and Koetzle 2000; Lambsdorff 1998). Yet, some authors broaden this definition to include rent extraction by public officials, which tends to subsume other activities of politicians, from direct embezzlement of funds for private gain through paying off political supporters to maximize their chances of re-election. An example of such work is Bueno de Mesquita et al. (1999). The authors formulate and empirically test a formal model in which politicians stay in office by offering public goods that benefit everyone and private goods that benefit only their core supporters. The latter is considered corruption and is influenced by institutions in both autocracies and democracies. For other arguments conflating corruption and targeted public spending, see Geddes (1994), Rasmussen & Ramseyer (1994), and Bicchieri (1995). We believe that it is analytically wrong to conflate corruption, an illegal activity, and politically targeted public spending, a legal activity. This is especially ill-advised when analyzing the impact of electoral rules on corruption. Some electoral systems, although enabling voters to monitor legislators behavior, also give incentives to politicians to use a legitimate means of political competition provision of narrowly targeted public services, or so-called pork-barrel projects. These are the dual issues of incentives and constraints that we discuss in this and the following section. First, electoral rules differ in the incentives they give to politicians to offer broad-based public goods or narrowly focused pork-barrel spending. An electoral system based on geographic 4
representation will encourage spending targeted to particular districts at the expense of more inclusive public goods. In contrast, when the competition for votes is more broad-based, candidates and political parties will find it more electorally beneficial to run on national public goods platforms. This is consistent with the existing theoretical models of public finance and electoral systems. Persson & Tabellini (1999) argue that in plurality systems, politicians only need to please swing voters in marginal districts, not the population as a whole. Hence, there will be disproportionately more geographically targeted pork-barrel and fewer universal public goods in plurality systems as compared to PR. The same theoretical result comes from a different model developed by Lizzeri & Persico (2001). To see the logic of this argument, consider our three stylized electoral systems: majoritarian/plurality with single member districts (PLURALITY); nationwide, closed-list proportional representation (CLPR) where party leaders rank candidates and voters only select political parties; and open-list proportional representation (OLPR) where voters can cast their votes for a particular candidate on a party list. Electoral competition in PLURALITY systems has a local, geographic basis. National issues such as war and peace or moral issues such as abortion may, of course, sway voters, but incumbents will also want to claim that they have brought home the bacon to their constituents. Unless national political parties are very strong, incumbents are likely to make a nonpartisan appeal for reelection arguing that they have been able to obtain targeted benefits. In contrast, under CLPR, politicians have an incentive to provide broad-based public goods so long as the parties constituencies are widely dispersed so that it is difficult to target narrow benefits to one s supporters. In OLPR systems, there are no ready-made geographical constituencies as in PLURALITY systems, but a candidate may try to appeal to particular group of voters by becoming their advocate within the party and later bragging about his or her success. Thus, there will be more targeting in OLPR than CLPR, but less than in PLURALITY. Furthermore, the type of targeting is likely to single out different groups. If pork-barrel spending is conflated with corruption, one would conclude that PLURALITY systems are more susceptible to corruption because they give incentives for pork-barrel spending. However, having made an explicit distinction between the two, we arrive at a different prediction. As we argue in the following section, the same features of electoral systems that create incentives for pork-barrel spending also constrain political corruption. 3. Monitoring Corruption Consider a conceptual framework with four types of actors. The first are the incumbent party leaders; the second are the individual rank-and-file legislators who are not in the leadership; the 5
