Evaluations of the quality of the representative channel and unequal participation

This article analyzes how individual-level assessments of the quality and functioning of the representative channel affect citizens’ likelihood to turn out to vote and to engage in alternative forms of non-institutionalized participation, and whether these relationships are moderated by individual resources as measured by education. Relying on novel data from the sixth round of the European Social Survey on how European citizens evaluate different aspects of democracy we show that negative evaluations of the quality of the representative channel discourage voting, but only promote participation in demonstrations among the highly educated. These findings highlight potential inequalities in citizens’ ability to voice their political demands: while highly educated individuals are likely to translate their negative evaluations of the institutional channel of representation into non-institutionalized forms of participation, in the presence of negative evaluations low educated individuals are simply more likely to withdraw from politics.


INTRODUCTION
Political participation is a crucial characteristic of democracies, since it constitutes the main tool for citizens to channel their demands to policymakers. This is manifested in numerous studies that analyze the determinants of political participation. One strand of this research focuses on the impact of characteristics of the electoral process on citizens' motivations to turn out to vote. This research has generally relied on macro-level factors, reflecting the competitiveness of elections or other characteristics of electoral systems, to account for the incentive structures surrounding specific elections (Blais and Dobrzynska, 1998;Franklin, 1996;Franklin and Hirczy, 1998). Following a similar logic, recent studies have analyzed how individual-level assessments of the integrity of the electoral process affect citizens' likelihood to participate in politics (Birch, 2010;Carreras and İrepoğlu, 2013;Norris, 2014).
Combining the insights of these two literature strands, in this paper we construct a measure that captures Europeans' assessments of the quality of the representative channel. Adapting the motivational approach to understanding political participation (see Norris, 2002, pp. 61-72) and Verba et al. (1995) civic voluntarism model we argue, first, that these assessments affect citizens' participation decisions by altering their motivations to engage in politics through different means, and, second, that individual resource inequalities play a moderating role in this process.
Citizens' subjective assessments of how much they can influence governments' composition and policymaking through elections should affect their motivations to express their demands through the representative channel. Hence, we expect that positive evaluations of the functioning of this channel will be positively related to participation in elections. When such evaluations are negative and, as a consequence, elections do not provide the proper means to influence policymaking, citizens might choose to engage in non-institutionalized forms of participation to voice their demands. Hence, positive evaluations of the quality of the representative channel should relate negatively to the likelihood of demonstrating.
However, from the literature on political participation we know that resources can moderate how and when motivations get translated into behavior, although differently for voting and demonstrating. As we argue below, these moderating effects might give raise to inequalities in political influence in the presence of a malfunctioning representative channel.
Our empirical analysis, based on novel data from European democracies, reveals that, on the one hand, there is a positive relationship between citizens' assessment of the quality of the representative channel and their likelihood of turning out to vote. On the other hand, we find a negative correlation between these assessments and participation in demonstrations.
Our hypotheses about the moderating role of individual resources are also confirmed. Results show that education is a significant moderator in the relationship between individual evaluations and participation in demonstrations, but not in the case of voting. An extension of this analysis reveals that when the representative channel is judged to be malfunctioning only those with higher education are more likely to resort to demonstrating as an alternative or supplemental form of expressing their demands, while those with lower levels of education are more likely to simply withdraw from politics. This paper proceeds as follows. We first lay out the theoretical framework for the analysis of citizens' assessments of the quality of the representative channel, and next we hypothesize how it relates to individual participation decisions. Next, we summarize the data and methods. Section four discusses the main results, while section five presents robustness checks. Section six concludes.

The quality of the representative channel
There are different aspects of a political system that can affect the quality of the representative channel. We consider that in established democracies this quality is a function of four characteristics: the degree to which elections are conducted freely and fairly; the capacity of organized opposition parties to effectively contest elections; the ideological differentiation of political parties; and the decisiveness of elections in determining the composition of governments. These are aspects that are likely to modulate citizens' capacity to transmit their political demands and affect policymaking through their vote. Hence, since participation is motivated by the will to exert influence over policymaking, citizens' subjective evaluations of these different aspects should affect their motivations to participate in politics (see below).
