Party system competitiveness and corruption

Abstract

This article examines why democratic competition sometimes fails to curb governmental corruption. We argue that in democracies party system competitiveness, which shapes the ability of voters to effectively select and control their politicians through elections, plays a critical role in conditioning the scope for corruption. For voters, governmental corruption is a classical principal–agent problem and its magnitude is mediated by the extent to which the competitiveness of a party system helps to make information and effective choices available to the electorate. Informed voters who can coordinate on credible alternatives to under-performing and corrupt incumbents, we argue, can select politicians who are likely to curb corruption and hold accountable those who do not. We test this argument through a controlled comparative analysis of corruption in 70 democracies around the world and find broad support for our hypotheses.

Keywords

Corruption empirical research global party systems theoretical critique

Introduction

Even in fully democratic political systems governmental corruption is strikingly common. This raises a series of questions: Why does democratic competition sometimes fail to curb malfeasance? Why can elections help corrupt politicians to power? and Why do voters often fail to punish corrupt or under-performing incumbents? These questions are puzzling because electoral competition should in principle allow voters to select politicians who will curb corruption and to replace representatives who do not. Yet, empirically, this is often not the case. This article examines under what conditions electoral competition can fail to secure clean government in the interest of the people. Our argument is that party system competitiveness, which shapes the effectiveness of elections as tools to select and control politicians, plays a critical role in conditioning the scope for governmental corruption.

In advancing this argument, we build on two distinct literatures in political science: empirical work that comments on the relationship between corruption and party system competitiveness, and positive democratic theory. Among the empirical literature, case-oriented work presents abundant evidence that party systems exert a powerful influence on corruption (della Porta, 2004; Geddes and Ribeiro Neto, 1992; Giliomee, 1998). Comparative work has also probed the implications of party system competitiveness for governmental corruption (Charron, 2011; Scheiner, 2005; Tavits, 2007). However, the theoretical claims and empirical results generated by this literature are contradictory. While some of this literature argues that the entry of effective opposition parties into a party system raises competitiveness and helps constrain corruption, other authors suggest that larger numbers of parties hamper effective accountability and give greater scope to corruption. As a result, no consistent argument has emerged about the effects of competitiveness on malfeasance.

In this article we draw on the insights of work in the positive theory tradition to address the contradictory conclusions that have characterized the empirical literature to date. Following the theoretical work, we conceive of corruption as a classical principal–agent problem that can arise between voters (the ultimate democratic principal) and politicians (their agents). The magnitude of this agency problem is conditioned by the information available to voters about their politicians and the effectiveness of their electoral choices in controlling their representatives. Informed voters, who can coordinate on credible alternatives to incumbents that permit or engage in corruption, we argue, can select better types of politicians and hold accountable those who turn out to be bad types. Party system competitiveness influences the scope for corruption because it shapes both the information and the effectiveness of the choices available to the electorate. By focusing on these two critical mechanisms, we are able to clarify conceptually how party system competitiveness is shaped by fragmentation and governing party dominance, and how it affects governmental corruption. Our main argument is that competitiveness can be expected to reduce the latitude for corruption, but its relationship to fragmentation is non-linear and governing party dominance constitutes a distinct dimension of competitiveness. We test these arguments through a controlled comparative analysis of corruption in 70 democracies around the world and find broad support for our hypotheses.

The article proceeds as follows. The next section introduces the findings of existing research on party system competitiveness and corruption. We then outline our theoretical approach and develop testable hypotheses before introducing the data, variable operationalizations and model choice. The final sections present our results and robustness checks before we conclude.

The literature on party system competitiveness and corruption

A striking feature of the work that examines the effects of competitiveness on accountability and corruption is that it follows two entirely divergent lines of reasoning. One set of arguments emphasizes the corruption-reducing effects of competition. From this perspective, competition is viewed as essential in giving rise to opposition parties that can inform the electorate about corruption and give voters the opportunity to ‘remove those in government who have abused their power’ (della Porta, 2004: 49). When effective competition from opposition parties is reduced – work on Italy, Japan and South Africa suggests – democratic accountability is at risk, and even deeply unpopular, under-performing and corrupt incumbents can survive in office for prolonged periods of time (Giliomee, 1998; Mershon, 2002; Scheiner, 2005).

A second line of argument, however, stresses the potentially corruption-enhancing effects of competition. From this perspective, clarity of responsibility and voter coordination are the critical mechanisms in ensuring accountability and curbing corruption. Work in this tradition argues that effective accountability can only operate if voters can clearly identify which parties are responsible for government performance and if they can coordinate on a clear alternative to the incumbents in order to remove them from office. More competitive multiparty systems, and the coalition governments that typically ensue, are thought to make it more difficult for voters to identify who is responsible for government policy and to coordinate on an alternative to the incumbent in elections (Anderson, 2000; Lewis-Beck, 1988). For both of these reasons systems that feature larger numbers of parties are expected to give greater latitude to corruption (Charron, 2011; Tavits, 2007: 221).

The empirical results reported by these studies replicate the apparent theoretical confusion. While Tavits (2007) finds that corruption rises with party system fragmentation and Charron (2011) reports that corruption is reduced by two-party systems (at least in the context of single-member district electoral systems), case-oriented work suggests that party systems with a low effective number of parties feature raised levels of corruption (Giliomee, 1998; Scheiner, 2005).

A problem besetting this literature is that it lacks conceptual clarity about the relationship between the number of parties that compete and the competitiveness of the party system. The assumption implicit in some of this work is that competitiveness rises in number of competitors, while the assumption of the remaining work is that the number of competitors is inversely related to competitiveness. Both sets of arguments conceive of the relationship between the number of competitors and the scope for corruption as linear, but it is clearly impossible for both assumptions to be correct. A more accurate and general account of how competitiveness affects the latitude for corruption must therefore clarify these conceptual relationships. In what follows we develop such an account.

