Voting Women, Protesting Men: A Multilevel Analysis of Corruption, Gender, and Political Participation

Abstract

en

Previous studies have been unable to establish the link between corruption perceptions and political participation. This is partly due to a disregard of different types of political participation, ignoring gender differences in how corruption perceptions affect political participation, and overlooking the importance of context. We therefore here examine gender differences in the links between corruption perceptions and three types of political participation: voting, institutionalized participation between elections, and noninstitutionalized participation between elections. We also examine how the context in the form of the national level of corruption affects these linkages. The data come from International Social Survey Program Citizenship II and includes 31 democracies, analyzed with multilevel regression models. Our results show that women become more likely to vote when faced with corruption, whereas men become more likely to engage in elite-challenging forms of participation when faced with corruption while women remain unaffected.

Related Articles

Caillier, James. 2010. “Citizen Trust, Political Corruption, and Voting Behavior: Connecting the Dots.” Politics & Policy 38 (5): 1015-1035. https://doi.org/10.1111/j.1747-1346.2010.00267

Lagunes, Paul F. 2012. “Corruption's Challenge to Democracy: A Review of the Issues.” Politics & Policy 40 (5): 802-826. https://doi.org/10.1111/j.1747-1346.2012.00384

McNally, Darragh. 2016. “Norms, Corruption, and Voting for Berlusconi.” Politics & Policy 44 (5): 976-1008. https://doi.org/10.1111/polp.12173

Abstract

es

Mujeres votantes, hombres que protestan: un análisis multinivel de la corrupción, el género y la participación política

Los estudios anteriores no han podido establecer el vínculo entre las percepciones de corrupción y la participación política. Esto se debe en parte a que se ignoran los diferentes tipos de participación política, se ignoran las diferencias de género en la forma en que las percepciones de corrupción afectan la participación política y se pasa por alto la importancia del contexto. Por lo tanto, aquí examinamos las diferencias de género en los vínculos entre las percepciones de corrupción y tres tipos de participación política: Voto, participación institucionalizada entre elecciones y participación no institucionalizada entre elecciones. También examinamos cómo el contexto en la forma del nivel nacional de corrupción afecta estos vínculos. Los datos provienen de ISSP Citizenship II e incluyen 31 democracias, analizadas con modelos de regresión multinivel. Nuestros resultados muestran que las mujeres tienen más probabilidades de votar cuando se enfrentan a la corrupción, mientras que los hombres se vuelven más propensos a participar en formas de participación que desafían a las élites cuando se enfrentan a la corrupción, mientras que las mujeres no se ven afectadas.

Abstract

zh

投票的女性、抗议的男性：关于腐败、性别和政治参与的多层分析

以往研究一直无法在腐败感知和政治参与之间建立联系。这部分归因于忽视不同类型的政治参与，在腐败感知如何影响政治参与一事上忽视性别差异，以及忽视情境的重要性。因此，我们分析了腐败感知和三种政治参与之间的联系所产生的性别差异，这三种政治参与分别为：投票、选举间隔期间的制度化参与、选举间隔期间的非制度化参与。我们还分析了以国家层面的腐败为形式的情境如何影响这些联系。数据源自国际社会调查项目公民权II （ISSP Citizenship II），包括31个民主国家，使用多层回归模型对数据进行分析。我们的结果显示，当面对腐败时，女性变得更有可能参与投票，而男性则在面对腐败时更有可能进行以挑战精英为形式的参与，但女性则未受到影响。

Corruption is often argued to have demobilizing effects on the political participation of ordinary citizens when it comes to voting (Chong et al. 2015; Dahlberg and Solevid 2016; Kostadinova 2009; Sundström and Stockemer 2015). However, scholars contend that it has the potential to mobilize citizens against corrupt power holders and that “the effects of corruption on voting are more complex than merely to say that malfeasance depresses voter turnout” (Kostadinova 2009, 707). This mobilizing potential may mainly concern political participation other than voting in elections (Bazurli and Portos 2019; Navot and Beeri 2017). This shows that much about how corruption affects the tendency to become politically active is still unknown.

We here examine a potential piece in this puzzle by examining how the links between corruption perceptions and political participation differ for men and women. Previous studies show that the effect of perceptions of corruption differ across social groups (Agerberg 2019; Bazurli and Portos 2019), but the differences between men and women remain unclear. It has been a cause for concern that previous research has found a consistent gender gap in political participation (Dalton 2017; Verba, Schlozman, and Brady 1995). It has traditionally been the case that women were less politically active, although some studies suggest that gender differences are evening out. Any remaining gender discrepancies in participation are explained by discrepancies between men and women in how demographic and attitudinal differences affect political behavior (Burns, Schlozman, and Verba 2001; Coffé and Bolzendahl 2010).

It is important to examine whether systematic differences exist in this regard since it has direct implications for how any worries over corruption are brought to the attention of political decision makers. Furthermore, there is reason to believe that there are important gender differences in how corruption perceptions affect political behavior since research suggests that men and women hold different views of corruption, with men being less judgmental and more permissive of corrupt behavior (Kravtsova, Oshchepkov, and Welzel 2017; Letki 2006; Swamy et al. 2001), which is likely to affect political behavior as well.

In the present article, we study differences between men and women in how corruption perceptions are linked to three different forms of political participation: voting, and institutionalized and noninstitutionalized activities. There are valid reasons to believe that important differences exist in how men and women transmit their concerns over corruption to political decision makers. We also study whether these gender differences in the relationships between corruption and political participation vary depending on the country-level indicator of corruption, as contextual differences are likely to play an important role.

We examine these research questions with multilevel regression using data from the International Social Survey Program (ISSP) Citizenship II and including 31 democratic countries (ISSP Research Group 2016). The results suggest that men and women react differently when faced with high levels of corruption in public services. While women are more likely to engage in institutionalized political participation, men are more likely to turn to noninstitutionalized forms of political participation.

The article proceeds as follows. First, we discuss the central concepts of the study to develop hypotheses on the associations between corruption perceptions, gender, and political participation. Following this, we present the data and variables before turning to the empirical analyses themselves. Finally, we discuss the conclusions drawn from these results.

Corruption, Gender, and Political Participation in Democracies

Corruption can be understood in a multitude of different ways as a problem or a “syndrome” across various types of societies as a consequence of differences in the capacity of public institutions and the competitiveness of political and economic markets (Johnston 2005). Hence, bureaucratic or petty corruption and related phenomena such as clientelism are more common in relatively poor and newly democratized countries, while advanced democracies are characterized by more ambiguous “structural” or “legal” forms of corruption and less outright bribery or embezzlement (Graycar and Monaghan 2015; Johnston 2005; Kaufmann and Vicente 2011).