third are political opponents, either individuals or parties; and the fourth are the voters themselves. The incumbents party leaders and/or individual legislators have opportunities to extract corrupt rents. However, they may also have incentives to monitor each other. In addition, they are monitored by the political opposition and by voters. We specify the objective functions of these four political actors as follows. First, incumbent politicians, both leaders and rank-and-file, care about individual wealth and re-election. They would like to maximize rents without being detected, since detection is associated with costs, both monetary (such as legal fees or even a prison term) and political (decrease in probability of re-election). Different electoral systems shape the incumbents opportunities to extract rents as well as their chances of re-election. Second, opposition candidates care about winning office. Their chances of winning increase as the chance of incumbents re-election decreases, so the opposition benefits from revelations of corruption involving incumbents. However, challengers incentives to work to uncover corruption depend on the electoral system. Finally, voters prefer honest to corrupt elected officials since their utility decreases when public resources are diverted for private gain. This is, of course, the reason why we assume that a corruption scandal lowers the incumbent s chance of reelection. Again, we claim that the voters ability and incentives to monitor corrupt politicians vary with the design of the electoral system. Corruption in our formulation is illegal self-regarding activity. If detected, a corrupt politician usually pays considerable costs, both legal and political. Hence, we assume that the probability of detection constrains political corruption, and, all else equal, the stronger the constraint, the lower we expect the equilibrium level of political corruption to be. We argue that electoral rules affect the probability of detection, and hence the anti-corruption constraints, by shaping the incentives and ability of political actors to monitor political corruption. Figure 1 depicts the oversight relations that are shaped by the electoral rules. 2 We explore them in detail below. Figure 1 about here To understand how electoral rules affect the monitoring of corruption, we need to explore three interrelated issues. First, how do the opportunities for private gain differ in different electoral systems? Second, who has the incentive to monitor potentially corrupt groups or individuals? Third, who has the ability to do so? The first salient difference between voting systems is the locus of control over corrupt rents, that is, the opportunities for private gain. Under CLPR the party leaders are very powerful vis à vis the rank and file because they determine a candidate s ranking on the list. 3 Thus the leadership faces most of the opportunities for private gain and can determine how the spoils are divided. The rankand-file party members in the legislature face few individualized corrupt opportunities. The situation is similar under OLPR, except that the leadership itself may be more divided and unstable. Individual 6
members can challenge the leadership by threatening to campaign against them through direct appeals to voters. 4 Compare the situation under PLURALITY rule. Here, individual incumbents may gain power within the party because of the strength of their local power base. Even where the party has the nominal power to assign candidates to districts, an incumbent who wins by a large margin has some independent bargaining power vis à vis the party. This means that such incumbents can demand a share of the rents of political power in the form of corrupt payoffs or in pork-barrel projects for their districts. Their own preference for payoffs or pork will be a function of the oversight by voters, the media, and citizens groups, on the one hand, and challengers, on the other. This discussion suggests that in terms of uncovering corruption, monitoring the leadership is relatively more important under PR, and monitoring the rank-and-file is relatively more important under PLURALITY rule. Of course, leaders may be able to garner corrupt rents under PLURALIY rule as well. Our claim is only that individual legislators have more opportunities for rent extraction under PLURALITY than under CLPR. Thus, in PR systems, especially closed-list ones, the opportunities for rent extraction are vested mainly in party leaders, and in PLURALITY they are divided between the party leaders and individual representatives. Now consider the incentives and ability of political actors to engage in monitoring. As Figure 1 shows, the oversight relations that interest us are mainly voters and political opponents monitoring of both party leaders and individual politicians. We also consider the possibility of intra-party monitoring, i.e., rank-and-file members monitoring of the party leadership and the leadership s monitoring of the rank-and-file. We are interested not in the overall level of oversight, but in the way the electoral system affects monitoring on the margin. In addition, we assume both that monitoring is costly, and that hiding one s malfeasance is a costly and imperfect process. 5 We consider each of the three monitoring relationships in turn. Monitoring by Voters As a normative matter, the rank-and-file need to be more closely monitored by voters under PLURALITY rule than under CLPR because the rank-and-file incumbents have more control over rents under PLURALITY rule than under PR. As a positive matter, we argue that this is what we would expect to happen. Voters do not have the ability to monitor the relevant actors equally in all electoral systems. We discuss two ways in which electoral systems may influence the voters ability to monitor corrupt politicians.the first involves the voters ability to observe malfeasance if it occurs. The second concerns the incentives of voters to organize for oversight. We consider these arguments in turn, although they are not independent of each other. Thus, voters may be more likely to overcome collective action problems if the benefits of engaging in oversight are relatively large. 7