Effective elections, the selection and control of politicians and corruption

We conceive of governmental corruption as the misuse of public office for personal or political gain, as well as the acquiescence in such misuse by bureaucrats (Key, 1936; Tavits, 2007). This definition encompasses all forms of governmental corruption – grand and petty forms of theft, bribery and rent-seeking by public officials. We therefore view governmental corruption as a public policy outcome for which politicians are in principle accountable – they may invest resources in fighting it, or tolerate it and possibly even engage in it.

Surveys around the world indicate that a majority of citizens (and thus the median voter) typically perceive corruption as a deleterious public policy outcome that results when their political agents – elected politicians – are unwilling or unable to redress malfeasance by public officials. This is true even in contexts in which corruption is endemic, so that citizens regularly experience it and benefit from corrupt transactions in individual circumstances, as is often the case in highly clientelistic political systems. For instance, in African democracies, as in other polities in which corruption and clientelism are widespread, decisive majorities of citizens typically regard corruption as morally reprehensible (Persson et al., 2010). The 2006 Afrobarometer makes clear that over 60 percent of all African respondents regard corruption by public officials as wrong and punishable (Afrobarometer, 2006). That is not surprising. Even when citizens are frequently involved in corrupt exchanges, and benefit from them in specific instances, this does not imply that they approve of governmental corruption to confer benefits on other groups of citizens, nor does it imply citizen tolerance for rent-seeking by politicians. Indeed, citizens clearly perceive a gap between their preferences and the actions of their representatives. As Transparency International’s Global Corruption Barometer, which polled 73,132 people in 69 countries and territories around the world indicates, 68 percent of all respondents perceived political parties as corrupt and 56 percent perceived their government’s efforts in fighting corruption as ineffective (Transparency International, 2009: 6, 33). Put differently, for citizens collectively, governmental corruption is an agency problem that arises when politicians deviate from the electorate’s interests.

We therefore apply a principal–agent approach and analyze governmental corruption as an adverse policy outcome that can result in the delegation relationship between voters (the principals) and their politicians (their agents). Voters delegate government power to politicians to attain certain goals – such as clean government – but self-interested politicians may abuse their office or tolerate malfeasance by bureaucrats when their interests are not well aligned with their electorate’s interests (Adsera et al., 2003: 447).

Much of the formal literature in classical democratic theory analyzes precisely the problem that politicians may not act in the interests of citizens. This work gives rise to two central insights. First, elections are a critical tool for voters to control politicians. Second, elections can only serve as a means to control politicians when they have two properties: First, they must be effective mechanisms of selection by which voters can select ‘good’ types of politicians whose aim is to serve the electorate (Fearon, 1999), and, second, elections must be effective sanctioning devices through which voters can reward or punish incumbents given their record in office (Banks and Sundaram, 1993; Barro, 1973; Ferejohn, 1986). Elections that are ineffective on either dimension give rise to two risks, adverse selection and moral hazard. Adverse selection is the risk that voters will elect politicians who do not have the motivation or skills to act in the interest of voters. Moral hazard is the risk that politicians will abuse their office by perpetrating or acquiescing in corruption because they cannot be effectively monitored and punished. As in other instances of delegation, these agency problems have their sources in asymmetries of information or limitations on the principal’s ability to translate information into effective choice.

Thus, the magnitude of these agency problems can be expected to vary with the information and effectiveness of the choices available to voters. When information is so poor that the electorate has difficulties telling the difference between corrupt and non-corrupt types of politicians, the risks of adverse selection and moral hazard are magnified (Fearon, 1999: 79). Indeed, as Ferejohn points out, ‘the greater the informational advantage that officials hold, the greater their ability to earn rents from office-holding’ (1986: 10). As a result, voters may elect representatives whose preferences diverge from their own or they may fail to punish politicians who allow corruption to flourish because they cannot discern the politicians’ true type.

Equally critical in limiting adverse selection and moral hazard is the availability of effective alternative electoral choices and the capacity of voters to coordinate on them. When credible challengers are not available or voter capacity to coordinate on them is reduced, the electorate may not be able to ensure the selection of good agents or to punish incumbents who have allowed or engaged in corruption (Ferejohn, 1986: 18, 1999: 133; Myerson, 1993: 119).

Thus, a principal–agent approach identifies information and effective electoral choice as critical variables in enabling voters to control their politicians. In modern representative democracies, of course, electoral choice and information are mediated by party systems. Parties inform and mobilize voters, and structure the choices of the electorate at the polls. The nature of the party system can therefore be expected to impact on the ability of voters to control their politicians. In the section that follows we argue that competitive party systems, which give voters information and effective choices to control their representatives, limit the scope for corruption. As we shall see, analyzing how party system competitiveness impacts on voter information and choice enables us to resolve, within a coherent theoretical framework, several of the contradictions that have characterized the empirical literature to date.

Party system competitiveness and the scope for corruption

Duverger’s early work defined party systems as the ‘forms and modes’ of competition for votes and governmental office among the constituent parties (Duverger, 1954: 203), and since then an extensive literature has noted party system competitiveness as a central dimension of variation among party systems. The competitiveness of party systems is typically analyzed along two separate dimensions: variation in the number of parties that compete and the extent to which competition is characterized by patterns of dominance (Bogaards, 2004; Duverger, 1954; Laakso and Taagepera, 1979). Party systems can be characterized as competitive when they (a) confront incumbents with well-defined opposition alternatives and when (b) these opposition alternatives represent effective choices so that voting for the opposition can be politically consequential and can alter who controls the power to govern. In competitive party systems, opposition politicians make greater efforts to mobilize support, voters pay more attention to politics and the pressure on incumbents to perform rises (Kitschelt, 2007).

