The Lesser Evil? Corruption Voting and the Importance of Clean Alternatives

Abstract

Surveys show that citizens in all parts of the world have a strong distaste for corruption. At the same time, and contrary to the predictions of democratic theory, politicians involved in the most glaring abuse of public office often continue to receive electoral support. Using an original survey experiment conducted in Spain, this article explores a previously understudied aspect of this apparent paradox: the importance of viable and clean political alternatives. The results suggest that voters do punish political corruption when a clean alternative exists, even when the corrupt candidate is very appealing in other respects. However, when only given corrupt alternatives, respondents become much more likely to tolerate a candidate accused of corruption—even when given a convenient “no-choice” option. I discuss how these results can help us understand corruption voting and why some societies seem to be stuck in a high-corruption equilibrium.

Keywords

corruption voting electoral accountability survey experiment

Introduction

Citizens in democracies around the world express considerable concern about corruption (Eurobarometer, 2014; Global Corruption Barometer, 2017). And rightly so, research shows that bribes and kickbacks are common, even in many developed democracies (Charron, Lapuente, & Rothstein, 2013; Warren, 2004). What is more, there is now a near consensus among researchers about the destructive consequences of corruption¹ and the societal benefits of clean government (Bardhan, 1997; Holmberg & Rothstein, 2011; Mauro, 1995; Mungiu-Pippidi, 2013; Rose-Ackerman, 1999; Rothstein, 2011). Yet, evidence shows that politicians involved with corruption do not always suffer electorally (Bågenholm, 2013; Chang, Golden, & Hill, 2010; Fernandez-Vazquez, Barbera, & Rivero, 2016; Peters & Welch, 1980). From the perspective of democratic theory, this looks like a paradox: Why would voters not “throw the rascals out” whenever they got the chance?

The growing literature on corruption voting² has provided different answers to this intriguing question. For instance, voters might simply be biased and not punish politicians from a party they identify with (Anduiza, Gallego, & Munoz, 2013). Or voters might even accept corruption if they get policies they like or competence in other domains from a corrupt political candidate (Konstantinidis & Xezonakis, 2013; Munoz, Anduiza, & Gallego, 2016; Rundquist, Strom, & Peters, 1977; Zechmeister & Zizumbo-Colunga, 2013). At the same time, other studies show that voters are highly likely to punish politicians when given credible information about their corrupt practices (Ferraz & Finan, 2008; Weitz-Shapiro & Winters, 2017; Winters & Weitz-Shapiro, 2013). In short, there exists no consensus among researchers regarding the electoral consequences of corruption (De Vries & Solaz, 2017).

This article identifies an understudied aspect of corruption voting: the role of clean political alternatives. Few, if any, argue that voters have a general preference for corrupt politicians. Rather, studies show that citizens, even in societies where corruption is endemic, express a very strong distaste for corrupt practices (Persson, Rothstein, & Teorell, 2013). Therefore, a crucial question to ask for the voter deciding whether to vote for a political candidate accused of corruption is: What are the alternatives? Is there a clean alternative where the voter can signal his or her disapproval of corrupt behavior?

I argue that the availability of clean political alternatives is a critical factor for understanding corruption voting. In a democracy, the decision to vote for a candidate is a decision to choose the candidate over other available options. Importantly, this is also the choice of voting rather than abstaining. Previous research designs studying corruption voting have not sufficiently considered what alternatives voters actually have when their favorite candidate faces a corruption allegation, and in particular, whether it is possible for the voter to switch to a clean option. By employing a survey-experimental design, including a “no choice” alternative, this study appropriately models the three options that are available to the voter: exit, voice, and loyalty (see Bauhr & Charron, 2018).

This article makes several contributions to the literature by exploring the previously understudied question of corruption voting and clean political alternatives (see De Vries & Solaz, 2017). The research design allows me to focus on how the availability of a clean political alternative, or the lack thereof, plays into voters’ punishment of corruption. The experiment, with over 2,000 Spanish respondents choosing between two candidates for city mayor, shows that voters express a strong distaste for corruption and are very likely to punish corrupt politicians (voice) when given a clean alternative—even if the corrupt candidate is very appealing in other regards. This goes in part against studies claiming that voters are willing to accept corruption if they get things like representation or policy in exchange (e.g., Munoz et al., 2016; Rundquist et al., 1977). At the same time, the results show that although voters in general are prone to punish corruption, they tend to vote for the “least bad alternative” to a relatively high degree (loyalty) when all candidates are corrupt—even when given a convenient exit option in the form of a “no choice” (not vote) alternative. This suggests that voters, despite having very strong anticorruption preferences, can become more willing to vote for corrupt politicians in a context where corruption is widespread and most political candidates are perceived as corrupt (see Corbacho, Gingerich, Oliveros, & Ruiz-Vega, 2016; Klasnja, Little, & Tucker, 2017; Pãvao, 2015, 2018). This article also provides evidence that widespread political corruption can dramatically depress voter turnout; both candidates being corrupt is by far the strongest predictor of voting abstention (exit). To my knowledge, this is the first study showing this in an experimental setting, confirming what other studies have shown at the aggregate level (e.g., Chong, De La O, & Wantchekon, 2015).

These results provide insight into when and why voters might be willing to support corrupt politicians. The results shed light on the question of why political corruption is so persistent in some contexts, even when corruption is almost universally condemned by citizens: When voters believe that all politicians are corrupt, a substantial share of the electorate will vote for the least bad corrupt political alternative, while other segments of the electorate (with the strongest distaste for corruption) will withdraw from political participation altogether. The broader implication is that the failure to punish corrupt candidates is not necessarily a failure of democratic accountability, but of the credibility and quality of political candidates.³ Repeated scandals can make voters infer that the overall candidate pool is of low quality and reinforce the belief that it is unlikely that any specific candidate will turn out to be clean (see Meirowitz & Tucker, 2013). This can in many situations allow corrupt politicians to get re-elected.

Do Voters Punish Corrupt Politicians?

Citizens view elected representatives as one of the major sources of corruption in society (Global Corruption Barometer, 2017). Yet, research shows that politicians do not always get punished electorally for involvement with corruption (Bågenholm 2013; de Sousa & Moriconi, 2013; Welch & Hibbing, 1997). Simultaneously, other studies have demonstrated that information about corruption among politicians can have substantial consequences for electoral outcomes (Ferraz & Finan, 2008). This has spurred an intense research debate on the relationship between corruption and democratic accountability, and on the question of why corrupt politicians are not always punished at the ballot box.

Winters and Weitz-Shapiro (2013) argue, based on a survey experiment in Brazil, that lack of information is often why voters fail to hold corrupt officials accountable. However, when given credible and reliable information about the misuse of public money, citizens are in fact, according to the authors, very likely to hold corrupt politicians accountable for their actions. This conclusion is supported by Weitz-Shapiro and Winters (2017) (see also Adsera, Boix, & Payne, 2003; Chang et al., 2010). Ferraz and Finan (2008) show, in line with this argument, that information dissemination about corruption, in the form of audit reports, increased the electoral punishment of incumbents in Brazil.

