A safety net against populism? An investigation of the interaction effect of political efficacy and democratic capacities on populist attitudes

ABSTRACT The paper interprets populism as a symptom of a mismatch between how the democratic polity operates and how citizens conceive their own aspirations, needs and identities vis-à-vis the polity. However, democracy requests certain attitudes and skills from citizens: political engagement, a reflective attitude, scrutiny of the power holders and balancing trust-based cooperation with critical reactions towards political authorities. In line with this, we investigate how external and internal political efficacy are associated with populist attitudes in the case of people who have and who do not have certain democratic capacities. Our findings drawing upon an original survey covering 15 European countries show that higher internal political efficacy is associated with more populist attitudes in the case of people with incomplete democratic capacities, but complete democratic capacities yield a ‘safety net’ against this effect. However, the negative relationship between external political efficacy and populist attitudes does not depend on these capacities: stronger dissatisfaction with the responsiveness of political elites leads to more populist attitudes irrespective of people’s democratic background. Nonetheless, our findings imply that a stronger emphasis on certain democratic practices and values in political socialization or civic education could prevent stronger political confidence would turn into populist views about politics.


Introduction
It is easy to understand how contextual and 'supply-side' factors in terms of representation gap, elite malfeasance, corruption, ineffective policies, and growing inequalities trigger frustration, anger and, potentially, populist choices from citizens.Facing the complexity of globalization, the plurality of interests and values in politics, the flooding of information to cope with and the sometimes painfully slow institutional responses to social problems one may easily feel frustrated and inefficaciousand, as a consequence, become either cynical or be tempted by the populist appeal.Discontent or frustrated citizens may withdraw from politics, or, alternatively become active and try to change politics.Passivity and cynicism, if pervasive, ultimately endanger the functioning of democracy (Agger et al. 1961).However, activism and participation may also be destructive, as populist mobilization (see Anduiza et al., 2019) illustrates: discontent and criticism without reflective engagement may lead to demagoguery, belief in conspiracy theories (Castanho Silva et al., 2017) and support for oversimplified solutions or even violent political action (Capelos & Dmertzis 2022).In this sense and from the citizens' perspective, populism might be interpreted as a specific mindset or coping strategy, a reaction to the perceived unresponsiveness of the political system and the lack of self-efficacy feeling in politics.Hawkins and Rowira Kaltwasser (2019) argue that populist attitudes are present in any society, but specific contexts and factors are needed to activate them.In other words, populism is a potentially negative consequence of declining trust in, and popular support for, institutions and policies (see Urbinati 2017).
However, these external factors result in populist responses only in certain segments of society, while others having the same frustration do not become open to populist arguments.The roots of populism are manifold.Apart from external factors and contexts individual-level characteristics may also help the activation of populist sentiments as there are differences between people in their receptivity to the populist appeal.
While in the study of populism there is a stronger emphasis on the supply-side and the investigation of populist political actors (e.g.Engesser et al., 2017;Rooduijn et al., 2014), an emerging sub-field seeks to understand the antecedents of citizens' populist attitudes (e.g.Bernhard & Hänggli, 2018;Spruyt et al. 2016).Our study aims to contribute to this stream of literature by linking populist attitudes to the subjective attitude of political efficacy on the one hand and objective democratic capacities, on the other.Political efficacy refers to the subjective perception of the role of the self in the political system and it is one of the key antecedent variables of political science that explain several aspects of citizens' political behaviour (Campbell et al., 1954).Its external variant relates to citizens' perceptions about the responsiveness of the political system, while internal efficacy captures how confident they are in their own political capabilities (Niemi et al., 1991).By democratic capacities we understand a set of practices, values and attitudes which are prerequisites of effective and democratic political participation according to normative democracy theories (Dahl, 1956;Dewey, 1954).Democracy puts burdens on the citizens as it requests from them certain skills, attitudes and investments: political engagement, a reflective attitude, scrutiny of the power holders and balancing trustbased cooperation with critical reactions towards political authorities (Caprara & Vecchione, 2017).
Theoretically, political efficacy can be easily connected to populist attitudes as the perception that the political system is not responsive to citizens' demand (lower external efficacy) is in accordance with the anti-elitist stance of populism, while strong political self-confidence (high internal efficacy) can be paired with its people-centric approach.However, we argue that democratic capacities function as a safety net that inhibits populism to take advantage of the mobilization power of this combination of political efficacy.Therefore, in this research we address the way democratic capacities interact with political efficacy feelings and how they relate to populist attitudes.We expect that low levels of external political efficacy as well as high levels of internal political efficacy predict stronger populist attitudes, but only for people who have incomplete democratic capacities.Complete democratic capacities are assumed to negate these associations.
We test and corroborate our hypotheses on the data of an original international survey covering 15 European countries.Our findings show that while lower external and higher internal political efficacy are associated with populist attitudes for people with incomplete democratic capacities, complete democratic capacities mitigate the effect of internal political efficacy.Put differently, stronger political confidence leads to more populist attitudes for many people, but less for those who have complete democratic capacities.This way democratic capacities can break the link between political confidence and the people-centric approach to politics and make populist attitudes dependent mainly on the perception of the political system instead of the self.

