In this article we assess how the stability of voting patterns over the course of two consecutive elections is affected by political dissatisfaction and populist party voting and whether the effects of the former are conditional upon the latter. First, as voting patterns have become increasingly volatile (Chiaramonte and Emanuele Reference Chiaramonte and Emanuele2017; Dassonneville and Hooghe Reference Dassonneville and Hooghe2017a), it is often argued that the tendency to switch one’s vote stems from an alleged ‘crisis in political support’. Several individual-level studies convincingly show that political dissatisfaction is correlated with unstable voting behaviour (e.g. Dalton and Weldon Reference Dalton and Weldon2005; Dejaeghere and Dassonneville Reference Dejaeghere and Dassonneville2017; Söderlund Reference Söderlund2008). Second, the alleged crisis in political support has also been linked to the rise of populist parties (Van Kessel Reference Van Kessel2015; Mudde Reference Mudde2013). Those expectations are also backed up by a large literature showing individual-level evidence of a relation between political dissatisfaction and voting for populist parties (e.g. Arzheimer Reference Arzheimer2009; Dalton and Weldon Reference Dalton and Weldon2005; Lubbers et al. Reference Lubbers, Gijsberts and Scheepers2002; Pauwels Reference Pauwels2014; Werts et al. Reference Werts, Scheepers and Lubbers2013).
Considering these literatures together, political dissatisfaction has been linked with two types of behaviour; the politically dissatisfied are expected to change their votes regularly and they are expected to be more likely to vote for populist parties. Those expectations might have been simultaneously true in the 1980s and 1990s, when electoral support for populist parties was strongly fluctuating in most Western democracies (Nedelcu Reference Nedelcu2015). However, the electoral support for populist parties increased strongly in several long-standing Western democracies from the 2000s onwards (e.g. Hartleb Reference Hartleb2015; Van Kessel Reference Van Kessel2015). As a result, the thesis that dissatisfied voters in general – and more particularly the dissatisfied voters of populist parties – are regular vote-switchers may no longer hold in political systems where dissatisfied voters can vote stably for viable populist party alternatives. In other words, the rise of populism necessitates that we revisit the relation between political dissatisfaction and vote switching.
The argument that we pursue in this article is that the stability of the vote of dissatisfied voters is conditional upon the type of party they previously supported: a populist party or a mainstream party. Because the rise of populist parties seems hardly explainable without such parties having attracted a larger core base of supporters, we argue that dissatisfied voters who switched their votes to populist parties may very well become rather stable voters again. In addition, we argue that politically dissatisfied citizens are more likely to develop attachments to populist parties in systems in which populist parties have already effectively established themselves in parliament.
In order to assess whether populist party support indeed conditions stable and volatile voting among politically dissatisfied citizens, this article first assesses whether voters for populist parties are more or less volatile than the electorate of mainstream parties. Secondly, we examine whether political dissatisfaction stabilizes voting behaviour among populist party voters. Thirdly, we look at how the viability and the effective presence of populist parties in parliament conditions the extent to which their voters are stable or volatile voters.
To investigate these research questions, we present an individual-level analysis of the stability of voting for (populist) parties in established democracies.Footnote 1 We make use of the data from modules 2, 3 and 4 of the Comparative Study of Electoral Systems (CSES) project, covering a wide range of 32 elections from 2001 to 2016 in 14 Western democracies. This data set makes possible an investigation of the stability of (populist) party choices, as well as the effects of political dissatisfaction on stable or volatile voting. We analyse these data by means of multilevel modelling.
How do the dissatisfied vote?
In democratic systems, citizens who are – temporarily – dissatisfied with specific authorities can punish officeholders by not voting for them. Those who previously supported the incumbent party, for example, can switch their support towards other (opposition) parties (Dalton and Weldon Reference Dalton and Weldon2005: 941; Zelle Reference Zelle1995). Vote switching as a result of short-term dissatisfaction with the performances of specific officeholders – particularly government (or parties) – applies to voters of all parties alike and can positively be understood as an indication that voters are performing their democratic task of critically evaluating officeholders (Van der Meer et al. Reference Van der Meer, van Elsas, Lubbe and van der Brug2015; Norris Reference Norris1999).
In addition, a more deep-seated attitude of political dissatisfaction towards political institutions and mainstream political actors in general is likely to affect voting behaviour and the stability of the vote as well.Footnote 2 Citizens who are structurally dissatisfied with the way democracy works can use their vote to continually reject incumbents regardless of who holds power (Dalton and Weldon Reference Dalton and Weldon2005; Zelle Reference Zelle1995). A fairly stable and general sense of political dissatisfaction is thought to make voters critical of every aspect of the political system (Söderlund Reference Söderlund2008: 222), which hinders the development of emotional bonding with specific parties and encourages a willingness to respond to short-term electoral factors (Dalton Reference Dalton2004). As a result, distrusters lack incentives to form stable party preferences, leading them to support new contender parties more often (e.g. Bélanger and Nadeau Reference Bélanger and Nadeau2005: 127; Dalton and Weldon Reference Dalton and Weldon2005: 940). Multiple studies have shown that political dissatisfaction is indeed related to higher levels of vote switching (e.g. Dalton and Weldon Reference Dalton and Weldon2005; Dejaeghere and Dassonneville Reference Dejaeghere and Dassonneville2017; Söderlund Reference Söderlund2008).