People in poorer countries often encounter corruption on a daily basis and are forced to pay bribes to receive services or goods to which they should be entitled as citizens. They may therefore become accustomed to such practices and even accept them as unavoidable. In wealthy countries, meanwhile, direct personal experiences of corruption may be rarer. Hence, people are likely to judge apparent unethical or unfair “shady” behavior brought to their attention as a sign of corruption, especially if they already distrust elites (Van de Walle 2008; Wroe, Allen, and Birch 2013). Graycar and Monaghan (2015) hence argue that the systemic and victimizing nature of corruption in poor countries means that some forms of corruption viewed as appalling in wealthy countries—such as influence peddling or favoritism—might be seen as a lesser evil or even something natural in less favorable circumstances. It is therefore necessary to employ a broad definition of corruption to compare countries where experiences of corruption vary. Here, following Johnston (2005), we choose to define corruption as the abuse of public roles or resources for private benefit. “Private benefit” refers here not just to individual benefits, but also to benefits for one's own in-group or clique, be it family or political party.

Corruption is often understood as irreconcilable with democracy and democratic values. Warren (2006) conceptualizes political corruption in democracies as “duplicitous exclusion,” by which he means covert actions that unjustifiably exclude certain citizens from collective decision-making processes where they have the right to express their will. This constitutes a “violation of the [democratic] norm of empowered inclusion of all affected” (Warren 2006, 804). It is therefore easy to see why perceptions of corruption are likely to affect ordinary citizens to participate in political activities. In recent years, several studies have explored the link between corruption perceptions and various forms of political participation (Bauhr and Grimes 2014; Bonifácio and Paulino 2015; Dahlberg and Solevid 2016; Kostadinova and Kmetty 2019; Miles 2015; Sundström and Stockemer 2015). Nevertheless, there is no agreement on how they are linked, especially when it comes to political participation between elections. This is important to understand since political participation, in elections and between elections, constitute the central mechanism through which citizens' concerns are transmitted to the formal political decision makers (Urbinati and Warren 2008). For this reason, it is important to examine whether corruption works to exclude certain groups from representation while enhancing the voice of others. Gender here constitutes a particularly essential social group since there are reasons to believe that corruption affects men and women unequally, as outlined shortly. We first discuss the links between corruption and different forms of political participation, before moving on to discussing the likely gender differences.

To reconcile the diverging interpretations of the link between corruption and participation, it is first necessary to establish a distinction between voting, and institutionalized and noninstitutionalized political participation between elections (Dalton 2008; Norris 2002). While voting may also be considered an institutionalized political activity, it differs from institutionalized activities between elections and is therefore often treated separately in empirical analyses (Bäck and Christensen 2016; Christensen 2013; Hooghe and Marien 2013). Institutionalized political participation is directly linked to the formal political institutions and aim to influence formal political outcomes. Noninstitutionalized activities such as demonstrations take place outside the formal political sphere and are frequently used to voice discontent. The association between corruption perceptions and participation is likely to differ depending on the type of participation under scrutiny.

The bulk of the evidence suggest that such assessments discourage turnout in elections on individual, national and regional levels (Bauhr and Grimes 2014; Chong et al. 2015; Dahlberg and Solevid 2016; Kostadinova 2009; Miles 2015; Sundström and Stockemer 2015). In this view, perceptions of corruption are expected to lead to public resignation (Bauhr and Grimes 2014). Corruption, it is argued, erodes confidence in public authorities and institutions, which in turn weakens their claim to legitimacy (Anderson and Tverdova 2003). A constant and chronic distrust in public authorities may even diminish belief in democracy and lead to a decline in system support (Linde and Erlingsson 2013). Consequently, citizens may refuse to take part in institutionalized participation both in elections and between elections to demonstrate their discontent.

However, citizens may engage in certain activities while they refrain from taking part in others when faced with corruption. The so-called “indignation-hypothesis” postulates that public sector corruption breeds indignation or anger among citizens, which in turn increases the demand for accountability and reform that may be channeled through either institutionalized or noninstitutionalized means of political participation (Bauhr and Grimes 2014). This line of research proceeds from the idea that corruption is universally understood as illegitimate and wrong (Rothstein and Varraich 2017). In the words of Kostadinova (2009, 707), the “outrage toward corrupt politicians may bring people to the polling booths” or, alternatively, to the city square to protest in demonstrations. Clientelism may also have a mobilizing effect since clientelistic practices can mobilize voters by distributing material incentives and personal favors in return for political support (Christensen and Utas 2008; Escaleras, Calcagno, and Shughart 2012; Hicken 2011; Vicente 2014). Such positive associations between corruption perceptions and participation have some empirical support as well. A natural experiment in Senegal showed that corruption perceptions increased the likelihood of voting, a result replicated using the Afrobarometer survey (Inman and Andrews 2015). A municipal-level study from Portugal found a positive association between number of corruption cases per municipality and turnout (Stockemer and Calca 2013).

This shows that perceptions may have a mobilizing potential, especially when it comes to noninstitutionalized forms of political participation that challenge the existing balance of power and are frequently used to express dissatisfaction with the responsiveness of authorities (Christensen 2013). When citizens perceive formal institutions to be corrupt, they are likely to infer that trying to influence their behavior through political participation is futile (Anderson and Tverdova 2003). Concerned citizens instead increasingly turn to protesting when corrupt practices clog the official channels for influencing political decision making. This line of reasoning is supported by Bazurli and Portos (2019) and Navot and Beeri (2017), who find a positive link between corruption perceptions and nonelectoral forms of participation, although they do not distinguish between institutionalized and noninstitutionalized participation, as we do here.

Others reason that the links between corruption perceptions and political participation depend on individual understandings of “abuse of public power” (Navot and Beeri 2017). Citizens who hold more judgmental conceptions of corruption and with less concern for extenuating circumstances are more likely to engage politically when faced with high levels of corruption compared to people with a narrower and permissive understanding of corruption (Navot and Beeri 2017). Hence, individual conceptions of corruption are important factors to consider when examining links between corruption perceptions and political participation. The effects of corruption perceptions may hinge not only on the form of political participation, but also on individual characteristics.

An aspect that may help explain the diverging results in previous literature is that the associations differ between men and women. The link between gender and corruption perceptions has already received attention (Frank, Lambsdorff, and Boehm 2011; Goetz 2007; Stensöta and Wängnerud 2018; Sung 2003), but how this affects differences in political behavior remains unclear. Corruption disproportionately affects women rather than men because they are often dependent on public services such as health care and education because of traditional gender roles (Hossain, Musembi, and Hughes 2010; Sida 2015). Moreover, women are often excluded from networks where corrupt exchanges take place (Bauhr, Charron, and Wängnerud 2019; Goetz 2007).

Studies show systematic gender differences in perceptions of and attitudes toward corruption and other kinds of dishonest behavior (Kravtsova, Oshchepkov, and Welzel 2017; Letki 2006; Swamy et al. 2001). Previous studies of gender gaps in broader sociopolitical attitudes have concluded that the most substantial and consistent gaps are in two areas: social compassion and traditional morality (Eagly and Diekman 2006). The authors argue that this gap is due to gender roles that “arise from the division of labor between men and women and encompass normative processes by which other people convey expectations based on gender” (Eagly and Diekman 2006, 27).