Free and fair elections is the first condition that a democracy, where citizens are meant to exercise influence over policymakers through their votes, must fulfill. If elections are tampered in any way, or they are not celebrated in an environment free from coercion, citizens' capacity to exercise influence over policymakers will be limited. As Birch (2010) argues, elections that are not free and fair are less meaningful and consequential. Hence, the quality of the representative channel depends, first, on whether elections are conducted in a free environment and the rules regarding the process are fairly applied.
Even if free elections are important to ensure that citizens can route their political demands through the representative channel this is by no means sufficient, since elections must also be contested. That is, opposition parties must be able to effectively compete with each other for votes to ensure that all of them have a real chance of winning office (Przeworski et al., 1996). To effectively compete in elections it is fundamental that all parties are free to criticize the government. If opposition parties are constrained in their capacity to criticize government actions, opposition parties will only have a slim probability of effectively running for office.
In established democracies citizens' capacity to transmit their demands through the representative channel not only depends on the integrity of the electoral process (i.e. elections being free and contested), but also on the ideological differentiation of political parties. One of the aspects that makes the choices in an election process meaningful is that the parties contesting it are distinguishable in terms of ideology (Wessels and Schmitt, 2008). In the presence of a sufficiently differentiated partisan offer it is likely that all sectors of society will be able to find a party that represents their interests and preferences (Norris, 2002). In its absence, some sectors of society will remain unable to express their true policy priorities through the conventional channel of representation, since they will find no party to vehicle their demands.
The different options offered to citizens, no matter how broad or narrow they are, become meaningless if citizens are not able to determine the composition of governments and reward and punish the incumbent government through their vote. That is, elections must be consequential and citizens must be capable of "throwing the rascals out" (Wessels and Schmitt, 2008). If this is the case, elections grant citizens the means to exercise control over political institutions and the political agenda through the representative channel (Morlino, 2009). Conversely, if elections are not decisive citizens' will not be able to sanction and hold the government accountable through their vote.
Empirically, earlier research documented the impact of characteristics of the electoral process such as the breadth of the partisan offer, the number of parties, the closeness of elections and the frequency of government change on aggregate turnout (Adams and Merrill, 2003;Blais, 2006;Jackman, 1987;Wessels and Schmitt, 2008). While this work analyzed how macro-level characteristics of the electoral process and the institutional system affect turnout by allegedly influencing individual incentives to vote, some recent studies have considered how individual assessments of aspects related to the integrity of the electoral process, mainly the freedom and fairness of elections, affect political engagement (Birch, 2010;Carreras and İrepoğlu, 2013;Hiskey and Bowler, 2005;Levin and Alvarez, 2009;McCann and Domıńguez, 1998;Norris, 2014). These studies have shown that when citizens judge that elections are conducted freely and fairly and the electoral process is not tampered they are more likely to vote and less likely to protest.
Although these studies constitute an important contribution because they moved from contextual factors to subjective evaluations of specific aspects of the electoral process, they have certain limitations. With the notable exceptions of Levin andAlvarez (2009) andNorris (2014), prior studies focus exclusively on voting, and disregard other forms of participation.
More importantly, all these studies focus on evaluations of the integrity and incorruptibility of the electoral process, and most of them rely on a single indicator about trust in elections or the extent to which elections are conducted freely and fairly. 1 Electoral integrity considerations might be more relevant for citizens' behavior in non-fully established democracies like those analyzed in most of these studies. 2 However, in a context of established democracies, where the prospects of elections being conducted in accordance with the highest democratic standards are high, we need to incorporate elements that go beyond electoral integrity and malpractice (e.g. the ideological differentiation between parties and the decisiveness of elections). 3 Moreover, it is also necessary to consider these processes in broader models of political participation that account for the potential moderating role of individual resources. Verba's et al. (1995) civic voluntarism model underlines the importance of motivations, resources and mobilization for participation decisions. In this paper we focus on the first two sets of factors, and begin by considering how subjective evaluations of the quality of the representative channel relate to motivations to participate in politics. The motivational or instrumental model of political participation sustains that citizens are rational actors who intend to affect the course of public policy through their actions (Franklin, 1996;Norris, 2002, pp. 61-72). Although it might appear naïve for individual citizens to expect to influence policymaking through their individual behavior, research has shown that the desire to influence policies is among the most relevant considerations motivating citizens' participation in elections and demonstrations (Verba et al., 1995). Hence, citizens' assessment of the probability that their actions will have any impact on policymaking should influence their decisions to participate, as well as the means through which they participate (see Birch (2010); Carreras and İrepoğlu (2013); Norris (2014) for a related view).

