“The Pure People” versus “the Corrupt Elite”? Political Corruption, Political Trust and the Success of Radical Right Parties in Europe

Abstract Contrasting “the pure people” with “the corrupt elite” is a prominent rhetorical figure of populist radical right parties and their proponents. However, scholars rarely consider political corruption as an explanatory factor for the electoral success of this party family. This study investigates the role of individual corruption experience in dealing with public officials for radical right party support in 12 European countries. As central argument, we contend that the relationship between corruption experience and voting radical right is essentially mediated by citizens' trust in public officials and political institutions. The empirical results support this argument, showing that exposure to corruption diminishes political trust which in turn leads to a higher propensity to vote for a radical right party. Additionally, we provide insight about country-specific factors that moderate the individual-level relationships. We find that the eroding effect of corruption experience on trust in public officials is less severe in contexts of low institutional quality. Moreover, radical right parties are particularly able to mobilize support from distrusting voters when in opposition.


Introduction
Right-wing radicalism is an established political and societal phenomenon in contemporary Europe. Parties of the "New Radical Right" (Kitschelt 1995) have gained political influence in several Western and Eastern European countries. Since its beginning in the 1980s, this third wave of the radical right has attracted increasing scientific interest, which has resulted in numerous theoretical models and empirical findings. There is some consensus in the literature that ethno-nationalism is a core characteristic of radical right parties (Mudde 2007;Rydgren 2007). Another key feature of this party family is its populist anti-establishment strategy emphasizing an "ideology that considers society to be ultimately separated into two homogeneous and antagonistic groups, 'the pure people' versus 'the corrupt elite'" (Mudde 2004, 543). However, while many authors examine the first characteristic and its implications for voting behavior that is driven by ethno-nationalist attitudes, the populist features of radical right parties and their electoral consequences have received far less scholarly attention.
In this study, we focus on the populist appeal of radical right parties and explore how individual experience with political corruption affects the probability to vote for a radical right party. We argue that voters take their experiences with political corruption to the ballots and vote accordingly. Since contrasting "the pure people" with "the corrupt elite" is a prominent rhetorical figure of radical right parties and their proponents (Caiani and della Porta 2011;Caiani, della Porta, and Wagemann 2012), citizens who are confronted with power abuse and malfeasance of public officials should be attracted to the pledges of these parties rather strongly. As mediating mechanism we contend that corruption experience essentially matters indirectly by eroding citizens' trust in public officials and political institutions. Accordingly, voters who are politically disenchanted and distrusting are more likely to opt for a radical political change, which parties on the far right tend to offer.
This study goes beyond previous research in several respects. First, prior studies have either focused on the link between political corruption and political trust or between political trust and (radical right) voting. We extend this perspective by connecting both linkages theoretically and test hypotheses regarding both direct and indirect effects. Second, previous research has mainly studied the effects of perceived political corruption (Mishler and Rose 2001;Anderson and Tverdova 2003;Slomczynski and Shabad 2012). Instead, we use data from a rotating module of the European Social Survey (ESS) 2004/2005 that enables us to study the effect of self-reported experienced corruption in dealing with public officials. This information is a comparatively well-suited indicator for examining causality since its reference to behavior makes it less contaminated with endogeneity than perception indicators. Third, we examine 12 Western and Eastern European countries in comparative perspective. In doing so, we investigate the role of contextual characteristics to explain country-specific differences regarding the relationship between experienced political corruption, political trust and radical right support.