How these differences are associated with political participation remains relatively unexplored. That women are less tolerant of corruption and more vulnerable to negative consequences could mean that they are more likely to take action to combat corrupt practices. This is in line with the study of Navot and Beeri (2017), who find that more judgmental attitudes can further participation, as discussed above. We therefore expect the associations between corruption perceptions and participation to differ between men and women. For institutionalized participation, the negative association is likely to be weaker for women since we expect them to remain more likely to use this channel when faced with corruption. For noninstitutionalized participation, on the other hand, we expect that women become more willing to take part in such elite-challenging participation, and the association between perceived corruption and participation is therefore stronger compared to men.

Hypothesis 1a: The negative association between corruption perceptions and voter turnout is weaker for women.

Hypothesis 1b: The negative association between corruption perceptions and institutionalized participation is weaker for women.

Hypothesis 1c: The positive association between corruption perceptions and noninstitutionalized participation is stronger for women.

But these linkages are likely to be influenced by the context in which they occur. We here examine if the associations between self-perceived corruption and participation depend on the general level of corruption in the country since previous studies suggest this to be the case.

A recent study by Dahlberg and Solevid (2016) shows that while individual perceptions of corruption depress turnout, this only occurs in countries with low-to-medium levels of corruption. The authors argue that this is due to the clientelistic networks that provide highly personal incentives to vote and can be found in countries perceived as highly corrupt (Dahlberg and Solevid 2016, 506-7). Other studies indicate that clientelistic strategies have an important gender dimension in that men are often their principal targets while women tend to be excluded from them (Vicente and Wantchekon 2009; Wantchekon 2003).

Bonifácio and Paulino (2015) and Kostadinova and Kmetty (2019) show that experiences with corruption increase the likelihood of engagement, at least with regard to nonelectoral forms of participation. These findings suggest that the mobilizing effect of corruption perceptions is higher in contexts where petty corruption is more common, and perceptions are more closely related to actual experiences. Moreover, this mobilizing effect could be stronger for women if they tend to be less tolerant of corruption.

All of this suggests that the level of corruption in society is an important moderator for the gender differences in how corruption perceptions affect political participation. The final hypothesis is therefore:

Hypothesis 2: High societal levels of corruption increase the gender-based differences in the association between corruption perceptions and political participation.

In the following, we outline how we examine these hypotheses before moving on to the empirical analyses.

Data, Variables, and Methods

Our research model that guides our empirical analyses is shown in Figure 1.

Figure 1

Research Model

We first establish the associations between individual corruption perception (ICP) and different forms of political participation. However, our main focus is on examining gender differences in these, as indicated by the dotted arrow from Gender to the full arrow between ICP and Political Participation. We also examine how these gender differences hinges on the Country-level Corruption Perception (CCP), as indicated by the dotted arrows from this box.

The individual-level data for our analyses come from ISSP study Citizenship II, covering 31 democracies (ISSP Research Group 2016).11 Russia and Venezuela are excluded since they are not categorized as democracies. Other countries are excluded from the analyses due to missing data. For Hungary, data are missing for all respondents on years of education while information on voting is missing from respondents in Great Britain. Sensitivity analyses revealed that South Africa constitute an outlier in all models. We therefore reran all models including a dummy variable for South Africa, but this did not affect the substantial results. This dataset is ideal for the purposes since it contains questions regarding corruption perceptions and political behaviors and attitudes and is common to use for examining such linkages (Agerberg 2019; Bazurli and Portos 2019), although these contributions do not examine gender aspects. We here present information on key variables while detailed information on question wordings and coding of variables are in the Appendix.

We construct three dependent variables to measure the extent of political participation: voting, institutionalized participation, and noninstitutionalized participation.

Voting

We here rely on a self-reported voting measure in ISSP 2016, where respondents are asked whether they voted in the last parliamentary/Upper House/presidential election. Although this involves a risk of overestimation of the actual turnout (Lahtinen et al. 2019), this is the best available measure for the current purposes and is customary to use for similar research endeavors. We coded respondents who were not eligible to vote at latest elections (N = 3,685) and those who did not answer (N = 1,595) as missing data, meaning the variable is dichotomous (voted Yes/No).

Institutionalized and Noninstitutionalized Participation

The data include the following indicators on institutionalized and noninstitutionalized political participation: contacted a politician, attend political meeting or rally, actively participate in political party, signing a petition, boycotting or deliberately buying products for political, ethical, or environmental reasons, joining an Internet political forum, and taking part in a demonstration. All items were dichotomized to indicate whether the respondent had performed the activity or not. As is customary in the literature on political participation (Bäck and Christensen 2016; Hooghe and Marien 2013), we conducted an exploratory factor analysis (principal component factoring with oblimin rotation) to assess the dimensionality and reduce the complexity of the data. The results are shown in Table 1.

Table 1. Factor Analysis of Participation between Elections

Variable	Factor 1	Factor 2
Sign petition	.7525	−.1321
Boycotting	.7511	−.1145
Internet participation	.4664	.2571
Demonstration	.6416	.2126
Contacting	.4917	.4474
Attend meeting or rally	.5217	.5021
Active in political party	−.0342	.8282
Eigenvalue	2.27	1.28
Proportion variance explained	.33	.18

Note:

• Entries are loadings from Principal component factoring with oblimin rotation.

The results indicate that we can reduce the items to two dimensions: the first dimension taps noninstitutionalized participation, and the latter institutionalized participation. Since some items load on both dimensions (as might be expected since activities such as attend meeting or rally can belong to either dimension), we use the factor loadings to predict two standardized variables that measure each dimension and constitute the other two dependent variables.22 There are some problems with the distribution of the index for institutionalized participation (kurtosis = 14.98, skewness = 3.05), which is caused by several of the activities loading onto this dimension being in infrequent use. To test the consequences, we dichotomized the index (below mean = 0, above mean = 1) and reran the models with substantially similar results. We are therefore convinced that this did not affect the results.

The independent variable of the study is ICP. This is operationalized with a question asking respondents how widespread they think corruption is in the national public sector (How widespread do you think corruption is in the public service in (COUNTRY)?). There are five response alternatives ranging from (1) “Hardly anyone involved to” (5) “Almost everyone is involved.” The variable is recoded to range from 0 to 1 (1 = highest level of perceived corruption).

We include two moderating variables that potentially moderate the relationships between ICP and the three forms of political participation. Gender is a dummy variable where 0 = female and 1 = male. At the macro level, we operationalize CCP with Transparency International's Corruption Perceptions Index (Transparency International 2013). We here use the corruption values from right before collection of survey data to ensure the direction of causality. This is the most widely used indicator of corruption worldwide and has been compiled on a yearly basis since 1995. It uses a scale of 0 to 100, where 0 is highly corrupt and 100 is very clean. We here have reversed and recoded the scale so that it ranges from 0 (very clean) to 1 (highly corrupt).