One reason voters do not punish corruption, even in the light of sufficient information, might be that they do not assign blame correctly. In general, people tend to attribute positive outcomes to their own in-group and blame out-groups for negative outcomes (De Vries & Solaz, 2017). Anduiza et al. (2013) show that partisan bias might play a substantial role in the process of blaming politicians for involvement with corruption. In a survey experiment conducted in Spain, the authors find that respondents view the same act as less corrupt when the accused is someone from the respondent’s preferred party (however, see Konstantinidis & Xezonakis, 2013).

Although information acquisition and blame attribution are important, De Vries and Solaz (2017) point to another condition that can be crucial for the punishment of corruption: behavioral response. Voters that are informed about an act of corruption and correctly attribute blame to some politician might still react in a number of different ways. First, voters might, if possible, switch to another (clean) party or candidate. Second, voters might simply abstain from voting in the face of corruption. Several studies have noted a negative association between corruption and voter turnout in general (Dahlberg & Solevid, 2016; McCann & Domínguez, 1998; Stockemer, LaMontagne, & Scruggs, 2013). In a field experiment in Mexico, Chong et al. (2015) provided corruption information to randomly selected voting districts. The authors observed a significant decrease in turnout in districts receiving the information treatment.

A third behavioral response might be to tolerate corruption. Even though voters in general do not approve of corrupt practices, other things might outweigh voters’ distaste for such acts. Rundquist et al. (1977) refer to this as “implicit trading,” where voters “trade” away honesty and integrity in government for other things that they like and prioritize. This might be economic performance (Choi & Woo, 2010; Zechmeister & Zizumbo-Colunga, 2013), or competence and good performance in another area (Munoz et al., 2016). In line with this, Konstantinidis and Xezonakis (2013) find that the perceived collective benefits from cutting taxes partly mitigate the electoral effect of political corruption. Fernandez-Vazquez et al. (2016) argue that voters ignore corruption scandals if they believe that there are “side benefits” (like increased economic activity) associated with the corrupt practices (see also Manzetti & Wilson, 2007). Voters might also “trade” corruption for issue positions that they like. Rundquist et al. (1977) found that voters were willing to ignore corruption charges when a candidate shared their stance on the Vietnam War. This finding is supported by Peters and Welch (1980) and Welch and Hibbing (1997).

Theory and Hypotheses

Voters tolerating corruption potentially poses a challenge to theories on democratic accountability. Can it really be the case that well-informed voters knowingly would choose to not punish a corrupt politician? I argue that to understand this question, we must consider the alternatives facing a voter when their favorite candidate is accused of corruption. Following Bauhr and Charron (2018), I divide a voter’s potential response to a corruption scandal involving an otherwise appealing candidate into three categories, using Albert Hirschman’s famous typology: voice, loyalty, and exit. Simply put, a voter has the option to either retain their support for the candidate (loyalty), to vote for a different (noncorrupt) candidate (voice), or to abstain from voting (exit).

The point of departure in this article is that voters strongly disapprove of corruption and therefore have a prima facie strong reason to choose the “voice” option in the light of a corruption scandal (punishing the corrupt candidate and rewarding the clean challenger). Survey studies repeatedly show that citizens around the world are highly concerned about corruption (e.g., Global Corruption Barometer, 2017), also in societies where corruption seems to be part of everyday life (Persson et al., 2013). In Spain, where my experiment was conducted, almost 80% of respondents say that taking a bribe can never be justified (World Values Survey, 2015).⁴ This percentage is higher than the same percentage for subjects like stealing, cheating on taxes, and benefit fraud (see Appendix A). Respondents from Spain also believe that politicians are among the most likely to engage in corruption (Eurobarometer, 2014, p. 26). As noted above, several studies show that voters are likely to punish political corruption when given credible information about corrupt activities (Botero, Cornejo, Gamboa, Pavao, & Nickerson, 2015; Ferraz & Finan, 2008; Weitz-Shapiro & Winters, 2017).⁵

Why should we expect voters to punish corrupt politicians just because voters themselves have a strong distaste for corruption? A central argument in the literature on retrospective voting is that voters rely on politicians’ past performance to make decisions (Fiorina, 1981). Ashworth (2012) presents a model where voters differentiate between candidates based on the voter’s beliefs about the candidates’ type. Voters look to the past to decide how to vote, but purely from prospective motives. Past actions provide voters with information about a candidate’s type and hence information about how the candidate is likely to behave in the future (Besley, 2006). Involvement with corruption thus sends a strong signal about a politician’s type; a politician who has been involved in shady dealings is more likely to be involved in such activities in the future. Theories of retrospective voting and democratic accountability thus give a clear prediction: Given voters’ very strong disapproval of corruption—as suggested by the survey evidence cited above, most people say corruption can never be tolerated—it is reasonable to expect voters to punish a corrupt candidate and opt for an existing clean alternative (voice), even when the corrupt candidate is appealing (e.g., when the candidate represents the respondent’s preferred party).

A core question is obviously to what extent a voter in a given situation can choose the “voice” option. What are the viable alternatives? Is there a clean alternative to switch to? Our understanding of how the viability of alternatives matters with regard to corruption voting is currently limited (De Vries & Solaz, 2017, p. 404). In a study of 24 European countries, Charron and Bågenholm (2016) show that voters who place themselves on the extremes of the ideological spectrum, and thus view fewer parties as viable, seem to be more likely to neglect political corruption accusations. Munoz et al. (2016), however, find that the lack of credible alternatives do not affect voters’ tendency to punish corruption. However, the authors express concern that their treatment for testing this hypothesis was not effective and that the null results (for this aspect of their experiment) should not be viewed as conclusive (Munoz et al., 2016, pp. 610-611).⁶

The availability of a clean alternative is an important factor to consider when it comes to the question of corruption voting; without such an alternative, the “voice” option is simply not available. However, most studies on corruption voting do not consider voting alternatives. For instance, experimental designs usually ask respondents to evaluate a single candidate involved with corruption, although varying the candidate’s general appeal (like competence or party) and occasionally also varying the corruption dimension (Anduiza et al., 2013; Botero et al., 2015; Klasnja & Tucker, 2013; Konstantinidis & Xezonakis, 2013; Weitz-Shapiro & Winters, 2017; Winters & Weitz-Shapiro, 2013, 2016). This is tantamount to only giving the respondent the option of “different degrees of loyalty.” I argue that this is a shortcoming of the literature. If not stated explicitly, we cannot know how respondents perceive their available options when asked whether they would vote for a specific candidate. My design, described below, aims at overcoming this problem. This discussion leaves me with my first hypothesis:

Hypothesis 1 (H1): When given a clean alternative, voters are likely to choose this option over a corrupt candidate, even when the corrupt candidate is appealing in other regards (voice).

