Populism between voting and non-electoral participation

Abstract The article focuses on a neglected aspect of populist mobilisation, i.e. non-electoral participation (NEP), and elaborates on the extent to which populist party voters engage politically outside the polling station. The study addresses the question that populist parties and movements of the left and right originate from distinctive political experiences, and drive different forms of engagement with politics. The hypotheses are tested on NEP, based on populist party vote, social values and issue preferences relying on a unique cross-national survey. The findings significantly broaden our understanding of populist mobilisation and refine a number of notions related to the different logics of participation of the left and right. While challenging common understandings of populism as inherently distrustful and apathetic, and protest as an exclusive practice of the left, the study critically places NEP at the heart of populism in general, and populist right politics in particular.

Notwithstanding a constant drive to push the scholarship on populism forward, a significant lacuna persists regarding non-electoral participation (NEP). Very simply put, we seem to know a lot about what populist parties advocate and the behavior of populist party voters (e.g. Rooduijn 2018;Van Kessel 2015), but we remain largely oblivious of whether, how and how much populist supporters mobilise outside the ballot box. By NEP, we mean every action besides voting that ordinary citizens resort to in order to influence political outcomes such as distribution of social goods and norms (e.g. Teorell et al. 2007; Vr abl ıkov a 2014). Anduiza et al. (2019) have recently made steps in this direction and elaborated on the effects of populist attitudes on different forms of political participation their assumption being that populist attitudes can prompt participation and moderate the impact of political inequality. With our study, we take yet another route and address whether, and to what extent, populist party voters engage in non-institutional modes of political participation.
Our enquiry moves from the notion that identification with populist parties does not mean being apathetical or apolitical. To the contrary, the populist repudiation of politics 'can lead to full engagement in politics' (Taggart 2018: 81, emphasis added). At the same time, the formation of an anti-establishment political identity is related to populist agents' ability to activate these sentiments through elite criticism. In contrast to disengaged non-partisans, populist supporters are essentially politicised citizens prone to anti-establishment appeals (Mel endez and Rovira Kaltwasser 2019). We, however, wish to move beyond understandings of populist mobilisation as a purely electoral phenomenon driven by protest and discontent (Van Hauwaert and Van Kessel 2018), and factor in the role of NEP in the way populist party voters engage in the political process. Whether informed by social values or issue positions, we argue that populist party voters engage more in NEP than non-populist party voters. Such undertaking resonates with the work on populist 'movement parties' as political agents (della Porta et al. 2017;Kitschelt 2006;Mosca and Quaranta 2017;Pirro 2019;Pirro and Castelli Gattinara 2018), and with pleas to focus resolutely on the intersection between protest and electoral politics (Castelli Gattinara and Pirro 2019;Hutter 2014;Portos et al. 2019).
Having taken stock of these insights, we ask: Do populist party voters engage in non-electoral forms of participation? Do value orientations and issue preferences affect their participation beyond voting? For the sake of our enquiry, we lay out three sets of hypotheses. The first set of hypotheses suggests that support for a populist party can lead to NEP. The second and third set of hypotheses reflect upon the possibility that value orientations and issue preferences, alongside vote for a populist party, could affect the degree of non-electoral engagement.
Drawing on unique data on populist party voters across nine European countries, this study elaborates on their political engagement outside the ballot box. We specifically come to five key findings. First, populist party voters on the whole mobilise more than non-populist party voters. Second, left-wing voters resort more to NEP than supporters of rightwing parties. However, right-wing voters are not a fully demobilised set; populist right voters indeed engage more in NEP than non-populist right votersand generally as much as left-wing voters. Third, populist right voters engage more in NEP than any other voters with authoritarian views, refining the common notion that participation outside the ballot box is primarily a left-libertarian affair. Fourth, while populist right voters mobilise more than non-populist voters on the basis of negative attitudes towards migrants, pro-migrant attitudes prompt populist left voters to participate more beyond the ballot box. Fifth, populist right voters' endorsement of socioeconomic redistributive and egalitarian issues leads to NEP levels comparable to those of left-wing (either populist or nonpopulist) voters. Through the non-institutional participation of populist right voters, we may be then observing an unprecedented, all-round politicisation of populist politics, hence opening major prospects for the study of activism and populism.

Populism and participation
Populist actors have gained increasing prominence across postwar Europe. Older and newer populist formations in Italythe Lega and the Movimento 5 Stelle (M5S)coalesced to rule the country between 2018 and 2019, relying on the support of half of the Italian electorate. The illiberal turn that is sweeping through Central and Eastern Europe cannot be dissociated from populist parties' rise to power. And seismic events like Brexit would be difficult to interpret without the influence exerted by the populist UK Independence Party (UKIP). As a result, populism has also become an integral part of our political jargon. Correctly or not, the label has been attached to politicians, parties, movements, as well as voters of varying social backgrounds and ideological persuasions.