A second body of literature studying the behaviour of generally dissatisfied citizens argues that they are more likely to vote for populist parties. Populists basically separate society into two homogeneous and antagonistic groups: ‘the pure people’ versus ‘the corrupt elites’ and argue that the former should reclaim popular sovereignty to prevent the corrupt elites from continually betraying them (e.g. Van Kessel Reference Van Kessel2015; Mudde Reference Mudde2004; Rovira Kaltwasser Reference Rovira Kaltwasser2012). Anti-elitism is thus one of the core elements of populism (Rooduijn Reference Rooduijn2018), and this anti-elitism is directed towards all mainstream political officeholders (including mainstream opposition parties). As such, anti-elitism is often measured by means of indicators of ‘generalized political dissatisfaction’ or ‘political disaffection’ (Rooduijn et al. Reference Rooduijn, Van der Brug and de Lange2016). While generalized dissatisfaction clearly originates from dissatisfaction with specific policies and institutions (Zelle Reference Zelle1995), it goes beyond specific distrust. In fact, focusing only on the latter would not allow a distinction to be made between citizens who are dissatisfied with mainstream politics in general and citizens who are only temporarily dissatisfied with the policies implemented by temporarily disliked officeholders (e.g. mainstream opposition supporters).
As dissatisfied voters and populist parties share a fierce criticism of the establishment (Bergh Reference Bergh2004), voting for populist parties may well appear as the most attractive electoral option for dissatisfied voters (Bélanger and Nadeau Reference Bélanger and Nadeau2005; Van der Brug et al. Reference Van der Brug, Fennema and Tillie2000), in particular when populist parties succeed in presenting themselves as political outsiders (Bélanger and Aarts Reference Bélanger and Aarts2006; Rooduijn et al. Reference Rooduijn, Van der Brug and de Lange2016: 34).Footnote 3 The existence of an individual-level relationship between generalized political dissatisfaction and voting for populist parties has been demonstrated in specific countries (Mayer and Perrineau Reference Mayer and Perrineau1992 (France); Swyngedouw Reference Swyngedouw2001 (Belgium); Schumacher and Rooduijn Reference Schumacher and Rooduijn2013 (the Netherlands); Lubbers and Scheepers Reference Lubbers and Scheepers2000 (Germany)), and in cross-national comparative studies (Arzheimer Reference Arzheimer2009; Betz Reference Betz1994; Dalton and Weldon Reference Dalton and Weldon2005; Lubbers et al. Reference Lubbers, Gijsberts and Scheepers2002; Norris Reference Norris2005; Pauwels Reference Pauwels2014; Werts et al. Reference Werts, Scheepers and Lubbers2013).Footnote 4
The stability of the populist vote: theory and expectations
Previous work has linked dissatisfaction with both instability in the vote and with voting for populist parties. It is therefore not surprising that the populist vote has generally been expected to be rather volatile (Arzheimer Reference Arzheimer2009: 259; Van der Brug et al. Reference Van der Brug, Fennema and Tillie2005). There are in fact good theoretical reasons for expecting instability in the electorate of populist parties. First, the attractiveness of populist leaders – which helps to attract votes – tends to dissipate over time (Schumacher and Rooduijn Reference Schumacher and Rooduijn2013: 126). Second, when a populist party becomes electorally successful and establishes parliamentary representation, it has more difficulty presenting itself as an ‘anti-elitist’ movement (Schumacher and Rooduijn Reference Schumacher and Rooduijn2013: 126). Finally, Anthony Downs’ (Reference Downs1972) issue-attention cycle theory states that voters will not focus on a single issue for a long period of time. At some point in time, hence, populist parties’ focus on anti-elitisms no longer attracts voters’ attention (Poznyak et al. Reference Poznyak, Abts and Swyngedouw2011: 675).
Empirical studies of (in)stability in voting for present-day populist parties are scarce. However, some empirical contributions relying on data from the 1990s have shown that even though populist parties temporarily benefited from dissatisfied voters, their support was rather unstable (Bergh Reference Bergh2004). In a study of the electorate of the populist French National Front in the 1990s, for example, the majority of those voters were described as ‘protest voters that come and go’ without really identifying with the party (Mayer and Perrineau Reference Mayer and Perrineau1992: 123). Such findings suggest regular switching between populist and mainstream (opposition) parties.
