Why a clean politician supports dirty politics: A game-theoretical explanation for the persistence of political corruption

https://doi.org/10.1016/j.jebo.2011.05.004Get rights and content

Abstract

Using a theoretical model of repeated political competition among two career politicians, I study the incentives of both the corrupt and clean politicians not to adopt a fully effective reform targeting political corruption. In the setup I study, each politician can credibly adopt the reform as part of his policy platform in the elections. Yet, when the level of political corruption is high, neither politician does so in a Nash Equilibrium. Intuitively, political corruption changes the zero-sum nature of political competition: the reform eliminates the illegal rents of the corrupt candidate and the competitive advantage of the clean candidate.

Highlights

► I show that when political competition is among a small set of career politicians with no term limits, political corruption produces a positive externality for the clean politician as well. ► Studying an infinitely repeated political duopoly, I identify the set of parameters under each politicians promises are credible, and then the subset under which neither politician promises the reform in a Nash Equilibrium. ► Compared to a (legal) rent-maximizer, a policy-motivated politician has stronger incentives to adopt the reform when the level of corruption is low; but, when the level of corruption is high, he too, will choose not to adopt the reform when the corrupt one did not adopt it.

Introduction

In a Turkish National Security Council meeting on February 19, 2001, President Mr. Sezer strongly criticized PM Mr. Ecevit for blocking corruption investigations about high level politicians, leading to a row between these two, as well as contributing to the onset of the most severe financial crises in modern Turkish history.1 Although most likely accurate, the accusation was surprising: Mr. Ecevit was a clean politician whose honesty was a major contributing factor to his electoral success in his more than three-decade political career as the head of a political party (four times PM of Turkey); see Tachau (2002, p. 114). A clean politician not enforcing or implementing effective anti-corruption policies, however, is not a phenomenon unique to Turkey; even in countries with some infamously corrupt politicians, an election victory by a clean politician has rarely been followed by implementation of effective anti-corruption policies (reforms).2

Using a formal model, I study the incentives of both corrupt and clean politicians to support an effective anti-corruption reform. The politicians’ incentives matter, because the very existence of political corruption indicates an agency problem: if a politician can steal when none of the voters want him to do so, then a politician can also choose not to adopt the reform even when the voters want him to do so.3 The conventional wisdom suggests that the incentives of politicians are easy to determine: only a corrupt politician benefits from the political corruption (so, only he may have incentives to block such a reform), while a clean politician always has incentives to support and implement policies that are effective against corruption. To some degree, the gap between the conventional wisdom and the casual observation can be explained by (i) the lack of fully effective anti-corruption reforms, (ii) the high cost of the reform (in terms of the resources and attention it diverts away from more pressing issues), and (iii) the voter apathy (Rose-Ackerman, 1977).4 These and other possible explanations, however, are not completely convincing when the level of political corruption is sufficiently high. Because, then, (i) at least one of the commonly proposed anti-corruption policies (higher salaries, higher penalties, and other institutional changes) is likely to reduce political corruption, (ii) the cost of the reform is justified by the excessive amounts of corruption and its harmful effects (Rose-Ackerman, 1999, Chap. 3), and, (iii) the voters rank political corruption as one of the biggest problems (so, they would support a clean politician in his anti-corruption campaign). I propose another explanation based on the observation that (when political competition is among a small set of career politicians) political corruption changes the zero-sum nature of political competition, producing a positive externality for the clean politician. Then, the clean politician, too, has incentives to not to adopt the reform, because by (effectively) turning his corrupt rivals in future elections into clean candidates, the reform eliminates the clean politician’s competitive advantage in future elections.