Populism and democracy
Populism is a complex political phenomenon that has different interpretations: as a communication style (Knight 1998), as a political logic (Weyland 2013), or as a 'thin-centered' ideology (Mudde 2004).Mainstream populism research is mostly following the latter approach defining populism as a thin-centered ideology, without a substantive leftwing or right-wing ideological core, along with the attitudes of people-centrism, antielitism, anti-pluralism and Manichean outlook on social conflicts.
Populism has an ambivalent relationship with liberal democracy.While some authors argue that populism is an essentially democratic phenomenon and has a corrective potential (Canovan, 1999;Mudde and Rovira Kaltwasser 2012), others point to those features of populism which are hardly compatible with the institutional and ideological complexity of liberal democracies: anti-pluralism, uncritical majoritarianism and a Manichaean worldview (Rummens 2017;Urbinati 2017).The potential tension between populism and democracy is not only a macro-level problem.Zaslove et al. (2021) argue that relatively little is known about the relationship between citizens' populist attitudes and their commitment to democracy.Their results support the democratic potential of populist attitudes in several dimensionsalthough they do not spur electoral participation (see also Nemčok et al 2022).However, Guinjoan (2022) demonstrates that commitment to democratic values is not general among people with strong populist attitudes and that there is a difference between right-wing and left-wing populism.It is probably fair to say that the debate has not been settled yet, but there are pieces of evidence showing that under certain conditions populism can become a threat to democracy (see Pappas 2014).
Our paper aims at contributing to the scholarship on the individual aspects of populism and does not wish to take a strong stance in the democracy and populism debate.Still, we adopt a moderately normative position accepting the thesis of normative democratic theory that the quality of democracy at least partly depends on the democratic quality of the political behaviour of the citizens as well as their commitment to democracy.We posit that populism is a borderline phenomenon that reflects on the problems of liberal democracy and while it has the potential of providing or triggering remedies to those problems, it also has the potential of undermining the fundaments of liberal democracy.We assume that this tension is reflected at the attitudinal level as well and, ultimately, we hypothesize that the set of attitudes, values and knowledge that we label democratic capacities interacts with political efficacy feelings in a way as to limit the populist appeal.If we are right our results may provide further arguments to those who hold that strong populist attitudes may be more a problem than an opportunity for the prospect of liberal democracy.

Populist attitudes
In recent years, populism research has expanded into two new directions.While before populism research has been focusing on defining populism and describing related meso-level political phenomena (populist movements, parties, leaders, and discourses), recent research has improved our understanding of populism in terms of policy making (e.g.Bartha et al. 2020) on one hand and the micro-or individual level of political attitudes, on the other.Developments concerning the individual level have been substantially advanced by the introduction of populism scales measuring attitudes that predict voting for populist parties.The first such scale was proposed by Hawkins et al. (2012) and further developed and popularized by Akkerman et al. (2014).Other, slightly modified versions have also been proposed in the literature (e.g.Castanho Silva et al. 2019).The scales are based on the ideational approach of Cas Mudde and their empirical validity and distinctness from other measures are proven (Castanho Silva et al. 2019;Geurkink et al. 2020).
Populist attitudes predict voting for populist parties, both left and rightit is actually the only measure to predict vote for these ideologically different parties (Geurkink et al. 2020).However, populist attitudes are widely present in today's societies and are not confined to the voters of populist parties (Hawkins and Rowira Kaltwasser 2019).That is why we use measures of populist attitudes as a dependent variable in our study instead of populist party support, unlike Krause &Wagner (2021), andMagni (2017).
Strangely, the literature is rather scarce on the roots of populist attitudes, as the focus has been more on what populist attitudes explain.Existing studies mostly concentrate on the social-economic antecedents.Evidences, however, are not straightforward: different studies show partly contradictory findings, while they generally agree that effect sizes are small.On a Swiss sample, Bernhard & Hänggli (2018) found that several variables have statistically significant, although rather weak effects on populist attitudes.Both education and income levels have negative, while age and unemployment as well as being male have a positive effect.However, the analysis of Rico et al. (2017) on a Spanish sample offers partly different results.They found also that education has a rather weak negative, while age has a similarly weak positive effect on populist attitudes, but income level and being unemployed is unrelated to them.Flemish data of Spruyt et al. (2016) confirm the (weak) negative effect of education and the (weak) positive effect of age, but political interest seems to weaken populist sentiments, while Bernhard & Hänggli (2018) on their Swiss sample found a positive influence of political interest on populism.Relative deprivation was found to enhance populist attitudes by Spruyt et al. (2016), but subjective social status has no effect in Abadi et al.' cross-country investigation (2020).Finally, a growing number of studies demonstrated the role of emotions, particularly that of anger on populist attitudes (Abadi et al. 2020;Magni 2017;Spruyt et al. 2016).All in all, socio-demographic factors seem to have limited predictive power, if any.Overall, populist attitudes appear as political, and maybe emotional, phenomena.In this sense, our endeavour to understand them in the light of political efficacy feelings and democratic capacities seems legitimate.