The first argument is derived from the career-concern model of Holmstrom (1982) and its extension to rents and corruption by Persson and Tabellini (2000, ch. 9). Voters prefer honest politicians to corrupt ones because of the costs that corruption imposes on citizens in terms of inflated budgets, low value public projects, etc. Hence, if voters can identify corrupt politicians, they will punish them by voting against them in the next election. Persson and Tabelliini argue that voting over individual candidates, as in a PLURALITY rule system, creates a direct link between individual performance and re-election. This, in turn, gives incentives to incumbents to avoid corruption (Persson, Tabellini, and Trebbi 2001). However, so far, this argument does not distinguish well between voting systems. In CLPR systems, the leadership is also known to the voters. Because it is precisely the leadership that needs to be monitored in CLPR, voters can identify those politicians most subject to corrupt incentives in both systems. However, it is not sufficient to identify those who might be corrupt. In addition, voters must be able to assess whether the politicians actually are engaging in malfeasance. They need to monitor the rank and file under PLURALITY and the leadership under CLPR. We argue that such monitoring is easier in the smaller districts produced by PLURALITY rule than in the larger districts characteristic of PR. Voters in a district with a small population are more likely to have direct contact with their representative and hence more information about the candidates they elect. Holding the freedom of press constant, we would expect that information about individual kick-backs to politicians would be less readily available to voters in CLPR systems than in PLURALITY systems. In short, we claim that voters are more likely to become aware of corruption under PLURALITY rule than under PR and that one reason for this is the smaller population of districts under PLURALITY rule. The second argument concerns the incentives of voters to organize to provide oversight. Free-rider problems are ubiquitous in political life, and the monitoring of corruption is no exception. In this case, only a small sub-set of the voters needs to organize to engage in monitoring. If they do, all voters benefit. In general, collective action problems among voters should be less severe in small groups than they are in large groups (Olson 1982). This suggests that citizens in smaller districts, measured by number of voters, ought to find it easier to overcome free-rider problems than those in larger districts. Holding other factors constant, PLURALITY rule produces districts with smaller numbers of voters than PR. Thus, overcoming free rider problems should be easier in such systems. In addition, it may be easier to organize groups willing to engage in monitoring simply because it is less difficult to observe corruption and its results in a PLURALITY system. Our discussion of monitoring by voters has so far left out OLPR systems. There, similar to CLPR, the leadership controls the rents. Yet OLPR shares one feature with PLURALITY systems: voters can cast their votes for particular candidates, albeit within a party. This strengthens the link between performing well in office and getting re-elected in comparison to CLPR systems and gives 8
the rank-and-file some power vis-à-vis the party leaders. However, the first feature of PLURALITY systems relatively small single-member districts is absent in OLPR systems. This disables grassroot monitoring by voters who, in addition, are likely to face considerable collective action problems. Furthermore, individual candidates are likely to have little say over their party s rent-extraction activities. On balance, then, OLPR systems fall in between CLPR and PLURALITY systems the link between performing well in office and being re-elected is stronger than in CLPR, but the voters cannot monitor their representatives as closely as in PLURALITY. Even if they did so, the integrity of the system would benefit little because rent-seeking opportunities are centered in the party leadership. Monitoring by Political Opponents The second piece of the puzzle is the competitiveness of the political system and its impact on the monitoring of incumbents by their political opponents. As we noted above, challengers have a direct incentive to uncover the malfeasance of incumbents because they increase their probability of winning office once the incumbent is discredited. They can do this under PR systems by investigating the integrity of party leaders, and under PLURALITY rule at the district level as well. Either might be a feasible strategy, but we suggest that there