Electoral systems, corruption and satisfaction with democracy Vincenzo Memoli Department of Political and Social Sciences University of Catania (Italy) memoli@unict.it Alessandro Pellegata Department of Social and Political Sciences University of Milan (Italy) alessandro.pellegata@unimi.it Paper presented at the XXVII Annual Conference of the Italian Political Science Association, University of Florence, 12-14 September 2013 Very preliminary draft. Please do not quote without permission 1
Abstract Corruption has been recognized as a detrimental factor of the citizens satisfaction with the way democracy works in their country and their confidence with political institutions. On the contrary, the direct impact of the institutional context on political support is less clear and previous studies present contradictory results. However, from a recent literature emerges that the institutional context plays a role in mediating the impact that the performance of the government authorities have on political support. This paper aims to analyze if and how the restraints posited by the electoral systems on political corruption affect the citizens satisfaction with democracy. The main argument advanced is that those characteristics of the electoral systems that help to constrain corruption among elected officials, strengthening the relation of accountability with the voters and favoring the monitoring by the opposition parties, weaken the negative effects of corruption on the level of political support expressed by citizens. Differently from previous studies that took into account the institutional context along the traditional distinction between majoritarian and proportional/consensual democracies, this paper focuses on specific aspects of the electoral system, such as the district magnitude, the ballot structure, the electoral formula and the level of vote-seats disproportionality. Several research hypotheses on the effect of corruption on political support conditional on the features of the electoral system are tested through a comparative multilevel design on a sample of 34 countries taken from the Module 2 of the CSES data. Results confirm that, even though there is not an electoral system that outperform the others, those features that reduce incentives of politicians to cultivate personal vote and extract rent from their position weaken the negative impact of perceived corruption on satisfaction with democracy. Keywords: comparative politics; corruption; electoral system; institutions; satisfaction with democracy. 2
Introduction In comparative political studies satisfaction with the way democracy works in a specific country is commonly considered an indicator of what Easton (1965; 1975) has defined as specific support. This indicator refers to the citizens satisfaction with the conduct of the political authorities and so incorporates an evaluations of the incumbents performance (Klingemann 1999; Norris 1999; Dalton 2004; Memoli 2011). Among the various aspects that citizens may consider in their performance evaluations we can include the estimation of the level of corruption among public officials. Much has been written on the detrimental effects of corruption on the macro-economic performance (Mauro 1995), the quality of governance (Kaufmann et al. 1999) and the political support (among others, Seligson 2002; Anderson and Tverdova 2003; Clausen et al. 2011). Although most of the researches consider as determinants of political support the political and economic performance, corruption included (Powell and Whitten 1993; Anderson and Tverdova 2003; Bellucci and Memoli 2012), and voters ideology and political preferences (Anderson et al. 2005; Kim 2009; Curini et al. 2012), only few studies take into consideration the potential role played by the institutions (Lijphart 1999; Norris 1999; Aarts and Thomassen 2008). Moreover, the empirical analyses conducted in these studies present contradictory results on the direct effects of institutions on support. This paper focuses on the interaction between political corruption and the institutional framework in affecting political support. More precisely, the ultimate purpose of this study is to assess if and how institutions, in particular the electoral system, mediate the impact of citizens perceptions of corruption on their level of satisfaction with the way democracy works in their country. Elections permit voters to hold their representatives accountable (Powell 2000). However, some electoral systems strengthen the relation of accountability between representatives and their voters, constraining politicians illicit rent-seeking, while 3
other systems make more difficult for voters to recur to retrospective voting. Therefore, our research question is: do electoral systems that contribute to constrain politicians malfeasances reduce the negative impact of corruption on democratic satisfaction? This paper presents two main differences with previous research. First, differently by other scholars which look at the direct effects of institutions on political support, this paper analyzes how the institutional framework, in particular the characteristics of the electoral system, interacts with political corruption. It has been proven that the impact of corruption on democratic satisfaction is lower among winners, i.e. those citizens that had voted for one of the government parties and among people that are ideologically distant from the government (Anderson and Tverdova 2003; Kim 2009; Curini et al. 2012). Nevertheless, we have little knowledge on how the effects of corruption change in different institutional contexts. Second, most of the research on the relationship between institutions and political support have been developed along the traditional distinction between Westminster and consensual democracies and/or majoritarian and proportional electoral systems (Norris 1999; Powell 2000; Anderson and Tverdova 2003; Criado and Herreros 2007). However, looking at these general institutional framework may be misleading. In fact, these are the results of somewhat different combinations of various elements that can present contrasting effects both on corruption constraining and on political support. Thus, in this work we prefer to advance several research hypotheses that look at the mediating effects on the impact of corruption on democratic satisfaction played by different aspects of the electoral system. These are the district magnitude, the ballot structure, the presence of open or closed party lists, the electoral formula and the level of vote-seat disproportionality. The conditional hypotheses advanced are tested making use of the Module 2 of the Comparative Study of Electoral System dataset (CSES) which includes data on elections and citizens attitudes toward governments between 2001 and 2006. The sample of analysis is 4