Political efficacy
The concept of political efficacy refers to citizens' subjective attitudes about their own role and position within the polity (see Campbell et al., 1954).The literature differentiates two variants of political efficacy: external political efficacy captures the extent to which people feel that the political system is responsive to ordinary citizens' demands, while internal political efficacy refers to the individuals' subjective conceptions about their own 'competence to understand, and to participate effectively in, politics' (Niemi et al., 1991(Niemi et al., :1407)).
External and internal efficacy are demonstrated to be crucial antecedents of political participation and behaviour (Abramson & Aldrich, 1982;Finkel, 1985).However, although political efficacy predicts political participation, it says nothing about the 'quality' of it: whether it is about democratic participation or populist, even radical, extremist mobilization.Political mobilization as such is not a cure for populism, because populism itself also mobilizes citizens (Anduiza et al., 2019).From a normative standpoint what is needed to make political efficacy truly democratic is reflective engagement which does not exclude a critical stance towards politics, but rather scrutinizes not only the power holders but their critiques as well.
That is, the participatory potential of political efficacy can be exploited by populist mobilization.However, the literature is relatively scarce concerning the relationship between political efficacy and populism.This is especially true for the relationship between political efficacy and populist attitudespopulism at the individual level was mostly measured with populist party support, although we know that populist attitudes are widely present in society.
A few studies focused on the association between external political efficacy and the support for populist parties (e.g.Geurkink et al., 2020;Krause & Wagner, 2021;Magni, 2017;Rooduijn et al., 2016) since the theoretical case seems compelling: dissatisfaction with the responsiveness of the political system is expected to manifest itself in populist anti-elitism.However, empirical evidences are contradictory.While Rooduijn et al. (2016) found a negative association between external political efficacy and populist party support, Magni (2017) argues that the relationship exists, but it is mediated by emotions, while Krause and Wagner (2021) demonstrated this association exclusively concerning non-established populist parties.Other studies did not find any relationship between external political efficacy and populist party support (Giebler et al., 2020;Van Hauwaert & Van Kessel, 2018).These results suggest that indeed, the populist attitudes scales and the political efficacy scale measure different phenomenaeven if external political (in)efficacy feelings seem to fall close conceptually to the essence of populist attitudes.
Even less is known about the effect of internal political efficacy on populism.Rico et al. (2020) argue that stronger internal political efficacy feelings predict greater openness to the populist appeal because, for someone who believes in their political competencies, it is easier to identify with populist arguments that emphasize the primacy of common people over the political elites.However, Van Hauwaert and Van Kessel (2018) did not find any relationship between populist party support and internal efficacy that they use as a control variable in their model.
Summing up, the relationship between populism and political efficacy remains unsettled.We assume that the conflicting findings are partly caused by the fact that the effect of external and internal efficacy on populist leaning is different among those who owe certain democratic capacities that are necessary for democratic participation from people who lack this background.
Democratic capacities. 1 It is intuitively compelling that some kind of knowledge, personal resources, and 'skills' are needed to participate effectively and democratically in politics (Kirlin, 2005).These are not only pragmatic considerations, but ideas founded in a normative theory of democracy that impose certain requirements not only at the level of the political system or the elite, but also at that of the citizens.According to these, ideal democracy can only work if citizens have certain democratic capacities.That is, our conceptualization of democratic capacities is largely based on normative democratic theory which emphasize that citizens should makeat least to some degreeinformed political decisions (Delli Carpini & Keeter, 1996), while accepting the constitutive values of democracy (Bene & Boda 2021;Brettschneider, 2006;Caprara & Vecchione 2017;Dahl, 1956;Finkel 2003;Verba et al. 1995).
First, some degree of information and knowledge is needed to substantively participate in politics since in democratic settings citizens have the responsibility to make decisions on the most important political questions (Delli Carpini & Keeter, 1996).Consequently, citizens need to have some general political knowledge, but also keep up with the day-to-day political processes.However, having political knowledge and being upto-date is not enough to make informed political decisions.Citizens need to be reflective and conscious about their decisions (Dewey, 1954) which may be jeopardized by strong emotional attachments to parties or leaders.Partisanship is an ambiguous phenomenon.
While it has the merit of spurring political interest and participation (Abramson & Aldrich 1982), recently the growing literature on political polarization and its pernicious effects on democracy has pointed to the dangers of extreme partisanship (e.g.McCoy & Somer 2019).Among other negative effects, a high level of partisanship may undermine reflexivity and introduce serious biases in the reasoning and evaluation of political information (Shapiro & Bloch-Elkon, 2008;Petersen et al, 2013).
Democracy as a moral enterprise cannot thrive without citizens who are committed to democratic ideals, values and norms (Caprara & Vecchione, 2017).In this perspective, democratic capacities should be more than instrumentally useful practices and competencies.The idea of democracy is based on some core values that citizens are required to embrace to make the democratic system work properly.First, citizens' political and legal equality is a prerequisite of democracy (Dahl, 1956).Second, the normative justification of democracy rests upon the premise that citizens can make free and autonomous decisions (Lakoff, 1990).Third, democracy is a social activity where individuals need to consider that others also have legitimate interests, opinions and values.Citizens must tolerate and listen to opinions different from their own because democracy is based on the idea of exchanging and confronting competing views and interests (Arendt & Kohn, 1968;Mutz, 2006).These values correspond to the three core values of democracy.Namely, the equality of interests, political autonomy, and reciprocity, as conceptualized by Brettschneider (2006) Summarizing the above arguments, we define democratic capacities as those political competencies, knowledge, information and democratic commitment which enable citizens to effectively participate in democratic politics.While numerous studies have investigated how the level of these factors shapes political outcomes, it is not their pure level that matters for our conceptualization of political capacities.We argue that all of these capacities are needed at least to a minimal degree for democratic participation.
There are two key implications for this conceptualization.First, we are interested in whether or not people have these capacities at a democratically satisfying minimum level instead of focusing on the overall level of their capacities.If we focus on the overall level, the difference between moderate and increased levels of capacities could be the same as the difference between low and moderate levels.However, from the perspective of normative democratic requirements, a moderate level of news consumption or knowledge can be appropriate (see, Schudson 1999), while a low level is rather problematic.In other words, when thinking about democratic capacities, we are interested in differences between people who do and do not comply with the normative requirements, rather than in differences between people who more or less meet with these criteria.This is why we opt for a dichotomizing approach with clear minimum criteria rather than focusing on the level of the dimensions of these capacities (Song et al., 2013).
The second implication of our conceptualization is that we treat these dimensions together as they can enable the normative citizenship ideal only if they work together: those citizens are considered to have complete democratic capacities who at least sometimes follow the news, at the same time have some political knowledge, are not extremely partisan and accept the democratic values of equality, autonomy, and tolerance.This conceptualization also implies that these dimensions could not be automatically lumped together by different data-collapsing methods.All of these criteria need to be met to realize the citizenship ideal that is required by normative democracy theory.
Overall, we assume that besides their 'quantitative' effect in terms of participation, a minimum level of democratic capacities can bring in a 'qualitative' aspect as well: they may protect populism and other potentially non-democratic political action.Democratic capacities imply an acceptance of tolerating political differences and a better understanding of complexity in decision-making, and these characteristics are hardly compatible with the anti-pluralism, uncritical majoritarianism and Manicheanism of populism.
More importantly, we expect democratic capacities to moderate or suppress the potentially positive effect of high internal and low external political efficacy on populist attitudes: democratic capacities may protect individuals to turn these feelings into populist attitudes.
In line with this argument, we hypothesize: H1. Democratic capacities moderate the effects of (a) external and (b) internal political efficacy on populist attitudes.
POLITICAL RESEARCH EXCHANGE -H1a: low external political efficacy predicts stronger populist attitudes for people having incomplete democratic capacities, but it is ineffective in the case of people with complete democratic capacities.
-H1b: a higher level of internal political efficacy predicts stronger populist attitudes for people having incomplete democratic capacities, but it is ineffective in the case of people with complete democratic capacities.