are features of PLURALITY rule that tip the balance in its favor. We offer two arguments why under PR it is more difficult for the opposition to monitor the incumbents: the first one is based on a free-rider problem among multiple parties, while the second one concerns the adverse effects of coalition politics on monitoring. We consider these two arguments in turn. First, we claim that the net benefits for an opponent of uncovering a scandal tend to be higher under PLURALITY rule than under PR. To see this, consider first the opposite argument, namely that competitive pressures are higher under PR and will lead to lower corruption (Myerson 1993; Persson, Rolland, & Tabellini 2000). Under PR, entry barriers are low so that multiple parties are common; under PLURALITY rule, when Duverger s law holds, only two major parties will compete in each district. The high cost of establishing a new party contributes to the lack of political competition in these models. Assuming that politics is multidimensional, voters find it more difficult to vote out corrupt politicians if honest candidates, whom the voters might like on other issues as well, find it difficult to enter into the competition for public office. Conversely, the less expensive it is for a challenger to take on the incumbent, the more likely one is to appear in response to voter dissatisfaction. This is an important argument, but it ignores the incentives that opponents have to uncover corruption independent of the party s ideology. The barriers-to-entry theory is an application of economic arguments from industrial organization to the realm of politics. The basic point that a competitive political system limits 9
corruption seems a valid one (Rose-Ackerman 1978). It is not obvious, however, that multi-member districts produce benefits similar to those of markets with many competing firms. What we question is the claim, implied by the empirical work, that competition is greater the larger is the district magnitude. True, a PLURALITY system with single-member districts will often produce only two parties. However, one opposition party with a credible chance of winning the election ought to be sufficient to give the incumbent an incentive to limit self-dealing, and the marginal cost of fielding a candidate is likely to be low for such an established party. Furthermore, multi-member districts selecting legislators by PR may indeed produce many more viable parties, but no party may have much incentive to monitor the corruption of incumbents because a scandal is a public good for all opposition parties. In short, the argument that corruption monitoring will increase with the number of political opponents does not seem convincing. 6 To the contrary, the incentive to free ride off the scandal-mongering of others could mean that the impact of competitive politics on corruption monitoring actually falls as the number of parties increases. 7 The second argument for PLURALITY being conducive to better monitoring by opposition focuses on inter-party relationships under PR and PLURALITY rule. With PLURALITY, coalition governments are possible but unlikely unless many regional parties exist. Under PR, they are common and in many countries parties do not sort themselves into two stable blocs. Instead, a party currently in opposition may expect to form a coalition with one or more of the incumbent parties sometime in the future. If this is so, opposition politicians will have little incentive to expose the corruption of those politicians that they might collaborate with in the future. The lack of a clear alternation between fixed groups of parties deters inter-party monitoring. 8 In sum, we argue that the number of parties is a poor proxy for the intensity of competition at least with respect to the control of corruption. The analogy to private market competition is misplaced. Although the party or candidate that uncovers the malfeasance gets some credit for diligence, the main benefit is an overall decline in support for the incumbent that benefits all opponents. Thus, a challenger has a larger incentive to bear this cost, the fewer the number of opposition parties. The challenger s incentive to uncover corruption is higher under PLURALITY rule than under PR so long as the number of effective parties is, in fact, lower and most districts are competitive. Comparison of Electoral Systems Table 1 summarizes our arguments about the oversight relations under the three electoral systems that we consider. The table presents our claims about the relative incentives and ability of political actors to monitor rent-extraction by politicians. They have no cardinal meaning. Table 1 about here 10