composed by thirty-four countries that were democracies at the time in which the elections were taken. The validity of our predictions is controlled for alternative explanatory factory of democratic satisfaction both at the individual and at the country level and the statistical technique used is a multilevel ordinal logistic regression. Empirical results confirm only some of the hypotheses advanced. These indicate that electoral systems are complex aggregations of various aspects that play different effects on corruption control and, as a consequence, on satisfaction with democracy. We cannot identify a specific electoral system that outperforms the others in weakening the negative impact of corruption. In the next section we elaborate the main theoretical framework and advance the research hypotheses, while in the following one we describe the sample of countries analyzed and the data and the methodology used. Then, we discuss the empirical results obtained from our analyses. Finally, in the last section we present the most important conclusions and the implications we can draw from our research. The role of the electoral system in mediating the impact of corruption Corruption is expected to negatively affect the citizens evaluations of the way democracy works in their country. When public agents are committed to extract illicit rents from their privileged positions, human, social, and economic resources are diverted from the public interest undermining the principles of democratic accountability and fairness and impartiality of political institutions (della Porta 2000). This implies that the public is compelled to pay the externalities of this misallocation process consisting in a distortion of public demand, an increase of the cost and a reduction of the quality of public services. In other words, corruption is unanimously recognized as a plague of good democracy (Morlino 2004). Thus, the more the corruption is widespread, the more citizens tend to negatively evaluate the performance of government authorities. 5
However, institutions, understood as the rules of the game, can act as a restraint on corruption interactions between politicians and business groups. Analyzing the relationship between institutions and corruption a broad recent literature have focused especially on the electoral system (Persson et al. 2003; Kunicovà and Rose-Ackerman 2005; Chang and Golden 2006). Elections allow citizens to hold their representatives accountable and to throw the rascals out (Powell 2000). The effectiveness of elections as an accountability mechanism depends primarily on the electoral system adopted. In fact, electoral systems affect the incentives and the ability of voters and opposition parties to monitor incumbents behavior. While voters prefer honest politicians, incumbents trade-off their re-election chances against potentially illicit personal enrichment. These illegal forms of rent-seeking can help incumbents to obtain the re-election, for instance increasing the amount of campaign financing. Nevertheless, at the same time, the revelation of their corrupt scandals by voters or opponents most likely reduce their re-election chances (Kunicovà and Rose-Ackerman 2005). According to the literature (Persson et al. 2003; Kunicovà and Rose-Ackerman 2005), majoritarian systems are more efficient in constraining corruption because generate a direct link between voters and their representatives. Moreover, attributing a clear majority to a single-party government or a minimal winning coalition, they favor a voters retrospective evaluations of incumbents performance. On the contrary, electoral systems that impose voters to vote on party lists reduce the monitoring abilities of voters and provide incentives to the incumbents to recur to corrupt interactions. Furthermore, the formation of large and volatile coalitions, typical in democracies that adopt PR electoral systems, makes more obscure for voters to identify the responsibilities for the policy adopted. The lack of a clear alternation at the government between fixed groups of parties deters inter-party monitoring. Present opposition parties have less incentives to unveil incumbents malfeasance because they may seat at the government with some of them in future coalitions (Pellegata 2012). 6
The main argument advanced in this paper is simple but straightforward. Electoral systems have an indirect impact on citizens support of political authorities through their mediating role on political corruption. The more an electoral system helps to constrain politicians illicit behaviors, the more it weakens the negative role that corruption has on democratic satisfaction. The degree of citizens satisfaction with the way democracy works in their country is intrinsically linked to the evaluation they make of the performance of government authorities. If they perceive a high level of corruption among politicians they most likely express a low level of political support. Nevertheless, in presence of an electoral system that allows citizens to monitor and, eventually, punish politicians malfeasance the impact of perceived corruption on political support should be less negative. Although corruption is present, voters who are able to punish corrupt politicians feel confident about the system. However, as several scholars argued (Lijphart 1994), electoral systems are composed by specific elements that regulate different aspects of the voting procedure and the translation of votes into seats. These elements can play a different, and sometimes contrasting, role both in constraining corruption and in shaping democratic satisfaction and institutional confidence. Therefore, differently from most of previous studies (Anderson and Tverdova 2003; Kunicovà and Rose-Ackerman 2005; Criado and Herreros 2007; Kim 2009) that focus on the main distinction between majoritarian and proportional systems(visions), this paper analyzes the main features that compose the electoral systems. The starting point of our theoretical framework is represented by the Persson et al. s (2003) research on the effects of the electoral rules on corruption. Relying on the seminal studies of Myerson (1993) and Carey and Shugart (1995), the authors identify three main features of the electoral systems that affect politicians incentives to cultivate personal vote and citizens ability to monitor their 7