Political Corruption and Radical Right Party Support
Explanations regarding electoral success of radical right parties have focused on both the demand side of the political process (i.e., characteristics of voters) and its supply side (e.g., features of the party system or political opportunity structures; see Rydgren 2007;Arzheimer 2009). Concerning the constituencies of radical right parties, individual voting behavior is mainly explained by the following three approaches: first, perceived immigrant threat and ethnic competition (Van der Brug, Fennema, and Tillie 2000;Lubbers, Gijsberts, and Scheepers 2002); second, relative economic deprivation and the "modernization loser thesis" (Betz 1994;Mughan, Bean, and McAlliser 2003); and third, political protest and discontent (Norris 2005). The latter approach largely focuses on how (lack of) political trust influences citizens' preference for a radical right party (Billiet and de Witte 1995;Norris 2005;Hooghe, Marien, and Pauwels 2011). The theoretical scope of our study builds upon this research by assuming that citizens' political discontent will make them particularly receptive to populist promises of radical right parties. However, behavioral sources of low levels of political trust have been largely overlooked so far. In this sense, we go beyond previous research by investigating the role of individual involvement in political corruption as experiential source.
Political corruption is defined as an illegal abuse of public office and authority for personal gain or the benefit of third parties. Taking place behind closed doors, political corruption undermines democratic principles, such as impartiality, equality and transparency; compromises the accountability of political actors and public officials; and fosters citizens' disenchantment and discontent (Anderson and Tverdova 2003). Several studies examine effects of (perceived) political corruption on voting behavior (Welch and Hibbing 1997;Krause and Méndez 2009;Chang, Golden, and Hill 2010;Costas-Pérez, Solé-Olléa, and Sorribas-Navarro 2012;Slomczynski and Shabad 2012). For the most part, incumbent politicians and parties lose votes when being perceived corrupt or actually have been charged with corruption. However, effects are contingent on transparency and media exposure of the corruptive acts (Chang, Golden, and Hill 2010;Costas-Pérez, Solé-Olléa, and Sorribas-Navarro 2012). More than that, several studies suggest that (perceived) corruption tends to lower electoral turnout (Caillier 2010;Stockemer, LaMontagne, and Scruggs 2013). Regarding the preference for radical right parties, however, there is no research known to us that examines the link between political corruption and citizens' preferences for the radical right. This is all the more surprising, given that a constituting element of radical right parties is their agitation against "the elite" -which is recurrently accused of being corrupt (Mudde 2004).
On the supply side, recent studies (Caiani and della Porta 2011;Caiani, della Porta, and Wagemann 2012) apply social movement theory and frame analysis in order to investigate the extent to which radical right parties make use of populist anti-establishment strategies. Using data from a broad content analysis of written statements from these parties, they find that ethno-nationalist and anti-establishment frames are indeed characteristic of the far right in several European countries. Analogous to Mudde (2004), the populist anti-establishment element of radical right discourses is also here reflected in contrasting "the" exploited and overlooked "people" with a generalized class of corrupt and selfish politicians.
There is an ongoing discussion as to which extent radical left parties use populist anti-establishment strategies (see Heinisch 2003, 93;Decker 2008). Yet the phenomenon of radical left-wing parties and left-wing populism has received only limited attention in the literature (but see March and Mudde 2005) and our empirical analysis provides no support that exposure to political corruption is directly or indirectly related to voting for a populist radical left party. 1 We therefore restrict the scope of our argument to parties and populist strategies on the far right.
At the individual level, we argue that individuals who have been involved in a corruptive act in dealing with public officials are more responsive to populist anti-establishment pledges radical right parties make. This claim finds support in arguments derived from research on social movements (Snow et al. 1998;Della Porta and Diani 1999). Taking part in an illegal and illegitimate activity such as corruption most likely appears as an external shock that disrupts everyday life routines and gives rise to feelings of frustration, insecurity and helplessness. The antagonistic political frames offered by radical right parties support individuals in attributing their feelings of confusion, insecurity and/or guilt externally toward political actors and institutions. Citizens affected by these frames are likely to blame not only specific corrupt officials or politicians, but also political elites and institutions in general, such as the government or the legal system. Casting a ballot for a radical right party might compensate for the accrued discontent, thereby contributing to re-establish a zone of privacy and control (Snow et al. 1998), which has been violated through the corruption experience. In a similar vein, Gingerich (2009) finds that exposure to corruption increases the willingness to voice political protest.