Data and method
We test our hypotheses on an original online survey taken in 2019 from 15 European countries (Germany, the UK, Czechia, Hungary, Italy, the Netherlands, Poland, France, Slovakia, Lithuania, Denmark, Turkey, Spain, Greece, and Bosnia and Herzegovina).Our desired representative sample size amounted to approximately 500 respondents per country, while quotas based on current census data were set up for gender, age and geographical region.The fundamental eligibility criterion for respondents was having lived in their current country of residence for at least 10 years, which we consider a sufficient time frame to feel at home in the country of residence.The questionnaire of the survey was developed by the researchers of the DEMOS project and the survey was administered by the University of Amsterdam. 2 To measure populist attitudes we use a slightly modified version of the Akkermanscale, based on existing items by Castanho Silva et al. (2019), consisting of People-Centrism (e.g.'Politicians should always listen closely to the problems of the people'), Anti-Elitism (e.g.'The political elites have failed to protect our cultural identity') and Manichaean Outlook (e.g.'You can tell if a person is good or bad if you know their political views').Each variable 3 was measured on a 7-point Likert-scale.Since the Cronbach-alpha value of the 9 items was only 0.48, we eliminated the most poorly performing item from each dimension.The remaining 6 items' 4 coherence was satisfactory (alpha = 0.58).We created an aggregate variable as a means of the 6 items (M = 5.06, SD = 0.81).
The concept of external political efficacy was measured through two items ('The political system in my country allows people like me to have a say in what the government does' and 'I cannot effectively influence politics in my country') on a 1-7 Likert-scale, but since their association was low (alpha = 0.48; Pearson's R = 0.31) we opted for the first item since it refers more to the system and less the respondent (M = 3.94; SD = 1.77).
Internal efficacy was measured through two items ('I am able to take an active part in groups involved in political issues' and 'I am confident that I am able to participate in politics'), measured on a 7-point Likert-scale.Since their internal coherence is high (alpha = 0.83; Pearson's R = 0.71), the variable was created as their mean (M = 4.03; SD = 1.56).
Based on our theoretical discussion, we differentiate between people with complete and incomplete democratic capacities.As our conceptualization requires to (a) use minimum criteria rather than scales for the components and (b) treat these dimensions together, we followed a 'meaningful grouping' strategy to create a composite variable from directly measured variables.Meaningful grouping is a 'nonstatistical combination of selected original variables based on the interpretation of the variables' values or scores' guided by theoretical considerations (Song et al., 2013).The outcome of this grouping is a set of theoretically defined meaningful categories.We consider people having complete democratic capacities if they (1) have a certain level of political knowledge; (2) consume political news at least sometimes; (3) are not extremely partisan and (4) identify with basic democratic values.Therefore, for measuring democratic capacities we included items on political knowledge, information, partisanship, and democratic values.Since we understand democratic capacities as something people can have or not have in the minimum form we settled, we defined each of them as dummy variables, where those who meet the pre-defined minimum criteria are considered to have a particular capacity.Naturally, defining such minimum criteria is not self-evident and is a somewhat arbitrary process.Nonetheless, we draw upon our conceptualization, and make efforts to translate the categories of our variables into the minimum criteria defined above: sometimes follow the news, at the same time have some political knowledge, are not extremely partisan and accept the democratic values of equality, autonomy, and tolerance.
We measured the information efforts of respondents through the frequency of following political news on a Likert-scale (1every day, 7never).Again, we created a dummyvariable, setting the threshold at 3 on the Likert-scale, so those were considered to have the democratic capacity of news consumptions who consume news at least 'some day' because this is what the closest to our conceptualization of 'sometimes' following the news (1 = 78%).Of course, we cannot know whether the news consumed by the respondents is from reliable sources or constitute what might be called 'fake news', the spread of which is a growing problem everywhere.We can only assume that our variable is a valid measure of political information-seeking behaviour.
The most common way to measure political knowledge is the application of factual survey questions referring to 'political institutions' formal arrangements and their decision-making procedures, current political/policy facts and politicians' party identification' (Strabac & Aalberg, 2011), but it is very difficult to apply in a cross-national survey reliably and validly.However, structural political knowledge measures if people can see the interrelatedness of political concepts (Hutchens et al., 2015).We measured political knowledge as a dummy variable through the answers given to two, mutually exclusive statements: according to the first the government should lower taxes, while the second stated that the government should increase welfare spending.The significant negative correlation between the answers signals that the majority understood that the two statements did not hold together; however, those who 'strongly agreed' (6, 7) or 'strongly disagreed' (1, 2) with both statements were considered as those respondents were qualified as lacking political knowledge (1 = 65%).We argue that the extreme agreement/disagreement with two statements which are commonly perceived as politically conflicting indicates that the level of political knowledge is suboptimal.
In a similar vein, we included several items measuring extreme partisanship ('When I speak about this party, I refer to them as 'my party'.' 'If this party does badly in the opinion polls, my day is ruined.''When people praise this party, it makes me feel good.') in 7-point Liker-scales (based on Bankert et al., 2017).The coherence of the partisanship items is very high (alpha = 0.86), therefore we used their means.As we want to differentiate only the extreme partisans and on a 7-point-scale the two highest values can be reasonably considered extreme cases, the mean value above 6 is coded as 0 (extreme partisan), while all other respondents have the democratic capacity of being non-extreme partisans (1 = 93%).
To measure democratic dispositions we asked for the respondents' agreement with three statements on equality 5 , tolerance, 6 and autonomy or personal freedom 7 (based on European Social Survey).As the coherence between the three items is relatively high (alpha = 0.73) we created a dummy-variable, where those were considered to accept constitutive democratic values who at least 'somewhat agree' (5 on the Likertscale) with each of the value statements (1 = 75%).This threshold was used because it still indicates acceptance in line with our conceptualization, while this is not true for the next category labelled as 'neither agree or disagree'.
Summarizing the democratic capacity measures: we created a dummy-variables and coded 1 the respondents who do not have contradictory expectations towards the government (political knowledge = 1), seek information on politics with some frequency (information = 1), are not extremely partisan (non-partisanship = 1) and accept basic democratic values (values = 1) (see, Table A2).It is important to highlight that these factors do not seem to be unrealistic democratic ideals.As our descriptives show each capacity under investigation characterizes the large majority of the sample.Overall, these seem to be rather common individual capacities in democratic societies.Naturally, fewer people have all these capacities, but their share is still remarkable: one-third (34%) of the total sample can be described as having complete democratic capacities according to our criteria, while two-thirds of them have incomplete democratic capacities (66%).We are mainly interested in the differences between people with complete and incomplete capacities as we argued that the lack of any of the capacities listed above result in less protection against populist appeals.
The following variables were used as controls: the strength of national identity 8 , worries about the future 9 , gender, age, type of settlement, subjective welfare, and education level.