behavior. These features are: (1) the district magnitude; (2) the ballot structure and (3) the electoral formula. The district magnitude lies at the basis of what Persson et al. (2003) call barriers-toentry effect. This argument posited that electoral rules promoting the entry of representatives of many parties reduce politicians rent seeking. All voters prefer to vote for honest candidates but disagree on their ideological preferences. If voters cannot find good alternatives to a dishonest candidate who shares their same ideological preference they are compelled to vote for her. Small electoral districts, together with majoritarian formulas that favor strategic voting, reduce the number of political forces represented increasing the barriers to entry in the electoral system and making more difficult to oust dishonest incumbents from office (Myerson 1993; Persson et al. 2003). On the contrary, with large districts, and PR electoral formulas, the availability of honest candidates for diverse ideological positions is more presumable. This reduces the barriers to entry and, consequently, the chances of reelection for dishonest candidates. The differentiation of political forces on the ideological spectrum is one of the main arguments posited by those who claim that proportional/consensual democracies are associated with higher level of democratic satisfaction (Lijphart 1999; Powell 2000). 1 Therefore, our first research hypothesis postulates that: H 1 : With increasing values of district magnitude the negative effects of the perceived corruption on the satisfaction with democracy weaken (barriers-to-entry effect). 1 However, more recent studies show that higher level of polarization and of differentiation of parties policy offerings increasing the gap around the position of the median voter reduce the level of democratic satisfaction (Kim 2009; Ezrow and Xezonakis 2011). 8
Career-concern effect rests on another important characteristic of the electoral system, i.e. the ballot structure. The main argument is that voting over individual candidates strengthens the relation of accountability that links the single representative to their voters. As Persson et al. (2003) argue, this creates a direct link between the performance of the single incumbent and his/her reappointment. To maximize the chance of being re-elected politicians are incentivized to maintain a honest behavior avoiding abuse of power. On the contrary when voters choose among party lists the relation of accountability is weakened because the number of seats won depends on the votes collected by the whole list, rather than the performance of each individual candidate. As a consequence, candidate are much more incentivized to behave as free-riders. The second research hypothesis claims that: H 2a : With an increasing share of representatives elected on an individual ballot, rather than on party lists, the negative effects of perceived corruption on the satisfaction with democracy weaken (career-concern effect). According to Kunicovà and Rose-Ackerman (2005) the weakest relation of accountability is generated by the presence of closed party lists. Whit open lists voters can express their preference to one or more candidates of the list. Thus, in order to be re-elected candidates aims to maximize the preferences for their person and this makes more costly behave as a free-rider. In presence of closed lists instead the individual s chance of re-election depends on his/her rank on the list and not on his/her performance (Personn et al. 2003). This is why we advance a variant of the second hypothesis that takes into account the potential corruption constraining role of PR-open list: 9
H 2b : With an increasing share of representatives elected on an individual ballot or PR-open lists, rather than on PR-closed lists, the negative effects of perceived corruption on the satisfaction with democracy weaken (open-lists effect). Following the same reasoning we expect that, among countries that adopt a PR electoral system, those that present open lists are more efficient in constraining the negative effects of corruption on satisfaction with democracy, than those with closed lists: H 3a : In presence of a PR electoral system with open lists the negative effects of perceived corruption on the satisfaction with democracy weaken (open-lists effect). However, Chang and Golden (2006), relying on the theoretical argument advanced by the literature on personal vote (see Carey and Shugart 1995), provide empirical evidence that the impact of PR electoral systems on corruption is different whether these are implemented through open lists or closed lists. The idea is that intra-party competition is stiffer in presence of open lists than with closed lists. The necessity to maximize personal vote provides incentives to single candidates to amass resources, even illicitly, to out-campaign their opponents during elections (Chang and Golden 2006). In an open list environment the importance of establishing a unique personal reputation, distinct from that of the party, grows as the number of candidates, namely the district magnitude, increases (Carey and Shugart 1995). In presence of closed list, by contrast, the relationship is the opposite. Single candidates are not incentivized to recur to illegal funding because the reputation of the whole party overshadows their personal one. Therefore, it is presumable that the role played by this different implementation of a PR electoral system in mediating the impact of corruption on 10
