Attitudes vs. actions? Direct-democratic preferences and participation of populist citizens

Abstract Direct-democratic instruments are often seen as a cure for the current disenchantment with representative democracy. While research has begun to investigate whether direct-democratic instruments match the procedural preferences of populist citizens, it is an under-researched question whether populist citizens and populist voters would actually participate in direct-democratic voting. This study advances existing research by focussing both on direct-democratic preferences and effective direct-democratic participation of different types of populist citizens in Germany—populist voters on the right and left and citizens with general populist attitudes. Analyses based on survey data from 2018 show that while populist citizens and right-wing populist voters display strong attitudinal sympathy for direct-democratic instruments, especially right-wing populist voters fail effectively to participate in direct-democratic voting. Those who strongly participate in direct-democratic procedures are sophisticated and postmaterialist citizens whose preferences hardly align with those of right-wing populist voters and citizens, widening rather than closing the ‘representation gap’.

is an under-researched question, however, whether populist citizens-i.e. citizens with general populist attitudes-and populist voters would actually participate in direct-democratic voting when respective instruments are available in real-world political practice. A discrepancy between attitudinal sympathy and effective participation in direct-democratic procedures would raise doubts whether democratic satisfaction of populist citizens and voters could be enhanced in the long run and whether the 'preference gap' and the 'representation gap' can be easily closed by introducing more direct-democratic instruments. Put differently, if populist citizens and voters fail to participate in direct-democratic votes, then resulting decisions might not reflect their substantive policy preferences and exacerbate their political disappointment (Bowler and Donovan 2019); and we do know from previous research that outcome perceptions shape citizens' attitudes towards political processes (Werner 2020). Moreover, a discrepancy between attitudes and action might also indicate that preferences for direct democracy merely reflect opposition to the institutions and actors of the representative system (see, e.g. Goldberg et al. 2020) rather than a real and authentic demand for more participatory forms of democratic governance.
Research on this topic is sparse. The study of Bengtsson and Christensen (2016) suggests that there is no discrepancy between procedural ideals and actual participation patterns in general, but it performs no separate analysis of populist voters and citizens, and results are limited to the Finnish context with very weak direct-democratic experience. Jacobs et al. (2018), in turn, report a strong relationship between populist attitudes and referendum preferences in the Netherlands, but they do not find that populist citizens are more (or less) likely to participate in a specific referendum, namely the 2016 Dutch referendum about the EU-Ukraine Association Agreement.
This research note seeks to advance this important topic, whereby we focus on direct-democratic preferences of different types of populist individuals in Germany-populist voters on the right and left as well as citizens with general populist attitudes-and their actual participation pattern in direct-democratic voting generally (and not limited to a specific referendum vote). Despite the fact that referendums are non-existent at the national level, Germany represents a highly illuminative case since there is an active use of direct-democratic instruments at the municipal level, and to some extent also at the level of the Länder. 1 Hence, German citizens are familiar both with the existence of direct-democratic procedures (a crucial precondition for not tapping into 'non-attitudes' in survey studies) and may also have the possibility to participate in direct-democratic voting. This provides a distinct methodological advantage, since it allows us to track the actual participation patterns of populist citizens and voters under real conditions, rather than only under hypothetical ones (namely willingness to participate in hypothetical direct-democratic votes). Moreover, the fact that many (populist) citizens in Germany had the possibility to participate in different direct-democratic votes in past years gives us a more comprehensive picture of their direct-democratic participation (compared to single events such as the Dutch 2016 referendum). Issue-wise, direct-democratic votes in Germany often dealt with education and culture as well as democracy, state organisation, and domestic policy at the Länder level and questions of public education, social services, transport, and economic projects at the municipal level (Rehmet 2019;Rehmet et al. 2020). In the following, we first discuss how populist visions of direct democracy look like and why we expect a gap between attitudes towards direct democracy and effective participation. We then present our data and the results of our statistical analyses, followed by a discussion of what our findings mean for the capacity of direct democratic instruments to close the representation gap felt by populist citizens and voters.