Findings
Descriptives Before testing our hypotheses, it is useful to examine the links between our dependent and independent variables whose interaction is the focus of this research.Figure 1 shows the correlations between populist attitudes, external political efficacy, internal political efficacy, democratic capacities, and its components.
In light of the literature, it is not surprising that internal and external efficacy are moderately correlated, however, the moderate effect size also confirms that the two constructs capture different dimensions of political efficacy.In line with our argument, political efficacy has little to do with democratic 'qualities' as its dimensions are weakly correlated with our democratic capacity variable: more efficacious people are somewhat more likely to have complete democratic capacities, but this association is far from being decisive.Extreme partisanship may cause stronger efficacy feelings as indicated by the weak/moderate negative correlation between these variables, and more efficacious people, especially those with higher internal efficacy, are more likely to consume news regularly.External political efficacy is weakly and positively correlated with political knowledge, however, internal political efficacy is not associated with it.The opposite pattern is found in the case of democratic values which are positively related to internal, but not to external efficacy.Overall, it seems that while some weak correlations exist between the dimensions of political efficacy and democratic capacities, they are far from being closely related to each other.
The theoretically defined democratic capacities are not associated with each other very well, moreover, they are sometimes negatively correlated.While all of them are important democratic 'qualities' as we discussed in the theoretical section and characterize most voters in democratic societies, only a minority embraces all of them.This fact justifies our decision to follow a 'meaningful grouping' strategy and treat all of these qualities as separate components of our composite democratic capacities variable.For instance, regular news consumption is democratically valuable only if it is paired with political knowledge and less extreme political bias, while these factors are negatively correlated with news consumption.As we argued all these capacities are needed to have for meaningful democratic behaviour.
As expected, populist attitudes are associated with external (negative) and internal (positive) efficacy, but these associations are weak.They are also weakly and negatively related to democratic capacities.Extreme partisans and people with limited political knowledge are more likely to have populist attitudes, but democratic values and news consumption are positively associated with populist feelings.
Figure 2 shows the share of people with low/moderate/high external/internal political efficacy levels across people with incomplete and complete democratic capacities.There are only minor differences, we can find a bit more people with low external and internal efficacy in the incomplete capacity group, while a moderate level is slightly more typical in the complete capacity group.This result suggests that the main concepts barely move together, therefore their combination can result in meaningful patterns.