This rather scattered evidence from the 1990s has strengthened the assumption that voters who vote for populist parties are unstable. We argue that this assumption should be reassessed because of the rise in electoral support for populist parties in several advanced democracies. The continued presence of populist parties in many national parliaments (Van Kessel Reference Van Kessel2015: 187; Mudde Reference Mudde2013) suggests that present-day populist parties might now have a core group of loyal voters. Of course, growing support for such parties implies that populists still attract new voters – who necessarily make a switch from a mainstream party (or from abstention). But the fact that populist support has continued to grow over a series of elections in several countries – or has stabilized at a high level – suggests that populist parties have built an electoral base of core voters.
The emergence of an increasingly stable populist electorate would also fit with the insights on the emergence of a new cleavage in established democracies. Several authors argue that a ‘transnational cleavage’ has emerged that has the capacity to (re)structure the voting behaviour of the ‘losers of globalization’ (Hooghe and Marks Reference Hooghe and Marks2018; Kriesi et al. 2006, Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2008). This cleavage mainly constitutes a conflict between the ‘winners’ of globalization (the middle classes who profit from open borders, European integration and international economic competition) and the ‘losers’, who are dissatisfied with the unquestioned support of the current elites for this ongoing process (Hooghe and Marks Reference Hooghe and Marks2018). As this new dimension crosscuts the traditional structural and political cleavages on which mainstream political parties continued to compete, the party system is more likely to respond to this structural change in voter preferences by means of the growth of new political parties instead of an incremental change in the established parties. Hence, with the winner/loser of globalization cleavage becoming more and more salient since the euro- and migration crises (Hooghe and Marks Reference Hooghe and Marks2018), new opportunities emerge for challenging (populist) parties to position themselves exclusively on the new divide as the opponents of the current elites (left- and right-wing populist parties) and the defenders of the sovereign nation state and national identity (mostly right-wing populist parties). The increased saliency of this divide subsequently provides populist parties with an improved structural basis to attract loyal support (Hooghe and Marks Reference Hooghe and Marks2018; Van Kessel Reference Van Kessel2015; Kriesi et al. Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2008; de Vries and Hobolt Reference de Vries and Hobolt2012).
In other words, it is very well possible that the electorates of populist parties are more stable now than they were before. Focusing on a period in which populist parties in several countries are electorally successful, we do not expect voters for populist parties to be particularly volatile.
Hypothesis 1: Voters for populist parties are not more likely to switch their votes than voters for mainstream parties (party switching and switching to abstention).
Even if those who vote for populist parties are not more volatile than those who choose mainstream parties, as we hypothesize (Hypothesis 1), it does not yet explain whether it is indeed political dissatisfaction that binds voters to populist parties.
Dissatisfaction with the political establishment is one of the most important individual-level attitudes on which populism thrives. However, because dissatisfaction is also correlated with vote switching, it is questionable whether generalized dissatisfaction with politics is compatible with loyally supporting a party, even if it is a populist one. In this regard, it should be noted that several scholars now argue that stable party support and generalized dissatisfaction with politics are not mutually exclusive. That is, those who are dissatisfied with politics in general are not necessarily unhappy with their own parties or representatives (e.g. Norris Reference Norris2011; Söderlund Reference Söderlund2008). We therefore argue that – besides attracting voters to populist parties – political dissatisfaction will also influence how dedicated and loyal the supporters of populist parties are (and thus discriminate between occasional and loyal populist party voters).
Thinking of dissatisfaction as a stabilizing factor of populist party support takes into account that not all populist party voters are strongly dissatisfied with politics. Some voters support populist parties because of specific policy issues (e.g. Van der Brug et al. Reference Van der Brug, Fennema and Tillie2000; Rooduijn Reference Rooduijn2018).Footnote 5 For them, dissatisfaction with politics might not be as strong as among the prototypical populist party voter. Issue-based populist party voters can choose from a broader set of appealing party options when deciding for whom to vote, as they can vote for mainstream contenders who advocate (less radical versions of) the same policies (Van der Brug et al. Reference Van der Brug, Fennema and Tillie2005). In contrast, the voting alternatives for those who are dissatisfied with the current establishment are more limited (to only true anti-establishment parties).
Turning to demand-side explanations for populist party support, dissatisfaction is found to be related to loyal populist party support (Schumacher and Rooduijn Reference Schumacher and Rooduijn2013). Recent findings also show that strong populist attitudes encourage voters to vote for populist parties.Footnote 6 In fact, populist attitudes (and distrust) are the sole determinants unifying the supporters of both populist radical left and populist radical right parties (Akkerman et al. Reference Akkerman, Zaslove and Spruyt2017). Those attitudes may even serve as a ‘motivational substitute’ when citizens’ policy preferences are (partly) incongruent with the agendas of populist parties (Van Hauwaert and Van Kessel Reference Van Hauwaert and Van Kessel2018). Finally, voters who support populist parties because they are politically dissatisfied might become even more dissatisfied as populist parties further fuel the discontent of their supporters (e.g. Bartels Reference Bartels2002; Van der Brug Reference Van der Brug2003; Rooduijn et al. Reference Rooduijn, Van der Brug and de Lange2016). For all these reasons, we expect that the most dissatisfied will also be the most loyal populist party voters:
Hypothesis 2: Political dissatisfaction increases the likelihood of a stable party vote among voters for populist parties.