From whatever angle we approach the issue, populism is concerned with popular sovereignty, and the identification of two ideal loci of control: the government with its powers and the people as the ultimate source of authority in the state (Ochoa Espejo 2011). Populist parties and movements seek to mend the degrading of popular sovereignty, which is perceived to be corrupted by evil and self-serving elites. Populism presents a dualist worldview juxtaposing a righteous people and a treacherous elite, and prescribes that politics should be an expression of the general will of the people (Mudde 2004), as if 'democratic politics needed to be conducted differently and closer to the people' (Rovira Kaltwasser et al. 2017: 4). Populists therefore place a moral and antagonistic distinction between 'the people' and 'the elite' at the heart of their worldview, and we use the occurrence of such notions to locate actors within the populist set (Mudde 2017).
While deeming this aspect central for classification purposes, we must acknowledge the 'chameleonic character' and 'empty heart' of populism (Taggart 2000). This caveat has two broad implications. First, we recognise populist actors' ability to emerge in different socioeconomic and political contexts (as well as adapting to changing circumstances), without however confounding it with political opportunism (cf. Betz 1994). Second, we appreciate populism's 'thinness', by which it can be combined with additional beliefs such as nativism, socialism, liberalism, etc. (e.g. Abts and Rummens 2007;Mudde 2004;Pirro 2018;Van Kessel 2015;Zaslove 2008). We therefore speak of populist parties, movements and supporters that could be located anywhere along the ideological left-right continuum (Mudde 2017: 39), irrespective of their context of origin. Populism's 'travelling capacity' (Sartori 1970) is indeed of the utmost importance in tackling engagement by voters of disparate populist parties.
Despite its repudiation of politics as the process for resolving conflict, populism is 'not without politics or apolitical' and 'is driven to engagement with politics but in a way that is at odds with that politics' (Taggart 2018: 81). While the success of anti-establishment appeals depends on populist actors' ability to convince their supporters that they are not part of the entrenched power structure (Barr 2009: 32;Mel endez and Rovira Kaltwasser 2019), political attitudes are also relevant for the populist party vote (Lubbers and Scheepers 2000;Van der Brug 2003;Van der Brug et al. 2000). However, populist political participation has been almost exclusively interpreted in terms of electoral participation (cf. Anduiza et al. 2019).
In this article, we argue for an all-round politicisation of populism, whereby it can lead to fulli.e. electoral as well as non-electoralparticipation. After all, populism is omnipresent wherever there is representative politics, and it is strongly committed to active and direct participation (Taggart 2000). Therefore, populism's democratic character also reveals through political participation, which is 'the elixir of life for democracy ' (Van Deth 2014: 350).
Our concern thus rests with political participation and its intersection with populism. Political participation serves as an umbrella concept for multiple activities (Huntington and Nelson 1976) ranging from electoral activities, such as turning out to vote and campaigning, to protest-oriented activities, such as joining a rally or boycotting certain products (Dalton 2008;Van Deth 2014;Verba et al. 1995). The consequences of political participation are manifold and largely dovetail with the idea that citizens with populist attitudes 'spare no effort in monitoring those in power' (Rovira Kaltwasser and Van Hauwaert 2020: 15): 'Through participation citizens voice their grievances and make their demands heard to the larger public; they also make governments accountable and politicians responsive. The venues open for such activities are multiple' (Teorell et al. 2007: 334).
We should stress that political participation is a ladder made up of several rungs of intensity, thus electoral and non-electoral participation represent a continuum in terms of political engagement (Hutter and Kriesi 2013;Quaranta 2013;Van Deth 2014). Different forms of participation are not mutually exclusive; to the contrary, those who participate in other ways are also likely to cast a ballot at elections (Dalton 2008;Verba and Nie 1972). In our study, we devote attention to NEP, which can be defined as engagement in political behavior other than voting (Vr abl ıkov a 2014). NEP includes a wide array of forms of participation, such as demonstrating, petitioning, political consumerism, and online activism (Braun and Hutter 2016;Earl et al. 2017;Graziano and Forno 2012;Quaranta 2013).
Despite the ever-burgeoning literature connecting NEP to micro-and macro-level correlates (Braun and Hutter 2016;Hooghe and Marien 2013;Giugni and Grasso 2019;Grasso and Giugni 2016; Vr abl ıkov a 2014), little attention has been paid to the participation of populist voters. This is a relevant question not only for the stances and claims of those undertaking such actions, but also timely, for one in four European voters had cast a vote for a populist party in recent times (Lewis et al. 2018).
