• 1 Introduction

  The success of radical left and radical right populist parties is rising throughout the world (de la Torre, 2019; Inglehart & Norris, 2016; Rooduijn et al., 2019). The 2019 elections in Belgium were no exception to that pattern. Compared to the 2014 elections, these elections saw a substantial shift in party preferences towards radical populist parties on both sides of the ideological spectrum (van Erkel et al., forthcoming). On the radical right-wing populist side, Vlaams Belang (VB) won a substantial number of seats (from 3 seats in 2014 to 18 seats in 2019 in the federal parliament). On the radical left-wing populist side, PVDA-PTB won a substantial number of seats (from 2 seats in 2014 to 12 seats in 2019).
  Previous research generally shows that both protest attitudes and ideological considerations drive the vote for radical populist parties – albeit, depending on the particular country and party, with varying degrees of importance for both determinants (e.g. Bélanger & Aarts, 2006; Oesch, 2008; Ramiro, 2016; Schumacher & Rooduijn, 2013; van der Brug, 2003; van der Brug et al., 2000). How did these determinants play out in the 2019 Belgian elections? To what extent do protest attitudes and ideological considerations drive the vote for the radical populist parties? And, more interestingly, do these determinants drive the vote for the radical right and for the radical left populist parties in the same way? This study contributes to these questions by analysing how protest attitudes and ideological considerations affect radical populist voting behaviour. In particular, we compare voters of radical right populist and voters of radical left populist parties to voters of the mainstream parties in one particular election, namely the 2019 general elections in Belgium. In doing so, we are able to not only disentangle the effect of protest and ideological considerations, but also show how they combine differently for radical right- and left-wing populist parties.
  First, we explore to what degree the votes that went to the VB and the PVDA in Flanders and to the PTB in Wallonia were protest votes. More specifically, we investigate whether voters who are most likely to express a protest vote, i.e. politically distrustful and dissatisfied voters, are more inclined to turn to these radical right and radical left populist parties, as compared to the ideologically more moderate mainstream parties. Second, we explore the role of ideological considerations. To investigate this, we first analyse the radicalness of voters’ policy preferences on radical populist parties’ core issues, i.e. migration for the VB and socio-economic issues for the PVDA-PTB (Delwit et al., 2011). We investigate whether voters with the most radical position on these issues are also most likely to vote for the VB and PVDA-PTB. Besides these insights into the positions on populist parties’ core issues, this study also gets a fuller grasp of the importance of ideological considerations by investigating voters’ positions on a broader range of topics. To do so, a measure of ideological congruence is created. We analyse whether voters whose positions on a broad range of topics are the most congruent with the positions of the VB/PVDA-PTB, are also the most likely to vote for these parties. We include these potential drivers in two multivariate analyses, i.e. one where we compare radical populist voters to all the mainstream parties, and one where we compare radical populist voters to the voters of the one mainstream party that represents the second-best choice for the radical populist voters. This way, our study shows how important protest attitudes and ideological considerations were in explaining voting behaviour for the two successful radical populist parties at the 2019 elections in Belgium, i.e. VB on the right and PVDA-PTB on the left.

• 2 Theory and Literature

  2.1 Voting to Protest

  Populist parties can be identified based on their shared ideology. This ideology

  “considers society to be ultimately separated into two homogeneous and antagonistic groups, ‘the pure people’ versus ‘the corrupt elite’, and which argues that politics should be an expression of the volonté générale (general will) of the people” (Mudde, 2004, p. 543 – original emphasis).