To ascertain the associations found, we include several control variables that have been known to influence political participation and therefore may confound the results (Bäck and Christensen 2016; Hooghe and Marien 2013; Verba, Schlozman, and Brady 1995). At the individual level, we include standard sociodemographic characteristics age (measured in years but recoded to vary between 0 and 1), education (years of education completed), marital status (married or not), and place of living (urban or rural). We also include several political attitudes that may influence the propensity for political participation since this is customary in the literature on political participation. These include an index measuring internal political efficacy (good understanding of important political issues + most people better informed than I am [reversed]), external political efficacy (no influence on what government does + government does not care what people like me think), and political interest (no interest at all—very interested).

We also include several control variables at the country level to control for contextual differences. While we are unable to include all possible aspects, we chose to control for economical and institutional aspects that previous studies suggest affect political participation (Braun and Hutter 2016; Quaranta 2018; Vráblíková 2014). These include GDP per capita, logged (Bolt et al. 2018; Coppedge et al. 2018), whether a country has compulsory voting or not (Comparative Study of Electoral Systems 2018), effective number of parties (Armingeon et al. 2019; Laakso and Taagepera 1979), and level of democracy (V-dem's polyarchy index, Coppedge et al. 2018).33 The inclusion of these variables entails problems with multicollinearity (VIF for CCP = 4.48, GDP/capita = 3.90, and level of democracy = 2.85). However, since we are not interested in the direct effects of Corruption Perceptions Index, we believe it is appropriate to include all relevant controls since it is unlikely to affect the substantial results of interest here. Furthermore, our robustness checks indicate that the substantial results are similar regardless of whether we include control variables. All data for the macro variables concern the year 2013.

All independent, moderating and control variables are coded to vary between 0 and 1 to make it easier to compare the results. Table 2 contains summary information on all variables.

Table 2. Descriptive Statistics

	Obs	Mean	SD	Min	Max
Dependent variables
Voting	41,737	.83	.38	.00	1.00
Institutionalized participation	41,082	.00	1.00	−1.19	5.90
Noninstitutionalized participation	41,082	.00	1.00	−1.96	2.58
Independent variable
Individual corruption perception	42,327	.50	.27	.00	1.00
Moderating variables
Gender	46,993	.47	.50	.00	1.00
Country-level corruption perception (CCP)	47,017	.47	.29	.00	1.00
Control variables
Age	46,879	.36	.21	.00	1.00
Education	44,690	.42	.14	.00	1.00
Marriage status	46,567	.53	.50	.00	1.00
Place of living	46,534	.68	.47	.00	1.00
Internal efficacy	43,304	.56	.21	.00	1.00
External efficacy	44,754	.41	.28	.00	1.00
Political interest	45,823	.47	.30	.00	1.00
GDP/capita	47,017	.63	.22	.00	1.00
Compulsory voting	47,017	.11	.31	.00	1.00
Level of democracy	47,017	.77	.25	.00	1.00
Effective number of parties	47,017	.28	.27	.00	1.00

We use multilevel regression analyses to take into account that respondents are nested in countries. For voting, the dependent variable is dichotomous, and we therefore use logistic multilevel regression, whereas for the two other dependent variables that are standardized indexes, we use multilevel linear regression. Since the data do not include comparable weights across countries, we use unweighted data, which means that the results are not necessarily representative.

We present the regression results in the tables, where, model 1 (M1) is fixed effects bivariate regressions between the dependent variable in question and ICP, M2 includes all control variables; M3 includes an interaction effect between gender and ICP as well as a random intercept for ICP while M4 includes a three-way interaction term between gender, corruption, and CCP with all constitutive terms to examine how the national extent of corruption affect the associations between gender, ICP, and political participation.

Following this, we discuss the implications for the hypotheses with the assist of plots of marginal means to clarify the implications in line with the recommendations of Brambor, Clark, and Golder (2006). Since traditional tests of significance can be misleading when it comes to interaction effects (Franzese and Kam 2007), we also visualize interaction effects where p > .05 to ascertain the implications.

Empirical Analyses

We start the analyses in Table 3 presenting country-level differences in political participation and corruption perceptions.

Table 3. Country-Level Differences

	Voting		Institutionalized Participation		NonInstitutionalized Participation		Individual Corruption Perception		Country-Level Corruption Perception
	Mean	SE	Mean	SE	Mean	SE	Mean	SE	Score
Australia	.98	(.00)	−.25	(.03)	.69	(.03)	.39	(.01)	.18
Austria	.87	(.01)	−.09	(.04)	.43	(.03)	.45	(.01)	.40
Belgium	.96	(.01)	−.14	(.03)	.57	(.03)	.39	(.01)	.29
Chile	.69	(.01)	.08	(.02)	−.46	(.02)	.62	(.01)	.36
Taiwan	.85	(.01)	−.08	(.02)	−.26	(.02)	.50	(.01)	.55
Croatia	.72	(.02)	−.08	(.04)	−.18	(.03)	.66	(.01)	.78
Czech Republic	.77	(.01)	.04	(.02)	.03	(.03)	.64	(.01)	.78
Denmark	.95	(.01)	−.14	(.02)	.64	(.03)	.25	(.01)	.00
Finland	.88	(.01)	−.18	(.03)	.39	(.03)	.33	(.01)	.04
France	.92	(.01)	−.22	(.03)	.67	(.03)	.46	(.01)	.36
Georgia	.83	(.01)	.31	(.03)	−.26	(.02)	.28	(.01)	.76
Germany	.85	(.01)	−.10	(.03)	.52	(.03)	.42	(.01)	.24
Hungary	.76	(.01)	−.05	(.02)	−.73	(.02)	.58	(.01)	.67
Iceland	.93	(.01)	−.07	(.03)	.57	(.03)	.49	(.01)	.24
India	.84	(.01)	.83	(.07)	.14	(.04)	.50	(.01)	1.00
Israel	.86	(.01)	.15	(.03)	−.10	(.03)	.62	(.01)	.55
Japan	.75	(.01)	−.28	(.02)	−.30	(.02)	.47	(.01)	.31
Korea (South)	.72	(.01)	−.30	(.02)	−.24	(.02)	.57	(.01)	.65
Lithuania	.74	(.02)	.04	(.03)	−.43	(.03)	.68	(.01)	.62
the Netherlands	.96	(.01)	−.03	(.03)	.36	(.03)	.31	(.01)	.15
Norway	.94	(.01)	.04	(.03)	.44	(.03)	.33	(.01)	.09
the Philippines	.87	(.01)	.41	(.04)	−.53	(.02)	.62	(.01)	1.00
Poland	.78	(.01)	−.12	(.01)	−.59	(.02)	.64	(.01)	.56
Slovak Republic	.74	(.01)	−.05	(.02)	.03	(.03)	.63	(.01)	.80
Slovenia	.77	(.01)	−.11	(.03)	−.19	(.03)	.66	(.01)	.62
South Africa	.76	(.01)	.56	(.03)	−.52	(.02)	.64	(.01)	.89
Spain	.85	(.01)	.04	(.03)	.27	(.03)	.55	(.01)	.58
Sweden	.96	(.01)	−.23	(.04)	.66	(.04)	.37	(.01)	.04
Switzerland	.74	(.02)	−.16	(.04)	.31	(.03)	.36	(.01)	.11
Turkey	.88	(.01)	.14	(.03)	−.65	(.02)	.60	(.01)	.75
Great Britain	–	–	−.24	(.02)	.29	(.03)	.38	(.01)	.27
United States	.72	(.01)	.38	(.05)	.37	(.03)	.56	(.01)	.33
Mean	.83	(.00)	.03	(.01)	.02	(.01)	.51	(.00)	.49