Political participation can be a means of socio-political change, identification with a group and/or expression of ideas and values (Klandermans and Mayer 2005;Tarrow 2011). In this respect, the analysis of intra-organisational activism showed that populist parties serve as mobilisation drivers and outlets (Pirro and R ona 2019;Whiteley et al. 2019), and recent studies on contemporary 'movement parties' have drawn similarities between the populist left and right regarding their origins and modes of participation (della Porta et al. 2017;Pirro and Castelli Gattinara 2018). From the West European populist left (e.g. the Spanish Podemos) to the East European populist right (e.g. the Hungarian Jobbik), there is no shortage of parties geared toward engagement in the protest arena and catalysing grassroots participation. The interest prompted by these organisations precisely rests in the successful mobilisation of supporters in and out of the polls, as if they were able to re-politicise modes of participation overlooked by mainstream parties of the left and right. At the same time, studies on protest politics indicated that radicals of both the left and right tend to contact officials, campaign, persuade others, cooperate and protest more than ideological moderates (Torcal et al. 2016;Van der Meer et al. 2009). With our contribution, we wish to move beyond the analysis of political radicals and delve deeper into the set of populist party voters, indeed arguing that they participate more at the non-electoral level compared to non-populist party voters.
While a few studies have examined the effects of ideological preferences on political action (e.g. Martin and Van Deth 2007;Teorell et al. 2007; Van der Meer et al. 2009), we take a step further and look at the nexus between populist vote and NEP. To be sure, reverse causality may be at play. This notwithstanding, the directionality of our expectations is grounded in theoretical and substantive knowledge. Political parties are certainly instruments of linkage between citizens and governments as much as channels of protest and revolution (Merkl 2005). In linking populism to political identity, Mel endez and Rovira Kaltwasser (2019: 522) argued that populist agency is instrumental in developing an antiestablishment political identity. In fact, unlike apathetic and politically disengaged individuals, anti-establishment voters are 'politically interested and prone to populist appeals' (Mel endez and Rovira Kaltwasser 2019: 530). Barr (2009: 37-8) also stressed that populists promote electoral or plebiscitary linkages to hold decision-makers accountable ('I can do it for you'). We however feel that this notion is part and parcel of representative politics and that populists also appeal to corrections based on increased citizen participation ('we can do it for ourselves'). Both linkages ultimately fit the populist direct-democratic ideal, though from differenti.e. top-down vs. bottom-upperspectives.
H1a: Voters of populist parties engage more in NEP than non-populist party voters.
While concerned with the mobilisation of populist party voters, we are equally interested to elaborate on differences between populist and nonpopulist voters within left and right blocs. As there is no such thing as the populist voter (Rooduijn 2018), we feel that intra-bloc variance can be quite revealing and should be investigated further. Moreover, a growing number of populist right organisations is spurring on from the movement sector and articulating contentious politics in the protest arena (Pirro and Castelli Gattinara 2018). We might expect their supporters to engage in non-institutional modes of participation, thus matching the participation levels of 'traditionally mobilised' left-wing voters (Torcal et al. 2016). If proven correct, this would challenge notions on NEP as the foremost prerogative of the left (Hutter 2014;Hutter and Kriesi 2013;Torcal et al. 2016), and indeed draw attention to the mobilising potential of the populist right (Castelli Gattinara and Pirro 2019; Mudde 2016).
H1b: Voters of populist left parties engage more in NEP than non-populist left party voters.
H1c: Voters of populist right parties engage more in NEP than nonpopulist right party voters.
Especially considering the patterns of NEP across the left-right divide, the analysis of protest politics has shown that, in order to mobilise citizens, the populist right prefers engagement in the electoral to the protest arena, while the political left promotes its claims in both at the same time (Hutter 2014;Hutter and Borb ath 2019;Kriesi et al. 2012). This may have to do with underlying values: right-wingers subscribe to authoritarian values, and favor orderly political action; left-wingers share libertarian values, hence preferring unconventional and protest activities (Hutter and Kriesi 2013: 293). As far as the emergence of new cleavages is concerned, the integration-demarcation argument places conflict over cultural liberalism as one of the foundational elements structuring political competition in the age of globalisation (Kriesi et al. 2008). The populist politics of the left and right have largely drawn upon this dimension of conflict (Mudde 2004;Taggart 2000) and we wish to test whether the value orientation of different populist party voters affects their NEP. On top of arguments on the 'normalisation' of protest, which suggest that political protest has become an integral part of contemporary life and is adopted by ever more diverse constituencies (Meyer and Tarrow 1998), it is worth noting that NEP also includes non-disruptive forms of participation (e.g. signing petitions or discussing political opinions online). Therefore, there are certainly reasons to probe the participation (levels) of populist voters beyond the realm of left-libertarian politics.