democratic satisfaction is affected by the district magnitude. An alternative third hypothesis is: H 3b : The mediating impact of the presence of open or closed lists depends on the average district magnitude. With low levels of district magnitude the presence of a PR electoral system with open lists has a better performance than PR-closed lists in weakening the negative effects of perceived corruption on democratic satisfaction. The relation is the opposite with high levels of district magnitude (intra-party-competition effect). As well as intra-party competition, also inter-party competition plays an important indirect role on satisfaction with democracy via corruption control. Persson et al. (2003) argue that real-world electoral systems can be generally subdivided into two main groups according to the electoral formula they adopt: majoritarian systems with majority or plurality rules and single-member districts and proportional systems with PR electoral formulas and multimember districts. The relation of accountability is stronger in the first type of systems because candidates of different parties are focused on winning a majority, not of the whole electorate, but of the voters of their districts. As these voters are more willing to change their preference in response to policy, single candidates are more prone to maintain an honest and disciplined behavior (Persson et al. 2003). The tougher inter-party competition typical of majoritarian systems should induce to a better corruption control. The perception of voters in these systems to have a direct control over the behavior of their representatives should positively influence the level of democratic satisfaction. The fourth research hypothesis is the following: H 4 : In presence of a majoritarian electoral formula the negative effects of perceived corruption on the satisfaction with democracy weaken (inter-party-competition effect). 11
Finally, there is another aspect of the electoral systems that is worth taking into consideration in our analysis, that is the degree of disproportionality in translating votes into seats (see Gallagher 1991). Two main arguments justify a positive impact of disproportionality level in weakening the effects of corruption on democratic satisfaction. First, a more disproportional translation of votes into seats tends to generate a bilateral competition and durable single-party governments or minimum-winning coalitions. A political-institutional framework of that kind is characterized by higher clarity of responsibility, and this increases the capacity of voters to monitor and evaluate the performance of the decision makers, and, eventually, punish those that are perceived as corrupt. Second, with high disproportional electoral systems even small shifts in the choices expressed at the electoral level generate substantial shifts at the legislative level. Thus, high level of disproportionality increases the possibility of alternation making governments more accountable to voters (Kaiser et al. 2002). Especially in those political systems in which voters perceive a widespread corruption, highly disproportional electoral systems should weaken the negative effect of corruption on democratic satisfaction. Our fifth and last hypothesis is: H 5 : With increasing values of vote-seats disproportionality the negative effects of perceived corruption on the satisfaction with democracy weaken (possibility-of-alternation effect). Each of these research hypotheses stresses a different mechanism according to which one or more particular aspects of an electoral system should indirectly impact the satisfaction with democracy attenuating the detrimental effects of corruption on it. Different aspects, such as district magnitude and the electoral formula, are combined in a systematic pattern in real- 12
world electoral systems (Persson et al. 2003). Majoritarian electoral systems, for instance usually combine single-member districts, individual ballot and plurality or majority rule, while proportional electoral systems combine multi-member districts, party lists and PR electoral systems. However, the theoretical expectations that link district magnitude and majoritarian formula (as well as the vote-seat disproportionality that is usually higher in majoritarian system) to corruption control and, consequently, to democratic satisfaction are divergent. As a consequence, we do not expect that an ideal type of electoral system outperform the other in increasing political support, via corruption control. Data and methods Sample To test the hypotheses advanced we make use of the final release of the CSES Module 2 dataset. This includes data on 40 elections taken in 38 countries between 2001 and 2006. 2 For the present study we have excluded one of the two studies conducted on the German Bundestag in 2002 and the study conducted on the Portuguese elections of 2002, while we have taken into account only the most recent (2005) of the two studies on Portugal. 3 Moreover, we have excluded from the analysis the non-democratic regimes (Albania, 2005; Hong Kong, 2004; Kirghizstan, 2005; Taiwan, 2001) taking into consideration only those countries that in the year analyzed have reached the status of Free according the Freedom in the World Index released by Freedom House. 4 Thus, our analyses employ data on 34 elections in 34 countries. 2 Data and documentation can be downloaded from http://www.cses.org/. 3 For Germany 2002, we have excluded the survey realized by phone and used the survey self-administered via mail. Taking into consideration both these surveys could over-estimate Germany in respect to the other countries. 4 See www.freedomhouse.org. 13
Dependent Variable Our dependent variable is the respondents level of satisfaction with democracy (SWD). This variable is gathered from the responses citizens give to the following question: On the whole, are you very satisfied (1), fairly satisfied (2), not very satisfied (3) or not at all satisfied (4) with the way democracy works in [country]? We reversed the original scale so that higher values correspond to greater satisfaction with democracy. Figure 1 plots the percentage of people in each sample country that are fairly or very satisfied with democracy. As we can see, the figure displays a high cross-national variation across the sample. Satisfied people range from less than 30 per cent in Bulgaria up to 95 per cent in Denmark. Figure 1 about here Independent Variables Our main independent variable is the level of corruption. The variable originates from the citizens perceptions obtained from the responses they gave to the following question: How widespread do you think corruption such as bribe taking is amongst politicians in [country]?: very widespread (1), quite widespread (2), not very widespread (3), it hardly happens at all (4). As we did for SWD, we reversed the original coding so that higher values represents perceptions that corruption is more widespread. To test the conditional hypotheses advanced in the previous section we make corruption interact with different features of the electoral system. The average district magnitude is computed as the overall number of seats allocated through elections over the number of districts present in the country. To test H 2a and H 2b we calculate the proportion of legislators elected on individual ballot rather than party list (H 2a ) and the proportion of 14