The supply and effectiveness of populist parties in a political system are evident contextual factors that may help to activate the transition from populist attitudes at the individual level into actual (stable) electoral support for populist parties. Voters have to be continually exposed to a party’s message to continue voting for them (Vliegenthart et al. Reference Vliegenthart, Boomgaarden and Van Spanje2012), which means that populist parties who already obtained substantial representation in parliament have a comparative advantage over more marginal populist parties. For populist parties that are only marginally (or not) represented in parliament, it might be more difficult to generate continuous visibility in the public debate. Strategic considerations may further stimulate vote switching among supporters of marginal populist parties, as voters could conclude that their party is not effective in influencing policies (Bos and Van der Brug Reference Bos and Van der Brug2010), let alone in governing (Blais et al. Reference Blais, Aldrich, Indridason and Levine2006; Rosema Reference Rosema2006). For pragmatic reasons, such voters could switch to a larger (mainstream) party with more power in parliament (Van der Brug et al. Reference Van der Brug, Fennema and Tillie2000). In a context where populist parties have already obtained substantial representation, strategic considerations may be less relevant. This leads us to expect that populist voters will be less likely to switch when populist parties are more prominently established political actors.Footnote 7
Nevertheless, it is uncertain whether this expection holds over the full range of populist parties, with different levels of representation. When a populist party’s representation in parliament becomes so substantial that it reaches ‘coalition potential’ (Sartori Reference Sartori1976: 122), the party may no longer be able to increase its support base. In some cases, populist parties are forced to stay in permanent opposition (cordon sanitaire), which might lead their supporters to consider a stable vote irrelevant in the long run (Pauwels Reference Pauwels2011). In addition, government participation by populist parties comes with electoral risks as well. Entering government means losing the ‘purity’ of their message, making compromises on their policy positions and cooperating with the political establishment (Heinisch Reference Heinisch2003; Van Spanje Reference Van Spanje2011). This leads us to expect that the supposed stabilizing effect of populist parties’ representation in parliament on the populist vote lessens (or even reverses) when populist party representation gets really substantial.
Hypothesis 3: The more strongly a populist party is represented in parliament, the more stable will the populist electorate be (with this effect flattening out for the cases with the highest levels of populist party representation).
Data, cases and selection of populist parties
Assessing voting patterns and sample of elections
To test our hypotheses, we analyse individual-level voting patterns in parliamentary elections in a broad selection of 14 Western democracies in which it was possible to provide a stable vote for a populist party using data from the second, third and fourth modules of the CSES (2007, 2013, 2017).Footnote 8 The CSES data are unique for their coverage of a large set of democracies while ensuring cross-national validity by means of identical question wording and coding schemes (Klingeman Reference Klingeman2012).
The CSES data are cross-sectional, but the inclusion of a recall question on vote choice in the previous election allows an investigation of citizens’ voting patterns in two consecutive elections. While it should be acknowledged that recalling might lead to an underestimation of volatility (e.g. Beasley and Joslyn Reference Beasley and Joslyn2001; Zaller Reference Zaller1992), previous work has shown that inferences drawn from such data are fairly similar to what holds when relying on panel data (Dassonneville and Hooghe Reference Dassonneville and Hooghe2017b). Furthermore, when focusing on the populist vote, the tendency to report overly consistent recalled votes may also be counterbalanced by the fact that certain social groups (e.g. women and higher educated) are expected to give socially desirable answers and thus under-report (previous) support for (populist) radical right parties (Arzheimer Reference Arzheimer2017: 283–285). Nevertheless, we cross-validated the data by comparing the recalled voting behaviour of the samples with actual election results. Only the election samples for which the recalled vote closely matched the election results were retained.Footnote 9 This step led us to drop seven elections, leaving a total of 32 elections in the sample.
Given that we investigate the voting patterns of voters who voted for a mainstream party or a populist party at the previous election, we distinguish three possible voting patterns over the course of two consecutive elections: (1) a stable party vote, (2) switching between parties, and (3) switching from a party vote to abstention. The aggregate voting patterns in our set of 32 elections are listed in the online supplementary information in Appendix 1B.Footnote 10 As we should not confuse the voting patterns of voters eligible to cast a vote at both elections with the voting patterns of voters who first entered the electorate at the current election, we limited the sample to respondents aged 23 and older.