In the formal model, two sufficiently forward looking party leaders (referred as politicians or candidates throughout the paper) compete repeatedly.5 The candidates differ in their (both ex-ante and ex-post) popularity due to either their personal characteristics, such as ethnicity, gender, charisma, religion, etc. or due to their social policies, such as the role of religion in government, the rights of a certain minority, etc. The fiscal policy each politician proposes in the elections is also different. The level of corruption is endogenously determined by the politicians’ fiscal policy platforms. In the status quo, one of the politicians choose to steal, while the other one chooses not to. There exists a policy (reform) that will eliminate corruption when implemented. To focus on the politicians’ incentives to support the reform, I assume away all other obstacles to reform: the reform I consider is fully effective, it is costless, and it is supported by a well informed electorate.6 All the reform needs is political support: it will be implemented only if at least one politician adopts the reform as part of his policy platform (and, wins the election).

Studying a game in which each politician simultaneously decides whether to adopt the reform as part of his policy platform or not, I identify the set of parameters under which neither the clean nor the corrupt politician adopts the reform in a pure strategy Nash Equilibrium (PSNE). Referred to as the bad PSNE throughout the paper, such an equilibrium exists when the level of corruption is high.7 Intuitively, for both corrupt and clean politicians, adopting the reform provides a short run benefit and has a long run cost. By (credibly) promising to adopt the reform, a politician increases the probability of winning the election in the period when the reform is an issue. More specific, unless the level of corruption is too low, deviating from the bad PSNE and single-handedly adopting the reform always guarantees an election victory. Yet, the reform eliminates the (current and future) illegal rents of the corrupt candidate, as well as the competitive advantage of the clean candidate (in future elections). This result holds when I consider several different motivations for the latter.

In all different versions of the model, how much (if, at all) a candidate steals is endogenously determined. In the benchmark case studied in Section 2, each candidates is a rent-maximizer, maximizing the present discounted value of his expected rents (the sum of his legal rents, such as ego rents, and illegal rents). In this benchmark case, one candidate stays clean for strategic reasons (he cannot afford to steal due to his lower ex-ante popularity). In the extensions studied in Section 3, the candidate who chooses to steal in equilibrium is still a rent-maximizer. The clean candidate in these extensions either dislikes money stolen from public (legal-rent-maximizer), or, he is committed to a social policy, doing all he can to move the equilibrium social policy closer to his ideal social policy (policy-motivated).

No matter which type of clean politician is considered, when the level of corruption is high, each politician’s (long-run) loss from adopting the reform exceeds his (short-run) benefit and none of the politicians proposing the reform is a PSNE. When the level of corruption is low, the benefit of proposing the reform exceeds the cost for a relatively impatient corrupt politician. In this case, a clean politician’s incentives to adopt the reform depends on whether he is interested in rents or policy. The rent-maximizer clean candidate will almost never deviate from the bad PSNE. Although for the legal-rent-maximizer the illegal rents (stolen money) provide no benefits at all, his incentives to deviate from the bad PSNE are not intrinsically different from those of the rent-maximizer clean candidate. The only difference is that the legal-rent-maximizer’s promise of reform is credible under a larger set of parameters, including lower discount factors. Thus, when he deviates from the bad PSNE, it is due to his impatience, not to his dislike of stolen money.

Compared to the first two types, the policy-motivated clean candidate has stronger incentives to adopt the reform when he is an “extremist,” i.e., when his most-preferred social policy is significantly different from that of the electorate. To have a reasonable chance of winning the election and implementing a policy that he strongly prefers to the most-preferred social policy of the (median) voter,8 an “extremist” has to moderate his social policy in the regular elections. An “extremist” would single-handedly adopt the reform when the level of corruption is low, because bundling his highly unpopular social policy with the reform provides him the only opportunity to implement his most-preferred social policy. Even an “extremist,” however, will not adopt the reform when the level of corruption is high. When competing against a highly corrupt opponent, he has a good chance of winning the election even when he proposes his highly unpopular social policy. Yet, when the reform is implemented, to win the election, the “extremist” will have to moderate his social policy significantly.