However, it is of course not always the case that such a clean alternative exists. Especially in contexts where corruption is widespread, no candidate might be able to credibly portray themselves as a noncorrupt challenger (Klasnja et al., 2017). When no clean alternative exists, voters should be more likely to cast their vote for a corrupt candidate, if the candidate is appealing in other respects (loyalty). Pãvao (2015, 2018) argues that corruption is a special electoral issue in that it tends to create cynicism among voters. When voters perceive corruption to be a constant among electoral alternatives, they are likely to overlook this aspect and base their vote on different issues. Although many voters might withdraw from the political process altogether in such a situation (presumably the voters with the strongest aversion to corruption), other considerations will become more salient for the remaining voters (De Vries & Solaz, 2017), increasing the likelihood of casting a vote for a corrupt but otherwise appealing candidate. Simply put, in such a situation, parts of the electorate are willing to consider the “least bad” political alternative (see also Rundquist et al., 1977).

Voting for the “least bad” alternative also makes sense from a “rational voter” perspective (Ashworth, 2012). If only given the option to choose between corrupt “bad types,” a rational voter should be expected to take other things than corruption into consideration and to select the candidate that matches the voter’s preferences in other regards. All else equal, it is better to have a corrupt politician in office with views similar to yours, than a politician that is both corrupt and politically distant to you.

Furthermore, research suggests that a belief that corruption is widespread might in itself make people more tolerant of corruption. In a study conducted in Costa Rica, Corbacho et al. (2016) show that citizens become increasingly willing to pay bribes when they are induced to perceive a higher overall level of corruption in society. The authors take a game theoretic perspective and argue that expectations about the behavior of others are central to this result. This is in line with the theoretical predictions of Klasnja et al. (2017): As politics becomes more corrupt, citizens in general become more tolerant of corrupt politicians, creating an expectation that makes it harder for voters to coordinate around a clean candidate. Repeated political corruption scandals, involving many different politicians, can thus reinforce voters’ belief that the general candidate pool is of low quality (Meirowitz & Tucker, 2013), and the belief that any specific candidate is unlikely to actually be clean. Overall, this makes voters less apt to punish corrupt behavior among politicians. Importantly, this situation can arise, and persist, even if citizens have strong anticorruption preferences.

Hypothesis 2 (H2): The lack of a clean political alternative increases a voter’s likelihood of selecting a corrupt candidate that is appealing in other regards (loyalty).

Finally, one of the most consistent findings in the literature on corruption and political behavior is a strong negative association between corruption and voter turnout (Chong et al., 2015; Dahlberg & Solevid, 2016; McCann & Domínguez, 1998; Stockemer et al., 2013; Sundström & Stockemer, 2015). In short, many authors argue that widespread corruption leads to political distrust and voter apathy, which in turn depress turnout. Corruption being a constant among political alternatives will hence have two effects: It will, on one hand, make parts of the electorate consider other political dimensions, resulting in a vote for the least bad corrupt political alternative. Other parts of the electorate will, on the other hand, feel such strong aversion to corruption that they will choose to not vote. Therefore, I also expect many voters to use the “exit” option and abstain when all political alternatives are corrupt, and “voice” is not an option. This aspect of corruption voting is very rarely considered in experimental studies, something that is discussed in more detail below:

Hypothesis 3 (H3): When only given corrupt alternatives, voters become more likely to abstain from voting (exit).

Research Design and Methods

To test these hypotheses, I conducted an original survey experiment in Spain. Spain is a country with moderate levels of corruption, scoring 58 on Transparency International’s scale ranging from 0 (highly corrupt) to 100 (very clean).⁷ Corruption scandals are common, especially at the local level (see Costas-Perez, Sole-Olle, & Sorribas-Navarro, 2012), making the topic relevant and the scenario presented to the respondents realistic. Several studies have taken advantage of this to study political corruption in the Spanish context (e.g., Anduiza et al., 2013; Fernandez-Vazquez et al., 2016; Munoz et al., 2016). The survey was administered online by the market research firm SSI⁸ in September and October 2017.⁹ In total, 2111 individuals completed the survey. The sample of respondents was based on nationally representative age and gender quotas. For details about the sample, see Appendix A.

The research design is based around a version of a conjoint survey experiment (Hainmueller, Hopkins, & Yamamoto, 2014). This is a tool, previously widely used in marketing research (see Green, Krieger, & Wind, 2001) that allows the researcher to identify the causal effects of many treatment components simultaneously. The experimental design puts the respondent in a position where he or she has to choose from or rate hypothetical profiles that combine multiple attributes. For the experiment at hand, the hypothetical profiles are two politician profiles (two candidates for city mayor). This mimics the situation a citizen faces in the voting booth, where the voter chooses between different political candidates. A major advantage of the conjoint design is that the respondent does not have to state directly how he or she views things like a corruption accusation (this is only one of several candidate attributes that are varied simultaneously), something that can be expected to reduce social desirability bias (see Hainmueller et al., 2014).

The respondents were first given a background survey with standard demographic questions, as well as questions about political attitudes and party identification. Next, respondents were presented with the choice task. The scenario was described as follows: “Two politicians, currently members of the city council in a city similar to yours, are running as their respective party’s candidate for city mayor. The race is close and they are the only candidates with a shot at winning the election.”¹⁰ The respondents were then presented with two candidate profiles, where six different attributes were varied.¹¹ First, the candidates’ gender was randomly varied (male or female). Second, four different factors affecting the general appeal of a candidate, and potentially a respondent’s propensity to punish a candidate for corruption,¹² were varied: party, experience/competence, religious denomination, and policy position. Party was randomized from the set of PP, Podemos, Cs, and PSOE.¹³ An indicator variable was coded as 1 if the party of a candidate matched the respondent’s preferred party (as stated in the background survey) and 0 otherwise. The candidates’ experience/competence was randomly varied between low experience, medium experience, and high competence/experience. This can be thought of as a dimension indicating a candidate’s overall quality as a politician, where the “highest” category describes the candidate as both competent and experienced, whereas the “lower categories” only refer to the candidate’s (limited) political experience. Religious denomination¹⁴ was varied with the candidate being either a practicing catholic, an atheist, or having no strong religious views. An indicator variable was coded as 1 if the respondent defined him or herself as a catholic and the candidate was described as a catholic. The variable was coded as 0 otherwise. Policy position was coded as 1 (and 0 otherwise) if the respondent and a candidate agreed on a political issue that was decided based on the background survey. The experiment was designed to guarantee that the candidates hold policy positions on an issue dimension that the respondent finds relevant. For details about this attribute, as well as details about the other attributes, see Appendix B.

For the corruption dimension, each candidate was randomly given one of the four different attributes. Two of the attributes said that the candidate was accused of corruption: The first stated the candidate was accused by a judge for accepting money in exchange for fast building permits, and the second stated that the candidate was accused by a judge for using public money to finance his or her election campaign. Varying between two slightly different corruption scenarios was expected to increase the realism of the experiment, by not displaying the exact same scenario over and over again. The two corruption scenarios were also designed to make the accusations credible. Therefore, the accuser was chosen to be a judge (see Botero et al., 2015; Costas-Perez et al., 2012; Weitz-Shapiro & Winters, 2017). The experiment was thus not designed to test the credibility of different sources of corruption information. Rather, the study tries to tap into voters’ different responses to credible corruption accusations because these cases potentially pose the most interesting challenge to democratic accountability. The corruption dimension also included two “clean” treatments: a “neutral treatment,” stating that the candidate had not been involved in any scandals, and a “positive treatment,” stating that the candidate has a reputation of working hard to fight corruption in the city. All dimensions and attributes displayed in the experiment are summarized in Table 1.