Political Trust as Mediating Variable
Studies on corruption and political trust suggest that perceived political corruption is associated with lower levels of political trust 2 (Della Porta 2000; Mishler and Rose 2001;Anderson and Tverdova 2003). Seligson (2002) demonstrates that also individual exposure to corruption has a negative effect on political trust. Regarding the underlying mechanism, Della Porta (2000) argues that corruption hampers the performance and allocative functioning of political institutions and reduces citizens' belief that the government is capable to address their demands. Political corruption may also serve as a signal concerning the moral standard of the institutional and social fabric a person lives in. Both "resource mismanagement" and "moral deviance" of political elites are thus likely to diminish citizens' political trust and erode the legitimacy of political authorities and institutions.
For political trust as the explanatory variable, numerous studies investigate the link between political discontent and voting behavior, including radical right voting (Billiet and de Witte 1995;Miller and Listhaug 1990;Norris 2005;Hooghe, Marien, and Pauwels 2011). In most cases, political (dis)trust turned out to have substantive impact on the probability to vote for a radical right party. Theoretically, this link is explained by accounts of protest voting, which refer to citizens' political discontent as a main motivation for radical right voting (Betz 1994;Norris 2005). Protest voters aim to teach established political elites a lesson by voting for the "outcast" of the political spectrum.
Integrating these linkages, we expect political trust to act as a crucial mediating mechanism that translates experience with political corruption into electoral support for radical right parties (see also Caillier 2010). Experience with political corruption first is assumed to decrease political trust and individuals who distrust political institutions are then more likely to vote for a radical right party.
In accordance with previous research, we consider political trust as a multidimensional phenomenon that refers to different objects. Rothstein and Stolle (2008) distinguish between trust in political actors on the representational side of the political process, such as government, politicians and parties; and political institutions on the implementation side, such as parliament, police, legal system and public officials. The former institutions are partisan and mainly responsive to their specific electorate. In contrast, implementation or "order" institutions are considered to be essentially impartial (Rothstein and Teorell 2008). They reflect fundamental features of a democratic system and have been regarded as crucial for its stability and functioning (Easton 1965). As the study's focus is on actual corruption practices in dealing with public officials, we expect trust-diminishing effects particularly for this group of political actors. Moreover, we expect that a trust-eroding effect further extends to political institutions more generally (e.g., the legal system) and thus consider institutional trust as additional mediating mechanism. 3

Explaining Country Differences
Scholars have investigated political, economic and cultural factors in order to explain country-level differences in radical right voting (see Rydgren 2007). Overall, these studies find that voters are more likely to support radical right parties in contexts where the mainstream party on the right tends to converge with the ideological positions of the mainstream left party (Carter 2005), and when radical right parties are able to utilize discursive opportunities on immigration and anti-establishment issues (Koopmans et al. 2005; Caiani and della Porta 2011). Moreover, high unemployment levels in conjunction with high immigrant rates were found to promote radical right support in some cases (Golder 2003); however, the relationship is further conditioned by welfare state regimes (Norris 2005;Arzheimer 2009). In Eastern European countries, radical right parties gain substantive electoral support subsequent to political success of ethno-liberal parties (Bustikova 2014). For the purposes of this study, we are interested in country characteristics that moderate the relationship between political corruption, political trust and radical right voting. We therefore focus on institutional and political conditions rather than economic or cultural factors.
First, we expect that the institutional context influences the degree to which corruption experience affects political trust. More specifically, we contend that involvement in corruptive practices should diminish trust less severely in contexts of poor institutional quality. A theoretical argument supporting this view is related to the fact that countries with high institutional quality are more likely to have implemented anti-corruption and compliance laws, which induce a more effective punishment of deviations (Treisman 2007). More than that, social norms regarding corruptive practices and bribing differ across institutional contexts. In countries with high institutional quality, political corruption and bribing are far less accepted than in countries of low institutional quality (Fisman and Miguel 2007;Batory 2012). In this vein, people accustomed to poor institutional performance or even pocketbook corruption of public officials will lose trust when exposed to corruption less severely than those who expect compliance and rule of law. Results from a recent experimental study (Klašnja and Tucker 2013) support this claim by showing that respondents from a high corruption country (Moldavia) react less negatively toward corruptive acts of local public authorities in terms of political support than those from a low corruption country (Sweden).