Evaluations of the quality of the representative channel, resources, and political participation
The quality of the representative channel modulates the extent to which citizens are able to vehicle their political demands through institutionalized means of participation. In the absence of free, competitive and decisive elections, or ideologically differentiated partisan alternatives citizens will have a low capacity to influence policymaking through the conventional channel. If citizens perceive that the representative channel does not work well, they should consider that their vote is less likely to be effective in transferring their demands to the political system. Hence, given that motivations to ultimately affect policymaking will play a central role in citizens' participation decisions, we expect that more positive evaluations of the quality of the representative channel will be positively associated with the likelihood of turning out to vote (H1).
A logical consequence derived from our first hypothesis is that those with negative evaluations of the representative channel will be less likely to turn out to vote. However, even when the evaluations of the representative channel are negative, citizens might still desire to influence the political process. Protest has become increasingly present in contemporary democracies, and it constitutes an important tool to exert influence over policymaking (Dalton et al., 2010). Historically, demonstrations have been a tool for those lacking access through the conventional channel of representation, and studies of protest argue that with limited means of conventional political access citizens' likelihood to demonstrate may increase (Dalton et al., 2010;Kitschelt, 1986;Marien and Christensen, 2013). Hence, as citizens perceive the conventional channels to be blocked or inadequate, they may opt to vehicle their demands through demonstrations, either in addition to or as an alternative to voting. As a consequence, we expect that more negative evaluations of the quality of the representative channel will be positively associated with the likelihood of participating in demonstrations (H2).
Although a citizen might or might not participate in elections and take part in demonstrations depending on how she evaluates the functioning of the representative channel, this choice is constrained by her individual resources and the different resource demands imposed upon her by each of the forms of participation. This implies that the role of motivations stemming from subjective assessments of the quality of the representative channel cannot be analyzed in isolation, and that one must also consider the role played by individual resources that are relevant for particular forms of political participation. Previous research has recognized the importance of personal resources to explain political participation. Resources such as education or income enable citizens to participate, since they provide the necessary skills and means to be active in politics (Verba et al., 1995).
Resources have not only been considered a direct correlate of participation, but also a moderating factor, affecting the relationship between political grievances or motivations and actual political actions. Following the argument put forward by Gamson (1968), the relationship between grievances and participation is considered to involve complex interactions (Levi and Stoker, 2000), since resources are assumed to be necessary for individuals to translate motivations into action. One strand within this literature has focused on the moderating effect of resources such as education or income (Chan, 1997;Citrin, 1977;Kriesi and Westholm, 2007), while others have predominantly focused on the moderating effect of political attitudes such as political interest, political efficacy, or regime support (Christensen, 2014;Craig and Maggiotto, 1981;Hooghe and Marien, 2013).
In comparison to other forms of participation, demonstrations are considerably more demanding in terms of resources (Dalton, 2006, pp. 73-74). As a consequence, not all citizens that asses the quality of the representative channel negatively will be equally likely to reroute their political demands through demonstrations. Protests and other direct action methods are considered high information activities, and, as such, the requirements to participate in terms of civic skills are higher than for other forms of participation (Dalton, 2000, pp. 929-930). These civic skills are fostered by citizens' education (Verba et al., 1995).
Thus, education is likely to affect citizens' capacity to grasp and exploit the opportunities to influence the policymaking process through demonstrations. We hence expect more negative evaluations of the quality of the representative channel to have a stronger effect on the likelihood to demonstrate for those who are more educated (H3). In fact, it might be that (when holding negative evaluations of the representative channel) only those who are more educated are able to add another form of participation to their political repertoire or to bypass the representative channel altogether to ensure that their demands are channeled into the political system. That is, a minimum level of education might be necessary for individuals to be able to resort to alternative means of participation.