PLURALITY rule scores the highest of the three stylized electoral systems on both the incentives and the ability of political actors to monitor rent-extraction. Districts with small numbers of voters mitigate the collective-action problems of voters and make it easier for them to observe the behavior of individual legislators likely participants in most corrupt deals in such systems. At the opposite extreme, under CLPR, collective action problems are more serious, and voters find it difficult to observe the behavior of party leaders the primary locus of corrupt deals in CLPR. Furthermore, PLURALITY rule will produce opposition parties with higher incentives to unveil the corruption of incumbents than in multiple party regimes. The opposition will reap more of the electoral benefits from such revelations and is not likely to want to form a coalition with the incumbents in the future. Because OLPR systems share features of both CLPR and PLURALITY systems, they occupy an intermediate category in monitoring corruption. In particular, voters are better able to monitor incumbents in OLPR than in CLPR because individual identifiability and accountability are greater, but transparency in the large OLPR districts is lower than in the smaller PLURALITY districts. Finally, OLPR is no different from CLPR in that political opponents have few incentives to expose corrupt incumbents due to the collective action problem among multiple challengers. Before deriving hypotheses from our analysis, we allow for the possibility of intra-party monitoring and check whether it changes our predictions. Intra-party Monitoring Intra-party monitoring is unlikely to be a sufficient constraint on corruption under any electoral system. The basic problem is collusion. If there is little outside monitoring by voters, opponents, or other aspects of civil society, leaders can collude with the rank-and-file to share corrupt rents. Internal party whistle-blowers can arise, but this is likely to be a risky role to play even under PLURALITY rule where individual members have more independent influence than under PR. First, consider the incentives of the rank-and-file legislators to monitor the leadership. Recall that the objective of the incumbent legislators is to maximize a function that includes both private rents and the probability of re-election. They may differ in their underlying commitment to honesty, however, with some opposed to all payoffs and others willing to balance the benefits and costs. Whether they will monitor the leadership depends on their own relative preference for payoffs, on whether they have access to rents themselves, and on whether they can increase their probability of re-election by exposing a corrupt party leadership. Under CLPR, individual members of the party have little access to rents, so the key variable for them is the probability of re-election. Individual members have little incentive to expose corrupt 11
leaders because their future careers depend directly on the leadership s decision to rank them on the party list or to give them other patronage positions in government or the party apparatus. 9 Hence, incentives for political entrepreneurship are low. Nevertheless, if a rank-and-file member does happen to observe corruption by the leadership, he or she may have the leverage to demand a share of the rents in return for silence. Alternatively, the potential whistle-blower could insist on a high place on the list. This blackmail potential gives the rank-and-file some bargaining power, but it seems likely that the result will be some kind of deal rather than a scandal because the cost to the whistle-blower of going forward is high. Unless he or she is planning on leaving politics in any case or is powerful enough to challenge the leadership for party control, a scandal will undermine the career of the whistle-blower along with all who are associated with the party. Under OLPR, individual legislators have more incentives to appeal to voters over the heads of party bosses, but accusing the leadership of corruption also risks destroying rather than reforming the party. Thus effective rank and file monitoring seems unlikely, although some gain sharing may occur. In a PLURALITY rule system, although risk is also present, it is somewhat weaker because the party is less centralized. Honest politicians have some incentive to blow the whistle on a corrupt leadership because the leadership does not entirely control their fate. Even those with no personal commitment to honesty may reveal the corruption of others if they believe that their constituents will reward them. Of course, even here, whistle-blowing is a risky strategy because it may lead to the fall of the government in a parliamentary system. However, on balance, if the rank-and-file members are honest, they are better equipped to constrain the rent-seeking behavior of the leadership in a PLURALITY rule system. In contrast, if they are corrupt, they can more easily flout attempts by the leadership to control their behavior and as in the PR cases, can demand a share of the leaders spoils. Relying on the leadership s incentives and ability to monitor rank-and-file appears similarly problematic. Regardless of the electoral rule, the leadership may have an incentive to monitor individual members because their potential corrupt practices, if exposed, may harm the name of the party and worsen its overall electoral prospects. In PLURALITY systems the leadership has the most incentives to monitor individual members because they have relatively more opportunities to extract rents, but unfortunately the leadership has relatively little ability to control individual members in comparison to PR systems. In sum, the incentives for intra-party monitoring do not seem very powerful under any electoral system. Therefore, we believe that monitoring by outside actors -- voters and political opponents is what truly differentiates electoral systems with regard to limiting corruption. However, where the intra-party monitoring does operate, it does not change our basic prediction about electoral systems influence on corruption. PLURALITY is superior to both PR systems on all counts except one the ability of the party leaders to monitor individual representatives. Although 12