Populist parties
To answer our research questions, we identified 25 parties from 14 countries as prototypical populist parties following the dichotomous classification of populist- and non-populist parties by Stijn Van Kessel (Reference Van Kessel2015: 13): ‘Political parties are classified as populist parties if they: (1) portray ‘the people’ as virtuous and essentially homogeneous; (2) advocate popular sovereignty, as opposed to elitist rule; (3) define themselves against the political establishment, which is alleged to act against the interest of “the people”.’ In order to decide whether a party satisfies all three conditions, Van Kessel relies on a study of secondary and party literature, and the assistance of country experts. Adopting this threefold criterion ensures that parties which only sporadically adopt a populist rhetoric are not grouped together with the prototypical cases of populism (Van Kessel Reference Van Kessel2015: 4, 34).Footnote 11
The populist parties in our sample are listed online in Appendix 2 and include right-wing, moderate and left-wing populist parties with different amounts of electoral success.Footnote 12 Only populist parties that won at least one seat in national parliaments in any election between 2001 and 2015 are included. For the elections in European countries in our sample we followed Van Kessel’s classification of populist parties.Footnote 13 Our election sample also contains six elections from three non-European advanced democracies (Australia, Israel and New Zealand). For these countries, we investigated the presence of populist parties in elections by applying Van Kessel’s classification method. We classify the following parties as populist parties: the One Nation Party (Australia), New Zealand First Party (New Zealand) and Yisrael Beitenu, Likud, Shas and Yesh Atid (Israel). We provide further motivations for these choices online in Appendix 2C.
We construct a variable indicating a vote for a populist party at the current elections and a variable indicating a vote for a populist party at the previous election. At the election level, we create a variable indicating the overall seat-share percentage of populist parties in the previous election. Data on populist party seat shares comes from the parliaments and government database (Döring and Manow Reference Döring and Manow2017).
Political dissatisfaction and other individual-level measures
As a measure of generalized dissatisfaction with politics we use a ‘satisfaction with democracy’ item. The question wording is: ‘On the whole, are you very satisfied, fairly satisfied, not very satisfied or not at all satisfied with the way democracy works in [country]?’ Although this item is somewhat contested (e.g. Canache et al. Reference Canache, Mondak and Seligson2001; Schedler and Sarsfield Reference Schedler and Sarsfield2007), it is considered to be a fairly good summary indicator of political support (Clarke et al. Reference Clarke, Dutt and Kornberg1993) that taps into individuals’ support for the performance of a democratic regime and its actors in general (Linde and Ekman Reference Linde and Ekman2003: 401; Söderlund Reference Söderlund2008). Lower levels of satisfaction with democracy have also been linked to the core of the populist message as an expression of political antipathy (Van Hauwaert and Van Kessel Reference Van Hauwaert and Van Kessel2018; Webb Reference Webb2013).
We control for different individual-level predictors that have been associated with volatility (Dejaeghere and Dassonneville Reference Dejaeghere and Dassonneville2017; Van der Meer et al. Reference Van der Meer, van Elsas, Lubbe and van der Brug2015), voting for populist parties, and with the quality of the recall vote. We control for voters’ level of education (whether or not they have a college degree), political knowledge, age, gender (0=female, 1=male), and employment status (1=employed, 2=unemployed, 3=school/student, 4=retired). To verify that our conclusions are not solely driven by voters for right-wing populist parties (the dominant type of populist-party), we also control for voters’ placement on the ideological left–right dimension (on a scale where left=0 and right=10).
Party identification is considered to be an important vote choice heuristic (Lachat Reference Lachat2007), and partisans are less likely to switch parties or abstain from voting. Others, however, argue that party identification is not exogenous to vote switching and should be considered as an intervening variable between voter characteristics and vote choice (e.g. Söderlund Reference Söderlund2008). We therefore estimate our main models without party ID, but control for the effect of party identification in an additional robustness check.
Party system control variables
Several party system characteristics have been shown to influence volatility. First, the number of viable party alternatives increases the likelihood of vote switching between parties at elections (Pedersen Reference Pedersen1979). A larger number of parties may also affect switching from a party vote to abstention. When there are more parties, it is more likely that one previously voted for a rather small party that turned out to be unable to have an impact on public policies (let alone participate in government). This may reduce the likelihood of a prolonged vote for such parties. Voters without other clear alternatives may switch to abstaining instead (Taagepera et al. Reference Taagepera, Selb and Grofman2014). We control for the number of viable party alternatives that voters have at each election in our sample using the index of the effective number of parties (ENEP) (Laakso and Taagepera Reference Laakso and Taagepera1979). Because the number of effective parties is related to the electoral system (Neto and Cox Reference Neto and Cox1997), we also control for the level of disproportionality in an electoral system (using Michael Gallagher’s least squares indicator). The data for both measures come from Gallagher (Reference Gallagher2017). Because the ideological distance between parties in a party system has an independent effect on vote switching (e.g. Dejaeghere and Dassonneville Reference Dejaeghere and Dassonneville2017: 110), we control for the level of polarization in a party system using Russell Dalton’s (Reference Dalton2008) polarization index measuring the spread of parties along the left–right scale. Furthermore, we control for the number of days between the previous and the current election as more time between elections leads to more switching (Birch Reference Birch2003). As compulsory voting obviously affects the probability that voters abstain from voting (Singh Reference Singh2015), we include a dummy variable for elections with compulsory voting and strict sanctions, and a dummy variable for compulsory voting without strict sanctions (see International IDEA 2017). We control for the GDP growth percentage in the election year because it has been argued that economic decline can animate both the electoral successes of populist parties (Arzheimer Reference Arzheimer2009) and volatile voting behaviour (Dassonneville and Hooghe Reference Dassonneville and Hooghe2017a). Finally, we include a linear and quadratic term for the year in which an election took place. This controls for the possibility of simultaneous but unrelated (time) trends between the predictor variables and trends in the outcome options (Fairbrother Reference Fairbrother2014: 124–125).