Second, we focus on the political status of radical right parties, namely whether they are in opposition or part of the incumbent government. The main argument here is that radical right parties attract support from politically disenchanted voters only if they operate out of opposition. By contrast, radical right parties that are part of the incumbent government are more likely to be perceived as collaborators of the established elites. Following a discursive opportunity structure framework (Koopmans et al. 2005;Caiani, della Porta, and Wagemann 2012), only parties in opposition can convincingly use the rhetorical figure of "the pure people" versus "the corrupt elite" and sell the frame of teaching the "top brass" a lesson. Similarly, Heinisch argues that once in government, radical right parties tend to tone down their populist agitation and accommodate to the conventions of mainstream party politics, which, in turn, "contains the danger of alienating core supporters, who can no longer distinguish the previous anti-system opposition from the political establishment" (2003, 102; but see Akkerman and de Lange 2012). Taking up this argument, we expect the mobilizing effect of political distrust on voting radical right to be stronger in contexts where radical right parties are in opposition.

Data and Variables
In order to test the hypothesized relationships empirically, we use data from different sources. All individual-level data come from the second round of the ESS conducted in 2004/2005. This data set is particularly suited because it contains items on voting behavior, experience with political corruption from a special rotating module, trust in different political objects and relevant control variables.
Concerning our dependent variable radical right vote, we use the following question in order to identify whether a respondent casted a ballot in the last national election: "Some people don't vote nowadays for one reason or another. Did you vote in the last [country] national election in [month/year]?" In a next step, we discern a person's radical right voting by referring to the question "Which party did you vote for in that election?" Voters of radical right parties were coded as 1, while all others were coded as 0. 4 Party coding was driven by previous conceptual classifications (Lubbers, Gijsberts, and Scheepers 2002;Mudde 2007;Arzheimer 2009). Unfortunately, the ESS data contain separate categories for radical right parties only in a limited number of countries. Table 1 shows the remaining countries in the sample, selected radical right parties, number and proportions of the radical right voters in the sample as well as actual electoral success in the latest national election prior to the survey. In several countries, reported electoral support for radical right parties is lower than the actual vote share in the preceding election. Under-representation of radical right voters due to social desirability is a constant feature of electoral surveys, either because supporters of this party family more often refuse to be interviewed on political issues in general or hide their true preferences when being interviewed because they perceive their opinion as socially undesirable (Breen 2000).
The main independent variable experienced political corruption is measured by the question "How often, if ever, have each of these things happened to you in the last five years? A public official asked you for a favour or a bribe in return for a service" with the response categories 1 "never" to 5 "5 times or more." We recoded this variable to 0 (never) and 1 (once or more often), although this choice does not alter the substantive meaning of the empirical results. Only a limited number of respondents reported experience with corruption. In our sample, about 4.1% of the respondents reported to have been asked to pay a bribe. This proportion varies considerably across countries with values between 1.3% (Switzerland) and 12.2% (Poland). 5 It is important to note that although referring to actual experience in dealing with public officials, the question did not ask whether a bribe was actually paid or not. Thereby potential corruptive acts are covered as well. Moreover, it gives respondents opportunity to report their experience even if they perceive their corruptive involvement as social undesirable or even fear legal consequences. However, this item may also leave room for interpreting non-corruptive behavior as corruptive. In order to test its convergent validity, we compare the country aggregates of the ESS corruption measure with those from Transparency International's Global Corruption Barometer (GCB), which asks about actual payment of a bribe within the last year. We use the country averages of the GBC data of 2004 and 2005 that cover 10 of the countries in our sample. A bivariate plot of the indicators appears in Figure A1 in the online appendix (see Supplemental material section). Both variables correspond to a high degree (r ¼ .86; p , .01), which provides some confidence that the ESS corruption item is sufficiently valid in order to indicate actual involvement in corruptive practices.