top-down intra-party monitoring appears to be problematic in PLURALITY systems, we believe that this deficiency is more than offset by the ability and incentives of both voters and political opponents to monitor individual representatives. At the same time, party leaders, who share the locus of rents with the individual representatives in PLURALITY, are relatively well monitored both inside the party and by outside actors. On balance, we predict that PLURALITY rule voting will do a better job at controlling corruption than PR, especially CLPR. The cost of PLURALITY rule may be a more particularized political system that focuses on providing benefits to narrow ranges of constituents in key districts, but that is simply the consequence of the more individualized nature of politics. Under PLURALITY rule, (1) corrupt opportunities will be concentrated in just those political actors who are best able to be monitored by voters, (2) the two-party system that frequently results will give opponents an incentive to uncover scandals at any level, and (3) the greater power of the rank-and-file means that, if the rank and file is honest, the leadership will be discouraged from corruption by credible rankand-file threats to reveal its malfeasance. 4. Presidential and Parliamentary Systems Other elements of constitutional design may be important for corruption control besides the differences in voting rules. The most interesting are those that are not merely controls but that interact with the voting rule to affect the opportunities for corruption and the incentives and opportunities for its control. We focus on one such feature: the distinction between presidential and parliamentary systems. Opportunities for corruption are enhanced by centralized control over government. Then those with power can, if they wish, create rent-seeking opportunities with little oversight inside government. A president who controls the executive branch has rent-creating possibilities that can be used for personal gain. In a presidential system, the leaders of the legislative parties are less powerful than under a parliamentary system (Shugart 1998, Mainwaring 1995). They must negotiate with the president to pass legislation and do not control the rents that arise within the executive. 10 Consider the framework that we introduced in the previous section with the addition of one more incumbent: the president. Figure 2 about here Among the incumbents, the president has access to the rents generated within the executive although some of these rents may be shared strategically with the legislature. The president s undivided power over many of these rents implies that diverting them for personal gain is likely to be easier than in a more collegial system of cabinet government. 11 He also has an interest in creating 13
additional rents through executive action. In spite of the legislature s incentive to monitor the President, his fixed term in office gives him considerable leeway subject only to the threat of impeachment. The legislature has no instrument similar to the vote of no confidence in parliamentary systems, where the legislature can remove the executive from power at any time between elections. Of course, legislatures do try to restrict presidential freedom, but their control is less direct than that exercised by a parliament over the Cabinet in a parliamentary system. In addition, in most presidential systems, US-style checks and balances are absent and presidents tend to have extensive legislative and non-legislative powers. This is not inherent in the nature of presidential systems but is an empirical reality. Using a large cross-section of countries, Kunicova (2001) shows not only that presidentialism per se is associated with higher corruption, but also that the more extensive the president s powers, the more corruption. Thus, monitoring of the executive by the legislature is in principle and in practice more difficult in presidential systems than in parliamentary ones where the vote of no confidence gives the legislature oversight powers over the executive. Presidents frequently need to cooperate with the legislature to get policy initiatives passed (Cox and Morgenstern 2001, Shugart 1998). A corrupt president may seek to enact statutes that incorporate rent-generating opportunities that can be exploited for personal gain. To pass such laws, however, the president needs to bargain with the legislature. 