It is worth noting that the point of this paper is not that political corruption persists only because of the clean politicians’ lack of support. Several obstacles mentioned above (such as the high cost of the reform and the voter apathy) are all likely to be significant. Yet, to design a successful anti-corruption reform, one needs to be aware of all possible obstacles. Further, being aware of this specific obstacle (both clean and corrupt politicians’ incentives and their power to oppose the reform) can be helpful when choosing among reforms that have similar costs and similar ex-post effectiveness. When, for instance, the politicians are (legal) rent-maximizer s, a reform that uses carrots (higher salaries, higher pensions, etc.) to eliminate corruption is more likely to be implemented than another one based on sticks (higher penalties, better auditing technologies, etc.), or on other institutional changes. Finally, using a formal model to study the incentives of the clean candidate helps to see that the clean politician’s loss from the reform is quite often larger (and, thus, his incentives to oppose the reform are stronger) than that of the corrupt candidate. Thus, the analysis here may help designers and supporters of the reform by noting that the clean candidate is not always a natural ally (he is not even always more likely to support the reform).

As Bai and Lagunoff (2011) notes “While an abundant literature in political economy studies the link from political power to policy, less is known about the ‘reverse causal link,’ i.e., from policy to power.” In the literature, several authors use analytical models of political agency to study which reforms would reduce political corruption when implemented; for a review, see Persson and Tabellini, 2000, Besley, 2006. Fewer studies use agency models to analyze the political will for the reform. Using an extension of the seminal incumbency model of political agency in Ferejohn, 1986, Ferejohn, 1999 examines why an incumbent politician would want to make himself more accountable. In his model, the political agency problem is due to asymmetric information: the voter cannot observe whether the low level of public good is due to corruption or an adverse cost shock. She only receives a noisy signal about the cost of the public good. Ferejohn (1999) explains the incumbent’s desire to increase the precision of this signal, by noting that only then will the voter trust him with more resources, i.e., a larger government. Ferejohn (1999) conjectures that if one considers the incentives of the infinitely many challengers as well, there would be even more accountability in the equilibrium. The main differences between the agency model employed in this paper and the incumbency model are as follows: in the model I study, candidates are not perfect substitutes for each other, and an incumbent’s political career does not end after an election loss. In the setup I study, a politician can risk losing elections for one period; he knows that he will still be the head of his party in the next election. As suggested by “the iron law of oligarchy” (Michels, 1999 [1911]), absent any term limits, the party leaders I consider entrench themselves in power.

In the seminal incumbency model, all incumbents are rent-maximizer s; the problem is moral hazard. Besley and Smart (2007) provide an agency model that has both moral hazard and adverse selection (and, term limits). They do not study the support for the reform, but in the setup they study a (clean) politician’s likelihood of adopting the reform would be even higher. This is because, when the adverse selection is an issue, all politicians will try to signal that they are clean. Adopting (and, later implementing the reform) helps them to do so.

In Caselli and Morelli (2004), the source of agency problem is not the imperfect information; they, too, study a setup in which the voters know which candidates are corrupt.9 They study factors that determine the quality (honesty or ability) of competing political candidates. They find that when the legal rents are low, the high-quality candidates will not run in the elections. They also note that the low-quality (unable or dishonest) incumbents will keep the legal rents low to keep the high quality candidates out of politics (there is a continuum of both types of candidates as well as of the offices that they can win). An honest candidate in the setup they study will always adopt the reform. Because, in their model, political corruption only produces a negative externality by reducing the social status of all the politicians.

Section snippets

The model

In this section, I first describe the status quo and then the reform. Then, I calculate the support for the reform when both the clean and corrupt candidates are rent-maximizers.