legislators elected on individual ballot and PR-open list rather than on PR-closed list (H 2b ). both these indicator range between 0 and 1. While the first assume value of 1 only in presence of the traditional first-pass-the-post system, the second assume value of 1 also in presence of PR electoral systems with open lists. 5 To operationalize the presence of a majoritarian formula we include a dummy variable that assumes value of 1 for those countries that adopt a pure majority or plurality electoral system and 0 otherwise. The level of vote-seats disproportionality is measured through the Gallagher s (1991) least squares index. 6 A dummy variable distinguishes, among countries that adopt a PR electoral systems, those that present open lists (1) from those that present closed lists (0). All the variables that refer to the electoral system are taken from the World Bank s Database of Political Institutions (DPI) or, if not available, are the result of personal calculations using data gathered from the Parline Database of the Inter-Parliamentary Union (IPU). 7 Table 1 lists the values presented by all the sample countries for each of the characteristics of the electoral system described. Table 1 about here Control Variables The validity of our research hypotheses are controlled for a set of alternative explanatory variables of the satisfaction with democracy both at the individual level and at the country level. At the individual level we include in the model three socio-demographic factors. Age measures the age of respondents in years excluding people that are less that sixteen years old, 5 For a more complete description of these two variables see Persson et al. (2003). 6 For semi-presidential systems, we estimate Gallagher s index following the method proposed by Lijphart (1999), i.e., by taking the geometric mean of the index computed for presidential and for legislative elections. 7 For information on DPI see Beck et al. (2001). For the IPU Parline database see http://www.ipu.org/parlinee/parlinesearch.asp. 15
while education is a categorical variable that measures their level of education. Gender is a dummy variable that assumes value of 1 for females and 0 for males. Moreover, we include two categorical variables that relate to the respondents political attitudes, namely, how much people think that democracy is better than any other forms of government, and that politicians in power can make a difference. Higher values represent a stronger feeling toward democracy and toward the political efficacy of those who are in power. In line with a consolidated literature, we include a dummy variable that distinguishes between winners and losers (among others, Anderson and Guillory 1997; Anderson and Tverdova 2003; Curini et al. 2012). Citizens who have voted for a government party tend to present higher levels of democratic satisfaction. This variable assumes value of 1 for those respondents that in the last elections voted for one of the parties at the government and 0 otherwise. Another variable measures the distance, in absolute values, between the left-right self-reported position of the respondent and the position of the government. The latter is measured as the median point of the range of the government coalition, obtained by the absolute distance between the left-right position of the two extreme parties. Finally, other two dummy variables measure whether respondents feel close to a particular political party and whether they usually join protests or demonstrations. People which identify themselves or feel close to a party should express greater satisfaction, while people that participate in political demonstrations tend to be less satisfied. At the country level we include a dummy variable that distinguishes between consolidated and newly established regimes. We expect lower levels of democratic satisfaction in new democracies because they tend to present a poorer political and economic performance. We consider a democratic regime as consolidated if did not experience any interruption in their status in the last twenty-five years according to the Przeworski s et al. (2000) classification. Among new democracies there are countries from East Europe, Central 16
and Latin America and East Asia (apart from Japan). We want to take into account also the quality of formal institutions because it is supposed to be positively correlated with democratic satisfaction. For this reason among the control variables we include an indicator that measures the enforcement of the rule of law in each country taken from the Worldwide Governance Indicators (Kaufmann et al. 1999). 8 Finally we take into account the economic performance of different countries. The better the economic performance of a political system, the higher the satisfaction with democracy. We include three different variables: the natural logarithm of the GDP per capita and the inflation rate, measured in the year before the elections, and the average growth in GDP in the five years before the elections. Data are gathered from the World Development Indicators collected by the World Bank. 9 Method Considering that our dependent variable, satisfaction with democracy, is an ordinal measure neither an OLS nor a multinomial logit or probit model are appropriate methods of analysis. We apply an ordinal regression model that has been developed for analyzing ordered response variable, as in our case (Long 1997). As we have seen our dataset combines information at the individual level of respondents nested with information at the aggregate level of the thirtyfour countries analyzed. To take into account the hierarchical structure of the data we compute heteroskedasticity consistent standard errors clustered at the country level (Rabe- Hesket and Skrondal 2012). 10 In any ordinal regression model the dependent variable is 8 Data and documentation can be downloaded from http://info.worldbank.org/governance/wgi/index.asp. 9 See http://data.worldbank.org/data-catalog/world-development-indicators. 10 As a robustness check we also estimate a model by dichotomizing the dependent variables and running a multilevel logistic regression model. The modalities not at all satisfied and not very satisfied are aggregated into the category not satisfied (0), while the modalities fairly satisfied and very satisfied are aggregated 17