H2a: Voters of populist left parties holding libertarian values engage more in NEP than non-populist left party voters.
H2b: Voters of populist right parties holding authoritarian values engage more in NEP than non-populist right party voters.
Another way to look at the structuring of populist conflict is considering specific issue positions. From a supply-side perspective, populist trademark issues fall within the areas of immigration (populist right) and wealth distribution (populist left), to the point that their criticism of elites and the political process can be funneled through these issues (March 2011;Mudde 2007). As far as the demand for populist parties is concerned, attitudes toward immigrants and income redistribution have consistently predicted the vote for populist right and populist left parties, respectively (Rooduijn 2018). These findings resonate with recent research on populist attitudes, according to which policy considerations on socioeconomic and redistributive issues, on the one hand, and anti-immigration, on the other, can moderate the effect of populist attitudes on populist party support (Van Hauwaert and Van Kessel 2018). As in the case of value orientations, we are interested in assessing whether given policy preferences, consistent with the ideological profiles of the populist left and right, and capable of mobilising voters at the electoral level, can prompt political participation beyond voting.
H3a: Voters of populist left parties endorsing socioeconomic redistributive and egalitarian issues engage more in NEP than non-populist left party voters.
H3b: Voters of populist right parties opposing immigration engage more in NEP than non-populist right party voters.

Data and operationalisation
For our empirical analysis, we use data from the LIVEWHAT survey, which was conducted in 2015 across nine European countries including France, Germany, Greece, Italy, Poland, Spain, Sweden, Switzerland and the United Kingdom (N ¼ 18,367). The data was collected ad hoc by the specialised polling agency YouGov, which administered online surveys between June and August 2015, using balanced country quotas in terms of sex, age, region, social class and education level in order to match national population statistics. This dataset has a unique advantage: it allows the building of a richer-than-usual dependent variable to capture the frequency of NEP while giving information on both prospective and retrospective voting for populist parties at two electoral levels (national and European), as well as information on authoritarian-libertarian values and specific issue positions.
The dependent variable: level of NEP Our dependent variable is the level of NEP. Following the standard approach in the literature, we conceive NEP as a matter of degree (Braun and Hutter 2016;Hooghe and Marien 2013;Quaranta 2013). In order to measure the level of NEP, we rely on information from 16 dummies that capture whether the respondent has engaged in non-electoral political activities in the last 12 months (Table 1). Our dependent variable includes some items that gauge non-electoral forms of political participation, which are similar to those used in the European Social Survey, the World Values Survey and the International Social Survey Programme. Overcoming one long-standing limitation of these datasets regarding the study of NEP frequencies, our survey extends the traditional repertoire of actions, considering different platforms (both online and offline) and degrees of disruption. Thus, it also asks respondents whether they have gone on strike, displayed or worn a logo or badge of a political campaign or organisation, resorted to ethical consumerism (i.e. buycotting), occupied spaces, used social media for political purposes, followed a political organisation on social media platforms, visited the website of a politician or party, used the Internet to search for political information, damaged public goods (e.g. broken windows, removed traffic signs, etc.) and used violence (e.g. clashed with the police). Although the actions above encompass a great deal of heterogeneity, we find that NEP compounds to a common, one-dimensional action repertoire (Braun and Hutter 2016;Hooghe and Marien 2013;Kern et al. 2015;Quaranta 2013; Vr abl ıkov a 2014). Following Vr abl ıkov a (2014: 212-3), we run a tetrachoric correlation matrix (Table A1, Online Appendix I), 1 and then perform a factor analysis. Our items load strongly on one factor (Eigenvalue ¼ 8.31), which accounts for as much as 62.56 per cent of the total variation, with all 16 factor loadings above the 0.60 threshold (Table A2, Online Appendix I). We thus build a summated rating scale for non-electoral political participation, which is robust (Cronbach's a ¼ 0.83). 2 The data and methodology used provide us with a unique advantage relative to the scant literature on the association between populism and NEP: we neither reduce non-electoral involvement to a participation/no-participation dichotomy nor do we limit our analysis to the most widespread modes of political participation (cf. Anduiza et al. 2019).