  In line with this definition, we focus on both ends of the ideological spectrum. Both ends are occupied by (radical) populist parties: VB on the right and PVDA-PTB on the left. The latter runs as PVDA (Partij van de Arbeid) in Flanders, and as PTB (Parti du Travail de Belgique) in Wallonia. In the literature, the VB is virtually always classified as a populist party (e.g. Hooghe et al., 2011; Rooduijn et al., 2019), and the PVDA-PTB is also often classified as such (e.g. Wauters & Pittoors, 2019), although its overall assessment of populism is less strong than for the VB. However, particularly in its 2019 campaign, the PVDA-PTB relied strongly on populist campaign messages. Our own analysis using the dataset of Lefevere and Van Aelst (2019) shows that 26% of the PVDA-PTB campaign slogans contained an anti-elite message, and in 13% of their slogans they referred to ‘the people’. This is clearly more compared to the campaign slogans of mainstream parties.¹x Among mainstream parties (i.e. CD&V, Groen, Open VLD, N-VA, sp.a), 0 to 8% of the slogans contained a reference to ‘the people’ while anti-elitist messages were even less prevalent (Lefevere & Van Aelst, 2019). Among Vlaams Belang, 71% contained a reference to ‘the people’ and 17% were anti-elitist. Mainstream parties, on the other hand, are defined in this study as “the electorally dominant actors in the centre-left, centre, and centre-right blocs on the Left-Right political spectrum” (Meguid, 2005, p. 348).
  We know from existing studies that a large share of the votes for populist parties are protest votes (e.g. Ford & Goodwin, 2010; Ford et al., 2012; Hooghe et al., 2011; Oesch, 2008), originating from voters’ political distrust and dissatisfaction with the political elite and the way politics and democracy work. In order to voice their discontent, protest voters choose parties that are “an outcast in the political arena” (Oesch, 2008, p. 353). These ‘outcasts’ extend the battleground because, compared to the standpoints of mainstream parties, their standpoints do not only focus on the substance of politics, but also include criticism on the political elite and the organisation of politics (Hayward, 1996). Overall, a consensus exists amongst the mainstream political elite about how politics and democracy should be organised. These parties rebel against that, which might be one of the core reasons why citizens decide to vote for them.
  In the quest of identifying these ‘outcasts’, the literature has mainly focused on radical right populist parties. However, as Schumacher and Rooduijn (2013) argue, protest attitudes can also explain why voters choose radical left populist parties. After all, it was found that

  support for radical right-wing populist parties can be explained by protest voting is not due to the radical right ideological nature of the party, but to the anti-elitist message in which the party attacks the ‘established political order’. (Schumacher & Rooduijn, 2013, p. 125)

  Accordingly, support for a radical left populist party or any other party that expresses such a message might just as well be explained by protest voting (Ramiro, 2016). This study responds to that by analysing both the populist radical right and the populist radical left.
  But what exactly are these protest attitudes that drive protest voting? The first attitude that has been linked to protest voting is political distrust (Bergh, 2004). This means that we can identify protest voters among the electorate of a particular party by studying whether distrustful voters turn to populist parties rather than to mainstream parties. This approach has been adopted by various studies in the past, and generally shows that this is indeed the case (e.g. Betz, 1994; Henry et al., 2015; Hooghe et al., 2011; Pauwels, 2014; Swyngedouw, 2001).
  A second attitude that has been linked to protest voting is political dissatisfaction. Political dissatisfaction refers to the idea that voters may be disappointed in the existing (mainstream) parties, or even the functioning of democracy. Political dissatisfaction can thus occur in various forms or types, and different types of political dissatisfaction have been shown to matter. Ford et al. (2012), for example, show that dissatisfaction with mainstream parties in Britain was an important driver for the UKIP vote in the 2009 European Parliament elections. Oesch (2008), on the other hand, identified dissatisfaction with the functioning of democracy as an important determinant of voting for populist parties in his cross-national comparative study. It proved to be the “single most important determinant of voting for the Flemish Block and the [Norwegian] Progress Party, and the second most important determinant for the [French] Front National” (Oesch, 2008, p. 368). Although these studies focus mainly on the radical right populist side, there is thus no reason to expect that protest attitudes will play a less important role in explaining the radical left populist vote. A handful of studies indeed already show that protest attitudes are also high among radical left populist voters (e.g. Akkerman et al., 2017; Ramiro & Gomez, 2017). Based on these findings, we formulate the following hypotheses for the 2019 Belgian elections:

  H1: The stronger the voters’ protest attitudes, the more likely they are to vote for the VB and PVDA-PTB, compared to mainstream parties.

  H1a: The higher the voters’ level of political distrust, the more likely they are to vote for the VB and PVDA-PTB, compared to mainstream parties.

  H1b: The higher the voters’ level of dissatisfaction with democracy, the more likely they are to vote for the VB and PVDA-PTB, compared to mainstream parties.