Populist visions of direct democracy
Two facets of populist ideology might generate a positive view of direct democracy. First, in the populist understanding of democratic rule, intermediate actors, especially established parties, can be circumvented by direct-democratic elements. Populism is directed against these actors since they stand between the government and the people (Mudde 2004). Second, direct democracy can be a procedural instrument to impose the will of the majority against checks and balances of liberal democratic systems (e.g. Pappas 2019). As long as populist parties are in opposition-as in the German case-direct-democratic instruments might bring fore the 'true' will of the people and unmask the alleged 'corrupt' decisions and behaviour by established elites. By contrast, if populist parties are in power, direct democracy may not be so much on citizen participation per se but on the acclamation of populist leaders (see Caramani 2017;van Crombrugge 2021). Indeed, as exemplified by Viktor Orbán, populist leaders in power might use direct democracy as 'plebiscites' to celebrate the authentic people and show that the leader is in close contact with 'his people' (Urbinati 2019: 119-20). 2 Notice that populist understandings of direct democracy-both the motive to bypass legacy institutions (and actors) of the representative system and the acclamation of the leader-are in strong contrast with a participatory vision of direct democracy (Barber 1984) which emphasises self-determination and initiative of citizens where empowered and informed citizens directly participate in politics.
There are, however, differences between right-wing and left-wing populism. Left-wing populism is associated with a more inclusionary conception of the 'people' than right-wing populism (Caiani and Graziano 2019) and calls for a broad transformation of politics and society to better include disadvantaged groups. Not only are calls for particular direct-democratic instruments more important for right-wing populist parties than for left-wing populist parties, with the former relying more on 'anti-elitist conspiracies' that direct-democratic voting could 'circumvent' (Mudde and Kaltwasser 2013: 163). Left-wing populists might also be more open to subscribe to a more participatory vision of direct democracy because of their more inclusionary view of participation.

Attitudes vs. action
Attitude-action (or behaviour) gaps are topical in social and political behaviour (e.g. in environmental protection; see Claudy et al. 2013). Against this background, we expect that populist citizens and voters exhibit strong procedural preferences for direct-democratic instruments but may fail to effectively participate in concrete direct-democratic voting. Regarding preferences, populist citizens in general may see direct-democratic tools both in an ideological and instrumental manner, namely as a possibility to bring the alleged 'true public will' to the fore (which in their view is subdued by 'corrupt' elites) as well as a way to circumvent or get rid of unloved institutions or actors of the representative systems which they think have not served their policy preferences well. Existing studies document that there is indeed a link between populist attitudes and preferences for direct democracy (e.g. Jacobs et al. 2018;Mohrenberg et al. 2021;Zaslove et al. 2021). We expect, however, some differences between right-wing and left-wing populist voters as well as between populist voters and citizens with general populist attitudes. While supporters of populist parties do not only vote for them because the latter support direct democracy, populist citizens favouring direct-democratic procedures may also vote for other parties. Thus, the association between voting for a populist party and support for direct-democratic instruments might be fuzzy. For instance, Bowler et al. (2017) do not find strong support for referendums among populist voters in Australia, New Zealand, and Canada. We therefore expect that the relationship between populism and support for direct democracy will be stronger for populist attitudes than for populist vote choice.