The mediating variables are trust in public officials and trust in political institutions. As indicator for trust in public officials, we use a question from the rotating module: "How much would you trust the following groups to deal honestly with people like you? Public officials", ranging from 1 "distrust a lot" to 5 "trust a lot" (recoded to 0 "low trust" to 4 "high trust"). Trust in political institutions entails trust in the country's parliament, the legal system and the police (each of these items range from 0 "no trust at all" to 10 "complete trust").
We also include several control variables that turned out to be influential in previous studies. Perceived immigrant threat is one of the most important explanatory variables; besides, male respondents and persons who are less educated, unemployed and older are particularly likely to vote for radical right parties (Lubbers, Gijsberts, and Scheepers 2002). In contrast, persons with migration background are expected to have a lower probability to vote for these parties. We use the following three statements as indicators of perceived immigrant threat: immigration is bad or good for one's country's economy; one's country's cultural life is undermined or enriched by immigrants; and immigrants make one's country a worse or better place to live. All items range from 0 to 10, with 10 indicating high threat. As further controls, we use dummy variables indicating respondents' migration status, gender and employment status as well as continuous variables for age and full-time education in years.
For testing the role of the country context, we employ rule of law as indicator of institutional quality (Worldwide Governance Indicators [WGI]; Kaufmann, Kraay, and Mastruzzi 2011). 6 We furthermore build a dummy variable that indicates whether or not a radical right party had been part of the government or supported it prior to the survey. This was the case in Austria, Denmark, Italy and Switzerland. 7

Method
The empirical analysis is conducted in two steps. First, individual-level relationships are estimated within a structural equation framework. Structural equation modeling is a very flexible method that allows for including latent variables accounting for measurement error as well as complex variable relationships, such as mediation (Kline 2011). In order to estimate direct and indirect effects of corruption on radical right vote, we estimate two separate models: one that includes trust in public officials and another with trust in political institutions (latent factor variable) as mediator. Since respondents of the survey are nested within country contexts, we use country fixed effects in order to account for country differences. The structural equation models are estimated in Mplus 7.11 (Muthén and Muthén 2012) using a weighted least squares means and variances adjusted estimator (WLSMV), which is robust to non-normally distributed data. 8 By default, standard errors for the indirect effects are here calculated via the Delta method (MacKinnon 2008). It is important to note that the standard errors differ only marginally when obtained via parametric bootstrapping (1000 samples).
Second, for testing the contextual hypotheses, we employ multilevel regression models (Snijders and Bosker 2012) in Stata 13 that include interaction terms between individual-level predictors and the country variable. Because of the low number of cases at the country level, we need to interpret the results carefully. We therefore estimate the interaction effect models also in a regression framework with country fixed effects and discuss significant findings against the backdrop of the country-specific findings. Figure 1 displays the results of the core variable relationships obtained from the structural equation models with country fixed effects. A table with all parameter estimates of the models appears in Table A1 in the online appendix (see Supplemental material section). 9 Model 1 includes trust in public officials as mediating variable. The results show that corruption experience has no significant direct impact on the probability to vote for a radical right party. The effect of trust in public officials on radical right support is, as expected, negative and significant. This also means that people who express low trust in public officials are more likely to vote for a radical right party. In order to get a sense of the effect size, we calculated country-specific predicted probabilities for minimum and maximum values of political trust, holding all other using the WLSMV estimator with a probit link for categorical outcomes (Mplus v. 7.11). Rectangles refer to manifest variables and circles to latent factor variables. Entries are unstandardized regression coefficients. Models include controls for perceived immigrant threat, age, gender, migration status, education and employment status. Italicized coefficient and standard error display indirect effect of corruption experience on radical right vote via trust in public officials/political institutions. * p , .05, * * p , .01, * * * p , .001 (two-tailed).