We see that there is considerable variation in how active the population are in the various political activities; for voting, turnout is greatest in countries with compulsory voting such as Australia and Belgium, but other countries such as Sweden and the Netherlands have similar levels of turnout. Institutionalized participation between elections is highest in countries such as India and South Africa, while it is lowest in South Korea and Japan. Although it is tempting to relate these differences to the quality of democracy, they are also affected by factors such as the vicinity of the latest election and domestic political events. For noninstitutionalized participation, there are noticeable differences between the countries, with Australia and France being the most active, while Turks and Hungarians are less eager to engage in such activities. For the two corruption measures, we also observe clear differences across countries, but the most interesting observation is maybe that the subjective perceptions at the individual level do not appear to be clearly connected to the expert observations at the country level. Citizens generally appear to be more prone to believe that corruption exist in the public sector, with some notable exceptions such as India, which receives the highest corruption score by the experts while citizens believe that it is fairly low (.50).

Table 4 displays the multilevel logistic results for voting, while Table 5 includes the results of multilevel linear regression analyses for institutionalized and noninstitutionalized participation.

Table 4. Multilevel Logit Regressions Results, Voting

	M1	M2	M3	M4
Individual corruption perception (ICP)	−.681****** p < .001; ** p < .01; * p < .05.	−.433****** p < .001; ** p < .01; * p < .05.	−.334***** p < .001; ** p < .01; * p < .05.	1.039****** p < .001; ** p < .01; * p < .05.
	(.059)	(.068)	(.118)	(.250)
Gender (male)		−.195****** p < .001; ** p < .01; * p < .05.	−.032	.344**** p < .001; ** p < .01; * p < .05.
		(.033)	(.077)	(.158)
Age		4.771****** p < .001; ** p < .01; * p < .05.	4.769****** p < .001; ** p < .01; * p < .05.	4.769****** p < .001; ** p < .01; * p < .05.
		(.305)	(.306)	(.306)
Age squared		−3.147****** p < .001; ** p < .01; * p < .05.	−3.146****** p < .001; ** p < .01; * p < .05.	−3.146****** p < .001; ** p < .01; * p < .05.
		(.400)	(.400)	(.400)
Education		1.039****** p < .001; ** p < .01; * p < .05.	1.027****** p < .001; ** p < .01; * p < .05.	1.017****** p < .001; ** p < .01; * p < .05.
		(.152)	(.152)	(.152)
Marriage status (married)		.425****** p < .001; ** p < .01; * p < .05.	.423****** p < .001; ** p < .01; * p < .05.	.422****** p < .001; ** p < .01; * p < .05.
		(.036)	(.036)	(.036)
Place of living (urban)		−.256****** p < .001; ** p < .01; * p < .05.	−.269****** p < .001; ** p < .01; * p < .05.	−.269****** p < .001; ** p < .01; * p < .05.
		(.038)	(.038)	(.034)
Internal efficacy		.936****** p < .001; ** p < .01; * p < .05.	.934****** p < .001; ** p < .01; * p < .05.	.934****** p < .001; ** p < .01; * p < .05.
		(.091)	(.091)	(.092)
External efficacy		.653****** p < .001; ** p < .01; * p < .05.	.637****** p < .001; ** p < .01; * p < .05.	.628****** p < .001; ** p < .01; * p < .05.
		(.068)	(.068)	(.068)
Political interest		1.434****** p < .001; ** p < .01; * p < .05.	1.443****** p < .001; ** p < .01; * p < .05.	1.443****** p < .001; ** p < .01; * p < .05.
		(.667)	(.067)	(.067)
Country-level corruption perception (CCP)		−1.314**** p < .001; ** p < .01; * p < .05.	−1.245**** p < .001; ** p < .01; * p < .05.	−1.735***** p < .001; ** p < .01; * p < .05.
		(.619)	(.584)	(.643)
GDP/capita		−1.425	−1.023	−1.341
		(.821)	(.778)	(.792)
Compulsory voting		1.230****** p < .001; ** p < .01; * p < .05.	1.228****** p < .001; ** p < .01; * p < .05.	1.253****** p < .001; ** p < .01; * p < .05.
		(.349)	(.332)	(.343)
Level of democracy		−.188	−.273	−.086
		(.615)	(.585)	(.598)
Effective number of parties		.597	.600	.466
		(.390)	(.371)	(.378)
Gender# ICP			−.294**** p < .001; ** p < .01; * p < .05.	.667***** p < .001; ** p < .01; * p < .05.
			(.123)	(.271)
ICP#CCP				1.373****** p < .001; ** p < .01; * p < .05.
				(.376)
Gender#CCP				.612**** p < .001; ** p < .01; * p < .05.
				(.259)
Gender#ICP#CCP				−.820
				(.453)
Constant	2.129****** p < .001; ** p < .01; * p < .05.	.512	.246	.645
	(.134)	(.834)	(.788)	(.828)
Random effects:
var(constant)	.515	.253	.205	.228
	(.135)	(.068)	(.061)	(.066)
var(ICP)			.139	.089
			(.072)	(.069)
Obs. (groups)	38,109 (31)	32,925 (30)	32,925 (30)	32,925 (30)
BIC	32,364.63	24,967.13	24,970.77	24,987.46
ICC	.135	.071	.059	.065

Note:

• Entries are coefficients from a multilevel logistic regression with standard errors in parenthesis.
• *** p < .001; ** p < .01; * p < .05.