Table 1.

Conjoint Experiment: Dimensions and Text Corresponding to Different Attributes.

Gender

Female

Male

(coded 1 for attribute, 0 otherwise)

Experience/competence

The candidate has very limited political experience

The candidate has had a few previous political appointments

The candidate has long political experience and has been very successful in attracting investments to the city

(coded 1 for attribute, 0 otherwise)

Party

Partido Popular (PP)

Partido Socialista Obrero Español (PSOE)

Ciudadanos (Cs)

Podemos

(coded 1 for match with respondent, 0 otherwise)

Religious denomination

The candidate is a practicing catholic

The candidate identifies as an atheist

The candidate has no strong religious views

(coded 1 for match with respondent [catholic], 0 otherwise)

Policy position

Dimension decided by question in background survey (Tourism, Refugees, or Carbon emissions/Growth)—three possible candidate positions

(coded 1 for match with respondent, 0 otherwise)

Corruption dimension

He or she has not been involved in any scandals (clean)

He or she has a reputation of working hard to fight corruption in the city (clean)

A judge has recently accused the candidate of corruption for accepting money in exchange for fast building permits (corrupt)

A judge has recently accused the candidate of corruption for using public money to finance his or her election campaign (corrupt)

(coded into categorical variable—see variable description)

The dimension labels (e.g., “Party”) were not displayed to the respondents. See Appendix B for additional information about the variable codings. The “judge” in the corruption scenarios should be thought of as the prosecutor leading the legal process. In the Spanish legal system a judge can take on this role in cases like this.

The main goal of this study—as indicated by the hypotheses above—is to explore different “corruption scenarios” (S) that voters might face. Is there a clean candidate to vote for? Or are all available candidates accused of corruption? Therefore, for the main part of the analysis, the corruption dimension was recoded into a categorical variable. A variable was first coded as 1 if a candidate was accused of any form of corruption and 0 otherwise (for the other two attributes). Next, I constructed a new variable indicating whether the other candidate, the challenger “running” against a specific candidate, was corrupt or not (coded in the same way as the dichotomous corruption treatment).¹⁵ The two corruption variables were then combined, indicating one of the four possible situations the respondent might be randomly assigned to when deciding whether to vote for a specific candidate. The final variable was coded as a categorical variable, $S$ , taking on values $s_{1}$ , $s_{2}$ , $s_{3}$ , or $s_{4}$ for each specific candidate displayed to the respondents.

After seeing the candidate profiles, the respondent was asked which of the candidates he or she would vote for in the election. The alternatives given were Candidate 1, Candidate 2, or I would not vote.¹⁶ The choice task was repeated three times, each time with different (randomly generated) candidates. Providing respondents with the “I would not vote” alternative is intended to give the respondents a choice task that is as close to the real-life decision situation as possible (in this case, voting in an election) (Haaijer, Kamakura, & Wedel, 2001). As discussed above, the inclusion of this alternative is an important part of the design. Most conjoint designs force the respondent to choose between the candidates, but as Hensher, Rose, and Greene (2005, p. 176) note: When forcing the respondent to make a choice, the design obliges the decision-maker to trade off the attribute levels of the available alternatives. Whether the respondent is willing to make such a trade-off is one of the aspects that the experiment is intended to test (and one of the interesting aspects of corruption voting in general).¹⁷ The design thus makes it possible to explore, for instance, what happens when both candidates presented are accused of corruption $(s_{4})$ : Do the respondents simply opt out and choose the “not vote” alternative in such a scenario, or are they willing to select one of the corrupt candidates?

Overall, I argue that these different aspects make the design more realistic than designs only asking the respondent to accept or reject one specific candidate. In a recent study conducted with Brazilian respondents, Boas, Hidalgo, and Melo (2019) were not able to replicate the negative effects of corruption found in vignette experiments (e.g., Winters & Weitz-Shapiro, 2013) in a field experiment disseminating corruption information. However, as the vignette in the study replicated previous survey experiments, the outcome question only asked about a specific candidate, without any reference to the available alternatives. Here, I think the design in the study at hand constitutes an alternative that potentially can provide higher external validity by putting respondents in a more realistic choice situation. Other studies suggest that the conjoint design in fact can track real-world decisions very closely. For instance, in a study of which immigrant attributes generate support for naturalization that compared conjoint survey experiments with real-world behavior of citizens, Hainmueller, Hangartner, and Yamamoto (2015, p. 2395) find that the conjoint design “performs remarkably well in recovering the effects of the same attributes in the [real world] behavioral benchmark.”

Results

To get an overview of the experiment, I first estimated the overall treatment effects for all candidate attributes, including the original corruption variable.¹⁸ I estimated the marginal effect of a given attribute on the probability of being selected, using a linear probability model with standard errors clustered by respondent. The dependent variable is thus “vote choice” $(Y)$ ; whether the respondent said he or she would vote for a specific candidate $(Y = 1)$ (as opposed to voting for the challenger or selecting the “not vote” option $[Y = 0]$ ). The results are presented in Figure 1. In total, over all profiles, the “not vote” alternative was chosen about 39% of the time. This is quite close to the actual abstention rate in the most recent Spanish local elections (2015), which was about 35%.

Figure 1.

All treatment effects.

In general, the effects go in the expected direction: Respondents are more likely to vote for candidates with long experience, candidates from their preferred party, and candidates representing their preferred policy position. The candidate representing the respondent’s preferred party shows the strongest positive effect among the different treatments: A respondent is almost 30% more likely to select a candidate in this case. Compared with the “no scandal” category, respondents were about 5.6% more likely to select a candidate with a reputation to fight corruption. The results show a strongly negative effect for the corruption treatments. On average, a candidate accused of corruption was more than 30% less likely to be selected.¹⁹ It is clear that the effects for the two different corruption treatments (“election” and “permits”) are very close in magnitude (the difference between the coefficients is not statistically significant). These results, hence, do not provide support for the argument that voters have a more benign view of potentially “welfare enhancing”²⁰ corruption (see Fernandez-Vazquez et al., 2016). Overall, this also shows that it is reasonable to code the corruption variable into a “clean” and a “corrupt” category, as discussed above.