Empirical Findings
predictors constant at their country-specific mean (see Table A2 in the online appendix, see Supplemental material section). 10 We find that moving from the maximum level of trust (4) in public officials to its corresponding minimum (0) usually doubles the probability of casting a ballot for a radical right party. Turning to trust in public officials as dependent variable, we find that exposure to corruption significantly decreases trust by approximately .48 units. Most importantly, we find a significant indirect effect of corruption experience on radical right voting via trust in public officials (b ¼ .04, SE ¼ .01). As expected, people who have experienced corruption in dealing with public officials tend to lose confidence in this group of political actors which in turn makes them more likely to vote for a radical right party.
Model 2 shows the estimates when considering trust in political institutions as the meditating variable. Again, we find no statistically significant direct effect of corruption experience on radical right voting. Similar to Model 1, trust in political institutions is negatively and significantly associated with radical right support. The respective probability ratios (Table A2) suggest that respondents showing a minimum (0) level of trust are about 2 (Switzerland), 3 (Belgium, Denmark, Italy, Norway), 4 (Austria, France, Netherlands, Poland, Slovenia), 5 (Germany) or even 7 (Greece) times more likely to vote for a radical right party than fellow citizens who reported maximum (10) institutional trust. Again, corruption experience substantively and significantly decreases trust. The estimated indirect effect indicates that experienced political corruption significantly translates into a higher probability to vote for a radical right party via trust in political institutions (b ¼ .06, SE ¼ .01).
Regarding the control variables, we find that in accordance with previous research (Lubbers, Gijsberts, and Scheepers 2002) perceived immigrant threat plays a major role in explaining individual differences in radical right support (Model 1: b ¼ .25, SE ¼ .01; Model 2: b ¼ .22, SE ¼ .01). Moreover, the impact of the other controls is largely in line with previous studies suggesting that younger, male and less educated persons have a higher probability to vote for the far right, while immigration status and employment situation have no significant impact.
In order to ensure that individual-level results are not driven by single countries or outliers, we re-estimated both models under a one-by-one exclusion of single countries. We have also considered excluding the two countries with the highest proportions of radical right voters (Italy and Switzerland) or corruption experience (Greece and Poland) at the same time. Moreover, we re-estimated the models without Poland and Slovenia to show that the findings do not depend upon the two post-communist countries of the sample. Throughout all of these alternative specifications, the substantive findings of the analysis remained robust.
In a next step, we test whether specific country-level characteristics moderate the direct and indirect relationships between corruption and radical right vote. 11 Estimates of the interaction terms and relevant constitutive coefficients are shown in Table 2. We first inspect whether the effect of corruption experience is contingent on institutional quality (rule of law). As the estimates in Model 3 show, the effect of corruption on radical right voting is not significantly moderated by rule of law.
Models 4 and 5 test whether the relationship between corruption and trust in public officials or trust in political institutions depends on levels of institutional quality. While the interaction term of corruption by rule of law is negative in both models, it is statistically significant only in Model 4. The marginal effects plot displayed in Figure 2 shows the effect of corruption on trust in public officials over different levels of rule of law. As theoretically expected, the effect of corruption is increasingly negative as the quality of institutions improves. Citizens who live under high institutional quality expect compliance and impartiality from public officials and react more sensitive toward violations of these principles compared to those living in contexts of low-quality institutions. However, while this holds true for trust in public officials, the negative effect of corruption experience on trust in political institutions is not significantly moderated by the institutional quality of the context.