Method
Our data set has a hierarchical structure, with 34,893 respondents nested in 32 elections, and these elections nested in 14 countries. Given the categorical nature of our dependent variable and our interest in explanatory factors at the individual and the election level – with the latter requiring multilevel modelling (Gelman and Hill Reference Gelman and Hill2007) – we estimate a series of multinomial multilevel regression models with random intercepts at both the election level and the country level. This type of complex multilevel model can be run via the runmlwin command in Stata (Leckie and Charlton Reference Leckie and Charlton2013). We obtain starting values for our models using second-order penalized quasi-likelihood estimation, and then estimate our final models using Markov Chain Monte Carlo (MCMC) methods (Leckie and Charlton Reference Leckie and Charlton2013). Model-fit statistics of models estimated by MCMC approximation can be assessed by looking at the Bayesian Deviance Information Criterion, with a lower value indicating a better model fit (Spiegelhalter et al. Reference Spiegelhalter, Best, Carlin and Van der Linde2002).
Results
We first investigate to what extent voters for populist and mainstream parties are dissatisfied with democracy. The average level of satisfaction with democracy among voters for populist parties in the pooled data set is 1.56 (on a 0–3 scale), which is considerably lower than the mean level of satisfaction among voters for mainstream parties (1.80). This finding, while in line with the existing literature, should be nuanced somewhat. Only 11% of the voters for populist parties report that they are ‘not at all satisfied’ with the way democracy works (6% among voters for mainstream parties), while 30.8% are ‘not very satisfied’ (20.8% among voters for mainstream parties). Dissatisfaction is thus certainly stronger among voters for populist parties, but not as omnipresent as sometimes suggested. Furthermore, a large majority (75%) of the actual voters who are ‘not at all satisfied’ with democracy still vote for mainstream parties.
Turning to the main analysis, we ran a series of models with the outcome options of: (1) stable vote for the same party, (2) switch to another party, or (3) switch to abstaining. We thus investigate the effects of our predictor variables on those three outcome categories. A stable party vote serves as the base outcome in our multinomial models.
First, we estimated an intercept-only model. The variance partition coefficients (VPC) of this model show that 7.7% of the variance in party switching (compared with stable voting) is at the election and country levels while 18.1% of the variance in switching to abstention is at the election and country levels. Approximately half of the total variance at the higher levels is between-country variance while the other half is between-election variance, but most variance in different voting patterns is thus at the individual level.
The results of our first substantive models are presented in Table 1.Footnote 14 The first set of estimates assesses the impact of the predictors on the outcome category of ‘vote switching’ relative to ‘stable party voting’. The second set of estimates assesses the impact of the predictors on ‘switching to abstention’ relative to ‘stable party voting’. Model I allows testing our first hypothesis by investigating to what extent voters who previously voted for a populist party are vote-switchers compared with voters who voted for mainstream parties. As the ratio of vote switching among voters for populist parties and voters for mainstream parties may vary from election to election, we include random slopes over the election level for the indicator whether one previously voted mainstream or populist. In line with our first hypothesis, the non-significant coefficient of the ‘populist voter previous’ (PVP) variable suggests that populist party voters are not more likely to switch their votes to other parties, nor are they more likely to switch to abstention compared with voters who voted for mainstream parties at the previous election.Footnote 15
Notes: Standard deviations in parentheses: *** p<0.01, ** p<0.05, * p<0.1. PVP=populist voter previous; SWD=satisfaction with democracy; SSPP=seat share populists previous; ENEP=effective number of electoral parties (seat-share based).
Interpreting the other variables in the model shows that increasing levels of satisfaction with the way democracy works significantly decrease the probability of both party switching and switching to abstention (relative to a stable party vote). Age negatively influences the probabilities of both types of vote switching. A university degree has a small positive effect on the probability of being a vote switcher relative to a stable voter, while it strongly reduces the relative probability of switching to abstention. Those on the right are a bit less likely to switch their votes to other parties than left-wing voters, while no ideological effects can be observed on the probability of switching to abstention. At the election level, the effective number of parties in a political system has a significant positive effect upon the probability of switching to abstention, while compulsory voting strongly and significantly reduces this probability.