  2.2 Ideological Considerations

  It has been argued in the literature that, besides protest voting, the populist vote is also driven by ideological concerns. This means that the proximity of one’s own policy positions to the policy positions of these parties also explains populist voting behaviour (Schumacher & Rooduijn, 2013). The populist ideology, as defined above, puts the antagonistic relationship between ‘the good people’ and ‘the corrupt elite’ at its core, connecting the populist vote to protest voting. Importantly, however, the populist ideology is a ‘thin-centred ideology’ that is always connected to other (‘thin or full’) ideologies, such as nationalism or socialism (Mudde, 2004, p. 544). This combination of the populist thin ideology – shared by all populist parties – with another ideology, explains the existence of many different variants of populist parties across the political spectrum. Radical right populist parties, such as the VB in Belgium, are generally known for their nativist and xenophobic viewpoints. As an example, the VB (at that time called ‘Vlaams Blok’) drafted in 1992 a 70-point proposal in reaction to what they perceived as ‘the problem of immigration’. As a lot of new parties do, they expand the marketplace of ideas by putting immigration issues on the agenda. Radical left populist parties, such as the PVDA-PTB, are generally known for their radical left socialist or communist policy stances (Bull & Heywood, 1994; Delwit, 2014). Rather than putting forward new issues, they offer a more radical view on existing socio-economic issues. The ideological differences between populist and mainstream parties, but also between different populist parties, are thus also expected to drive the populist vote.
  Several existing studies find evidence to support this argument. For instance, Schumacher and Rooduijn (2013) find that, besides protest attitudes and evaluations of party leaders, policy preferences, particularly the closeness of one’s own policy positions to the policy positions of a populist party, was one of the most important reasons to vote for both left- and right-wing populist parties in the 2006 and 2010 Dutch elections. Similarly, Ivarsflaten (2008) shows that grievances over immigration were the single most important predictor for the success of seven right-wing West European populist parties (2002-2003) (see also van der Brug, 2003; van der Brug et al., 2000, 2005). Moreover, Van Hauwaert and Van Kessel (2018) find that left-wing socio-economic issue positions drive support for left-wing populist parties, and authoritarian and anti-immigration issue positions drive support for right-wing populist parties. Based on these findings, we expect ideological concerns also have played a role in the populist vote in the 2019 elections in Belgium. We expect these ideological considerations to play out in two ways.
  First, we expect radical populist party voters to have different, and particularly more radical, policy positions than mainstream party voters. In particular, we expect these voters to have more radical positions on their core issues. Populist parties are widely known for campaigning on one or two core issues. Generally, these core issues are migration for radical right populist parties and socio-economic issues for radical left populist parties, which relate strongly to their other attached ideology (Mudde, 2004). Populism and radicalism have typically been associated in Western Europe (Rooduijn & Akkerman, 2017). The radicalness particularly relates to their core issues and not so much to other issues that are not central to their right- or left-wing ideology (e.g. Akkerman et al., 2016). We thus expect the VB voters to have more radical positions on migration than the voters of mainstream parties. We also expect this for the PVDA-PTB voters on socio-economic issues, for which we particularly investigate voters’ positions on economic redistribution. This leads to the formulation of two hypotheses:

  H2a: The more radical right-wing the voters’ positions are on migration, the more likely they are to vote for the VB, compared to mainstream parties.

  H2b: The more radical left-wing the voters’ positions are on economic redistribution, the more likely they are to vote for the PVDA-PTB, compared to mainstream parties.

  Second, we expect that the populist vote is also driven by ideological congruence, i.e. the proximity of one’s own policy positions to the policy positions of a political party on a broad range of topics. When ideological congruence is high, voters tend to agree with the policy positions of a certain political party on a broad range of topics. In this study, we expect that voters whose own positions are most in line with the VB or the PVDA-PTB are also most likely to vote for these parties. In line with this reasoning, we formulate our last hypotheses:

  H3a: The more voters’ positions are congruent with the positions of the VB, the more likely they are to vote for VB, compared to mainstream parties.

  H3b: The more voters’ positions are congruent with the positions of the PVDA-PTB, the more likely they are to vote for PVDA-PTB, compared to mainstream parties.