Thus competitiveness depends on two conditions. First, in order for party competition to revolve around well-defined alternatives, citizens must have information about the performance of the incumbent and about the choices offered by the opposition. Unless citizens have that information, they will have difficulty distinguishing the options before them, including corrupt and non-corrupt offerings. Second, voters ‘must have a fair opportunity to cast a meaningful vote for or against the policymakers’ (Powell, 2000: 51). Reduced competitiveness in party systems that limit the choices on offer, or give rise to coordination problems for honest challengers and voters, can powerfully undermine the ability of voters to cast meaningful votes against the incumbent. Thus, the conditions that restrict party system competitiveness also restrict the electoral control of politicians and can, therefore, be expected to broaden the scope for corruption.

This analysis suggests that the relationship between competitiveness, party system fragmentation and dominant party systems may be more complex than previously understood. In particular, increasing the number of parties which compete in a party system cannot unambiguously be expected to raise the information and the political effectiveness of the opposition choices available to voters. The next section develops this argument in more detail.

Party system fragmentation

Scholars have long recognized that the number of competitors is an important characteristic of party systems. Consequently, party systems were initially classified as two- or multiparty (Duverger, 1954), and then by the use of continuous measures of the level of fragmentation, most notably the effective number of parties (Laakso and Taagepera, 1979). The level of fragmentation varies extensively among party systems. At one extreme lie party systems that offer a very restricted range of effective choices to voters. Namibia illustrates this type of party system well. Reflecting the dominance of SWAPO as the party that ushered in independence, Namibia throughout the 2000s had an average of just 1.7 effective electoral parties. At the other extreme lie democracies with highly fragmented party systems that offer voters extremely diverse choices and few clues about who is likely to emerge as the winner (Coppedge, 1998). In 2007, that group included Colombia, Belgium and Brazil (with between 8.6 and 11 effective electoral parties).

Party system fragmentation affects the electoral control of politicians because it shapes the effectiveness of the choices and the quality of information available to voters. At the lower bound, a very restricted number of effective parties signals a degree of electoral hegemony that in itself presents formidable barriers to achieving a politically consequential vote for the opposition. The level of mobilization, opposition coordination and vote switching required to unseat an incumbent in such systems is so extensive that a rejection of the incumbent by voters is often not followed by the actual loss of policy-making power. This can be expected to raise the risks of adverse selection and moral hazard and, from a theoretical perspective, both Myerson (1993: 119) and Ferejohn (1986: 18) anticipate that voter control over politicians is reduced when the number of parties is restricted. However, the increase in fragmentation required to make a party system significantly more competitive at this lower bound is not big. A rise from 1.7 to between 2 and 3 electoral parties (the range of fragmentation that characterized the party systems of the US, Uruguay, Hungary and Spain in 2007) is often sufficient to give rise to genuine electoral competitiveness and meaningful voter choice, which can be expected to improve the electoral control of politicians and thus reduce the scope for corruption.

But as the number of parties continues to rise, both information costs and coordination problems can be expected to offset and reverse the beneficial effects of increased competition. Other things equal, highly fragmented party systems make it more costly for voters to gather sufficient information to assess the record of incumbents and the promises of potential challengers, which can compromise their ability to distinguish between honest competitors and those who are corrupt or do not see curbing corruption as a priority. In addition, high levels of party system fragmentation accentuate the coordination problems for clean challengers and voters in replacing corrupt or under-performing incumbents. If an electoral challenge is to be successful, voters must converge on the opposition party or coalition most likely to oust the government (Ferejohn, 1986: 22). By presenting voters with a broad choice of potential challengers, highly fragmented party systems tend to split the opposition vote, making coordination on any one challenger less likely. This in turn reduces the probability that a rejection of the incumbent by voters will result in the loss of policy-making power. Moreover, as Kurer (2001) makes clear, these problems are mutually reinforcing – the greater the information failures, the smaller the probability that voters will be able to coordinate effectively. In sum, we anticipate that high levels of party system fragmentation accentuate adverse selection and moral hazard problems for voters.

By focusing on the implications of fragmentation for voter information and the effectiveness of voter choice – the two critical determinants of the electoral control of politicians – our approach clarifies that fragmentation can be expected to have a non-linear effect. This insight not only departs from earlier work, but also helps to resolve some of the theoretical contradictions that characterized it. Our first hypothesis, then, is that:

Hypothesis 1: Corruption initially improves as the effective number of parties rises, but this effect reverses at high levels of fragmentation.

The second dimension of variation in party system competitiveness arises from patterns of dominance.

Dominant party systems

The concept of dominant party systems captures situations in which multiple parties exist, but the existence of this range of political choices has, for various reasons, little or no political consequence. As a result, the opportunities for political forces to win or lose power through elections are reduced. The ‘uncommon’ democracies of Italy and Japan where one party dominated the country’s politics throughout most of the post-war period by virtue of its majority status or as a plurality party controlling the parliamentary pivot (Mershon, 2002; Scheiner, 2006), as well as many African democracies, which feature multiparty elections but no changes in government, are cases in point (Bogaards, 2004; Bratton and Van de Walle, 1997).

Dominant party systems are defined by the protracted and dominant position of a party or coalition in government (Bogaards, 2004). Their origins typically lie in a combination of voter cleavages and the competitive strategies adopted by parties (Arriola, 2011; Mershon, 2002). Thus, the exceptionally high prevalence of such systems in African democracies is often traced to voter loyalties commanded by parties that led the independence or democratization struggles (such as the ANC in South Africa), but it also appears to reflect the use of patronage by governing parties to undermine the incentives for opposition politicians to coalesce and form an effective electoral opposition (Arriola, 2011). In Europe, dominant party systems are similarly based on voter cleavages, as well as parties’ competition strategies, and tend to arise when a party successfully positions itself ideologically as the core party, which implies its membership in all possible coalitions. In the Netherlands, Dutch Christian Democracy (the KVP/CDA) occupied such a position; in Italy the Christian Democrats held core party status from 1946 through 1992 (Mershon, 2002: 12-3.).