By the same token, we expect the relationship between populism and effective participation in direct-democratic voting to be negative. First, it has been repeatedly emphasised that populist voters and citizens do not necessarily call for more political participation in terms of effective behaviour, which is what the use of direct-democratic procedures would require in practice. For instance, Mudde (2004: 558) points to the populists' focus on democratic output rather than on the opportunities for political participation: 'The current heartland of the populists does support democracy, but they do not want to be bothered with politics all the time' . Populist citizens rather prefer a passive role in politics (Heinisch and Wegscheider 2020;Stoker and Hay 2017) which is supported by empirical evidence on the limits of mobilization of populist voters by populist parties in Western Europe (e.g. Huber and Ruth 2017;Leininger and Meijers 2021). However, supporters of left-wing populist parties might be more active in direct-democratic voting than citizens supporting right-wing populist parties. In alignment with left-wing populist parties' ideology, they might favour broader political participation of citizens generally (e.g. de la Torre 2010). Such a 'participatory' vision of direct democracy could motivate them to effectively participate in direct-democratic voting when respective opportunities are offered. Finally, direct-democratic procedures in Germany might not attract citizens with populist attitudes or populist vote intentions since these participation opportunities may not be in line with the populist understanding of direct democracy. They are organised within the current representative system, and the procedures are often initiated by parties or movements of the established system, to which populists stand in opposition. From the perspective of populist citizens, participation in these procedures might not be suitable for circumventing the country's established political elites.

Data, constructs, and indicators
We use data from the 2018 German General Social Survey (GESIS 2019) and focus on two dependent variables. The first variable measures attitudinal support for direct democracy via the following question: 'It is not possible at present for citizens to vote directly on political issues at the federal level. Using this list please tell me how much you would be in favour of or opposed to introducing such direct citizen participation at the federal level.' Respondents could select an answer from a six-point scale ranging from 'very opposed' to 'very in favour' of this suggestion (more detailed information on the variables used in this study can be found in the supplemental material, Tables A-1 and A-2). The second variable focuses on the reality of direct-democratic participation. It indicates whether a respondent has taken part or not in a referendum/ plebiscite. Mean support for direct democracy at the federal level is rather high (3.64 on a scale from 0 to 5). Regarding patterns of actual participation, 40 percent of respondents indicate that they have already taken part in a referendum or plebiscite.
In order to measure populist orientations, five items are employed. They cover three core components of populism, namely anti-elitism, people-centrism, and the distinction between 'good' and 'evil' in politics (Mudde 2004). These elements are also addressed in established measures of populist attitudes (Akkerman et al. 2014;Schulz et al. 2018;Spruyt et al. 2016). While the data used in this study does not contain a very large number of items measuring populism, the wording of the available items is based on tried and tested instruments. In a principal component analysis, all five items load on one single factor (eigenvalue of 2.63, 53 per cent explained variance). Hence, we construct an additive index of populist attitudes, yielding a Cronbach's alpha of 0.77 which indicates good reliability.
The party-oriented facet of populism is operationalised using the respondents' likelihood to vote for the AfD and the Left Party. The focus on the AfD and the Left Party is backed by classifications of these two parties as populist parties in expert surveys (e.g. Rooduijn et al. 2019). Recent studies, furthermore, show that various elements of their respective programmes contain populist ideas (e.g. Hough and Keith 2019; Olsen 2018). 3 Populist orientations are positively correlated with the reported likelihood to vote for the AfD (Pearson's r: .28), but negatively correlated with the probability of casting a ballot for the Left Party (Pearson's r: −.09) or one of the other four parties currently represented in the German Bundestag. Thus, these correlations do not only illustrate differential relationships between populist attitudes and likelihood to vote for the right-wing and the left-wing populist party but are also indicative of existing differences between populist attitudes and actual voting for populist parties.
Several control variables are added to each of these models (see Table  A-2 in the supplemental material): political interest, level of education, postmaterialist values, satisfaction with democracy, external efficacy, trust in political parties, an individual's abstention in recent elections, left-right orientation, age, and gender.

Results
We first discuss the results of the model predicting attitudinal support for direct democracy (Figure 1). As expected, the stronger the populist attitudes people exhibit, the more they are in favour of direct citizen participation (at the federal level). Compared with the impact of the other variables in the model, the positive effect of populist attitudes on attitudinal support for direct democracy is very strong. This result fits in with recent findings on the relationship between populist attitudes and preferences for direct democracy. Figure 1 also shows that the more likely individuals are to vote for the AfD, the more they support direct democracy in Germany. As expected, the effects of populist attitudes on support for direct democracy are stronger than the effects of populist vote intention. This result is also in line with previous studies arguing that populist parties do not only attract people with populist beliefs but also people with other orientations who might be less in favour of referendums (Akkerman et al. 2014;Jacobs et al. 2018). This may be particularly the case for potential voters of the Left Party because there we find no link between support for this party and support for direct democracy.