In contrast to demonstrations, voting is one of the most common and least demanding forms of participation, since the act of voting makes only modest demands on citizens in terms of cognitive and material resources (Verba et al., 1995). In comparison to other forms of participation, voting has a "low-cost" nature (Aarts and Wessels, 2005, p. 81). Research on the determinants of voting in Europe has shown that there is barely any educational effect for voting and that, as a consequence, voting can be considered one of the most democratic forms of participation (Marien et al., 2010). Topf (1995) argued that since the 1960s all Europeans appear to possess the skills to participate in national elections. Hence, while educational attainment generates pronounced unequal participation patterns in most non-institutionalized forms of political participation, people of all educational levels participate at similar rates in elections (Marien et al., 2010, p. 197;Teorrell et al., 2007, p. 395). 4 As a consequence, we do not expect educational attainment to moderate the association between respondents' evaluations of the representative channel and their likelihood to turn out to vote. Hence, we should not find any differences in the effect of evaluations of the representative channel on the likelihood of voting for individuals with different levels of education (H4). That is, the impact of negative evaluations on the likelihood of withdrawing from electoral participation should be the same across individuals with different levels of education.
If confirmed, our first two hypotheses imply that negative evaluations of the quality of the representative channel should not be considered a threat for the correct functioning of contemporary democracies. Those who hold negative evaluations would not withdraw from politics altogether, but they would just be more likely to adjust the way in which they channel their demands into the political system. The "critical citizens" thesis argues that dissatisfied citizens may eschew institutionalized forms of participation to engage instead in protest activities. Within this framework, a critical outlook towards the functioning of political institutions is not seen as a symptom or precursor of political alienation, but as a healthy attitude, which, even if it may discourage participation through conventional means, it is also likely to motivate citizens' to remain vigilant and engage in alternative forms of political participation (Hofferbert and Klingemann, 2001;Norris, 1999;Rosanvallon, 2008). Although this conclusion would be reassuring, this might not always be the case.
Our third and fourth hypotheses imply that more negative assessments of the functioning of the representative channel would entail that all citizens, independently of their level of education, would be less likely to vote. However, negative evaluations would only imply a greater likelihood to demonstrate for those who are more educated. As a consequence, for those who are less educated, negative evaluations would entail an increasing likelihood of withdrawing from politics (i.e. political alienation). Conversely, for those who are more educated, negative evaluations imply that these individuals are more likely to participate in demonstrations, and this could be done as an alternative to voting or in addition to it.
It is possible that those who demonstrate more as their perceptions of the representative channel worsen still participate in elections, since citizens can also express dissatisfaction through voting (e.g. by casting a vote for protest parties). In fact, recent studies suggest that protest might be an instrument that some citizens add up to their participation repertoires, instead of being a tool predominantly used by those who decide to withdraw from conventional politics (Saunders, 2014). In terms of the participation outcomes we study, we expect that given their lack of resources individuals with lower levels of education will simply withdraw from politics when they have negative perceptions of how the representative channel works. At the same time, we expect those with higher education to be more likely to adapt their behavior either by only demonstrating, or by incorporating this form of participation to their repertoire as a way of adding strength to their voices in the presence of a malfunctioning representative channel. Hence, the joint consideration of both motivations and resources leads us to expect that only a resourceful fraction of the population will behave as the ideal "critical citizen", who does not withdraw from the political process in the presence of a malfunctioning representative channel.

DATA AND METHODS
Our empirical analyses draw on data from the European Social Survey (ESS), a crossnational survey frequently used to study political participation. In its sixth round, conducted The main independent variable (individual evaluations of the quality of the representative channel) is operationalized with these four survey items. This operationalization is consistent with the discussion in the theory section, which summarizes the theoretical rationale underpinning the aggregation of these different indicators. The empirical analysis confirms that these indicators can be combined into a single measure. An exploratory factor analysis (table 1) yields a one-factor solution, with only one factor extracted with an eigenvalue higher than one, and with all indicators loading strongly on this single dimension. 7 The Cronbach's alpha for these indicators equals 0.72. We estimate our main independent variable based on the factor scores, which take higher values for better evaluations of the quality of the representative channel. The resulting index ranges between (-1.7) and (0.9), with Kosovo being the country with the lowest/worst average evaluations and Sweden the country with the highest/best. 8

<TABLE 1>
With regard to our dependent variables, voting takes the value 1 for those who voted in the last national election and 0 for those who did not. 9 Following Saunders' (2014) recommendations, we restrict our analysis to participation in demonstrations without incorporating to our measure any other non-conventional activity. The demonstration variable takes the value 1 for those who participated in lawful demonstrations in the last 12 months and the value of 0 for those who did not. These two variables are combined to generate our third dependent variable which classifies respondents in four different categories: neither votes nor demonstrates, only votes, only demonstrates, votes and demonstrates.