To illustrate the robustness of these findings, we re-estimated all models in a regression framework with country fixed effects (not shown). The results from these models are virtually congruent to the multilevel regression estimates. Moreover, we used a two-step procedure by plotting the country-specific coefficients of corruption on trust in public officials against the rule of law indicator. The results are displayed in Figure A2 in the online appendix (see Supplemental material section). The relationship suggests that indeed the (negative) effect of corruption on trust in public officials is stronger in contexts of high institutional quality (r ¼ 2.52, p , .10, n ¼ 12). Besides institutional quality, we expect that radical right parties in opposition are more successful to attract politically disenchanted and distrusting voters than those that are part of the government. Model 6 tests whether the effect of individuals' experience with political corruption on radical right support is contingent on whether this party is in opposition or in government. The estimate is not significant and thus provides no empirical support that a direct effect of corruption experience depends on the political status of radical right parties. Note: Unstandardized coefficients obtained from multilevel regressions (maximum likelihood). Logit link function is used for radical right vote as outcome variable. All models control for perceived immigrant threat, migration status, gender, education, age and unemployment. The variables trust in political institutions and perceived immigrant threat are included as mean indices. * p , .05 (two-tailed). * * p , .01 (two-tailed). * * * p , .001 (two-tailed).
Models 7 and 8 examine whether the effect of trust in political institutions or public officials on radical right support depends on the political status of the radical right party. The non-significant interaction coefficient for Model 7 indicates that the relationship between trust in public officials and radical right support is not systematically conditioned by the political status of radical right parties. The interaction effect in Model 8 is significant and positive. Figure 3 presents the corresponding average marginal effects. The graph shows that trust in political institutions is systematically associated with radical right support only in countries in which the far right is in opposition. In countries where the far right supports the government, this relationship is less substantive and statistically not significant. In other words, distrust in political institutions is effectively translated into a higher probability to vote for a radical right party only in countries where radical right parties are in opposition.
Again, a re-estimation of the interactive relationships within logit regression models including country fixed effects yield very similar results. Moreover, by looking at country-specific results (not reported), we indeed find small and non-significant effects of trust in political institutions on radical right voting in three out of four countries (Austria, Italy and Switzerland) in which the radical right was part of the government. However, in the remaining eight countries where the radical right parties are in opposition, we find significant associations in Belgium, Germany, Greece, the Netherlands and Norway.

Discussion and Conclusion
The main aim of this study has been to explore the link between political corruption and voting for radical right parties in Europe both theoretically and empirically. Our initial argument was that populist radical right parties use an anti-establishment strategy in order to attract voters who are politically disenchanted and frustrated about (actual or perceived) abuses of political and/or bureaucratic power. Consequently, we investigated individual-level mechanisms between exposure to political corruption, political trust and electoral support of radical right parties. As expected, experience with political corruption in dealing with public officials appears to lower trust in this group of political actors. However, a trust decreasing effect is not limited to public officials. This sort of experience tends to diminish also trust in fundamental political institutions. Scholars have pointed out that citizens' trust in basic political institutions is particularly important for the viability of a democratic political system and that this type of trust is not easily restored (Easton 1965;Dalton 2004). In this sense, corrupt public officials gamble away important prerequisites of citizens' support of the political system. Regarding the role of political trust, we find that low levels of trust in public officials and political institutions are associated with a higher propensity to vote for a radical right party. This suggests that politically disenchanted citizens' are more likely to opt for fundamental political change and be attracted to populist anti-establishment frames radical right parties offer. Ultimately, the findings reveal that political corruption substantively increases radical right support through diminishing political trust.
Moving beyond the individual-level logic, we also examined relevant country characteristics that potentially condition the relationship between corruption experience, political trust and radical right party support. While we observe that the two mediating trust variables perform similarly at the individual level, their contextual dependency differs to some extent. We find that the relationship between corruption and trust in public officials is moderated by institutional quality. Political corruption translates into lower trust in public officials especially in contexts of high institutional quality, while this association is much weaker in contexts where institutional quality is low. This relationship might be driven by a less pronounced legal and normative framework regarding compliance in contexts of poor institutional quality or simply because people are more accustomed to pocketbook corruption of officials under these circumstances. However, we find no such interactive relationship for trust in political institutions. This indicates that corruption experience generally translates into lower trust in basic political institutions, regardless of the institutional context. Moreover, the political status of radical right parties moderates the effect of trust in political institutions on radical right support, while we find no corresponding effect for trust in public officials. In opposition, radical right parties are able to effectively utilize populist frames in order to attract votes from people who distrust basic political institutions. However, the political status of the radical right parties seems to be a less important criterion for citizens' voting decision if their distrust is primarily related to public officials.