Dominant party systems can be expected to accentuate adverse selection and moral hazard problems for two reasons. First, the dominant presence of one party in government creates incentives for other coalitionable parties to collude with it, because entering government requires them to enter into coalition with the dominant party. Thus, in Italy, the long era of dominance by the Christian Democrats was accompanied by ‘a strong tendency toward inter-party collusion’ (della Porta, 2004: 51). Collusion was typically reinforced by agreements that gave the other parties a stake in tolerating malfeasance by the dominant party, for instance through the ‘fixed distribution of public contracts among firms based on the colour of their political protection’ (della Porta, 2004: 51). Collusion affects the capacity of voters to control their politicians because it compromises the flow of information and, thus, their ability to distinguish between clean and corrupt types of politicians and to punish the corrupt ones. Second, the mechanisms by which dominance emerges – be that the positioning of a party or coalition in the ideological core of the party system, or the use of patronage and clientelism – limit the effectiveness of voter choice. As Arriola notes, in Africa, dominance is often achieved by incumbents who deliberately use patronage to enhance the coordination problems for opposition parties in mounting an electoral challenge (Arriola, 2011). In Japan, the heavy reliance of the LDP on clientelism in the context of a fiscally centralized governmental structure, equally, had the effect in fostering opposition party failure and reducing the probability that opposition parties would successfully challenge the clientelist regime (Scheiner, 2006). Similarly, core parties are insulated to a large degree from the effects of electoral punishment by their ideological position, which tends to secure their inclusion in government even if they are reduced in size. Thus, the mechanisms that give rise to dominant party systems can be expected to blunt the threat of electoral punishment.

For both of these reasons we anticipate that patterns of dominance correlate with higher levels of corruption. From a theoretical perspective, Ferejohn and Myerson note that mechanisms which limit successful challenges help to ‘maintain collusive opportunities for officeholders of the established party’ (Ferejohn, 1986: 23; see also Myerson, 1993: 119). Case-oriented work on South Africa and Italy supports these theoretical expectations. Thus, Giliomee finds that South Africa’s dominant party system contributed to widespread corruption through the abuse of state patronage (1998: 129). Similarly, Mershon observes that the Christian Democrats’ core party status in Italy enabled ‘corruption of unprecedented scale and reach’ (2002: 184). We therefore expect that:

Hypothesis 2: Corruption is more pronounced in dominant party systems.

In sum, our argument is that these systemic features of party systems – fragmentation and dominance – condition competiveness by shaping the quality of information and the effectiveness of the choices available to voters. To the extent that these party system features reduce the flow of information, or restrict the ability of the electorate to make effective electoral choices and coordinate on opposition alternatives, they broaden the scope for corruption. We test this argument in the section that follows, but, before we do so, one further concern must be addressed.

The above discussion raises the question how far fragmentation and dominance constitute distinct aspects of party system competitiveness. Specifically, dominance is often thought to be highly correlated with very low levels of fragmentation. Yet, from a theoretical perspective, there is no reason to expect a strong correlation between the effective number of parties in a political system and the emergence of a dominant party system. As we have seen, dominance is based on an interaction between voter cleavages and competitive strategies, such as the use of patronage, clientelism or core party positioning, which are not only capable of giving rise to prolonged incumbency in the context of multiparty systems, but often also reinforce opposition fragmentation. Indeed, extant work finds little correlation between dominance and low numbers of effective parties. Both Italy and the Netherlands have party systems with an intermediate number of parties and, as Bogaards notes, in Africa dominant party systems feature anything ‘from one to more than three effective parliamentary parties’ (2004: 188). As is consistent with the theory and findings of previous work, our data suggest that the correlation between dominance and party system fragmentation is weak and not statistically significant (r = –0.14, p-value = 0.26).¹ Theoretically and empirically, it is therefore clear that the two dimensions of competitiveness we identify are distinct.

Data and dependent variable

Since the question why voters may fail to control their politicians is of interest only in functioning democracies, and not where electoral manipulation and fraud foil the democratic process (Kurer, 2001: 65), we test our hypotheses about the effects of party system competitiveness on governmental corruption only in fully democratic polities that rank 6 or higher on the Polity Index of Democracy. Our unit of analysis is the country and our data cover 70 democracies observed in 2007 (see Appendix 1 for a list of the countries included in the analysis).

One of the most widely used and accepted measures of corruption is Transparency International’s Corruption Perceptions Index. This index gauges the essentially hidden phenomenon of corruption via surveys of international and domestic business people, risk analysts and residents of a country, and aims to capture the extent to which public power is exercised for private gain, including petty and grand forms of corruption, as well as ‘capture’ of the state by elites and private interests. To arrive at robust indicators, Transparency International averages the standardized values from these surveys and provides corruption perception scores only for countries for which a minimum of three surveys are available. The two critical advantages of this index are its breadth of coverage and the variety of sources employed, which reduces its susceptibility to survey-specific or question-specific idiosyncrasies.

Despite these advantages, though, working with these data poses challenges because the data record corruption perceptions, rather than the frequency or seriousness of actual corruption. Unfortunately, no measures of actual corruption exist for a sufficiently large number of cases to enable cross-national analysis. Surveys that gauge corruption experiences come closest to providing such a measure, but their coverage of countries and years is as yet too limited. Corruption perceptions, though, do appear to capture the underlying frequency of corrupt interactions. As Treisman reports, the correlation between the Transparency International measures of corruption perceptions and the main survey measures of corruption experiences are reassuringly high and statistically highly significant, with correlation coefficients that range from 0.64 to 0.79 (Treisman, 2007: 218). To address any residual concerns, we robustness test our results using a different measure of perceived corruption drawn from the World Bank’s Governance Indicators. To the extent that both measures yield the same result we can have greater confidence in our findings.

Independent variables: Measures and measurement validity

To gauge party system fragmentation, we employ the effective number of electoral parties (ENEP) as measured by the Laakso–Taagepera Index. The majority of these data are drawn from Gallagher and Mitchell (2008) and augmented using Golder (2005), with remaining missing values calculated by the authors. We take the natural logarithm of ENEP because the marginal effect of each additional party can be expected to decrease as the number of parties rises. To capture high levels of party system fragmentation we include the quadratic term of the logged variable, the expectation being that the effective number of parties will initially improve perceived corruption, but the quadratic term should have the opposite effect.