regarded as a latent variable that is estimated as a linear function of the independent variables and a set of thresholds defined as cutpoints. The probability of observing a given outcome therefore corresponds to the probability that the estimated linear function is within the range of the cutpoints estimated for the outcome. Empirical results Table 2 reports the results of seven statistical models that test the validity of the research hypotheses advanced. Every model includes a two-way interaction (three-way in Model 6) between the citizens perception of corruption and a characteristic of the electoral system, and all the control variables, both at the individual and at the country level, described above. Given that multiplicative interaction models aim to test conditional hypotheses, regression coefficients cannot be interpreted as in additive linear models (see Brambor et al. 2006). Thus, results reported in tables 2 are in some sense uninformative for the control of the validity of our conditional hypotheses. This is why we compute the marginal effects of corruption on democratic satisfaction (and the relative confidence intervals) as the different characteristics of the electoral system change and we have summarized the results in Table 3 and 4 and in Figures 2 through 6. Table 2 around here Figure 2 plots four graphs with the marginal effects of corruption on the four different modalities that the dependent variable, satisfaction with democracy, can assume as the average district magnitude increases. As we can note, H 1 is confirmed by the results obtained. into the category satisfied (1). Results obtained are very similar to the ones presented below and are available upon request. 18
With increasing values of district magnitude the positive effect of corruption on being not at all or not very satisfied with democracy does not change or slightly decrease, while the negative effect of corruption on being very satisfied with democracy weaken. In all but one graph the relationship is significant for all the possible values of district magnitude. These results are obtained excluding Israel which presents a unique nationwide district (average district magnitude=120) that makes it an outlier in the distribution. However, we have run Model 1 and computed marginal effects also including Israel and the results remain the same. 11 Figure 2 about here Figure 3 reports four graphs with the marginal effects of corruption on the four outcomes of the dependent variable as the proportion of legislators elected on individual ballot increases. The career-concern effect advanced in H 2a is confirmed by these results. With an increasing share of representatives elected on individual ballot the positive impact of corruption on the probability of being not at all or not very satisfied decrease, while the negative effect on the probability of being fairly and, above all, very satisfied weaken. However, it is worth noting that the relationship is significant only until a proportion of representative elected on individual ballot around.5-.6. This means that pure plurality or majority systems in which the whole set of representatives is elected on individual ballot does not contribute in weakening the detrimental effect of corruption on democratic satisfaction. It rather seems that this positive mediating role is played by mixed electoral systems in which around a half of legislators are elected in single-member districts. 11 Results are available upon request. 19
Figure 3 around here On the contrary, H 2b is not confirmed by the results reported in Figure 4 in which the mediating variable is the proportion of legislators elected on individual ballot or PR-systems with open lists. With increasing values of this variable the positive effect of perceived corruption on the probability that respondents are not at all or not very satisfied increases or remains the same, while its negative impact on the probability that they are fairly or very satisfied does not change or becomes stronger. As a consequence, PR systems with open lists not only have a worse performance in reducing the negative effect of corruption on democratic satisfaction than majoritarian systems, but also seem to exacerbate it. Figure 4 around here Even though we focus only on countries which adopt PR electoral systems, as reported in Model 5 in Table 2, the presence of open lists contributes to worsen the detrimental effect of corruption on satisfaction with democracy. This result is evident from the marginal effects of corruption on the dependent variable in presence of closed or open list reported in Table 3. For three out of four outcomes of the dependent variable with both the variants of party lists corruption displays a significant positive correlation with the probability of being not at all or not very satisfied and a negative correlation with the probability of being fairly or very satisfied. However, switching from a PR electoral system with closed lists to a system with open lists the relation between corruption and democratic satisfaction becomes stronger. Therefore, H 3a is not confirmed by the empirical results. Table 3 around here 20