Hypothesis testing
For the analysis, we used OLS regression.Because of the nested (country-level) data structure countries were built into the model as fixed effects.As a robustness check, we ran the model with clustered standard error and as a multilevel random intercept model, but it yielded the same results (in terms of the sizes and significance of the effects), therefore we opted for the OLS + country-level fixed effect alternative.
Table 1 shows the model summaries.Model 1, which does not contain interaction terms, suggests that in line with our theory while the increase in external efficacy depresses the likelihood of populist attitudes, the higher level of internal efficacy result in more populist views.All else being equal, a person with a maximum level of external Table 1.OLS regression models on populist attitudes.efficacy is less populist with 0.22 points on the 7-point populist attitude scale than a respondent with the minimum level.In contrast, someone with a maximum level of internal political efficacy is more populist with almost 0.5 points (0.47) than others with a minimum level.Overall, we can see that these effect sizes are not too large, but still remarkable and significant.
At the same time, it is also confirmed that people with complete democratic capacities have lower populist scores than those with incomplete capacities which yields a sort of construct validity for our variable.However, the effect size is rather weak again: all else being equal, people with complete democratic capacities will be less populist with 0.14 points than respondents with incomplete capacities.
Our hypotheses are directly tested in Model 2 where the respective interactions are added.We can see that H1a is rejected since the effect of external political efficacy on populist attitudes is not moderated by democratic capacities.Low external political efficacy indeed predicts stronger populist attitudes, but we cannot say that it is ineffective for people with complete democratic capacities.Democratic capacities do not seem to have a mitigating effect on populist attitudes in case of low external efficacy feelings: people who are dissatisfied with the political system are more likely to identify with populist appeals irrespective of their democratic capacities.
In contrast, H1b is mostly supported: a higher level of internal efficacy predicts stronger populist attitudes, but it is less effective for people with complete democratic capacities.However, despite the significant interaction effect, we need to be careful here.As Figure 3 and Figure 4 indicate, higher internal political efficacy still significantly increases populist attitudes in the case of people with complete democratic capacities, but the effect size is significantly smaller vis-á-vis people with incomplete capacities.Internal political efficacy can explain 0.6 points difference in the 7-point populist attitude scale in the case of respondents with incomplete capacities, while this variation between the less and most efficacious people is only 0.16 points for people with complete capacities.So, in contrast to our hypothesis, we cannot say that internal political efficacy is ineffective in the case of people with complete democratic capacities, but this association is remarkably weaker in their case.
Overall, it seems that the main function of democratic capacities in this equation is to counteract the effect of internal political efficacy on populist attitudes and connect them mainly to the level of external efficacy.As a result, these capacities constitute to some extent a 'safety net' against the negative effect of political confidence.This way the key in the explanation of populist attitudes will be the external political efficacy: for these people, more trust in the responsiveness of the political system will result in less populist attitudes irrespective of their democratic capacities.In contrast, the effect of internal political efficacy is largely turned off by democratic capacities.Consequently, the combination of a high level of external efficacy and complete democratic capacities is extremely beneficial against populism at any level of internal efficacy.