Together with the evaluations of the representative channel, education is a key independent variable. The ESS includes two measures of education. A categorical variable capturing the highest level of education achieved by a respondent, and a continuous variable measuring the number of years a respondent spent in full time education. Although the latter has been extensively used in political science research, survey and education research questioned its use in cross-national analyses (Müller, 2008;Schneider, 2007). As a consequence, we rely on the ISCED categorical education variable to group respondents in three categories: primary education or less, secondary education, and university education.
All models include a control variable that identifies respondents that support any of the parties in government. It is important to account for the potential confounding effect of "winners and losers", since being a winner affects citizens' assessment of the fairness of elections (Birch, 2008), while at the same time it might also alter citizens' decisions to join

RESULTS
Our first hypotheses refer to the association between evaluations of the quality of the representative channel and the likelihood of voting and demonstrating. Table 2  political interest. The analysis reveals that the estimated change in the probability of voting is higher than the one associated to moving from being hardly interested in politics to being quite interested in politics.

<FIGURES 1 AND 2 >
In the case of participation in demonstrations the change in the adjusted prediction is smaller when moving from one extreme of the evaluation index to the other. The adjusted prediction of demonstrating is 0.10 for those with the worst evaluations, and it decreases to 0.06 when the evaluation index takes its maximum value, a change of just 0.04 points. Hence, while H1 is clearly confirmed by these results, H2 is only weakly supported. It is possible that the marginal effect of the evaluations of the representative channel is smaller when explaining participation in demonstrations than voting because, as we hypothesized above, in the case of demonstrations we expect this effect to vary according to educational levels, with flatter slopes for those with low levels of education.
The third and fourth hypotheses focus on the moderating effect of education on the association between evaluations of the representative channel, voting and participating in demonstrations. Models 3 and 4 in table 2 summarize the results of the two interactive models specified to test these hypotheses. In both cases the evaluations of the quality of the representative channel have been interacted with education levels (with the level of primary education or less set as the reference category). The coefficients reported in model 4 reveal that in the case of participation in demonstrations the interactive effect between the evaluations and secondary and tertiary education are both significant at least at the 0.01 level.
However, these interactive terms fail to reach conventional levels of significance in the model in which voting is specified as the dependent variable (model 3). These results provide preliminary support for H3 and H4. However, since interactive effects in logistic regression models are not easily interpreted by raw coefficients we turn to figures 3 and 4 for a better assessment of these results. combination of these two activities they perform allows us to investigate further the impact of evaluations on political involvement, and, more importantly, to determine if negative evaluations can be considered a mobilizing or an alienating factor depending on citizens' resources. Table 3 summarizes the results of two multinomial logistic models in which only votes is set as the base outcome.

LIMITATIONS AND ROBUSTNESS CHECKS
Like most studies analyzing attitudes and behavior, our analyses are susceptible of being affected by endogeneity. 20 Respondents might rationalize and edit their answers to the attitudinal questions according to their behavior. For example, it is possible that respondents who did not vote provide worst evaluations of the representative channel to appear consistent, avoid cognitive dissonance, or justify a socially undesirable behavior (Birch, 2010;Norris, 2014). The act of voting itself might also reinforce citizens' evaluations of the functioning of the representative channel. Although Birch (2010) showed, using UK panel data, that prior perceptions of electoral fairness affect subsequent voting decisions, in our case endogeneity might bias some of our results, especially in the case of voting.
If endogeneity biases our findings, it is more likely to affect some of the variables of our index of the quality of the representative channel than others. Of the four questions we use to operationalize evaluations of the representative channel only one directly refers to the electoral process (elections being conducted freely and fairly). The remaining three questions ask respondents about their opinion on elements that are related to the functioning of the representative channel, but without explicitly mentioning elections. As a consequence, the likelihood of respondents rationalizing and editing their answers according to their behaviors should be lower for these three questions. We exploit this feature of the dataset in order to assess the robustness of our findings.