Explanatory variables and controls
We use one main predictor, i.e. populist party vote. We create a fourcategory indicator that takes into account the ideological stance of the party (1 ¼ populist right party voter; 2 ¼ non-populist right party voter; 3 ¼ non-populist left party voter; 4 ¼ populist left party voter), considering the retrospective vote at the 2014 European Parliament (EP) election and prospective vote at the national level. 3 Given the cases at hand and the timeframe of analysis, we identified a total of 21 populist parties, indeed assuming that it is possible to distinguish between populist and non-populist parties in each of the nine countries included in the study (Van Kessel 2015). Our classification of populist parties is consistent with the latest state of the art on populist parties in Europe (Rooduijn et al. 2019). 4 In order to differentiate between populist and non-populist parties of the left and right we also rely on Chapel Hill Expert Surveys (Polk et al. 2017) and the validation of country experts contacted ad hoc. The full list of parties and relative categorisations is reported in Online Appendix I (Table A8). 5 In a preliminary specification to test H1a, we use a simple populist vs. non-populist party voter dummy predictori.e. with no party breakdown in terms of leftright ideology. 6 As individuals have different views in terms of democratic freedoms and rights, we additionally consider how authoritarian or libertarian an individual is, drawing on the cultural liberalism dimension of conflict (Kriesi et al. 2008). After performing a factor analysis, we built a summated rating scale for authoritarian-libertarian attitudes through placement of one's views along five different items. 7 Finally, we distinguish between those cultural and economic dimensions associated with support for populist parties along the left-right ideological divide (Hutter and Kriesi 2013;Van Hauwaert and Van Kessel 2018). An 11-point scale measures support for anti-immigrant attitudes. 8 Following Van Hauwaert and Van Kessel (2018), we use five indicators to create an economic issue positions index. 9 In line with the literature on the determinants of NEP, we include a number of control variables in order to strengthen our arguments against alternative explanations. Among these variables, we distinguish different groups: biographical availability, grievances, political values, and network exposure (Dalton et al. 2010;Schussman and Soule 2005; Vr abl ıkov a 2014). Table 2 summarises the descriptive statistics.
First, we know that enhanced resources are associated with increased prospects for citizens' engagement in political action (Dalton 2008;Verba and Nie 1972;Verba et al. 1995). Specifically, certain biographical features and personal constraints reduce the costs and risks of NEP, such as respondent's age, education, gender and the presence of children (Schussman andSoule 2005: 1088;Verba and Nie 1972;Verba et al. 1995). Gender is measured through a dummy variable (1 ¼ female; 0 ¼ male). Similarly, we use a dummy to capture whether the respondent (or their partner) currently has children under 18 years old living in the household (1 ¼ yes; 0 ¼ otherwise). A continuous indicator captures the age of the respondent. 10 A 9-point scale measures the highest educational level attained by the respondent, ranging from 'primary school or less' to 'PhD or equivalent'.
Second, the respondent's personal experiences of hardship and adverse economic circumstances could determine not only whether they decide to punish the incumbent by voting for the challenger (Lewis-Beck and Stegmaier 2000) but also whether they engage in non-electoral forms of action (Portos et al. 2019). A 1-10 increasing scale is used to capture the net monthly income level of the respondent's household. Although material conditions may be relevant, grievances are socially constructed, thus considering how they are perceived is also important (Kern et al. 2015). Hence, a 0-11 summated indicator with both objective and subjective aspects accounts for the worsening job conditions of the respondents. Third, certain political attitudes related to political engagement could similarly predict NEP (Beissinger 2013: 575;Dalton et al. 2010). Political interest is captured through an ordinal variable that consists of a 1-4 increasing scale. We then consider satisfaction with democracy, which is measured through an increasing 0-10 scale. 11 Based on a 5-point scale that measures the extent of agreement with the statement 'people like me don't have any say about what the government does' (1 ¼ strongly disagree; 5 ¼ strongly agree).
Finally, social capital and network exposure are important predictors of participation (Beissinger 2013;Schussman and Soule 2005;Verba et al. 1995). A 1-4 ordinal variable that ranges from 'less than once' to 'almost every day' captures how often the respondent met with friends who do not live in the same household during the last month. We control for party membership through a dummy variable, plus belonging to social movement organisations through an 11-point summated index (see Online Appendix II). We also include a dummy variable on the status of the populist party, assuming that their role as incumbent could dissuade their voters from engaging in forms of grassroots mobilisation (Torcal et al. 2016). Finally, we account for the specific (e.g. institutional, historical, cultural) characteristics of each country that might be correlated with NEP through country fixed effects.