Because governing party dominance is a party system feature that is conceptually complex, we describe not only our measure but also the tests we performed to establish its validity. We measure governing party dominance by the number of years a governing party has spent in office consecutively, coded from the International Parliamentary Union Database and the Psephos Election Archive.² Since years-consecutively-spent-in-office is a variable with a distribution that is heavily skewed to the right, we take the natural logarithm. This measure captures the initial effects of ordinary incumbency, say a government’s first and second term in office – about which we have no theoretical expectation – as well as party systems in which a governing party has established long-term dominance. To differentiate between these two effects, we include the main and quadratic term of the logged variable. Our expectation is that long-term governing party dominance, captured by the quadratic term, accentuates corruption.

The question that arises is how well this measure captures the complex concept of governing party dominance. We therefore examine its construct validity by probing how far it correlates with two other features that often characterize dominant party systems – high levels of opposition fragmentation and high vote-shares of the dominant governing party (Bogaards, 2004). As anticipated, our data show that long-time ruling parties tend to face much more fragmented oppositions than ordinary incumbents. On Beck et al.’s (2001) opposition fragmentation index, which ranges from 0 to a maximum of 1, average opposition fragmentation for countries in which the longest serving governing party spends less than eight years in power (e.g. less than approximately two terms) is 0.48, but for countries which feature incumbents who spend 12 years or more in power (e.g. approximately three consecutive terms in office or more), average opposition fragmentation rises to 0.55. Similarly, long-term incumbents tend to win significantly larger vote-shares, as expected. Ordinary incumbency of up to eight years in office (e.g. approximately two terms) is associated with an average vote-share of the largest governing party of only 33 percent (as measured by Beck et al. (2001)). However, where parties serve 12 or more years in power, the largest party wins on average fully 53 percent of the vote – more than an absolute majority. Thus, our measure appears to capture the concept of dominant party systems well.

Control variables

We employ two sets of control variables which have been shown to affect corruption in previous cross-national work. The more parsimonious set of controls includes the age of democracy, constitutional, economic, social and regional factors. The age of democracy is usually expected to reduce corruption because, over time, the re-election imperative can, all else equal, be expected to incentivize politicians to reduce governmental corruption (Montinola and Jackman, 2002). We measure the age of democracy by the number of consecutive years that a country has been rated 6 or better on the Polity Index of Democracy (Marshall et al., 2010). In addition, different democratic constitutions are thought to vary in the extent to which they limit opportunities for corruption and rent extraction. Thus, constitutions which decentralize power and those that feature executive presidents increase the number of institutional veto players in the policy process, which some scholars argue reduces transparency and increases opportunities for rent extraction and governmental corruption (Gerring and Thacker, 2004; Kunicova and Rose-Ackerman, 2005). We gauge decentralization using Beck et al.’s (2001) measure, which records the extent to which a country has autonomous, locally elected, governments, and includes an indicator for democracies that feature executive presidents (all presidential and semi-presidential regimes = 1, otherwise 0) based on Svolik’s (2008) coding.

Economic conditions have been shown to have a powerful influence on corruption. Thus economic development can be expected to curb corruption because it ‘increases the spread of education, literacy, and depersonalized relationships – each of which should raise the odds that an abuse will be noticed and challenged’ (Treisman, 2000: 404). Additionally, the ability of officials to extract rents in the domestic market should be reduced when that market is open (Gerring and Thacker, 2005; Treisman, 2000). We measure economic development using the natural logarithm of real GDP per capita (in constant 2000 US$), reported as part of the World Bank World Development Indicators. Trade openness, also drawn from the World Bank’s Development Indicators, is measured by the sum of a country’s imports and exports as a share of GDP – missing country-years were completed using import and export data as reported in the IMF’s International Financial Statistics.

In addition, we take account of the fact that concerns about corruption may not be equally salient for voters across different societies. Although voters may well prefer clean to corrupt politicians, other competing considerations can outweigh this concern at the polls. This argument is made particularly frequently for ethnically and linguistically divided societies in which opportunities for advancement, material, cultural and political benefits are accessed on the basis of ethnic and linguistic affiliation (Chandra, 2006: 418). In such contexts ethnic and linguistic loyalties are often expected to override voter concerns about corruption because even a voter who believes that her co-ethnic politicians are corrupt and will promote her interests very imperfectly may still choose to elect them since politicians of other ethnic groups cannot be expected to act as her agents at all (Padro i Miquel, 2007). To account for this possibility we use Alesina et al.’s (2003) index of ethno-linguistic fragmentation.³

Finally, we include a series of regional indicators for the Former Soviet Union, the Middle East, Central and Latin America, Asia and Africa in the analysis to account for unobserved regional influences on perceived corruption.

The more extensive set of controls additionally captures electoral system features which - like party system competitiveness – shape electoral information and choice. Thus, plurality electoral systems are often expected to make it easier than PR lists for voters to attribute responsibility (Kunicova and Rose-Ackerman, 2005). In addition, PR systems, in particular in combination with open lists, are thought to induce politicians to focus on personal reputations in order to differentiate themselves from their co-partisans and ‘to use illegal proceeds to fund electoral competition’ (Chang and Golden, 2006: 119) thus skewing the choices available to voters toward corrupt types. To account for the nature of the electoral rules, a series of indicator variables is used to record whether a country employs Proportional Representation, Plurality and Open Lists, as well as an interaction to capture Open List PR systems (Open List*Proportional Representation).⁴ The electoral systems data are drawn from Beck et al. (2001) and augmented using Regan and Clark (2010) and Golder (2005). Since the electoral system variables can be expected to influence not just voter information and choice, but also the scope for governmental corruption, their inclusion can spuriously obscure the relationship between party system effects and corruption. However, as we shall see, despite these confounding effects party system competitiveness is associated with governmental corruption in a manner that is consistent with our expectations.