Tables in appendix B replicate our models with an evaluation variable generated from a factor analysis that excludes the free elections question. Overall, our findings are robust to the use of this alternative specification. Only in the case of the model in which voting is specified as the dependent variable the effects of the evaluations weaken, but still remain significant. This might suggest that, for voting, a share of the direct effects we estimate could be endogenous. However, the interaction effects are not modified. For participation in demonstrations the results are also unaltered by the different specification of the main independent variable. Lastly, in the case of the multinomial logistic analysis, the results are only slightly weaker. Hence, in spite of the inherent limitations of cross-sectional data to address potential endogeneity biases, these analyses increase our confidence in the robustness of our findings, by showing that the exclusion of the question most susceptible of being affected by this bias does not substantially alter our findings.
Another limitation of our paper stems from the fact that we consider only one form of non-conventional participation (demonstrations), and citizens may rely on other forms of non-conventional participation to channel their demands to policymakers. Although different forms of non-conventional participation may not be entirely comparable (Saunders, 2014), we re-specify our demonstration variable to include a larger number of non-conventional actions, and we re-estimate all our models. 21 The results (available upon request) are very similar to the ones obtained with the measure based on demonstrations only. Negative evaluations of the quality of representative channel are associated with a greater likelihood of engaging through non-conventional forms of participation only for respondents who are more educated.

CONCLUSION
With this paper we contribute to the field of political participation studies by implementing a measure that captures one of the determinants of citizens' motivations to engage in politics: their evaluations of the quality of the representative channel. Our initial hypothesis contended that those citizens who evaluate positively the functioning of the representative channel should be more motivated to vote. At the same time, those who evaluate it negatively should be more likely to choose extra-institutional forms of participation as a mechanism to channel their demands into the political system. Our empirical results support these initial hypotheses but with certain caveats, namely that individual resources play an important moderating role in the case of participation in demonstrations.
In line with studies analyzing attitudes on electoral integrity (e.g. Carreras and These findings also have broader implications for the functioning of European democracies. For those who are more educated, negative evaluations of the quality of the representative channel are less likely to imply a withdrawal from the political process, because these citizens have a greater likelihood of adapting the repertoire through which they vehicle their demands into the political system. Conversely, for those with low levels of education, negative evaluations are more likely to imply a withdrawal from the political process altogether. For these citizens, who have fewer resources to engage in demanding forms of political participation, negative evaluations are not translated into a greater likelihood to engage in demonstrations and, in the same way as for those who are more educated, they are associated with a lower likelihood to vote. This finding qualifies the optimistic view of the "critical citizens" thesis, which contends that in post-industrial societies negative orientations towards the political system might not be problematic for the functioning of democracy, since those who are dissatisfied, disenchanted or critical are more likely to change their repertoire of political actions but they will not withdraw from politics.
Our findings show that, whenever the channel of representative politics is judged to be malfunctioning, only the most resourceful citizens are likely to reroute their political demands through alternative channels. Hence, a low quality of the representative channel is more likely to politically alienate those with fewer resources.
Given that political participation is one of the main mechanisms linking citizens' preferences to the policymaking process, the logical implication of these findings is that when perceptions of the representative channel are negative not all citizens are equally likely of making their voices heard. This would violate democracy's ideal that all citizens' needs and preferences should be given equal consideration, since there is evidence that policy makers are likely to neglect the preferences of those groups that are less likely to participate (Bartels, 2008). There are, however, alternative forms of political participation other than demonstrating that could mitigate these inequalities in the presence of negative evaluations, as long as engagement in them is not conditional on individual resources. A succinct analysis of other forms of participation included in the ESS indicates that inequalities are also apparent in them, but further research should analyze other emerging forms of participation (e.g. online participation). Besides considering other forms of political participation, further extensions of these analyses could consider the role played by contextual factors (e.g. the strength of mobilization agents or characteristics of the political opportunity structure) in how negative perceptions of the representative channel relate to participation decisions, and how this relationship is moderated by individual resources like education.

Notes
1 Although still focusing on the integrity of the electoral process Carreras and İrepoğlu (2013) and Norris (2014) rely on more than one indicator to operationalize their electoral integrity/malpractice measures. 2 The studies by Hiskey and Bowler (2005), and Levin and Alvarez (2009)  Likewise, Armingeon and Schädel (2015) recently argued that there are remarkable cross-country and temporal differences with respect to voting inequalities related to education. In any case, Gallego (2015: 25) points that in most countries the overall differences in turnout rates for citizens with different levels of education are moderate in size, and Armingeon and Schädel (2015) identify an average difference in turnout rates between those with the highest and the lowest education of just 4.9 percent (for the 1999-2009 decade). 5 Our final sample includes 27 countries. We exclude Russia and Ukraine because they cannot be considered fully democratic. None of these countries had a score above 6 in the Polity IV dataset .We exclude countries that are not fully democratic because voting and demonstrating, as well as answers to questions related to the functioning of democracy, might be distorted by the non-democratic character of these regimes. 6 The question wording and descriptive statistics of all items used in this paper can be found in appendix A. 7 The same factor solution is obtained when factor analyzing these indicators in each of the countries separately.