Figure 5 reports the results of the test of H 3b which advances the intra-partycompetition effect developed by Carey and Shugart (1995) and is alternative to H 3a. The figure plots four graphs. The two graphs on the left side of the figure show the marginal effects of corruption in presence of a PR electoral system with open lists on being not at all satisfied with democracy (at the top) and being very satisfied (at the bottom) as the average district magnitude changes. The two graphs on the right side display the same marginal effects with the only difference of the presence of closed lists. What emerges from the analysis of Figure 5 is that in presence of PR electoral systems with closed list, rather than open list for each value of district magnitude the size of the impact of corruption on satisfaction with democracy is generally lower or, at least, not higher. However, with increasing values of district magnitude the presence of closed list weakens the negative effect of corruption on democratic satisfaction. These findings do not confirm H 3b which lies on the Carey and Shugart s (1995) prediction according to which closed lists have a better performance in reducing intra-party competition and illegal rent seeking only with high levels of district magnitude. In fact, closed lists have a stronger impact on satisfaction with democracy, via its corruption-constraining role, with every potential value of district magnitude. Figure 5 around here The role of majoritarian formula in mediating the effect of corruption on satisfaction with democracy is not clear. Table 4 reports the marginal effects of corruption on the dependent variable in presence or not of a majoritarian formula. As we can note, in line with our prediction, adopting a majoritarian formula the effect of corruption on the probability of 21
being not at all satisfied is less positive as well as the effect on being fairly satisfied is less negative. Nevertheless, switching from a non-majoritarian formula to a majoritarian one we have a worse performance in weakening the positive effect of corruption on being not very satisfied and its negative effect on being not at all satisfied. As a consequence we cannot confirm the validity of H 4. Table 4 around here Finally, Figure 6 plots four graphs reporting the marginal effects of corruption perception on democratic satisfaction at different values of the level of disproportionality. As predicted by what we call the possibility-of-alternation effect advanced in H 5, with increasing values of disproportionality the positive impact of corruption on being not at all or not very satisfied decreases, while its negative impact on being fairly or very satisfied weakens. The relationship is significant only until a disproportionality level more or less equal to 10. However, this does not invalidate our result because only 4 out of 34 countries included in the sample (11%) display a level of disproportionality higher than 10. Figure 6 around here Looking at the Table 2 presented above we can note that some of the control variables tested are significantly correlated with the dependent variable. At the individual level the two variables that reflect the citizens political attitude positively influences their level of satisfaction with democracy. People who think that democracy is better than any other forms of government, and that politicians in power can make a difference tend to be more satisfied. Also winners, i.e. citizens who had voted for one of the government parties, express higher 22
level of satisfaction in all but one of the model tested. On the contrary, the regression coefficient of the distance between the ideological preferences of the citizens and the government, though presents always a negative sign, is never significantly related to satisfaction with democracy. Finally, people who feel close to a political party tend to express higher levels of satisfaction, while people who usually join protests or demonstrations are more unsatisfied. Socio-demographic variable, instead, are never significantly related to democratic satisfaction. Among the alternative country-level explanatory factors, newly established democracies only in some models are associated to lower levels of satisfaction with democracy. The quality of formal institutions, instead, positively impacts democratic satisfaction. In all but one of the models tested the rule of law variable is positively and significantly associated with the dependent variable. Among the three variables chosen to operazionalize the economic development of different sample countries, the natural logarithm of GDP per capita is never significantly correlated with satisfaction. On the contrary, the average growth of real GDP in the five years before the elections is significantly associated to higher levels of satisfaction with democracy. In conclusion, inflation rate is inversely correlated with the dependent variable. Concluding remarks The present paper investigates the role of the electoral system in mediating the negative impact of corruption on citizens satisfaction with the way democracy works in their country. While, it has been proven that corruption has detrimental consequences on democratic satisfaction and institutional confidence, we have little knowledge about the direct effects of the political and institutional context on political support. However, a broad literature explains how electoral systems can help to constrain corruption, understood as an 23
indicator of (bad) performance of the political authorities. In this study we hypothesize that those characteristics of the electoral system that contribute to constrain corruption and strengthen the relation of accountability between representatives and their voters weaken the negative effects of corruption on satisfaction with democracy. More precisely, we advance several research hypotheses which test the effects of corruption on democratic satisfaction conditional to different aspects of the electoral system. These characteristics are the district magnitude, the ballot structure, the electoral formula, the presence of open or closed party lists and the level of vote-seat disproportionality. The validity of the research hypotheses have been tested through the CSES Module 2 dataset which includes data on 34 elections in 34 consolidated and newly-established democracies employing a multilevel ordinal logistic regression. Empirical results confirm only some of the hypotheses advanced. The negative effects of corruption on democratic satisfaction are weakened by increasing levels of district magnitude and vote-seat disproportionality, and a moderately high share of legislators elected on individual ballot rather than with party lists. The implementation of a PR electoral systems with open lists, instead, does not help to reduce the negative consequence of corruption. Irrespective of the average level of district magnitude present in a country, the presence of closed lists has a better performance than open lists. Finally, the adoption of a majoritarian formula seems to play any significant role in mediating the impact of corruption on satisfaction with democracy. From these findings we can draw two main general conclusions. Firstly, we find evidence that citizens levels of satisfaction with democracy are strictly connected to an evaluation of the performance of the political authorities, represented here by their perception of the spread of corruption among public officials. However, not all those electoral systems mechanisms that constrain corruption automatically help to weaken its impact on satisfaction 24