Model II tests our second hypothesis. It allows investigations into how political dissatisfaction among voters for populist parties influences vote switching and whether the impact of dissatisfaction is different among voters for populist parties compared with voters for mainstream parties. For this purpose, an interaction term between the PVP variable and satisfaction with democracy (SWD) is included. In Figure 1, we graphically present the estimated marginal effects of satisfaction with democracy on the two types of vote switching (relative to a stable party vote).Footnote 16 Among voters for mainstream parties there is a significant negative effect of SWD on the relative probabilities of party switching and switching to abstention. Satisfaction with democracy thus stabilizes the vote among voters for mainstream parties. However, among voters for populist parties, there is a significant positive effect by SWD on the relative probability of party switching. Voters for populist parties who are relatively satisfied are thus more likely to switch their vote to another party while political dissatisfaction reduces their probability of party switching relative to a stable vote. The relative predicted probabilities in Figure 2 show that the most dissatisfied voters for populist parties are significantly less likely to switch parties than the most dissatisfied voters for mainstream parties. In contrast, very satisfied voters for populist parties are significantly more likely to be party-switchers than the very satisfied voters for mainstream parties. Those findings are in line with our second hypothesis.
The second set of estimates from model II shows that political dissatisfaction among voters for populist parties has no significant impact on the relative probability of switching to abstention while such a negative effect exists among the voters for mainstream parties. The predicted probabilities of switching to abstention furthermore show that strongly dissatisfied voters for mainstream parties are more than twice as likely to abstain compared to strongly dissatisfied voters for populist parties, this difference is statistically significant at the p < 0.05 level (Cumming and Finch Reference Cumming and Finch2005). The findings from model II are also in line with our expectation (Hypothesis 2) that it is political dissatisfaction which primarily binds voters to populist parties and makes them stably vote for them.
To test our expectation that stable party voting among voters who previously voted for a populist party is more likely when a populist party has already succeeded in establishing itself as a considerable force in parliament (Hypothesis 3), we included an interaction between PVP and seat-share populist previous (SSPP). Furthermore, as we argued that the effect of SSPP on its capacity to obtain loyal support might follow a curvilinear pattern, those models also contain a quadratic term of SSPP as well as a three-way interaction term of SSPP with PVP. Comparing the model-fit statistics of a model with only the PVP*SSPP interaction (not shown in the tables) and our model III shows a better model-fit by the latter. The results in model III and the marginal effect plots in Figure 3 show a negative effect of the previous level of populist party representation in parliament upon party switching among voters for populist parties. The average marginal effect is not statistically significant over the whole range of the SSPP variable, but this might be because the effect (as expected) flattens out towards the higher end of the SSPP distribution.Footnote 17 Interpreting the relative predicted probabilities in Figure 4 shows a significant drop in the likelihood of populist party voters switching their votes to another party at the lower ranges of the SSPP variable. The probability of party switching among voters for populist parties starts off relatively high (at almost 0.5) when the populist party they voted for is rather small. But when a populist party’s previous size increases to a seat share of around 15%, the probability of switching is nearly halved. When a populist party has a seat share of around 20% and more, its voters become less likely to switch to other parties than the average voters for mainstream parties. At even higher values of SSPP the effect flattens out and becomes insignificant. As our analysis contains only three elections after a government period with a populist party in office, we may have underestimated the curvilinear shape of the effect of populist party size on vote switching among populist party voters. In summary, with regard to party switching, our findings provide good support for our Hypothesis 3. For abstention, however, no clear pattern emerges and voters for populist parties are not very different from voters for mainstream parties.
Robustness tests
We have verified the robustness of our results when controlling for ‘party identification’ and ‘employment’ in the models (models B in the online Appendix). Unsurprisingly, party identification reduces the probabilities of all types of vote switching. Unemployment and being a student increase the probability of party switching, while unemployment and being retired increase the probability of switching to abstention. Adding those control variables does not substantially affect our main findings, though the confidence intervals of the relative predicted probabilities are a bit larger, rendering the predictions of party switching at opposite ends of the SWD variable no longer significantly different when we compare the voters for populist parties with the voters for mainstream parties. The findings regarding the third hypothesis are strengthened after the inclusion of party identification and employment, as the overall marginal effect of the populist seat-share variable becomes significant.
We have also verified whether a stricter exclusion of elections for which vote recall measures might be biased affects our results. Changing the threshold to exclude deviations in the recall vote from 7.5 percentage points to 5 percentage points leads us to drop six additional elections from the sample. Estimating the models on this reduced sample (models C in the online Appendix) leads to substantially the same results (the findings regarding the third hypothesis are even strengthened).