My general argument holds that whether a voter actually has the opportunity to vote for a clean alternative is an important part of understanding corruption voting. Therefore, I continue the analysis using the categorical corruption variable (S) described in the previous section. H1 states that voters in general should punish corruption when they have the chance, even when the corrupt candidate is appealing in other regards. To test this, I first ran a model with “vote choice” as the dependent variable and the categorical corruption variable as the independent variable.²¹ The corruption variable was also interacted with candidate experience, candidate party, and candidate policy position. This was done to provide a flexible model that can account for the potentially heterogeneous effect of corruption conditional on different attributes and, conversely, the heterogeneous effect of different attributes in different corruption scenarios (Berry, Golder, & Milton, 2012). I chose these specific interactions based on the previous literature and attributes that have been suggested to mitigate the negative effect of corruption (see literature review), as well as the number of observations in each treatment group. I therefore excluded the “catholic treatment” from the analysis because the matched group was too small to estimate an interaction effect with any precision.²² The full model specification is shown in Appendix A. My aim here is thus to estimate $E [Y_{(i)} | x (s), S = s]$ , where $Y_{(i)} (i \in 1, 2)$ indicates if a specific candidate is selected, $x$ is a vector of candidate attributes that are allowed to vary with $S$ , and $S$ denotes one of the four different corruption scenarios discussed above $(S \in s_{1}, s_{2}, s_{3}, s_{4})$ .

Next, I computed the predicted probability to vote for a candidate in two different scenarios: when the candidate is clean and the challenger is clean $(s_{1})$ and when the candidate is corrupt but the challenger is clean $(s_{3})$ .²³ I estimated the vote probability in these scenarios for three different candidates, with varying levels of appeal. The first candidate was simply an average candidate (the predicted probabilities conditional on the corruption variable, averaged over all other candidate attributes).²⁴ The second candidate is a candidate from the respondent’s preferred party. Anduiza et al. (2013) have argued that voters are more tolerant of corruption when the accused represents their preferred party (see also De Vries & Solaz, 2017). As shown in Figure 1, party is also one of the strongest predictors of candidate choice. Finally, I computed the vote probability for an ideal candidate—a candidate that (apart from the corruption dimension) in general should be maximally appealing to the respondent. This candidate was defined as representing the respondent’s preferred party, having high competence/experience, and agreeing with the respondent’s policy position. The goal here was to include essentially all candidate characteristics that have previously been argued to reduce the negative effect of a corruption scandal (De Vries & Solaz, 2017). The predicted vote probabilities are displayed in Figure 2.

Figure 2.

Voting for candidate—challenger clean.

The estimates in Figure 2 thus displays the probability of voting for a candidate (as opposed to choosing the challenger or the “not vote” option), conditional on the challenger being clean (averaging over all other attributes for the challenger). This corresponds to the left column of Table 2 ( $s_{1}$ and $s_{3}$ ).

Table 2.

Possible Corruption Scenarios.

$(s_{1})$ Candidate clean—Challenger clean	$(s_{2})$ Candidate clean—Challenger corrupt
$(s_{3})$ Candidate corrupt—Challenger clean	$(s_{4})$ Candidate corrupt—Challenger corrupt

In general, the estimates indicate a very strong negative effect of corruption: When the average candidate goes from being clean to corrupt, the voting probability drops from about 38% to 8%. For the candidate from the respondent’s preferred party, the same number drops from 75% to 21%. In absolute terms, the drop is largest for the ideal candidate: from 88% to 28%. In sum, it is clear that respondents do punish corruption to a significant degree, even when the candidate is very appealing.

Are respondents switching to the clean alternative when their preferred candidate is accused of corruption, or do they abstain from voting altogether? To test this, I computed predicted vote probabilities (based on the same statistical model) when two specific candidates are facing off. First, a corrupt candidate from the respondents preferred party versus an average candidate and then a corrupt ideal candidate versus an average candidate.²⁵

As shown in Figure 3, a substantial share of respondents do switch to the clean alternative when their preferred candidate is accused of corruption. The graph displays estimates for two candidates “running” against each other. The left-hand graph shows that respondents prefer a clean challenger with average attributes to a corrupt candidate from their preferred party by a large margin. Even when the corrupt ideal candidate is running against a clean average candidate, respondents are more likely to choose the latter alternative (about 41%, vs. 28% for the corrupt ideal candidate). The remaining respondents in this case (about 31%) are predicted to choose the “not vote” alternative. In line with H1, these results suggest that respondents have a strong inclination to choose the voice option—when they are presented with a clean alternative and have the chance to do so.

Figure 3.

Two candidates running against each other.

The results thus far show that respondents do punish corruption, when given the opportunity. This is hence in line with the theoretical predictions: When voters have clear information about poor performance (corruption) and have the possibility to switch to a better performing candidate, they take this opportunity (Ashworth, 2012; Fiorina, 1981). What happens when this opportunity is more limited, when the respondent does not have a clean candidate to vote for? As argued above (H2), I expect respondents to be substantially more likely to accept a given corrupt candidate when both candidates are corrupt—despite their strong preference for clean politicians. To test this argument, I ran the same interaction model as above and estimated the predicted probability of selecting a specific candidate accused of corruption in two different scenarios: when the challenger is clean $(s_{3})$ and when the challenger is corrupt $(s_{4})$ .

The left-hand graph in Figure 4 shows the predicted support for a corrupt candidate, when the challenger is either clean or corrupt. The share of respondents selecting the corrupt candidate goes up from 8% to 17% when the challenger goes from being clean to corrupt. When the corrupt candidate represents the respondent’s preferred party, the voting probability increases from 21% to 36% in the same scenario.²⁶ The voting probability in the case of the ideal candidate increases from 28% to 48%. These results are interesting because the design does not “force” respondents to select one of the corrupt alternatives. Rather, respondents always have the option to abstain, to choose the “not vote” alternative. This suggests that respondents are much more likely (almost twice as likely) to choose loyalty and retain their support for an otherwise appealing candidate, when both candidates are described as corrupt.

Figure 4.

Voting for corrupt candidate.

One reason that this might be the case is that other dimensions become more salient when corruption is a constant among electoral alternatives (De Vries & Solaz, 2017; Pãvao, 2015). To see how the effect of different treatment components vary, I computed the marginal effects (for each attribute that was interacted with the corruption variable) conditional on the different corruption scenarios. The analysis, presented in Appendix C, supports the notion that other attributes become more important when both candidates are corrupt. The marginal effect for each attribute is significantly larger in the latter scenario, compared with a scenario when only one candidate is corrupt.

Overall, these results lend support to both H1 and H2: Voters in general express a strong distaste for corruption and punish corruption to a considerable extent, even when the corrupt politician is very appealing in other regards. However, when both candidates are corrupt and respondents are given no clean political alternatives, the willingness of many respondents to consider a corrupt alternative goes up.

At the same time, many voters are expected to prefer the exit option in the latter situation, as predicted by H3. To test the effect of corruption on abstention, I model each choice task as one observation with the outcome being either “vote” (if one of the candidates was chosen, coded as 0) or “not vote” (if this option was selected, coded as 1). The independent variables in this case consist of the candidates’ joint attributes. I consider four different variables that might be predictive of abstention: both candidates having low competence/experience, neither candidates matching the respondent’s preferred policy position, neither candidates matching the respondent’s preferred party, and both candidates being corrupt. I then regressed the vote/not vote variable on these four predictors and calculated the probability of abstention when a specific variable is equal to 1 (averaging over the other variables).