with democracy. Of course corruption is not the only aspect that citizens take into account in their evaluations of the governments performance. There are several other elements of political performance, such as the state of the national education system and the welfare state, that play an important role in affecting their degree of democratic satisfaction and presumably counterbalance the impact of corruption (Criado and Herreros 2007). Secondly, empirical results confirm the complexity of electoral systems in affecting the perceptions that citizens have of the general government performance and the quality of the democratic system. What emerges is that there is not a specific electoral system that outperforms the others in weakening the negative impact of corruption on the citizens satisfaction with democracy. The adoption of a pure plurality or majority electoral rule does not guarantee a better performance on democratic satisfaction via its corruption-constraining role than the adoption of a proportional rule. Our analysis rather confirms that single aspects, that are generally combined in a specific electoral system, may present contrasting effects and should be analyzed separately. In fact, it seems that to weaken the impact of corruption on democratic satisfaction a country should adopt a combination of single-member districts in which the representative is elected on individual ballot and multi-member districts regulated by a PR electoral system with closed lists. Moreover, this kind of system should imply a moderately high level of vote-seat disproportionality. Looking at the real-world electoral systems, this combination, though in different variants, is implemented by mixed-member electoral systems. Eight out of thirty-four countries in our sample have a mixed electoral system. Most of them have reformed (or emended) the electoral system adopting this formula during the nineties. However, it is difficult to derive from our results that adopting a mixed system would contribute to improve the quality of contemporary representative democracies (Massicotte and Blais 1999; Shugart and Wattemberg 2001). Although, mixed systems help to 25
reduce the negative effects of corruption on satisfaction with democracy, further research that analyzes their impact on different elements of democratic quality are required. References Anderson, Christopher J., and Christine A. Guillory (1997) Political Institutions and Satisfaction with Democracy: A Cross-National Analysis of Consensus and Majoritarian Systems, American Political Science Review, 91(1): 66-81. Anderson, Christopher J., and Yuliya V. Tverdova (2003) Corruption, Political Allegiances and Attitudes toward Government in Contemporary Democracies, American Journal of Political Science, 47(1): 91-109. Aarts, Kees and Jacques Thomassen (2008) Satisfaction with Democracy: Do institutions matter?, Electoral Studies, 27(1): 5-18. Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer and Patrick Walsh (2001) New Tools in Comparative Political Economy: The Database of Political Institutions, World Bank Economic Review, 15(1): 165-76. Bellucci, Paolo, and Vincenzo Memoli (2012) The determinans of support in Europe, in David Sanders, Pedro C. Magalhaes and Gabor Toka (eds.), Citizens and European Polity. Mass Attitudes toward European and National Polities. Oxford: Oxford University Press. Brambor, Thomas, William R. Clark and Matt Golder (2006) Understanding Interaction Models: Improving Empirical Analyses, Political Analysis, 14(1): 63-82. Carey, John M. and Matthew S. Shugart (1995) Incentives to Cultivate a Personal Vote: A Rank Ordering of Electoral Formulas, Electoral Studies, 14(4): 417-39. Chang Eric C. and Miriam A. Golden (2006) Electoral Systems, District Magnitude and Corruption, British Journal of Political Science, 37(1): 115-37. 26
Clausen, Bianca, Aart Kraay, Zsolt Nyiri (2011) Corruption and Confidence in Public Institutions: Evidence from a Global Survey, The World Bank Economic Review, 25(2): 212-49. Criado, Henar, and Francisco Herreros (2007) Political Support. Taking into Account the Institutional Context, Comparative Political Studies, 40(12), 1511-32. Curini, Luigi, Willy Jou and Vincenzo Memoli (2012) Satisfaction with Democracy and the Winner/Loser Debate: The Role of Policy Preferences and Past Experience, British Journal of Political Science 42(2): 241-61. Dalton R.J. (2004) Democratic Challenges, Democratic Choices: The Erosion of Political Support in Advanced Industrial Democracies, Oxford: Oxford University Press. della Porta, Donatella (2000) Social Capital, Beliefs in Government, and Political Corruption, in Susan Pharr and Robert Putnam (eds.), Disaffected Democracies: What s Troubling the Trilateral Countries. Princeton, NJ: Princeton University Press, 208-28. Easton, David (1965) A Systems Analysis of Political Life. Chicago: University of Chicago Press. Easton, David (1975) A Re-assessment of the Concept of Political Support, British Journal of Political Science, 5(4): 435-57. Ezrow, Lawrence and Georgios Xezonakis (2011) Citizen Satisfaction with Democracy and Parties Policy Offerings, Comparative Political Studies, 44(9): 1152-78. Gallagher, Michael (1991) Proportionality, Disproportionality and Electoral Systems, Electoral Studies, 10(1): 33-51. Kaiser, Andreas, Matthias Lenhert, Bernhard Miller and Ulrich Sieberer (2002) The Democratic Quality of Institutional Regimes: A Conceptual Framework, Political Studies, 50(2): 313-31. 27