Sensitivity analysis
To gain a deeper understanding of the complex association between efficacy feelings, democratic capacities and populist attitudes we also ran some sensitivity analysis.First, while we define the concept of 'democratic capacities' in a deductive, theory-guided way, it is important to uncover which elements of the concept matter in this 'safety net' effect.Therefore, we re-ran the models with different reduced forms of the 'democratic capacities' variable by omitting one or some of the variables (Table 2).
Findings show that if we leave out any of the components, the alternative interaction effects remain very similar to the overall effect.Also, in separation, it is only the political knowledge that produces a similar effect.Both the lack of extreme partisanship and political knowledge can result in significant effects when they are combined with any other components.However, if any of these two variables are not involved, the remaining two components (democratic values, and news consumption) cannot yield a significant 'safety net' effect in any composition.
Based on this analysis it seems that the 'safety net' effect does not require to have complete democratic capacities: a similar effect can be reached via only political knowledge or the combination of the lack of partisanship with any other factors.But in any case, it is confirmed that a stronger 'safety net' effect is mostly the result of the combination of certain democratic capacities.In a second sensitivity analysis, we tested if this effect differs across the different dimensions of populism (Table 3).In each of the three dimensions, there are significant negative interaction effects, but a stronger effect was found in the case of Manicheism.Consequently, democratic capacities can be particularly effective when the positive associations between internal efficacy and Manicheism need to be neutralized.The explanation is that this is probably the least democratic dimension of populism.However, it is also important to note that while the interaction effect is weaker for anti-elitism and people-centrism, in these cases our H2b is completely confirmed since the effect of internal political efficacy is insignificant for people with complete democratic capacities, while it positively affects populist attitudes of our respondents with incomplete capacities (see, Figure 5).