As shown in Figure 5, both candidates being corrupt is by far the strongest predictor of vote abstention. When this variable is equal to 1, 67% of respondents are predicted to abstain. This can be compared with, for instance, 47% when neither candidates represent the respondent’s preferred party. This lends support to H3: When only given corrupt alternatives, the propensity to choose the exit option goes up dramatically.²⁷

Figure 5.

Predicted probability of abstention.

Discussion and Conclusion

Using an experimental design, this study explores several important aspects of corruption voting, many of which are underresearched or absent in the previous literature. The study establishes that when given unambiguous and credible information about political corruption, voters are likely to punish a candidate involved in shady dealings. The results suggest, contrary to some of the previous literature, that this is true even for candidates that are very appealing in other respects. As shown above, many respondents prefer the voice option in this situation, even if this means switching to a candidate that is less appealing on other dimensions.

This finding is obviously to some degree a result of the conditions in this particular experiment, with clear information about corruption from a credible source. Other experimental studies have shown that information quality and source credibility can matter for the extent of punishment (Botero et al. 2015; Weitz-Shapiro & Winters, 2017). The experimental setting hence provides an idealized scenario where information is clear and unambiguous. In this sense, the results provide an estimate of how voters would act in cases where they are well-informed about the background and true credibility of available political alternatives. Needless to say, the information signal in real-world elections is oftentimes substantially more noisy, where the accused candidate might fight tooth and nail to deny any wrongdoing and try to obscure media reporting by launching counter-campaigns. In settings where the information quality is poor (and negative information hence more easily discounted), it is reasonable to expect traditional political factors like partisanship to play a more important role than a corruption accusation (Anduiza et al., 2013; Solaz, De Vries, & de Geus, 2018). At the same time, the setting with clean, credible information is arguably the most interesting from an accountability perspective: Given voters’ strong general preference for clean politicians, failing to punish corruption when such an alternative exists would put a serious challenge to democratic accountability theory. In this sense, “failing” to punish corruption when information is ambiguous or inadequate is much less of a paradox (Winters & Weitz-Shapiro, 2013). Existing studies of real cases of political corruption show strong heterogeneity in the effects (Costas-Perez et al., 2012; Fernandez-Vazquez et al., 2016). Although many corrupt politicians essentially go unpunished, several cases of substantial punishment for corruption are also reported, with electoral losses that resemble the estimates in the paper where a corrupt candidate is facing a clean alternative.²⁸ What the real-world cases with strong corruption-effects seem have in common is high quality of information, where the accusations are viewed as credible due to extensive press coverage (Chang et al., 2010; Costas-Perez et al., 2012; Ferraz & Finan, 2008). It is reasonable to view the experimental setting in the study at hand as resembling one of these situations.

The voter’s ability to hand out electoral rewards and punishment is crucially a function of available political alternatives. This dimension is understudied in the literature on corruption voting (De Vries & Solaz, 2017). In part, this can be traced to limitations in many research designs, often providing respondents with only a single candidate or party to evaluate. Here, I argue that the conjoint experiment with the inclusion of a “not vote” exit-option provides a design alternative that mimics the real-life choice situation facing voters well. Recent studies showing great similarities between respondent behavior in conjoint experiments and real-world choices gives credence to these designs (Hainmueller et al., 2015). The main contribution to the literature is that the design allows me to focus on how the available political alternatives play into voters’ punishment of corruption. Although voters, in line with theories of retrospective voting and democratic accountability, are very prone to punish corruption when a clean alternative exists, the absence of such an alternative leads some voters to become significantly more willing to show loyalty to the corrupt candidate. This is consistent with models emphasizing the increased salience of other dimensions in such a situation (De Vries & Solaz, 2017; Pãvao, 2015, 2018) and citizens’ increased tolerance of corruption in high-corruption environments (Corbacho et al., 2016). In the light of respondents’ strong aversion to corruption, and given the fact that respondents in the experimental setup were always provided clear information and allowed to choose the “not vote” alternative, this finding is informative.

The results also provide causal evidence that many voters will withdraw from the political process when facing several candidates accused of corruption. This is another aspect that most previous experimental designs have not considered. As the negative relationship between corruption and voter turnout is one of the most consistent findings in the literature on corruption and political behavior (e.g., Dahlberg & Solevid, 2016; McCann & Domínguez, 1998; Stockemer et al., 2013), this is an important omission. The results show strong support for this notion: Both candidates being corrupt is a very strong predictor of vote abstention—much stronger than, for instance, neither candidates representing the respondent’s preferred party. When political corruption is rampant, many voters will opt for the exit option.

These results shine light on the question of why corruption seems so persistent in some contexts (Mishra, 2006). Klasnja et al. (2017) explore a model where some countries or regions are stuck in a “high-corruption equilibrium”—a corruption trap. The model describes how the decisions made by politicians and voters interact in ways that can lead to more or less permanent “political corruption traps.” In a context of high corruption, many voters will think that all political entrants are likely to be corrupt and will have no incentives to vote based on corruption. They will also have less reason to believe that other voters will vote based on corruption, thus failing to coordinate to “throw the rascals out.” In such a situation, voters become more “tolerant” of corruption—even though they in general have a strong preference for clean and competent politicians. As suggested above, voters with the strongest distaste for corruption might also simply abstain when only given corrupt alternatives. If voters are less apt to punish corrupt behavior, then politicians will be more free to engage in corruption, which in turn reinforces voters’ belief about most politicians engaging in corrupt activities (Meirowitz & Tucker, 2013). Situations like this can hence create very persistent political corruption traps. The results from this study are largely consistent with this model.

Can a belief among voters that “all alternatives are corrupt” be part of the explanation for why many corrupt politicians get re-elected in a developed country like Spain? I think the answer to this question is yes. As noted above, local corruption scandals are commonplace in Spain and often get extensive coverage in the media (Costas-Perez et al., 2012; Fernandez-Vazquez et al., 2016). This has led many Spanish citizens to have a very negative view of politicians. As noted above, many believe politicians to be the most likely group in society to engage in corruption (Eurobarometer, 2014). In line with the argument in Meirowitz and Tucker (2013), this potentially reflects the fact that citizens are updating their belief about the overall quality of the candidate pool; repeated scandals make voters skeptical of the possibility that the next candidate running for office actually will turn out to be of a “good type.” When asked in the background survey about the prevalence of corruption in local politics, over 40% of respondents said that “most” or “almost all” local politicians are involved in corruption (see Appendix A). These statistics suggest that many citizens might feel that they do not have a clean and credible political alternative to vote for in many situations and might therefore show loyalty to a corrupt but otherwise appealing candidate. In the Supplemental Material, I also show—in line with the main results—that voters with high corruption perceptions are significantly more tolerant of corruption. I argue that this can be part of the explanation for why politicians, in Spain and elsewhere, often do not suffer significant electoral losses when involved in corruption scandals. Overall, this suggests that the failure to punish corruption often can be ascribed to the low credibility of the candidate pool.