Discussion
In this article, we have reassessed the claim that political dissatisfaction is simultaneously related to voting for populist parties and instability in the vote. Using data from the CSES project, we tested three main hypotheses on the stability of voting for populist parties. Our analyses offer evidence that is in line with each of these hypotheses: we find that populist party voters are not more volatile than mainstream voters, that higher levels of political dissatisfaction increase the likelihood of stably voting for a populist party and that stable voting for a populist party is more likely when there is an offer of strong populist parties – though this effect flattens out.
Our findings suggest that in the current era of a ‘populist zeitgeist’ (Mudde Reference Mudde2004) the populist electorate can no longer be considered a group of volatile voters that come and go. Populist parties attract a core group of voters who are just as loyal to their parties as voters for mainstream parties. This suggests that compared to the past, current-day populist parties may be better at neutralizing dissipating leadership effects and they may be more successful in establishing themselves as political outsiders – despite their more permanent presence in the party system. An important key to this success is that populist parties succeed in stably binding a specific part of those voters who are very dissatisfied with politics. Furthermore, it seems that a more permanent presence by strong populist parties in the party system removes some of the initial hesitation that may have existed against (stable) populist party support.
Nevertheless, when investigating the role of dissatisfaction in voting behaviour, we should not focus on populist parties only. A large majority of the politically dissatisfied still vote for mainstream parties or abstain from voting. Carsten Zelle’s (Reference Zelle1995) logic that ‘frustrated voters’ are volatile voters still holds among voters for mainstream parties. This reveals that the ‘dissatisfied voter’ – as an undifferentiated subtype of voters – does not exist. Citizens can be dissatisfied with mainstream politics, while at the same time rejecting populism.
First, not all voters who are dissatisfied with the inefficiency of political institutions and the inherent opportunism in today’s political processes may be attracted by the easy solutions that populists propose for such practices. Instead, some ‘dissatisfied democrats’ might see the populist message – with its glorification of the common people and its usage of simplistic language – as a further example of the opportunistic strategies of political parties. Second, dissatisfied voters (with full populist attitudes) can decide not to support populist parties because of the (extreme) issue positions of populist parties (e.g. Van Hauwaert and Van Kessel Reference Van Hauwaert and Van Kessel2018). Third, the element of ‘people-centrism’ in populism (Akkerman et al. Reference Akkerman, Mudde and Zaslove2014; Rooduijn Reference Rooduijn2018) can scare away voters. Dissatisfied members of minority groups are indeed unlikely to vote for populist parties because these parties often argue that minorities do not belong to ‘the pure people’. Finally, the dissatisfied can decide to abstain as well. Roy Kemmers (Reference Kemmers2017) has argued that this is especially likely for those who think that the political system is not very responsive. In summary, there are good theoretical reasons for explaining why not all dissatisfied mainstream party voters switch to populist party support, and for explaining that party switching and switching to abstention are more common among dissatisfied mainstream party voters. However, with such explanations only recently being developed in the literature, there is plenty of opportunity to improve upon those views and to put them to an empirical test.
Our analyses offer important and new insights into the stability of populist electorates, but we should also acknowledge a number of limitations, which relate mostly to the available data. We do not dispose of panel data for a varied set of democracies and have to rely on the recall questions that are available in the CSES data to model the determinants of vote switching. In addition, we are limited to observing the dynamics of voting behaviour in two consecutive elections. As scholars have shown that distrust is not only the cause of populist voting, but also one of its consequences (Van der Brug Reference Van der Brug2003; Rooduijn et al. Reference Rooduijn, Van der Brug and de Lange2016), a potential mechanism for the correlation between dissatisfaction and stable support for a populist party – which we cannot test – is that populist party support further fuels political discontent. We leave it to future research, perhaps on a more limited number of countries, to explore this.
Despite these limitations, our results offer important insights with implications for the debate on the impact of populism on democracy. Presuming that populism has some features that strengthen as well as others that weaken aspects of democracy (Mudde and Rovira Kaltwasser Reference Mudde and Rovira Kaltwasser2012), our study shows that one of the supposed positive aspects of populism – that populist parties succeed in stably channelling discontent – functions increasingly well in many Western democracies. This channelling of discontent in a rational, regulated fashion might lead to a sense of democratic fulfilment among populist party voters. Whether this is desirable, however, also depends on the question to what extent political systems with an increasing supply of populist parties succeed in counterbalancing the imminent anti-institutional and often exclusionary tensions inherent in populism (e.g. Kemmers Reference Kemmers2017; Rovira Kaltwasser Reference Rovira Kaltwasser2012).
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/gov.2018.24
Acknowledgements
An earlier version of this paper was presented at the 2017 ECPR general conference. We thank the discussants and panel participants for their feedback. We would also like to thank Tom van der Meer and the anonymous reviewers for their constructive comments on earlier versions of the manuscript. Remko Voogd gratefully acknowledges financial support from the Netherlands Organization for Scientific Research for his entire PhD project (grant number 407-13-029). Ruth Dassonneville acknowledges support from the Fonds de Recherche du Québec-Société et Culture (grant number 2017-NP-199028).