Table 5. Multilevel Linear Regression Results, Institutionalized and Noninstitutionalized Participation

	Institutionalized				NonInstitutionalized
	M1	M2	M3	M4	M1	M2	M3	M4
Individual corruption perception (ICP)	−.114****** p < .001; ** p < .01; * p < .05.	−.011	.047	.009	−.122****** p < .001; ** p < .01; * p < .05.	.046**** p < .001; ** p < .01; * p < .05.	.004	.081
	(.022)	(.023)	(.037)	(.075)	(.020)	(.020)	(.041)	(.080)
Gender (male)		.136****** p < .001; ** p < .01; * p < .05.	.186****** p < .001; ** p < .01; * p < .05.	.197****** p < .001; ** p < .01; * p < .05.		−.072****** p < .001; ** p < .01; * p < .05.	−.121****** p < .001; ** p < .01; * p < .05.	−.205****** p < .001; ** p < .01; * p < .05.
		(.011)	(.023)	(.039)		(.009)	(.020)	(.034)
Age		.235**** p < .001; ** p < .01; * p < .05.	.233**** p < .001; ** p < .01; * p < .05.	.232**** p < .001; ** p < .01; * p < .05.		.974****** p < .001; ** p < .01; * p < .05.	.981****** p < .001; ** p < .01; * p < .05.	.985****** p < .001; ** p < .01; * p < .05.
		(.103)	(.103)	(.103)		(.089)	(.089)	(.089)
Age squared		.119	.123	.123		−1.518****** p < .001; ** p < .01; * p < .05.	−1.526****** p < .001; ** p < .01; * p < .05.	−1.528****** p < .001; ** p < .01; * p < .05.
		(.131)	(.131)	(.131)		(.113)	(.113)	(.113)
Education		.065	.069	.069		1.201****** p < .001; ** p < .01; * p < .05.	1.198****** p < .001; ** p < .01; * p < .05.	1.196****** p < .001; ** p < .01; * p < .05.
		(.048)	(.048)	(.048)		(.041)	(.041)	(.041)
Marriage status (married)		−.017	−.016	−.016		−.038****** p < .001; ** p < .01; * p < .05.	−.037****** p < .001; ** p < .01; * p < .05.	−.037****** p < .001; ** p < .01; * p < .05.
		(.012)	(.012)	(.012)		(.010)	(.010)	(.010)
Place of living (urban)		−.104****** p < .001; ** p < .01; * p < .05.	−.105****** p < .001; ** p < .01; * p < .05.	−.105****** p < .001; ** p < .01; * p < .05.		.039****** p < .001; ** p < .01; * p < .05.	.040****** p < .001; ** p < .01; * p < .05.	.041****** p < .001; ** p < .01; * p < .05.
		(.012)	(.012)	(.012)		(.011)	(.011)	(.011)
Internal efficacy		.172****** p < .001; ** p < .01; * p < .05.	.170****** p < .001; ** p < .01; * p < .05.	.169****** p < .001; ** p < .01; * p < .05.		.526****** p < .001; ** p < .01; * p < .05.	.527****** p < .001; ** p < .01; * p < .05.	.530****** p < .001; ** p < .01; * p < .05.
		(.031)	(.031)	(.031)		(.027)	(.027)	(.027)
External efficacy		.155****** p < .001; ** p < .01; * p < .05.	.154****** p < .001; ** p < .01; * p < .05.	.154****** p < .001; ** p < .01; * p < .05.		.159****** p < .001; ** p < .01; * p < .05.	.168****** p < .001; ** p < .01; * p < .05.	.166****** p < .001; ** p < .01; * p < .05.
		(.022)	(.022)	(.022)		(.019)	(.019)	(.019)
Political interest		.722****** p < .001; ** p < .01; * p < .05.	.721****** p < .001; ** p < .01; * p < .05.	.721****** p < .001; ** p < .01; * p < .05.		.701****** p < .001; ** p < .01; * p < .05.	.694****** p < .001; ** p < .01; * p < .05.	.693****** p < .001; ** p < .01; * p < .05.
		(.022)	(.022)	(.022)		(.019)	(.019)	(.019)
Country-level corruption perception (CCP)		.262	.295	.299		−.455	−.536	−.631**** p < .001; ** p < .01; * p < .05.
		(.214)	(.229)	(.234)		(.280)	(.313)	(.313)
GDP/capita		−.681**** p < .001; ** p < .01; * p < .05.	−.704**** p < .001; ** p < .01; * p < .05.	−.703**** p < .001; ** p < .01; * p < .05.		−.420	−.747	−.746
		(.286)	(.306)	(.306)		(.375)	(.418)	(.415)
Compulsory voting		−.038	−.051	−.051		.226	.211	.211
		(.118)	(.125)	(.125)		(.155)	(.172)	(.171)
Level of democracy		−.076	.023	.022		.654**** p < .001; ** p < .01; * p < .05.	.787**** p < .001; ** p < .01; * p < .05.	.792**** p < .001; ** p < .01; * p < .05.
		(.213)	(.229)	(.229)		(.280)	(.313)	(.465)
Effective number of parties		−.021	−.037	−.038		.085	.083	.084
		(.132)	(.142)	(.142)		(.174)	(.194)	(.193)
Gender# ICP			−.101**** p < .001; ** p < .01; * p < .05.	−.080			.100***** p < .001; ** p < .01; * p < .05.	.108
			(.040)	(.085)			(.034)	(.074)
ICP#CCP				.058				−.097
				(.126)				(.140)
Gender#CCP				−.042				.253****** p < .001; ** p < .01; * p < .05.
				(.074)				(.064)
Gender#ICP#CCP				−.005				−.155
				(.139)				(.120)
Constant	.057	−.255	−.355	−.351	.114	−1.300****** p < .001; ** p < .01; * p < .05.	−1.134***** p < .001; ** p < .01; * p < .05.	−1.112***** p < .001; ** p < .01; * p < .05.
	(.046)	(.289)	(.309)	(.310)	(.076)	(.378)	(.420)	(.420)
Random effects:
Var (Cons)	.063	.031	.034	.034	.183	.054	.066	.065
	(.016)	(.008)	(.009)	(.009)	(.046)	(.014)	(.018)	(.017)
Var(residual)	.980	.948	.947	.947	.815	.707)	.705	.705
	(.007)	(.007)	(.007)	(.007)	(.006)	(.005)	(.005)	(.005)
Var(ICP)			.012	.012			.030	(.028
			(.006)	(.006)			(.011)	(.011)
Obs. (groups)	37,710 (32)	33,201 (31)	33,201 (31)	33,201 (31)	37,710 (32)	33,201 (31)	33,201 (31)	33,201 (31)
BIC	106,417.1	92,754.91	92,760.29	92,789.99	99,502.55	83,029.86	83,010.38	83,009.53
ICC	.061	.032	.035	.035	.183	.071	.085	.084

Note:

• Entries are coefficients from a multilevel linear regression with standard errors in parenthesis.
• *** p < .001; ** p < .01; * p < .05.

Figure 2 displays the estimated effect of ICP on the probability of voting without taking into account gender differences, but controlling for other factors obtained in M2.

Figure 2

Estimated Effect of Individual Corruption Perception on Probability of Voting

For voting, the significant negative coefficient (B = −.433, p = .000) in M2 for ICP entails that the expected probability of voting on average decreases from about 85.7 percent when an individual has no suspicions of corruption to about 80.7 percent when convinced that corruption is widespread. Even if the effect is relatively small, this result still shows the negative association found in previous literature (e.g., Bauhr and Grimes 2014; Kostadinova 2009; Miles 2015).