When it comes to the other factors that might affect procedural preferences for direct democracy, the analysis points to the importance of disaffection (e.g. Donovan and Karp 2006): Dissatisfaction with the current state of democracy and perceptions of a lack of external efficacy of the current system foster support for direct democracy. Attitudinal and demographic factors related to the 'new-politics'-approach (Dalton et al. 2001) are less important for understanding direct-democratic preferences. The more educated people are, the more sceptical they are about direct citizen participation at the federal level (a finding in line with Anderson and Goodyear-Grant 2010).  Second, we focus on effective participation of populist citizens in direct-democratic votes in the past (Figure 2). Since there are considerable differences between the Länder regarding the number and the precise design of direct-democratic procedures (Grotz and Lewandowsky 2020), we ran multilevel models with 16 Länder as level-2 units (see model 1 in Table A-4 and the empty model equally shown in Table  A-4, in the supplemental material online). Results in Figure 2 show that AfD voters have a lower proclivity to participate in direct-democratic voting-despite the fact that the AfD is the most fervent partisan supporter of more direct-democratic procedures in Germany. For citizens with populist attitudes, the effect is negative but not statistically significant, while supporters of Die Linke have a higher proclivity to participate. Even though populist citizens are the most ardent supporters of direct-democratic procedures (even more so than sophisticated citizens; see Figure 1), they nonetheless do not have a higher propensity to participate in direct-democratic votes when such opportunities are offered in reality. We have run robustness checks by inserting a 'supply' variable in our multilevel models, which represents a rough approximation of the frequency of direct-democratic voting in the Länder. 4 It shows that the supply of direct-democratic procedures varies considerably between the Länder (with 0 procedures in Saarland and more than 1800 in Bavaria). Controlling for the 'supply' of direct-democratic votes, we find that 'supply' matters-the higher the possibility of direct-democratic voting, the higher the frequency of participation-but results for populist citizens and voters do not change (see model 2 in Table A-4). Overall, our study partly corroborates the finding from Jacobs et al. (2018)-focussing on a singular case, the Dutch 2016 referendum-showing that populist citizens in Germany do not have a higher proclivity to participate in direct-democratic voting than other citizens; we demonstrate, however that supporters of right-wing populist parties even participate less.
Zooming in on the societal groups who actually do participate in direct-democratic voting at higher propensities, we find a corroboration of the 'new-politics'-approach. It is citizens with higher political interest, higher education, and postmaterialist values who flock the direct-democratic urns. With regard to higher education, these are even citizens who do not have much attitudinal sympathy for direct democratic tools (see above).

Discussion
This study reveals an important paradox: while direct-democratic instruments are attitudinally popular for populist citizens in Germany, they are effectively used by others, namely sophisticated citizens. We find that especially voters of the AfD fail to participate effectively in direct-democratic voting when respective opportunities are offered. An important exception, however, are supporters of the populist left-wing party (Die Linke) who also have a high proclivity to participate in direct-democratic voting. This nuance notwithstanding, our findings not only challenge previous research suggesting a close link between procedural ideals and actual participation patterns (Bengtsson and Christensen 2016). They may also be indicative of an instrumental support for direct democracy especially by right-wing populist voters, whereby direct-democratic instruments are seen as a means to bypass or to get rid of unloved institutions and actors of the representative system, and/or of a non-participatory understanding of direct democracy. Overall, our results also point to a potential trade-off: an expansion of the range of direct-democratic procedures in Germany could possibly bring populist voters and citizens closer to the current political system in the first instance, satisfying their normative considerations on the use of these procedures (see Werner and Jacobs 2021) and enhancing democratic satisfaction. By the same token, new dissatisfactions could arise due to lower rates of participation in such procedures in the second instance, especially for right-wing populist voters. Those who actually participate in direct-democratic procedures are sophisticated and postmaterialist citizens whose preferences hardly align with those of right-wing populist voters and citizens, a pattern which eventually widens the 'preference gap' .