12 In such cases, electoral rules affect the strength of legislative parties and their bargaining power in dealings with the President. Theoretical claims about electoral rules and party strength have two prongs. First, as discussed above, parties are stronger under CLPR than under PLURALITY rule. Second, parties are weaker in presidential systems than in parliamentary systems because in the former they do not need to organize themselves to form a government (Mainwaring 1995; Shugart 1998, 1999) Taken together, these arguments imply that parties will be weakest in presidential systems with PLURALITY-rule legislatures and strongest in parliamentary systems operating under CLPR. However, the other two possibilities cannot be clearly ranked except to note that they fall between the extremes. One cannot say a priori if parties are stronger under a parliamentary system with PLURALITY voting or under a presidential system with the legislature chosen by CLPR. The relation to the expected level of corruption is complex. On the one hand, we expect that presidential systems will be more corrupt than parliamentary systems that use similar voting rules for the legislature. On the other hand, we expect that CLPR systems will be more corrupt than PLURALITY systems. The combination of CLPR and presidentialism is likely to create unfortunate synergies. This happens, we argue, because several factors converge. In line with our previous discussion consider, first, the bargaining relationship between the President and the legislature and the scope for intra-party monitoring. If he wishes, the President has the power to generate high levels of corrupt rents through his control of the executive. Under 14
CLPR, he can form an alliance with the party leaders to share rents at relatively low transaction costs because the rank and file does not have to be included. Even if the parties are somewhat weaker than under a parliamentary system, they still control their members. The rank and file, as we argued above, can often be kept in the dark and even if they learn of a corrupt deal, they have little incentive to blow the whistle on party leaders. Aspects of this corruption-prone scenario do not carry over into other combinations. If a president faces a PLURALITY-rule legislature whose parties are weak, he may have to bargain with a multitude of individual members in order to form an alliance. Any alliance will be difficult to form in the first place because of the number of legislators involved; it is not enough just to include the party leaders. Furthermore, it may be difficult to sustain if marginal members threaten to defect. The greater number of individuals involved increases the risk that one of them will report the scandal. In addition, even rank-and-file members of a party that is nominally allied with the president have an incentive to be integrity entrepreneurs to enhance their own reputation. The very difficulty of forming an alliance with the legislature should deter corruption. This discussion suggests that the greater corrupt opportunities created by presidentialism should be especially evident in CLPR systems. OLPR ought to be an intermediate case. The parties are less powerful bargaining units than under CLPR, but individual members have little to bring to the table. To complete the argument we need to consider the possibility of monitoring by voters and opposition politicians. These groups have an incentive to monitor the president, but their ability is limited. For voters, the situation is similar to that of their monitoring of party leaders the collective action problems are large in nation-wide districts and evidence of corruption is hidden in public contracts and individualized decisions that are hard to monitor. This is part of the basic argument for expecting more corruption in presidential systems. In contrast, legislators from parties different from the president s do have an incentive to monitor, and if they control the legislature, they may have the power to enact laws that constrain the president. The issue is then whether the role of opposition parties ought to differ between PR and PLURALITY systems. We propose the following way of thinking about this issue: Suppose that the corrupt president tries to preempt future oversight by giving many potential opponents a share of the spoils, thus neutralizing them as monitors. This is a more feasible option if only party leaders have any real power and in presidential systems this is most likely to be true under CLPR (Carey&Shugart 1995, Seddon et al. 2001). In contrast, consider PLURALITY rule in a case where only two major parties are likely to exist, one of which is likely to be the president s own party. 13 As we argued above, PLURALITY rule lessens the power of party leaders, including the president s party, and gives individual politicians incentives to be integrity entrepreneurs, on the one hand, or free-lance rent seekers, on the other. The problem is not simply to get a majority 15