Models and results

Our model choice and specification is driven by three considerations. First, the dependent variable – the Transparency International Corruption Perception Index – is a continuous variable, ranging from 2.3 to 9.4 in our data, which makes an OLS regression model the appropriate choice. Second, to account for potential sources of unobserved heterogeneity between countries we estimate robust standard errors. Third, in examining the effect of party systems on corruption the possibility of reverse causation is an important concern. Party systems themselves result – at least in part – from the choices of politicians who may wish to protect corrupt practices by limiting the information and choices available to voters to identify and avoid or replace corrupt politicians. To address this concern about the direction of causality, we lag all of our explanatory variables which capture aspects of the party system, as well as all time-varying control variables, by a period of seven years.

Table 1 presents the results of the analysis.⁵ Model 1 uses the parsimonious set of control variables, while Model 2 employs the more extensive set of controls. Across both specifications, the pattern of party system coefficients reflects the expectations we derive from our analysis of corruption as a principal–agent problem between voters and politicians. Unless otherwise noted, the discussion of the results is based on Model 1 (for all predicted changes in corruption scores described in this section, see Table 2).⁶

Table 1.

Party system competitiveness and corruption perceptions.

	(1)	(2)
	Parsimonious controls	Full controls
Party system competitiveness
H1 Fragmentation (ln ENEP)	2.654**	2.306**
	(0.796)	(0.790)
Fragmentation sq	–0.763**	–0.680**
	(0.201)	(0.203)
H2 Governing party dominance	0.723**	0.896**
	(0.258)	(0.255)
Governing party dominance sq	–0.134*	–0.176*
	(0.065)	(0.078)
Controls
Age of democracy	0.590**	0.543*
	(0.215)	(0.245)
Decentralization	–0.040	–0.162
	(0.203)	(0.214)
Executive president	–0.144	0.021
	(0.360)	(0.350)
GDP per capita (ln)	0.929**	1.139**
	(0.158)	(0.214)
Trade openness	0.044	0.113
	(0.267)	(0.267)
Ethnolinguistic fragmentation	0.131	0.647
	(0.763)	(0.825)
Former Soviet Union	0.614	0.630
	(0.631)	(0.624)
Middle East	–1.229**	–1.460**
	(0.365)	(0.385)
Central and Latin America	–1.106*	–1.171*
	(0.466)	(0.477)
Asia	–0.093	–0.168
	(0.530)	(0.500)
Africa	0.496	0.308
	(0.549)	(0.527)
Proportional representation		0.036
		(0.375)
Plurality (first-past-the-post)		1.073*
		(0.488)
Open lists		–0.356
		(0.401)
Open lists*proportional representation		–0.298
		(0.791)
Constant	–7.039**	–8.590**
	(2.140)	(2.283)
Observations	70	70
R-squared	0.847	0.861

Robust standard errors in parentheses +significant at 10%; *significant at 5%; **significant at 1% (two-sided).

Table 2.

Magnitude of predicted changes in corruption scores.

	One SD increase from	Change corruption score	Percentage change	SE	95% confidence interval
Fragmentation (ENEP)	Min	3.05	43%	0.92	1.23	4.81
Fragmentation (ENEPsq)	Min	–3.94	55%	1.04	–5.85	–1.99
Governing party dominance sq	Min	–1.20	17%	0.55	–2.24	–0.15

All predicted changes are based on Model 1. Effects for all logged variables are reported in the original metric. Other variables were kept constant – continuous variables at their mean, dichotomous variables at their mode.

Table 2 makes clear that corruption scores respond strongly to changes in the competitiveness of the party system. We anticipated that party system fragmentation initially improves perceived corruption levels because it enhances competition and choice for voters. As expected, an initial increase in the effective number of parties is always associated with a significant reduction in corruption perceptions. This effect appears large in Model 1 – a one standard deviation increase from the minimum of 1.7 to 5.3 reduces corruption scores by 43 percent. As expected, the effect of fragmentation is highly non-linear and reverses at high levels of party system fragmentation. Thus, a one standard deviation increase in the squared number of effective parties yields a very significant 55 percent rise in corruption perceptions, which is consistent with our hypothesis that highly fragmented party systems raise the information costs for voters and create coordination problems in the effort to punish corrupt incumbents (H1).

Long-term governing party dominance, captured by the quadratic term, also has a powerful effect. A one standard deviation increase is associated with a 17 percent deterioration of corruption scores, as is consistent with our argument that long-term dominant party systems foster collusion and blunt the tool of electoral punishment (H2). Note that the main effect of governing party dominance, which captures ordinary incumbency and about which we had no expectation, is positive and reduces the perceived levels of corruption significantly.

Turning to the controls, the age of democracy consistently correlates with significantly improved corruption scores, as anticipated. The coefficient for decentralized constitutions always has the negative sign Gerring and Thacker (2004) would expect – indicating that they tend to correlate with worse corruption scores. But as previous research has suggested (Treisman 2007), this effect, as well as the effect of executive presidents, is fragile and neither coefficient is precisely estimated. Wealth is – as previous work overwhelmingly suggests – associated with significantly improved levels of perceived corruption. The coefficients for trade openness and social structure, however, fall short of statistical significance. The region dummies suggest that Middle Eastern and Central and Latin American democracies tend to have worse corruption perception scores, while the indicators for Africa, Asia and the Former Soviet Union do not reach statistical significance.

In Model 2, which includes the expanded set of controls for electoral system features, the party system variables retain their significance and substantively sizeable effects. The additional controls have the expected signs. While plurality electoral systems correlate with significantly improved perceptions of corruption, corruption appears to be worse in the context of open-list proportional electoral systems (captured by the interaction). However, the effects of the interaction and the indicators for PR and open-list systems are fragile, as previous work suggests, and do not reach conventional levels of statistical significance.

Thus, whether we use the more parsimonious or the fuller set of controls, the variables of theoretical interest always have the expected signs and are statistically significant. The level of explained variance across the two models is high with R-squared statistics of 0.85 and 0.86.