Empirical results and discussion
We start from the premise that electoral and non-electoral participation are indeed part of the same continuum (Model 1, Table A3, Online Appendix I). The histogram in Figure A1 (Online Appendix I) then shows that the NEP dependent variable is not normally distributed. As most people's engagement in non-electoral activities is limited, mean values of the NEP index are low (see Tables 1 and 2). Also, data are strongly skewed to the right. Further evidence confirms that there is overdispersion, thus a negative binomial regression seems the most appropriate modeling strategy. 12 We take voting in the next national election to build the four-category populist electoral support variable (Table 3). 13 In Models 1 and 2, we present the full additive specification, including authoritarian-libertarian values, as well as covariates related to biographical availability, grievances, political attitudes and social capital and networks. While in Model 1 we use the populist/non-populist party vote dummy variable, we use the main (four-category) indicator of populist  (Table 3). On top of the variables included in the full additive specification, we incorporate the interaction between authoritarian-libertarian values and populist party vote in Model 3 (Table 3). Finally, we add the interactive models between populist voting and attitudes toward migrants and the economic issue positions index in Models 4 and 5, respectively (Table 3). 14 Besides educational level, age and worsening job conditions, we find that some variables related to social network exposure and political attitudes are positively associated with the intensity of NEP. Meeting often with friends, party and social movement organisation membership, 15 political efficacy, democratic dissatisfaction and especially political interest increase the levels of NEP; these results are robust across all model specifications. We also see that libertarian attitudes are associated with higher levels of NEP (Model 2, Table 3). These results are consistent with earlier findings (e.g. Hutter and Kriesi 2013;Grasso and Giugni 2016;Torcal et al. 2016).
Broadly speaking, we observe that populist voters tend to engage more in NEP activities, in line with H1a. 16 However, in light of the results from Model 2 (Table 3; see also Figure 1), we cannot confirm that voters of populist left parties engage more in NEP than non-populist left voters (H1b). Populist party support is conversely a crucial factor in deciphering  Table 3).
NEP in the case of right-wing voters; voters of populist right forces participate more at the non-electoral level compared to non-populist right voters (H1c). While the predicted values in the NEP rating scale is 1.86 for voters of populist right parties, it decreases to 1.53 for supporters of non-populist right forces (Figure 1). This finding echoes the burgeoning literature on movement parties, which calls attention to the investments in grassroots politics by populist radical right parties and the prospects they might offer in terms of political socialisation (Pirro and Castelli Gattinara 2018;Pirro and R ona 2019).
Populist party vote, however, acquires its own standing when put in interaction with authoritarian-libertarian values and/or coherent issue positions, thus suggesting that the relationship between populism and NEP is generally better understood when vote is bundled with one's values and policy preferences. Holding issue positions that are consistent with the ideological orientation of the populist party of preference crucially enhances our understanding of the level of mobilisation outside the ballot box.
As anticipated, our second set of hypotheses concerns the interactive effects of populist party vote and authoritarian-libertarian values on NEP. While not entirely independent from positioning in terms of left and right, we find strong evidence in support of the authoritarian-libertarian dimension as a predictor of NEP. Although this finding resonates with previous research findings (e.g. Hutter and Kriesi 2013;Grasso and Giugni 2016), it is worth noting that our authoritarian-libertarian index is based on five items and thus offers a more exhaustive outlook on social values than proxies based on cultural liberalism and immigration alone. The interaction between populist party vote and authoritarian-libertarian values goes a long way in explaining the degree of NEP. Populist left voters holding libertarian views clearly engage more in NEP than nonpopulist right supporters but not more than non-populist left voters. Empirical evidence is thus mixed and we cannot confirm H2a. Conversely, populist right voters subscribing to authoritarian values resort more to NEP than non-populist voters of the right (as well as left-wing voters), hence confirming H2b.
Reporting extreme authoritarian values is associated with higher values of NEP among populist right voters vis-a-vis non-populist party voters (or populist left, for that matter; see Figure 2). Far from demobilised, our results suggest that populist rightists are the subset of voters mobilising the most beyond the ballot box among those holding authoritarian views. Yet, this effect does not work the same way for those holding libertarian values, as both populist and non-populist party supporters of the left report very high levels of NEP. While the levels of the NEP index remain constant along the authoritarian-libertarian dimension for populist right supporters (from 1.78 to 1.73 when the values of the authoritarianlibertarian axis are at 0 and 3.6, respectively), they increase dramatically as both non-populist and populist left voters rank higher on libertarian values (from .99 and 1.30 to 2.69 and 2.83, respectively). This finding nicely complements our additive model, demonstrating that we cannot quite speak of a neat partition between a libertarian left that mobilises in the protest and electoral arenas and an authoritarian right that solely embarks on institutional participation (cf. Hutter and Kriesi 2013;Van der Meer et al. 2009). We are therefore compelled to refine notions on the left-wing dominance of the protest arena (Torcal et al. 2016) factoring in vote for populist right parties and linked value orientations.