behind the corrupt initiative, but also to assure that minority legislators do not blow the whistle to enhance their own political standing with the voters. Members can gain political capital by presenting themselves to their own constituents as defenders of honest politics. It appears that what matters in presidential bargaining with the legislature is not the number of parties that the president needs to work with, but the extent to which these parties are capable of voting as unified blocs. In this sense, CLPR seems to be most conducive to corrupt deals. 14 5. Hypotheses From our discussion in the previous three sections, we derive the following three hypotheses: H1 Existence of a relationship between electoral rules and corruption. Ceteris paribus, we expect CLPR systems to be more corrupt than OLPR and PLURALITY systems. This should hold controlling for other institutional factors, as well as for background factors such as level of economic and political development. H2 Interaction effects. Ceteris paribus, presidential PR systems are expected to be more corrupt than their parliamentary counterparts. We predict that CLPR presidential systems will be especially corrupt relative to other types of government structures. H3 Causal mechanism. Ceteris paribus, the population size of districts and the number of parties in the opposition should be negatively associated with corruption control. 6. Data, Measurement, and Econometric Methods We use cross-country data sets that characterize countries in terms of the level of corruption and that record the way in which their legislatures are selected. To this basic data we add information on whether a separately elected president exists and include other background political and economic variables. We use average perceptions of corruption in each country, based on a measure that captures high-level political and bureaucratic corruption as perceived by outsiders, mostly business people. Although these perceptions combine two kinds of corruption, the indices are mostly a measure of corruption at the top of government where politicians are likely to be involved. High-level bureaucratic corruption in, say, the award of contract or other favors, almost always has a strongly partisan cast. 15 On the other side of the ledger, we have summary measures of the voting rules in 16
each country. We cannot distinguish empirically between opportunities for corruption and the ability and incentives to monitor it. However, we claim that the convergence of these three factors ought to mean that PLURALITY systems are less corrupt on balance. We also provide a first cut at testing some of the causal mechanisms that we proposed in the theoretical section. 6.1 Corruption Data Corruption is difficult to define, systematically observe, and measure. The most comprehensive cross-country data on corruption are based on perceptions, not concrete measures of payoffs. This raises the possibility that corruption may be perceived to be high because people take the trouble to uncover it. In the extreme, there could be an inverse relationship between the underlying amount of corruption and perceptions of its prevalence. For example, the underlying level of corruption may have been higher in Italy before the Clean Hands investigations than after it, but public perceptions of its level could have increased as a result of the investigations. However, although this can occur in an individual country as policy changes over time, we doubt that it would represent a stable result. We assume that politicians act strategically. If they predict that their corruption is likely to be uncovered, they will engage in less of it. Then perceptions and reality would converge. Political systems that encourage corruption would be more corrupt, and those who deal with the state at high levels would also perceive corruption to be high. The surveys that we use are mostly based on the perceptions of such well-placed observers. In recent years, several indices have been developed that attempt to capture the abuse of political and bureaucratic power across countries. We rely on two indices that both measure perceptions of corruption, but use different aggregation methodologies: the Corruption Perception Index (CPI), compiled by Transparency International (Lambsdorff 1998), and the Control of Corruption Index (CORRWB), also known as GRAFT, compiled by the World Bank (Kaufamann, Kraay, and Zoido- Lobaton 1999). We prefer the CORRWB measure, but to check the robustness of our results, we run all our models on CPI as well. Transparency International (TI) has published its annual CPI ranking of countries since 1995. TI aggregates surveys of perceived corruption across countries based on the views of business people, risk analysts, investigative journalists, and the general public. Notice that many of the respondents are likely to have first hand knowledge of state operations not available to ordinary voters. The index aggregates corruption scores from up to 17 different polls for every country, including Wall Street Journal, Gallup International, Economist Intelligence Unit, World Bank, World Economic Forum, and others. These polls ask questions based on the concept of corruption as the misuse of public power for private benefit; specifically, the focus is on kickbacks in public procurement, the embezzlement of public funds, and the bribery of public officials. 17