Robustness

To test the robustness of these results further, we take two different approaches (models not presented). First, we re-run our models substituting the control of corruption dimension of the World Bank Governance Indicators (Kaufmann et al., 2004) for Transparency International’s Corruption Perception Index as the dependent variable. The surveys and method of aggregation used by the World Bank differ from those used by Transparency International, and to the extent that these two different measures of corruption as perceived by a range of different actors yield substantively the same results, we can have greater confidence in the robustness of our findings. This test shows that our findings are robust to the use of the World Bank measure as an alternative dependent variable. Second, because the shared variation between the main and quadratic terms for party system fragmentation and governing party dominance is high, it is difficult to distinguish in these models the separate effects of these terms. For this reason, we residualize the two quadratic terms to render them uncorrelated with the main terms and then replicate the analysis. The results of these analyses confirm that fragmentation has a highly non-linear effect and that there are no numerical problems in estimating the quadratic terms for both fragmentation and governing party dominance, which retain their negative signs, size and significance.

In sum, we find that party system competitiveness has powerful effects on levels of corruption as perceived by citizens, domestic and international business people and risk analysts.

Conclusion

In this article we have drawn on work in the tradition of positive democratic theory to resolve some of the contradictions in the empirical literature on party system competitiveness and corruption. Applying a principal–agent approach, we argue that the scope for corruption is powerfully conditioned by the ability of voters to control their politicians, which in turn is shaped by the competitiveness of the party system. Party system competitiveness both in terms of fragmentation and governing party dominance, we propose, impacts on the control of politicians because it structures the information and choices available to voters in controlling their politicians. Our empirical results are consistent with these expectations – when party system competitiveness enhances the information and effectiveness of the choices available to the electorate, the scope for corruption is reduced. However, the latitude for corruption increases when high levels of party system fragmentation and patterns of governing party dominance reduce the information available to voters and their ability to coordinate on credible alternatives to the incumbents.

Our analysis and findings give rise to a more accurate and nuanced understanding of party system competitiveness and its relationship to governmental corruption in two respects: First, while much of the extant literature on corruption and party system features focuses either on the number of parties competing or on patterns of governing party dominance, our work makes clear that competitiveness is a two-dimensional phenomenon that is shaped by the number of parties and the relative dominance of the governing party, both of which impact on the scope for corruption. Second, our work clarifies that the relationship between the number of parties that compete and the competitiveness of the system is non-linear – competitiveness initially increases in the effective number of parties, but that effect reverses at higher levels of fragmentation. In both ways, this article helps to resolve the theoretical and empirical contradictions that have characterized much of the empirical work on party system competitiveness and corruption to date.

The findings presented here open up wider research agendas in two respects. Our results clearly make only a first step towards causal inference. Given the large-N character of this study, we are able to demonstrate a constant conjunction and correlation between party system competitiveness and corruption that is consistent with the expectations we derive. These results, though, will need to be probed further through case studies that can examine the causal mechanisms we outline and investigate counterfactuals (Brady, 2008: 218-9.). Moreover, party system competitiveness is of course only one aspect of a wider range of party system features and, given our results in this study, it is not unreasonable to expect that party system institutionalization and programmatic structuration, too, may affect the scope for corruption. These are interesting and important questions which we reserve for future research.

Footnotes

Acknowledgements

We gratefully acknowledge the very helpful comments on earlier versions of this article from Nic Cheeseman, Philip Keefer, Herbert Kitschelt, Mona Lyne, Edward Morgan-Jones, Simon Persico and Nicolas Sauger.

Funding

Dr Schleiter’s research for this project was supported by a British Academy grant (grant reference number SG090658).

Notes

Appendix:

Democracies included in Analysis.

Argentina	Cyprus	India	Mexico	Senegal
Australia	Czech Republic	Indonesia	Mongolia	Slovakia
Austria	Denmark	Ireland	Namibia	Slovenia
Belgium	Dominican Rep.	Israel	Nepal	South Africa
Benin	El Salvador	Italy	Netherlands	South Korea
Bolivia	Estonia	Jamaica	New Zealand	Spain
Botswana	Finland	Japan	Nicaragua	Sweden
Brazil	France	Latvia	Norway	Switzerland
Bulgaria	Germany	Lithuania	Panama	Trinidad and Tobago
Canada	Greece	Macedonia	Paraguay	Turkey
Chile	Guatemala	Madagascar	Phillippines	Ukraine
Colombia	Guyana	Malawi	Poland	United Kingdom
Costa Rica	Honduras	Mali	Portugal	United States
Croatia	Hungary	Mauritius	Romania	Uruguay

References

Adsera

Boix

Payne

(2003) Are you being served? Political accountability and quality of government. Journal of Law, Economics, and Organisation 19(2): 445–490.

Afrobarometer (2006) Citizens and the State in Africa. Working Paper 61. The Afrobarometer Network.

Alesina

Devleeschauwer

Easterly

Kurlat

Wacziarg

(2003) Fractionalization. Journal of Economic Growth 8(2): 155–194.

Anderson

(2000) Economic voting and political context: A comparative perspective. Electoral Studies 19(2-3): 151–170.

Arriola

(2011) Patronage circulation and party system fragmentation in Africa. Working Paper.

Banks

Sundaram

(1993) Adverse selection and moral hazard in a repeated elections model. In: Barnett

Schofield

Hinich

(eds) Political Economy: Institutions, Competition and Representation. Cambridge: Cambridge University Press, pp. 295–312.

Barro

(1973) The control of politicians: An economic model. Public Choice 14(1): 19–42.

Beck

Clarke

Groff

Keefer

Walsh

(2001) New tools in comparative political economy: The database of political institutions. World Bank Economic Review 15(1): 165–176.

Bogaards

(2004) Counting parties and identifying dominant party systems in Africa. European Journal of Political Research 43(2): 173–197.

10.