For institutionalized participation between elections, there is also a negative coefficient in M1 (B = −.114, p = .000), but the coefficient grows insignificant when including controls in M2 (B = −.011, p = .644). This is also visible in the plot, where there is a weak decline in the predicted rate of participation as the individual perception of corruption increases, but this remains around zero and the estimate is surrounded by uncertainty as indicated by the wide confidence intervals.

For noninstitutionalized participation, we observe the expected positive association after including controls in M2 (we return to the reversal of effects compared to M1 later). The association entails that the expected rate of participation increases from .05 to .1 as ICP moves from minimum to maximum.

It is worth noting that although the estimates are weak on average, they have fairly broad confidence intervals, which suggests that there may be important differences between individuals, as posited by our hypotheses. Next, we show the gender differences obtained in M3 in Figure 3.

Figure 3

Gender Differences

For voting, the significant interaction effect (B = −.294, p = .017) entails that the differences in predicted turnout between men and women increase when the individual perception of corruption increases. As predicted by Hypothesis 1a, the negative association is weaker for women, who are less affected by a belief that corruption is widespread.

For institutionalized participation between elections, there is also a significant interaction effect (B = −.101, p = .011). The implications are even more pronounced since the associations between ICP and institutionalized participation differ markedly for men and women, which explains the nonsignificant estimate found in M2. Although men are more likely to be active regardless of ICP, the differences decrease with rising perceptions of corruption since women become more likely to be active (from −.09 when ICP = .0 to −.04 when ICP = 1.0) whereas men become less likely to be active (from .10 when ICP = .0 to .05 when ICP = 1.0). Hence these results support Hypothesis 1b, although not only is the association weaker for women, it is actually reversed to be positive.

Finally, for noninstitutionalized participation, we also find a significant interaction effect (B = .100, p = .004). The results show that while women are generally more active in noninstitutionalized activities, there are no substantial differences in their predicted rates of participation across levels of ICP. For men on the other hand, while they are generally less active, they become more likely to participate in noninstitutionalized activities when believing that corruption is widespread. This result contradicts Hypothesis 1c since men rather than women use noninstitutionalized activities to express dissatisfaction with widespread corruption.

Finally, we examine Hypothesis 2 in M4 by including three-way interactions between ICP, gender, and CCP. Figure 4 shows the predicted differences for men and women depending on a situation with low (CCP = 0) and high (CCP = 1) corruption for the three forms of participation.

Figure 4

Predicted Differences for Men and Women

For voting, the estimate for the three-way interaction term is slightly above the .05 threshold (B = −.820, p = .070), but the interaction effects with gender (B = .612, p = .018) and ICP (1.373, p = .000) are significant and show that the context may well affect the associations of interest here. Figure 4 shows that the associations for men and women are similar when corruption is low, although women are slightly more likely to vote regardless of the level of perceived corruption at the individual level. In a situation of high corruption, on the other hand, we see a more complex pattern since women become more likely to vote with increasing levels of ICP, whereas men become less likely to vote as they perceive corruption to be widespread. Since these patterns differ remarkably from what we previously found, this shows that it is important to consider the country-level corruption to understand how ICP and gender interact to shape voting patterns.

The differences are less pronounced for participation between elections. For institutionalized participation, all estimates for interaction effects involving CCP are weak and none of them come close to p < .05. Consequently, we can also observe that the patterns for men and women are similar for both situations of low and high corruption, and generally resemble what we found for Hypothesis 1b.

For noninstitutionalized participation, the estimate for the three-way interaction term is again not significant (B = −.155, p = .197), but the interaction effect between gender and CCP suggest that the context may play a role in shaping the associations (B = .253, p = .000). Figure 4 shows that in less corrupt situations, we see the pattern observed for Hypothesis 1c, since men become more likely to engage while the differences for women are negligible. In more corrupt situations, the predicted rates of participation are the same, regardless of level of ICP. In other words, perceived corruption only mobilizes males when it occurs in a situation where corruption is relatively uncommon.

Discussion of the Results

This article extends our understanding of gender differences in how corruption perceptions are linked to different forms of political participation and thereby how men and women channel their concerns over corruption to formal political decision makers. Whereas previous studies have identified a negative association between corruption perceptions and political engagement, this study shows that there are important differences in how men and women voice their concerns. Furthermore, what forms of political participation men and women engage in also depend on the country-level corruption. In the following, we discuss some of the noteworthy results we found for each form of participation.

In line with previous research (Bauhr and Grimes 2014; Kostadinova 2009; Miles 2015; Sundström and Stockemer 2015), we generally find that perceptions of corruption lead to lower turnout for both men and women. However, as predicted by our Hypothesis 1a, the demobilizing effect is weaker for women and, when we consider the country-level corruption, we even find that the association for women is positive in highly corrupt countries. Hence, women who perceive widespread corruption in the public services tend to be mobilized to cast a ballot in highly corrupt societies, whereas men are more likely to refrain from voting. In other words, gender differences in the association between corruption perceptions and voting seem to be greater in high-corruption contexts, as suggested by Hypothesis 2. These contradictory effects could explain why Dahlberg and Solevid (2016) only found a negative effect of ICPs in low and medium-corrupt countries.

Although the differences were less pronounced for institutionalized participation between elections and hence do not support Hypothesis 2 for this activity, we find important gender differences that support our conclusion that institutionalized participation is for women a way to voice concerns over corruption, as predicted by our Hypothesis 1b. However, not only was the negative effect of corruption perceptions in this case weaker for women; it actually reversed to become positive. For men, on the other hand, it becomes less attractive to engage in such activities, presumably because they are considered less effective means of achieving influence when corruption is perceived as widespread.

A likely explanation for the mobilizing effect with regard to elections is offered by Bågenholm, Dahlberg, and Solevid (2016), who argue that the effect of corruption perceptions hinges on a political party politicizing the issue of corruption and clean government, which is more likely to occur when corruption is widespread. Similarly, one would suspect that civil society groups, community leaders, the media, and other mediating or linking institutions could play an equally important role in shaping and focusing perceptions in ways that mobilize groups such as women on public service-related issues. It might also well be the case that these linking institutions differ for men and women, which could potentially help to explain the observed differences. Studies have for instance noted that the relatively small gender gaps in voting in areas such as Sub-Saharan Africa might reflect that international organizations such as the UN encourage women to become more active in formal politics in order to improve the quality of their lives (Coffé and Bolzendahl 2011). Elections and other institutionalized activities offer a safe and low-cost path to improve the quality of essential public services, which may explain why women are more likely to engage in such activities.

Our results for noninstitutionalized participation differ from this pattern, which shows that it is important to differentiate between different modes of political engagement. Contrary to our Hypothesis 1c, we find that men become more likely to engage in elite-challenging forms of participation when faced with corruption while women remain unaffected. Furthermore, these gender differences are more pronounced in countries with low corruption whereas the effect is marginal in societies where corruption is widespread, which is contrary to the expectation in Hypothesis 2. These findings are in line with those reported by Bazurli and Portos (2019), who also find that perceptions of endemic corruption are likely to engage citizens to take part in extra-institutional behavior such as boycotts and public protests.