Discussion and conclusion
Arguably 'democracy is a moral enterprise that rests largely upon the public morality of its citizens' (Caprara & Vecchione, 2017, x).If there is a mismatch between how the democratic polity operates and how citizens conceive their aspirations, needs and identities vis-à-vis the polity, this situation might be blamed either on the polity and/or the citizens.This paper shed some light on the citizen side.We showed that certain individual characteristics rooted in efficacy feelings as well as certain democratic capacities may protect against the populist appeal.This is not to deny that social problems and other contextual circumstances may exist and, sometimes understandably, trigger anger or even populist action.Addressing the populist challenge may require institutional changes and better involvement of citizens in politics.However, populism cannot be properly addressed by citizen mobilization and participation alone since populist politics is also about mobilizing dissatisfied citizens.The complex relationship between populism and political participation is reflected in our data in the varied association between populist attitudes and political efficacy feelingsthe latter being a recognized predictor of political participation.
One may argue that citizen participation in politics should be meaningful in substantial terms and give a sense of true influence to neutralize the populist appeal.We cannot but agree with this proposal.However, to realize it changing the operation of the political system (e.g.providing more fora for participation) might not be enough if citizens are not prepared to cope with the complexity of democratic politics.The starting point of our study was that certain democratic capacities of the citizens help them to meaningfully participate in politics which is good for democracy from a normative point of view and which reduces the populist appeal.
Our findings show complex interactions between political efficacy and populist attitudes.The complexity of this relationship may explain the somewhat contradictory results of previous studies.In line with our theoretical expectations, low external and high internal political efficacy is associated with more populist attitudes.However, when we consider the moderating effect of democratic capacities, the overall picture has changed.In contrast to our expectations, the effect of external political efficacy on populist attitudes is very similar for those having complete or incomplete democratic capacities.If people feel that the political elite is not responsive to their demands, they will be more open to populist arguments irrespective of their democratic practices and values.However, democratic capacities yield a safety net against the populism-boosting effect of internal political efficacy.Stronger political confidence leads to more populist attitudes to a lesser extent if people have democratic capacities.For this reason, in case of these people, it is mostly the perception of the political elite that matters, but less their belief in their political capabilities.Overall, these findings suggest that self-confidence in politics is not necessarily good and can lead to strong populism if it is not combined with democratic capacities.
We argue that this finding has both theoretical and practical implications.In theoretical terms, it brings further nuances to the debate on the democratic nature of populism.While it does not offer an ultimate argument against the democratic aspects of populism, it certainly shows that from a normative point of view, there are concerns about them beyond the usual claims about the conflict of populism with liberal values and institutions, since self-confident, but democratically poorly capacitated people are especially affected by the populist appeal.If democracy ideally needs citizens who possess a certain level of political knowledge and information and who are committed to the values of pluralism and respect for other opinions, populism might be seen as a potential threat to the democratic ideal, since it has a mobilizing potential toward people with weak democratic capacities.Populism may endanger liberal democracy not only through the institutional, constitutional reforms it proposeswhich point towards a majoritarian democracy with weak checks and balances and the lack of protection for minority rightsbut through mobilizing people who support this limited version of democracy.
In practical terms, our finding suggests that developing the democratic capacities of citizens may protect against the populist appeal.It goes beyond the scope of this paper to investigate whether and how democratic capacities can be developed.While there is a recurrent debate on the effectiveness of 'citizen education', evidence shows that people's political competencies and democratic attitudes can indeed be developed (e.g.Finkel, 2003).However, besides formal citizen education the democratic atmosphere of the school as well as other possibilities of democratic participation seem to be more influential in this respect (Perliger et al., 2006).It seems that while cognitive political skills can be taught, democratic capacities and attitudes are best developed through practice and experience (Werner, 2020).
Of course, our study has its limitations.Establishing causal relationships is always problematic, especially through the cross-sectional analysis that we apply.Although we formulated hypotheses for the investigation, we consider our study more like an exploration of new ideas than hypothesis testing in the strong sense.A further limitation is the sample.While we use a relatively large international sample, data come exclusively from European countries.Further research should explore whether the relationships and patterns we found between political efficacy, democratic capacities and populist attitudes hold in different political cultural contexts as well.Future research may also bring political behaviour into the picture, investigating how the different attitudinal constellations lead to specific political choices, like voting for populist parties or participating in politics.

Notes
1.The following section is based on Bene and Boda (2021).2. DEMOS -Democratic efficacy and the varieties of populism in Europe was an EU H2020 project, see https://demos-h2020.eu/en.We would like to thank David Abadi and Agenta Fischer from the University of Amsterdam for administrating the survey and finalizing the dataset.3. See Table A1 in Appendix for the operationalization and descriptives of all variables.4. People-centrism: (1) "Politicians should always listen closely tot he problems of people'; (2) "The will of people should be the highest principle of a country's politics.Anti-elitism: (3) "The government is pretty much run by a few big interests looking out for themselves'; (4) 'Quite a few of the people running the government are crooked'.Manichaen outlook: (5) 'You can tell of a person is good or bad if you know their political views'; (6) 'The people I disagree with politically are just misinformed'.5. 'I think it is important that every person in the world be treated equally.' 6. 'It is important to listen to people who are different from me even if I disagree with their views.' 7. 'It is important to me that I am free to make my own decisions, independently of others.' 8. 'Being [my nationality] is an important part of who I am.' 9. 'I am anxious about what the future will bring.'

Figure 2 .
Figure 2. The share of people with different political efficacy levels within the groups of incomplete and complete democratic capacities.

Figure 3 .
Figure 3. Marginal effects for the interaction term of internal political efficacy and democratic capacities on populist attitudes (model 2).

Figure 4 .
Figure 4. Effect sizes for the interaction term of internal political efficacy and democratic capacities on populist attitudes (model 2).

Figure 5 .
Figure 5.Effect sizes for the interaction term of internal political efficacy and democratic capacities on different dimensions of populist attitudes.

Table 2 .
The coefficients of interaction between internal political efficacy and democratic capacities in models with alternative democratic capacities variable.

Table 3 .
The coefficients of interaction between internal political efficacy and democratic capacities in models with alternative populist attitude variables