At the same time, the results show some promise with regard to electoral accountability as a tool for curbing political corruption. When actually presented with a clean candidate, respondents were very likely to reject the corrupt candidate, even if the accused candidate was very appealing in other regards. This suggests that politicizing corruption, and credibly presenting yourself as a clean political alternative, can be a viable electoral strategy (see Bågenholm, 2013). The results suggest that voters’ punishment for corruption can be harsh under the right conditions, and that anticorruption strategies can attract votes. Of course, building the image of a “clean outsider” might be easier said than done in a context where corruption is widespread and political distrust and cynicism is deeply rooted (Pãvao, 2015). Research suggests that a radical change in the information environment, with an aggressive free press tirelessly reporting on existing cases of political malfeasance, can be what is needed to break a corrupt equilibrium and pave the way for a credible and clean alternative (Chang et al., 2010).

Spain is a context with moderate levels of corruption. The most obvious next step for future research would be to conduct similar studies in contexts where corruption is more widespread and endemic. Are patterns of electoral punishment similar where corruption, to an even higher degree, is part of everyday politics? The corruption scenarios presented in the survey are all relatively clear and unambiguous. As noted above, this is of course not always the case in the real world. Future studies should explore how information uncertainty and message credibility interact with the existing set of political alternatives. How credible does information about corruption have to be for voters to switch from a corrupt but otherwise appealing candidate to a clean and average challenger? These are all projects that can advance the study of corruption voting even further.

Supplemental Material

Supplementary_materials2 – Supplemental material for The Lesser Evil? Corruption Voting and the Importance of Clean Alternatives

Supplemental material, Supplementary_materials2 for The Lesser Evil? Corruption Voting and the Importance of Clean Alternatives by Mattias Agerberg in Comparative Political Studies

Footnotes

Appendix A

Appendix B

Appendix C

Acknowledgements

The author would like to thank Lena Wängnerud, Nicholas Charron, Anne-Kathrin Kreft, Rasmus Broms, Jonna Hellström, the Quality of Government Institute, and the three anonymous reviewers at Comparative Political Studies for their valuable comments and suggestions on earlier drafts of this article.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Support for this research was provided by The Quality of Government institute and Wilhelm & Martina Lundgrens Vetenskapsfond.

ORCID iD

Mattias Agerberg

Supplemental Material

Supplemental material for this article is available online at the CPS website

Notes

Author Biography

Mattias Agerberg is a PhD candidate at the Department of Political Science, University of Gothenburg. His research focuses on corruption, political behavior, and public opinion.

References

Adsera

Boix

Payne

(2003). Are you being served? Political accountability and quality of government. Journal of Law, Economics, & Organization, 19, 445-490.

Anduiza

Gallego

Munoz

(2013). Turning a blind eye: Experimental evidence of partisan bias in attitudes toward corruption. Comparative Political Studies, 46, 1664-1692.

Ashworth

(2012). Electoral accountability: Recent theoretical and empirical work. Annual Review of Political Science, 15, 183-201.

Bågenholm

(2013). Throwing the rascals out? The electoral effects of corruption allegations and corruption scandals in Europe 1981-2011. Crime, Law and Social Change, 60, 595-609.

Bardhan

(1997). Corruption and development: A review of issues. Journal of Economic Literature, 35, 1320-1346.

Bauhr

Charron

(2018). Insider or outsider? Grand corruption and electoral accountability. Comparative Political Studies, 51, 415-446.

Berinsky

A. J.

Huber

G. A.

Lenz

G. S.

(2012). Evaluating online labor markets for experimental research: Amazon.com’s Mechanical Turk. Political Analysis, 20, 351-368.

Berry

W. D.

Golder

Milton

(2012). Improving tests of theories positing interaction. The Journal of Politics, 74, 653-671.

Besley

(2006). Principled agents? The political economy of good government. Oxford, UK: Oxford University Press.

10.

Boas

T. C.

Hidalgo

F. D.

Melo

M. C.

(2019). Norms versus action: Why voters fail to sanction malfeasance in Brazil. American Journal of Political Science, 63, 385-400.

11.

Botero

Cornejo

R. C.

Gamboa

Pavao

Nickerson

D. W.

(2015). Says who? An experiment on allegations of corruption and credibility of sources. Political Research Quarterly, 68, 493-504.

12.

Chang

E. C. C.

Golden

M. A.

Hill

S. J.

(2010). Legislative malfeasance and political accountability. World Politics, 62, 177-220.

13.

Charron

Bågenholm

(2016). Ideology, party systems and corruption voting in European democracies. Electoral Studies, 41, 35-49.

14.

Charron

Lapuente

Rothstein

(2013). Quality of government and corruption from a European perspective: A comparative study of good government in EU regions. Cheltenham, UK: Edward Elgar.

15.

Choi

Woo

(2010). Political corruption, economic performance, and electoral outcomes: A cross-national analysis. Contemporary Politics, 16, 249-262.

16.

Chong

De La O

A. L.

Wantchekon

(2015). Does corruption information inspire the fight or quash the hope? A field experiment in Mexico on voter turnout, choice, and party identification. The Journal of Politics, 77, 55-71.

17.

Coppock

(2018). Generalizing from survey experiments conducted on Mechanical Turk: A Replication Approach. Political Science Research and Methods, 1-16.doi:10.1017/psrm.2018.10

18.

Corbacho

Gingerich

D. W.

Oliveros

Ruiz-Vega

(2016). Corruption as a self-fulfilling prophecy: Evidence from a survey experiment in Costa Rica. American Journal of Political Science, 60, 1077-1092.

19.

Costas-Perez

Sole-Olle

Sorribas-Navarro

(2012). Corruption scandals, voter information, and accountability. European Journal of Political Economy, 28, 469-484.

20.

Dahlberg

Solevid

(2016). Does corruption suppress voter turnout? Journal of Elections, Public Opinion and Parties, 26, 489-510.

21.

de Sousa

Moriconi

. (2013). Why voters do not throw the rascals out?—A conceptual framework for analysing electoral punishment of corruption. Crime, Law and Social Change, 60, 471-502.

22.

De Vries

C. E.

Solaz

. (2017). The electoral consequences of corruption. Annual Review of Political Science, 20, 391-408.

23.

Eggers

A. C.

Vivyan

Markus

(2018). Corruption, accountability, and gender: Do female politicians face higher standards in public life? Journal of Politics, 80, 321-326.

24.

Eurobarometer. (2014). Special Eurobarometer 397: Corruption (Tech. rep.). TNS Opinion & Social at the Request of the European Commission. Available at http://ec.europa.eu/commfrontoffice/publicopinion/archives/ebs/ebs_397_en.pdf

25.

Fernandez-Vazquez

Barbera

Rivero

(2016). Rooting out corruption or rooting for corruption? The heterogeneous electoral consequences of scandals. Political Science Research and Methods, 4, 379-397.

26.

Ferraz

Finan

(2008). Exposing corrupt politicians: The effects of Brazil’s publicly released audits on electoral outcomes. The Quarterly Journal of Economics, 123, 703-745.