With regard to the third set of hypotheses, we cannot confirm that voters of populist left parties endorsing socioeconomic redistributive and egalitarian issues engage more in NEP than non-populist left voters (H3a; Model 5, Table 3). However, supporters of populist right parties endorsing anti-migrant attitudes clearly do so vis-a-vis non-populist right voters, thus corroborating H3b (Model 4, Table 3; Figure 3). Levels of NEP dramatically increase for (populist or non-populist) left supporters as they turn more pro-migrant. In sharp contrast, populist right supporters' endorsement of redistributive and egalitarian issues increases engagement    Table 3).
in NEP in relation to non-populist right voters (Figure 4). Therefore, NEP increases as more 'leftist' economic policy preferences are shared by left-wing supporters. Similarly, populist right supporters who report 'leftist' economic policy preferences participate more than non-populist right voters outside the ballot box. Non-populist rightists simply do not mobilise around such issues, suggesting that centre-right parties are still perceived as advocates of free enterprise and deregulation. At the same time, populist right voters' NEP levels might indicate a progressive demand for economic paternalism and (selective) welfarism, which would be consistent with the policy proposals of some populist (radical) right parties across Europe (e.g. Pirro 2017;Otjes et al. 2018).
These are notable pieces of evidence challenging the different logics underlying mobilisation of the populist left and right, which have so far seen supporters of radical political alternatives grounding their decision to (not) participate on the basis of discording issue preferences (Hutter and Kriesi 2013). 17 Essentially, while generally confirming previously held notions on the political participation and activism of the left, our findings provide novel and compelling evidence for a politicisation of the populist right beyond the electoral arena.

Conclusion
The scholarship on populism is at the crossroads. The consolidation of the ideational approach (Mudde 2017) brought a significant degree of consensus at the definitional level, which will hopefully lead to progressively less contentious classifications of populist parties and movements around the globe. There are also important developments to note with regard to the analysis of the 'demand side' of populism. While the familiarity with the electoral politics of populism is therefore at an unprecedented high, attention to non-electoral forms of mobilisation is still rare. We attribute this lacuna to the most direct and visible consequences of populist party exploits (i.e. their ever-larger share of seats in parliaments) and the view that populist party voters are motivated by distrust and protest, while remaining largely apathetic. On the one hand, recent evidence suggested that populist party voters are driven by specific socio-economic and cultural concerns consistent with the agendas of populist parties of preference (Van Hauwaert and Van Kessel 2018). On the other, contributions focusing on the political participation of the left and right have almost unequivocally interpreted mobilisations in the protest arena as a leftist/libertarian trademark (Hutter and Kriesi 2013). We took up these cues to explore value-and issue-related aspects underlying the demand side of populism beyond the ballot box.
In our investigation of the non-electoral politics of populism, we particularly focused on populism's ability to prompt full political engagement (Taggart 2018). Hence, we had been not only concerned with the aspect of NEP and whether populist party voters engage at the non-electoral level, but also with the actual degree of their participation. Our analysis returned a number of key findings on the relationship between populism and NEP, indicating that populist party voters value casting votes in elections as much as unconventional non-electoral politics. In actual fact, populist party voters tend to engage more in NEP than non-populist party voters. Delving deeper into the type of populist party voted for, as well as individual value orientations and issue preferences, we found that left-wing people tend to mobilise more at the non-electoral level. In essence, the evidence presented corroborates previous work on left-wing participation outside the ballot box (Torcal et al. 2016).
Our study, however, unveils previously unexplored participation patterns within the populist right. We have indeed demonstrated that populist right voters: (a) engage more in non-institutional participation than non-populist right votersand generally as much as left-wing voters; (b) are the most mobilised subset among voters holding authoritarian views; (c) resort to more NEP than non-populist voters when positions on immigration are considered; and (d) also participate more vis-a-vis non-populist rightists on the basis of economically leftist orientations. While providing initial empirical evidence of the broad-ranging politicisation and non-institutional participation of populist right voters, these aspects prompt us to dig into the movement strategies of the populist right and its investments at the grassroots level (Pirro and Castelli Gattinara 2018;Castelli Gattinara and Pirro 2019).