In summary, the results strengthen the status of variables of political discontent with regard to the radical right voting literature. More than that, this study heightens the status of citizens' experience with political corruption as a crucial behavioral source explaining low levels of political trust. In doing so, it sheds light on a previously overlooked way corruption may harm democracy, by encouraging electoral support for radical anti-system parties. This yields serious implications for policymakers and those concerned with the quality of democracy, in particular in transition states where the implications could be very destabilizing. In positive terms, the findings suggest that politicians and public officials may effectively reduce citizens' political discontent by keeping their own houses in order concerning standards of transparency and compliance.
Still, some limitations need to be addressed. Concerning the relative importance of corruption and political trust, we find that while political trust (and indirectly political corruption) has a significant impact on radical right support, it rarely outperforms perceived immigrant threat as a predictor. Moreover, it is quite possible that the relationship between political trust and radical right support varies over time (Werts, Scheepers, and Lubbers 2013) and is affected by further political, economic and social contexts. While we identify a novel substantive individual-level mechanism with regard to radical right voting, we only consider a small fraction of potentially important contextual factors. An important direction for future research is to consider additional contextual conditions, including the role of party-specific mobilization strategies or ways populist frames are presented, transmitted and perceived by the electorate. Clearly, the collection of comparative high-quality data about the discourses of radical right parties as already commenced by Caiani et al. (2012) will serve as a crucial prerequisite in this regard. competition between different motivations in actual voting behavior. Therefore, non-voters were excluded from the sample. In doing so, the reference category is slightly smaller, which might increase the statistical power of our test statistics. A re-analysis including non-voters reveals very similar findings. 5. Besides this "passive" corruption indicator, the ESS contains a question whether respondents have offered a bribe themselves within the last five years. Only 1.3% of the respondents reported to have been actively involved in corruption. Since we expect this question to be plagued by social desirability bias, we focus on the role of passive corruption. Again, a re-estimation of the empirical models using a composite indicator of passive and active corruption leads to virtually the same empirical findings. 6. Incorporating alternate indicators of institutional quality (e.g., WGI Government Effectiveness, WGI Control of Corruption or International Country Risk Guide Quality of Government) leads to substantively similar results. 7. In the case of Denmark, the Danish People's Party was not part of the government but provided constant parliamentary support. Considering Denmark as "not in government" does not alter the substantive meaning of the results. 8. The results are not sensitive to this choice. Using a maximum-likelihood estimator with robust standard errors (Huber-White) and a probit link for the radical right vote equation leads to similar results. 9. The reported RMSEA values indicate that our models have a "close fit" to the data (RMSEA , .05; Hu and Bentler 1999). Since included country dummies are additional parameters which explain less variance than a regular explanatory variable, the goodness of fit is comparatively higher (RMSEA , .05; CFI . .95) in models without country fixed effects. 10. The probit model expresses the probability of y given as P(y¼1|x)¼F(2t+b×x), where y¼1 represents voting for a radical right party, b is the raw trust coefficient and t stands for the probit threshold (not reported; see Muthén and Muthén 2012, 492, for details). For the computation of the countryspecific probability values, parameters of the dummy variables are included as country-specific intercepts. 11. An analysis of the variance components (intraclass correlations) indicates that about 21% of the variance of radical right voting can be attributed to country differences. In the case of trust in political institutions, 19% of the variance is accounted for by the country level, while for trust in public officials this proportion is about 6%.