Brady

(2008) Causation and explanation in social science. In: Box-Steffensmeier

Brady

Collier

(eds) The Oxford Handbook of Political Methodology. Oxford: Oxford University Press, pp. 217–270.

11.

Bratton

van de Walle

(1997) Democratic Experiments in Africa: Regime Transitions in Comparative Perspective. Cambridge: Cambridge University Press.

12.

Chandra

(2006) What is ethnic identity and does it matter? Annual Review of Political Science 9(1): 397–424.

13.

Chang

ECC

Golden

(2006) Electoral systems, district magnitude and corruption. British Journal of Political Science 37(1): 115–137.

14.

Charron

(2011) Party system, electoral systems and constraints on corruption. Electoral Studies 30(4): 595–606.

15.

Coppedge

(1998) The dynamic diversity of Latin American Party Systems. Party Politics 4(4): 547–568.

16.

della Porta

(2004) Political parties and corruption: Ten hypotheses on five vicious circles. Crime, Law and Social Change 42(1): 35–60.

17.

Duverger

(1954) Political Parties: Their Organization and Activity in the Modern State. London: G. Allen and Unwin.

18.

Fearon

(1999) Electoral accountability and the control of politicians: Selecting good types versus sanctioning poor governance. In: Przeworski

Stokes

Manin

(eds) Democracy, Accountability and Representation. Cambridge: Cambridge University Press, pp. 55–97.

19.

Ferejohn

(1986) Incumbent performance and electoral control. Public Choice 50(1): 5–25.

20.

Ferejohn

(1999) Accountability and authority: Toward a theory of political accountability. In: Przeworski

Stokes

Manin

(eds) Democracy, Accountability and Representation. Cambridge: Cambridge University Press, pp. 131–153.

21.

Gallagher

Mitchell

(2008) The Politics of Electoral Systems. Oxford: Oxford University Press.

22.

Geddes

Ribeiro Neto

(1992) Institutional sources of corruption in Brazil. Third World Quarterly 13(4): 641–661.

23.

Gerring

Thacker

(2004) Political institutions and corruption: The role of unitarism and parliamentarism. British Journal of Political Science 34(2): 295–330.

24.

Gerring

Thacker

(2005) Do neoliberal policies deter political corruption? International Organization 59(Winter): 233–254.

25.

Giliomee

(1998) South Africa’s emerging dominant party regime. Journal of Democracy 9(4): 128–142.

26.

Golder

(2005) Democratic electoral systems around the world: 1945–2000. Electoral Studies 24(1): 103–121.

27.

Kaufmann

Kraay

Mastruzzi

(2004) Governance matters IV: Governance indicators 1996–2004. Work Bank Policy Research Working Paper 3630: World Bank.

28.

Key

(1936) Techniques of Political Graft in the United States. Chicago, IL: University of Chicago Press.

29.

Kitschelt

(2007) Party systems. In: Boix

Stokes

(eds) The Oxford Handbook of Comparative Politics. Oxford: Oxford University Press.

30.

Kunicova

Rose-Ackerman

(2005) Electoral rules and constitutional structures as constraints on corruption. British Journal of Political Science 35(4): 573–606.

31.

Kurer

(2001) Why do voters support corrupt politicians? In: Jain

(ed.) The Political Economy of Corruption. London: Routledge, pp. 63–86.

32.

Laakso

Taagapera

(1979) Effective number of parties: A measure with application to West Europe. Comparative Political Studies 12(1): 3–27.

33.

Lewis-Beck

(1988) Economics and Elections: The Major Western Democracies. Ann Arbor, MI: University of Michigan Press.

34.

Marshall

Jaggers

Gurr

(2010) Polity IV Project: Political regime characteristics and transition, 1800–2010. Center for Systemic Peace. Available at: http://www.systemicpeace.org/polity/polity4.htm.

35.

Mershon

(2002) The Costs of Coalition. Stanford, CA: Stanford University Press.

36.

Montinola

Jackman

(2002) Sources of corruption: A cross-country study. British Journal of Political Science 32(1): 147–170.

37.

Myerson

(1993) Effectiveness of electoral systems for reducing government corruption: A game-theoretic analysis. Games and Economic Behavior 5(1): 118–132.

38.

Padro

Miquel

(2007) The control of politicians in divided societies: The politics of fear. Review of Economic Studies 74 (4): 1259–1274.

39.

Persson

Rothstein

Teorell

(2010) The failure of anti-corruption policies: A theoretical mischaracterization of the problem. QoG Working Paper Series 2010:19. Department of Political Science, University of Gothenburg.

40.

Powell

(2000) Elections as Instruments of Democracy: Majoritarian and Proportional Visions. London: Yale University Press.

41.

Regan

Clark

(2010) The Institutions and Elections Project. Available at: http://www2.binghamton.edu/political-science/institutions-and-elections-project.html.

42.

Scheiner

(2005) Pipelines of pork: Japanese politics and a model of local opposition party failure. Comparative Political Studies 38(7): 799–823.

43.

Scheiner

(2006) Democracy Without Competition: Opposition Failure in a One-Party Dominant State. Cambridge: Cambridge University Press.

44.

Svolik

(2008) Authoritarian reversals and democratic consolidation. American Political Science Review 102(2): 153–168.

45.

Tavits

(2007) Clarity of responsibility and corruption. American Journal of Political Science 51(1): 219–229.

46.

Tomz

Wittenberg

King

(2003) CLARIFY: Software for Interpreting and Presenting Statistical Results. Stanford University, University of Wisconsin, and Harvard University.

47.

Transparency International (2009) 2009 Global Corruption Barometer. Full Report available at: http://www.transparency.org/policy_research/surveys_indices/gcb/2009

48.

Treisman

(2000) The causes of corruption: A cross-national study. Journal of Public Economics 76(3): 399–457.

49.

Treisman

(2007) What have we learned about the causes of corruption from ten years of cross-national empirical research? Annual Review of Political Science 10(1): 211–244.