In all countries only one factor with an eigenvalue higher than one is extracted, and in all cases all indicators have a factor loading above the 0.3 threshold. 8 If instead of relying on the factor scores we rely on an index obtained through the sum of the four indicators we obtain very similar results that lead us to the same conclusions for all the analyses presented below (results available upon request). 9 Respondents not eligible to vote have been excluded from all the analyses.
10 This choice is motivated by the fact that it is not possible to directly measure winner/loser status according to the party voted by the respondent, because this variable predicts success perfectly in non-linear models in which voting is specified as the dependent variable. 11 To ensure that our results are not driven by our model estimation decisions we replicate all the analyses using random-intercepts logistic and multinomial logistic models. Empty random-intercepts models reveal that the amount of variance at the country level is 8.6 percent for voting, and 13.7 percent for demonstrating. Following recent analysis of political participation (Braun and Hutter, 2014;Dalton et al., 2010;Marien and Christensen, 2013) we introduce in these models a country-level control for the openness of the political system. To operationalize this variable we follow Dalton et al., (2010) who rely on the World Bank rule of law indicator to measure system openness. This choice is motivated by this being the only system openness proxy (among the ones used in previous studies) that is available for all the countries in our sample. These multilevel models also include a country-level variable measuring the enforcement of compulsory voting in national elections. These models, which can be found in appendix C, do not alter the substantive results and lead us to the same conclusions. 12 Independent variables are rescaled so that numeric inputs represent the effect of the mean ±1 standard deviation. Binary predictors are not rescaled. 13 We have re-estimated these predictions relying on adjusted predictions at representative values (APRs) instead of average adjusted predictions (AAPs). We have estimated APRs of the likelihood of turning out to vote for a young individual with a low level of political interest. In this case the adjusted predictions of voting change from 0.46 for an individual with the worst evaluations to 0.62 for and individual with the best evaluations. That is, in this case the change in the adjusted prediction of turning out to vote is of 0.16. 14 In the case of voting a contrast of the statistical significance of the average marginal effects of the evaluations reveals that there is no statistically significant difference in the effect of the evaluations between the three education groups. 15 In the case of demonstrating a contrast of the statistical significance of the average marginal effects of the evaluations reveals that there are statistically significant differences in the effect of the evaluations between all these three education groups. The negative average marginal effects of the evaluations are statistically significant at the 0.001 level for those with secondary and university education. The positive average marginal effect of the evaluations fails to reach conventional levels of statistical significance for those with primary education. 16 The high proportion of voters in the sample is likely to be caused by turnout overestimation due to social desirability bias.
17 Even if only 503 respondents fall in the only demonstrates category, it is meaningful to separate those respondents from those who both vote and demonstrate since their attitudinal profile is likely to be quite different (e.g. they should have more negative evaluations of the functioning of representative channel than those who both vote and demonstrate).
18 For those with university and secondary education, the average marginal effects associated to a one-unit change in the evaluations factor indicate that, for them, more negative evaluations statistically significantly increase the likelihood of both voting and demonstrating, while this is not the case for those with primary education. 19 As in the previous case, for those with university and secondary education, the average marginal effects associated to a one-unit change in the evaluations factor indicate that, for them, more negative evaluations statistically significantly increase the likelihood of demonstrating only, while this is not the case for those with primary education. .  -Topf, R., 1995. Electoral Participation, in: Klingemann, H.-D., Fuchs, D. (Eds.), Citizens and the State, Beliefs in Government. Oxford University Press, Oxford ;New York. -Verba, S., Schlozman, K.L., Brady, H.E., 1995. Voice andequality: civic voluntarism in American politics. Harvard University Press, Cambridge, Mass. -Wessels, B., Schmitt, H., 2008. Meaningful choices, political supply, andinstitutional effectiveness. Elect. Stud. 27, 19-30.