Alternative motivations for the clean candidate

The clean politician studied in the previous section was only interested in rents, be they legal or illegal. Even in highly corrupt countries, not all clean politicians can be convincingly modelled as pure rent-maximizer s. Models of purely selfish politicians who have no scruples are quite common in what is sometimes called “the rational choice approach to politics.”30

Conclusion

To summarize, I find that when the political agency problem is due to imperfect competition in a world with career politicians, a bad PSNE in which neither the clean nor the corrupt candidate adopts a fully effective (and, costless) reform exists when the level of corruption is high. When the level of corruption is low, the bad PSNE fails to exists unless the candidates are sufficiently patient. This result holds not only when the clean candidate is a rent-maximizer who cannot steal due to his

Acknowledgement

I am grateful to Sule Atahan-Evrenk, Darlene Chisholm, Dhaval Dave, Timothy Lambie-Hanson, Ben Koskinen, Dilip Mookherjee, Rasim Ozcan, Onur Tas, the participants at the Economics seminars at Bentley and Suffolk Universities and the participants at PET 2010 meeting at Bogazici University, Istanbul, and especially to an anonymous associate editor for very helpful comments. Unfortunately, all the remaining errors are my own.

References (30)

  • T.J. Besley et al.

    Fiscal restraints and voter welfare

    Journal of Public Economics

    (2007)
  • F. Caselli et al.

    Bad politicians

    Journal of Public Economics

    (2004)
  • R. Myerson

    Effectiveness of electoral systems for reducing government corruption: a game-theoretic analysis

    Games and Economic Behavior

    (1993)
  • J. Andreoni et al.

    Tax compliance

    Journal of Economic Literature

    (1998)
  • J.H. Bai et al.

    On the faustian dynamics of policy and political power

    Review of Economic Studies

    (2011)
  • R. Barro

    The control of politicians: an economic model

    Public Choice

    (1973)
  • T.J. Besley

    Principled Agents? The political economy of good government

    (2006)
  • D. de Ferranti et al.

    How to Improve Governance

    (2009)
  • A. Downs

    An Economic Theory of Democracy

    (1957)
  • Evrenk, H., 2009a. A duopoly model of political agency with applications to anti-corruption reform. The B.E. Journal of...
  • H. Evrenk

    Three-candidate competition when candidates have valence: the base case

    Social Choice and Welfare

    (2009)
  • H. Evrenk

    Erratum: three-candidate competition when candidates have valence: the base case

    Social Choice and Welfare

    (2009)
  • J. Ferejohn

    Incumbent performance and electoral control

    Public Choice

    (1986)
  • J. Ferejohn

    Accountability and authority: toward a theory of political accountability

  • D. Fudenberg et al.

    The folk theorem in repeated games with discounting or with incomplete information

    Econometrica

    (1986)
  • Cited by (8)

    • The role of international anti-corruption regulations in promoting socially responsible practices

      2021, Journal of Economic Behavior and Organization
      Citation Excerpt :

      Çule and Fulton (2009) study the inspection game, showing that it has multiple equilibria, and that corruption is persistent in many of these, meaning penalties can have perverse effects. Balafoutas (2011) studies how perceived public beliefs affect corrupt behavior in a repeated game, while Evrenk (2011) models a repeated game between a clean and a corrupt competition, showing that when corruption is high, in equilibrium, no politician adopts a political reform targeting corruption. Examining other aspects and settings, Zhu (2012) explains why higher penalties may fail to deter corruption; Accinelli and Carrera (2012) study how corruption is driven by imitation and Accinelli et al. (2017) analyze how politically active citizens can prevent the spread of corruption; Cerqueti and Coppier (2016) examine corruption in relation to pollution inspection issues; Litina and Palivos (2016) report the existence of multiple self-fulfilling equilibria in which different levels of corruption emerge; Strîmbu and González (2018) show that higher transparency increases the size of bribes paid.

    • Voter heterogeneity and political corruption

      2020, Journal of Economic Behavior and Organization
      Citation Excerpt :

      The setup of their paper is different to ours because we assume that voters can choose to elect non-corrupt politicians. Also related in this literature is Evrenk (2011), who considers a model with voter bias where a clean and a corrupt politician compete in an election and as part of their manifesto they can choose to eliminate corruption. In his study, although politicians could eradicate corrupt institutions, both the corrupt and non-corrupt politician have an incentive to preserve corrupt institutions, because these institutions provide rents to the corrupt, but also a competitive advantage to the clean politician.

    View all citing articles on Scopus
    View full text