All these results demonstrate that it is important to consider gender differences when examining how corruption perceptions are associated with different forms of political participation. Moreover, while we only find partial support for Hypothesis 2, the results still show the importance of taking contextual differences into account.

Since examining why men and women participate is beyond the present aspirations, we can only speculate why these differences exist. Nevertheless, it is interesting to note that other studies relate their findings to the anti-establishment populist attitudes often found among the so-called “losers of globalization” (Bazurli and Portos 2019, 6), who are often believed to be male. Other studies demonstrate that men are more likely to vote for right-wing populist parties (Harteveld et al. 2015; Spierings and Zaslove 2017) and hold more populist attitudes; that is, belief in a cleavage between ordinary people and political elites, where power should lie with the former (Spierings and Zaslove 2017, 825). Such attitudes have clear affinities to corruption perceptions since distrust in political elites may also lead people to believe that corruption is widespread. The differences we observe may therefore form part of more general trends in society. This interpretation is supported by our finding that the mobilizing effect for males to engage in noninstitutionalized activities is only found where corruption is uncommon, which underlines that corruption perceptions form part of a general distrust of the political system rather than a reaction to genuine cases of corruption (Van de Walle 2008).

Future studies should aim to substantiate this “populist” explanation by incorporating populist attitudes to see whether these can explain our findings. Another important task that remains is to explore the role of the mentioned mediating or linking institutions and their interactions with different societal groups in shaping political behavior. It is also important to examine whether the associations we find can be corroborated by more robust time series analyses that make it possible to settle the direction of causality more firmly than what we can do with our cross-sectional data.

Appendix

Table A1. Questions Wording

Item	Question	Answer Alternatives
BVQ_27. VOTE_LE—Did respondent vote in last general election [voting]	Some people don't vote nowadays for one reason or another. Did you vote in the last [country] national election in [month/year]?	(0) Not eligible to vote in the last election
		(1) Yes
		(2) No
		(7) Refused
		(9) No answer
Q13-17, Q20 Political actions: [institutionalized and noninstitutionalized participation]	Here are some different forms of political and social action that people can take. Please indicate, for each one, whether you have done any of these things in the past year, whether you have done it in the more distant past, whether you have not done it but might do it or have not done it and would never, under any circumstances, do it.	(1) Have done it in the past year
(1) Attended a political meeting or rally		(2) Have done it in the more distant past
(2) Contacted, or attempted to contact, a politician or a civil servant to express your views		(3) Have not done it but might do it
(3) Signed a petition		(4) Have not done it and would never do it
(4) Boycotted, or deliberately bought, certain products for political, ethical or environmental reasons		(8) Can't choose
(5) Took part in a demonstration (any kind of demonstration)		(9) No answer
(6) Expressed political views on the internet. Expression could be public or private in nature
Q23 Status of belonging to: political party [institutionalized participation]	People sometimes belong to different kinds of groups or associations. For each type of group, please indicate whether you, belong and actively participate, belong but don't actively participate, used to belong but do not any more, or have never belonged to it.	(1) Belong and actively participate
		(2) Belong but don't actively participate
		(3) Used to belong but do not any more
		(4) Never belonged to it
		(8) Can't choose
		(9) No answer
Q67 Public service: involvement in corruption [individual corruption perception]	How widespread do you think corruption is in the public service in [COUNTRY]?	(1) Hardly anyone is involved
		(2) A small number is involved
		(3) A moderate number is involved
		(4) A lot of people are involved
		(5) Almost everyone is involved
		(8) Can't choose
		(9) No answer
BVQ_01. SEX [gender]	Sex of respondent	(19 Male
		(2) Female
		(9) No answer
BVQ_02a. AGE	Age of respondent	(15) 15 years
		(102) 102 years
		(999) No answer
BVQ_03. EDUCYRS [education]	How many full years of schooling or education have you had? Please include primary and secondary schooling, university and full-time vocational training, but do not include repeated years.	(0) No formal schooling, no years at school
		(1) One year
		(83) 83 years
		(98) Don't know
		(99) No answer
BVQ_34. MARITAL [marriage status]	What is your current legal marital status?	(1) Married
		(2) Civil partnership
		(3) Separated from spouse/civil partner (still legally married/still legally in a civil partnership)
		(4) Divorced from spouse/legally separated from civil partner
		(5) Widowed/civil partner died
		(6) Never married/never in a civil partnership, single
		(7) Refused
		(9) No answer
BVQ_37. URBRURAL [place of living]	Would you describe the place where you live as…	(1) A big city
		(2) The suburbs or outskirts of a big city
		(3) A town or a small city
		(4) A country village
		(5) A farm or home in the country
		(8) Don't know
		(9) No answer
Q39 Agree: good understanding of important political issues [internal efficacy 1]	To what extent do you agree or disagree with the following statements? I feel I have a pretty good understanding of the important political issues facing [COUNTRY].	(1) Strongly agree
		(5) Strongly disagree
		(8) Can't choose
		(9) No answer
Q40 Agree: most people better informed than I am [internal efficacy 2]	To what extent do you agree or disagree with the following statements? I think most people in [COUNTRY] are better informed about politics and government than I am.	(1) Strongly agree
		(5) Strongly disagree
		(8) Can't choose
		(9) No answer
Q37 Agree: no influence on what government does [external efficacy 1]	To what extent do you agree or disagree with the following statements? People like me don't have any say about what the government does.	(1) Strongly agree
		(5) Strongly disagree
		(8) Can't choose
		(9) No answer
Q38 Agree: government does not care what people like me think [external efficacy 2]	To what extent do you agree or disagree with the following statements? I don't think the government cares much what people like me think.	(1) Strongly agree
		(5) Strongly disagree
		(8) Can't choose
		(9) No answer
Q43 Level of personal interest in politics [political interest]	How interested would you say you personally are in politics?	(1) Very interested
		(5) Not at all interested
		(8) Can't choose
		(9) No answer

Biographies

• Fredrik G. Malmberg is a postdoctoral researcher at the Department of Political Science at Åbo Akademi University, Turku, Finland. In 2019, he defended his doctoral thesis on corruption tolerance. He currently leads a project funded by SLS where he examines perceptions of and attitudes toward corruption among Swedish-speaking Finns. His research interests include corruption, good governance, public opinion, and political behavior. He can be reached at [e-mail]

• Henrik Serup Christensen is Senior Lecturer in Political Science at Åbo Akademi University in Turku, Finland. He currently leads a project funded by the Academy of Finland concerning the impact of democratic innovations on political legitimacy in Finland. His research interests include the connection between attitudes and political participation and has published on these and related topics in journals such as European Political Science Review, European Union Politics, European Politics and Society, European Societies, European Political Science, and Comparative European Politics.