Of course, this study has a number of limitations. First, we need to further disentangle different types of populist citizens with a closer eye on who is in favour of participatory or 'plebiscitary' conceptions of direct democracy. Indeed, there might be differences between populists emphasising an anti-elitist position and those who actually value the people's will (see Mohrenberg et al. 2021). Second, our study does not take into account specific policy issues that might affect the participation of populist voters and citizens (see Werner 2020). As mentioned before, previous direct-democratic procedures have often dealt with issues that may not always be extremely salient to populist citizens (such as education, social services, transport, or economic projects). The participation of generally more passive right-wing populist voters in direct-democratic voting might be enhanced if topics such as immigration were on the agenda (which are particularly salient for populist citizens and entail the hope that they can win them), or when direct-democratic voting is about the acclamation of the leader or the party. Third, our study does not shed light on the question whether distrust of populist citizens in those 'elites' who organised the referendum (or initiative) hinders their willingness to effectively participate in direct-democratic voting. Future research will need to take an in-depth look at distrust and participation in an actual referendum vote. Fourth, we would expect differences in direct-democratic voting between the subnational level (where referendums in Germany take place) and the national level. Nation-wide and salient topics in referendum votes, their politicisation by populist parties, and higher media attention might trigger higher participation rates by populist voters. But this is speculative since it is highly improbable that Germany will introduce referendums at the national level for the foreseeable future. It will be worthwhile to explore such differences in countries where referenda take place both at the national and subnational level. To conclude, it is certainly not an existential problem for a democracy if attitudes and actions do not match. Citizens can still have 'ownership' feelings even if they do not participate (see MacKenzie and Moore 2020). However, the imbalance found here between attitudes and actions of populist citizens (and right-wing populist voters in general) is to some extent problematic because it underlines that the channels of direct citizen participation, which are often seen as a remedy against political disenchantment, have their limits in mobilising and integrating distant citizens (see Webb 2013). In fact, direct democracy (in Germany) turns out to be the playground for sophisticated, not populist citizens and voters. Notes 1. Between 1956 and 2018, at least 63 direct-democratic procedures took place at the level of the Länder-24 of which were initiated bottom-up, 39 were obligatory referendums (Rehmet 2019). Different forms of direct democracy are also applied at the municipal level. Between 1956 and 2017, a total of about 3,800 procedures took place at this level-when citizen initiatives and council referendums are counted ). 2. For instance, Viktor Orbán's government initiated the 2016 Hungarian migrant quota referendum (on the resettlement of non-Hungarian citizens to Hungary) to secure confirmation for his anti-EU course from the population. The result of the referendum was not valid as voter turnout was below the 50 per cent threshold. A similar motive is brought forth by observers as a reason for Orban's July 2021 announcement of a referendum on the country's anti-LGBT-law. 3. Of course, recent party studies equally report differences between the two parties in terms of their populist tendencies (e.g. Zaslove 2020, 2021;Olsen 2018: 81-82). 4. The supply variable is a measure of the number of direct-democratic procedures in each of the 16 Länder that took place between 1956 and 2017. It includes referendums at the level of the Länder (procedures initiated bottom-up and obligatory referendums) and at the municipal level (procedures initiated bottom-up and council referendums). Most of these procedures took place at the municipal level, however. Since we are not in a position to assign these cases to the respondent's municipal 'residency' , we refer to the Land level as an approximation of the level of supply.
preferences and the challenges of mapping and measuring deliberation and political communication. He has recently published in Science, the British Journal of Political Science and with Oxford University Press. [andre.baechtiger@sowi. uni-stuttgart.de]