The article additionally opens avenues for further enquiry. More research is necessary to unveil the association between populist behavior and NEP, including not only the role that income and education play as moderating variables (Anduiza et al. 2019), but also political sophistication and social network exposure. Although more exhaustive evidence is necessary to exclude the reverse causality hypothesis, our findings are substantiated theoretically and are robust across a number of model specifications. In conclusion, our study critically places NEP at the heart of populism in general, and populist right politics in particular, enhancing an otherwise partial understanding of one of the pressing phenomena of our times. While challenging the common notions of populism as inherently distrustful and apathetic, and protest as an exclusive practice of the left, populist politics is expanding beyond the electoral arena and to the right. No matter how reluctantly political, populism is gaining a foothold through non-electoral participation and displaying margins for all-round politicisation. Notes 1. Since we are dealing with dummy variables, standard Pearson's r procedures that assume a normal distribution do not fit our data. 2. Alternatively, we also built a simple summated scale resulting from adding the (up to 16) forms of non-electoral political participation that a given individual might have engaged inmeasured through yes/no variables. Results do not change in any substantial way (Table A5, Online Appendix I). 3. While acknowledging differences between the two types of elections, we believe that drawing on vote recalls and voting intentions provides a comprehensive picture of populist vote and NEP. Conversely, we did not rely on vote spells at previous national elections as they took place at varying time points, making cross-country comparisons problematic. Moreover, a number of relevant populist parties did not compete and/or had little support in the previous elections, e.g. Podemos in Spain. 4. The Italian M5S combines left-libertarian issues with mild nativist positions thus keeping an ideologically ambiguous profile (Pirro 2018 Table A7, Online Appendix I), none of the interactive hypotheses can be confirmed, indeed suggesting that the M5S constituency behaves differently from supporters of either populist left or right parties. 5. We have also excluded a number of more or less marginal extreme-right parties from our analysis on the basis of their opposition to the democratic constitutional order (visa-vis populist right parties) and the ensuing problems in associating them to the non-populist right (i.e. mostly, centreright parties). See Table A8, Online Appendix I. 6. The M5S is included in the models with populist/non-populist party vote dummies (Model 1, Table 3; Model 1, Tables A4-A6, Online Appendix I). 7. These five items reflect the respondent's personal views on 11-point scales.
Overall, the authoritarian-libertarian scale of our dataset is richer than any other we could build through existing datasets. 8. The level of correlation between anti-immigrant attitudes and the authoritarian-libertarian index is moderate (Pearson's r ¼ 0.37). Therefore, we report alternative specifications. 9. Using information on five statements measured through 0-10 scales, we run a Principal Component Analysis (PCA) and build a weighted summated index (Eigenvalue ¼ 2.00, one-component solution; Cronbach's a $ 0.6; see Online Appendix II). 10. We have also controlled for potential quadratic effects, without finding any significant results. 11. Besides political interest and perceived efficacy, trust in political institutions can be an important predictor of extra-institutional political participation (Braun and Hutter 2016). Our survey has information on ten indicators related to institutional trust, so we can build a weighted summated scale (Cronbach's a ¼ 0.91; Eigenvalue ¼ 5.60, one-item solution; 55.99 per cent of total variance explained; see Online Appendix II). While we do not include the institutional trust index in the reported models due to concerns of over-specification, its inclusion does not change our findings. Alternatively, we created an index to control for internal political efficacy. The three indicators combined in the index happen to be inter-correlated (0.51 < Pearson's r < 0.65). The PCA conducted offers a one-component solution (Eigenvalue ¼ 2.19; 73.12 per cent of total variance explained). The scale reliability coefficient falls within acceptable standards (Cronbach's a ¼ 0.81). While the internal efficacy index is not correlated with the indicator of external political efficacy (Pearson's r ¼ À0.03), it is highly correlated with political interest (Pearson's r > 0.50), thus we omit it due to possible multicollinearity. 12. We run a test of the overdispersion parameter alpha. Overdispersion happens when the conditional variance exceeds the conditional mean. When the overdispersion parameter is zero, the negative binomial distribution is equivalent to a Poisson distribution (a ¼ 0). As alpha is significantly different from zero, we conclude that the Poisson distribution is not the most suitable modelling strategy. 13. As a robustness check, we replicate the negative binomial specifications with OLS regressions (estimated with robust standard errors; Table A4, Online Appendix I). 14. For the average of authoritarian-libertarian values, anti-immigration and economic issue position scales by populist vote, see Figures A3.1-A3.3 (Online Appendix I). 15. As the level of correlation between party and organisational membership is moderate (Pearson's r ¼ 0.46), we do not include it in the main models due to endogeneity concerns, but we include it in Models 3-5 (Table A3, Online Appendix I). We report the double interaction between populist voting and attitudes towards migrants, and populist voting and economic issue positions in Model 2, Table A3, Online Appendix I. 16. For the average values of NEP by populist vote, see Figure A2 (Online Appendix I). 17. Marginal effects are plotted in Figures A8-A10 (Online Appendix I). If we replace the populist voting predictor in the next national election with vote recalls in the last election for the EP, our results do not change in any substantial way. See Table A6 (Online Appendix I); Figures A4-A7 (Online Appendix I).

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
Mart ın Portos' work was supported by the project 'Living with Hard Times: How Citizens React to Economic Crises and Their Social and Political Consequences' (LIVEWHAT), funded by the European Union's Seventh Framework Programme (Grant agreement no: 613237).