27.

Fiorina

M. P.

(1981). Retrospective voting in American national elections. New Haven, CT: Yale University Press.

28.

Global Corruption Barometer. (2017). People and corruption: Citizen’s voices around the world (Tech. rep.). Transparency International. Available from https://www.transparency.org/whatwedo/publication/people_and_corruption_citizens_voices_from_around_the_world

29.

Green

Krieger

Wind

(2001). Thirty years of conjoint analysis: Reflections and prospects. Interfaces, 31, 56-73.

30.

Haaijer

Kamakura

Wedel

(2001). The “no-choice” alternative in conjoint choice experiments. International Journal of Research in Marketing, 43, 93-106.

31.

Hainmueller

Hangartner

Yamamoto

(2015). Validating vignette and conjoint survey experiments against real-world behavior. Proceedings of the National Academy of Sciences, 112, 2395-2400.

32.

Hainmueller

Hopkins

D. J.

Yamamoto

(2014). Causal inference in conjoint analysis: Understanding multidimensional choices via stated preference experiments. Political Analysis, 22, 1-30.

33.

Hanmer

M. J.

Kalkan

K. O.

(2013). Behind the curve: Clarifying the best approach to calculating predicted probabilities and marginal effects from limited dependent variable models. American Journal of Political Science, 57, 263-277.

34.

Hensher

D. A.

Rose

J. M.

Greene

W. H.

(2005). Applied choice analysis: A primer. Cambridge, UK: Cambridge University Press.

35.

Holmberg

Rothstein

(2011). Dying of corruption. Health Economics, Policy and Law, 6, 529-547.

36.

International Social Survey Program. (2014). Citizenship II - ISSP 2014 (Tech. rep). ISSP Research Group. Available at http://w.issp.org/data-download/by-year/

37.

Kaufmann

Kraay

Mastruzzi

(2011). The worldwide governance indicators: Methodology and analytical issues. Hague Journal on the Rule of Law, 32, 220-246.

38.

Klasnja

Little

A. T.

Tucker

J. A.

(2017). Political corruption traps. Political Science Research and Methods, 6, 413-428. doi:10.1017/psrm.2016.45.

39.

Klasnja

Tucker

J. A.

(2013). The economy, corruption, and the vote: Evidence from experiments in Sweden and Moldova. Electoral Studies, 32, 536-543.

40.

Klasnja

Tucker

J. A.

Deegankrause

(2016). Pocketbook vs. sociotropic corruption voting. British Journal of Political Science, 46, 67-94.

41.

Konstantinidis

Xezonakis

(2013). Sources of tolerance towards corrupted politicians in Greece: The role of trade offs and individual benefits. Crime, Law and Social Change, 60, 549-563.

42.

Long

J. S.

(1997). Regression models for categorical and limited dependent variables. London, England: Sage.

43.

Manzetti

Wilson

C. J.

(2007). Why do corrupt governments maintain public support? Comparative Political Studies, 60, 471-502.

44.

Mauro

(1995). “Corruption and growth. The Quarterly Journal of Economics, 110, 681-712.

45.

McCann

J. A.

Domínguez

J. I.

(1998). Mexicans react to electoral fraud and political corruption: An assessment of public opinion and voting behavior. Electoral Studies, 17, 483-503.

46.

McDonough

Barnes

S. H.

Pina

A. L.

(1988). Social identity and mass politics in Spain. Comparative Political Studies, 21, 200-230.

47.

Meirowitz

Tucker

J. A.

(2013). People power or a one-shot deal? A dynamic model of protest. American Journal of Political Science, 57, 478-490.

48.

Mishra

(2006). Persistence of corruption: Some theoretical perspectives. World Development, 34, 349-358.

49.

Mullinix

K. J.

Leeper

T. J.

Druckman

J. N.

Freese

(2015). The generalizability of survey experiments. Journal of Experimental Political Science, 2, 109-138.

50.

Mungiu-Pippidi

(2013). Becoming Denmark: Historical designs of corruption control. Social Research: An International Quarterly, 80, 1259-1286.

51.

Munoz

Anduiza

Gallego

(2016). Why do voters forgive corrupt mayors? Implicit exchange, credibility of information and clean alternatives. Local Government Studies, 42, 598-615.

52.

Pãvao

(2015). The failures of electoral accountability for corruption: Brazil and beyond (Doctoral dissertation). University of Notre Dame, IN.

53.

Pãvao

(2018). Corruption as the only option: The limits to electoral accountability. The Journal of Politics, 80, 996-1010.

54.

Persson

Rothstein

Teorell

(2013). Why anticorruption reforms fail: Systemic corruption as a collective action problem. Governance, 26, 449-471.

55.

Peters

J. G.

Welch

(1980). The effects of charges of corruption on voting behavior in congressional elections. The American Political Science Review, 74, 697-708.

56.

Rose-Ackerman

(1999). Corruption and government: Causes, consequences, and reform. Cambridge, UK: Cambridge University Press.

57.

Rothstein

(2011). The quality of government: Corruption, social trust and inequality in international perspective. Chicago, IL: The University of Chicago Press.

58.

Rundquist

B. S.

Strom

G. S.

Peters

J. G.

(1977). Corrupt politicians and their electoral support: Some experimental observations. The American Political Science Review, 71, 954-963.

59.

Solaz

De Vries

C. E.

de Geus

R. A.

(2018). In-group loyalty and the punishment of corruption. Comparative Political Studies, 52, 896-926.

60.

Stockemer

LaMontagne

Scruggs

(2013). Bribes and ballots: The impact of corruption on voter turnout in democracies. International Political Science Review, 34, 74-90.

61.

Sundström

Stockemer

(2015). Regional variation in voter turnout in Europe: The impact of corruption perceptions. Electoral Studies, 40, 158-169.

62.

Warren

M. E.

(2004). What does corruption mean in a democracy? American Journal of Political Science, 48, 328-343.

63.

Weitz-Shapiro

Winters

M. S.

(2017). Can citizens discern? Information credibility, political sophistication, and the punishment of corruption in Brazil. Journal of Politics, 79, 60-74.

64.

Welch

Hibbing

J. R.

(1997). The effects of charges of corruption on voting behavior in congressional elections, 1982-1990. The Journal of Politics, 59, 226-239.

65.

Winters

M. S.

Weitz-Shapiro

(2013). Lacking information or condoning corruption? When will voters support corrupt politicians? Comparative Politics, 45, 493-504.

66.

Winters

M. S.

Weitz-Shapiro

(2016). Who’s in charge here? Direct and indirect accusations and voter punishment of corruption. Political Research Quarterly, 69, 207-219.

67.

World Values Survey. (2015). World Values Survey. Wave 6: 2010-2014 (Tech. rep). World Values Survey Association. Available at http://www.worldvaluessurvey.org/WVSDocumentationWV6.jsp

68.

Zechmeister

E. J.

Zizumbo-Colunga

(2013). The varying political toll of concerns about corruption in good versus bad economic times. Comparative Political Studies, 46, 1190-1218.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.24 MB