• 3 Data, Method and Operationalisation

  To test our hypotheses, we rely on the RepResent panel survey. This panel survey has been collected by the EOS RepResent consortium, a group of scientists from five Belgian Universities.²x University of Antwerp, Vrije Universiteit Brussel, KU Leuven, Université libre de Bruxelles and UCLouvain. The consortium aims at investigating the relationship between popular democratic resentment and the functioning of representation. It conducted pre- and post-electoral surveys in the scope of the elections of 26 May 2019. Data for the first wave were gathered between 5 April and 5 May 2019. Data for the second wave were collected between 28 May and 18 June 2019. In total, 3,910 respondents completed both surveys (1,978 in Flanders, 1,429 in Wallonia and 503 in the Region of Brussels). The surveys were filled in online and the respondents were provided by the survey company KANTAR TNS, who drew a sample out of their online panel. The company used quota for sex, age and education level.³x A detailed description of the sample can be found in the introduction of this special issue.
  Our analyses are conducted with data from both waves. As we want to explain the vote for the radical populist parties, we rely on the second wave to measure our dependent variable. We use the respondents’ self-reported vote choice for the federal elections in the two main regions of Belgium, i.e. Flanders and Wallonia. For practical reasons, we exclude respondents who casted their vote in Brussels. We include respondents who indicated that they voted for the CD&V, Groen, N-VA, Open VLD, PVDA, sp.a, and VB in Flanders, and the cdH, DéFI, Ecolo, MR, PS and PTB in Wallonia. We exclude respondents who reported to have voted for an ‘other’ party, who did not vote, as well as those who were not allowed to vote or who don’t remember. We also exclude respondents who reported casting a ‘blank/invalid vote’, except for our analysis of protest attitudes (see e.g. Hooghe et al., 2011). Last, we exclude respondents who voted for the Parti Populaire (PP) in Wallonia, because this party did not win a seat.⁴x We present analyses for PP in Appendix E. This study explains the drivers behind the vote of the successful populist radical parties during the 2019 Belgian elections. The reason to include the PP analyses in appendix is that PP in Wallonia is, like VB in Flanders, a populist radical right party (Rooduijn et al., 2019). Although unsuccessful during the 2019 elections – the party lost the one seat they had in the federal parliament and was dissolved by unanimous vote of the party members in June 2019 – we expect the same reasoning to apply for PP voters in Wallonia as for VB voters in Flanders. This results in a total of 1,797 respondents for Flanders (58.8% men, mean age = 52.2) and 1,115 for Wallonia (49.1% men, mean age = 46.8).
  To investigate the role of protest attitudes and ideological considerations in explaining citizens’ vote choice, we rely on items measured in the first wave of the survey. We recognise the interactive nature of the relationships between the different variables and that we should be careful with drawing too strong conclusions about the causality of these relationships. Yet, as the measurement of our independent variables in Wave 1 precedes the measurement of our dependent variable vote choice in Wave 2, the panel structure of our data allows us to make careful predictions about the direction of the results. To tap into voters’ protest attitudes, we use two indicators: political trust and satisfaction with democracy. We rely on a battery of three items to measure political trust: trust in political parties, in the federal parliament and in politicians. These items run from 0 to 10, where 0 indicates ‘no trust at all’ and 10 indicates ‘complete trust’. The three items are combined into a single additive index (Cronbach’s alpha = 0.942). Satisfaction with the way democracy works in Belgium is measured based on one item on a reversed 5-point scale ranging from 1 (‘very dissatisfied’) to 5 (‘very satisfied’). The exact wording of the question reads as follows: “Overall, how satisfied are you with the way democracy is working in Belgium?”
  Regarding ideological considerations, we are interested, first, in the positions of the respondents on populist radical parties’ core issues, i.e. migration for the VB and economic redistribution for the PVDA-PTB. To capture the positions of the respondents on the migration issue we rely on their answers to the following question:

  Some believe that non-western foreigners should be able to live in Europe while maintaining their own culture. Others believe that they must adapt to European culture. Where would you place yourself on a scale from 0 to 10, where 0 means that non-western foreigners must be able to fully retain their own culture and 10 that they must fully adapt to European culture?

  To measure the respondents’ position on the economic redistribution issue, we focus on their position on the statement “Great fortunes must be taxed more” with answers ranging from 1 (‘completely disagree’) to 4 (‘completely agree’).⁵x As H2a and H2b are about being radical on ideological issues, we repeated our analyses and introduced the items with squared effects. For the measure of migration, we kept the item on cultural assimilation measured on an 11-point scale. As regards economic redistribution, our original item was a 4-point scale on taxing great fortunes. We re-ran the analyses with an alternative item measured on a similar 11-point scale to be able to add square effects (item on state intervention in the economy). The results of these analyses are very similar to the ones presented in the article and are available on request.
  Second, we measure ideological congruence on a broader range of issues. In the RepResent survey, 18 statements were asked about different topics (e.g. the economy, environment, abortion, etc.; see Appendix A for the list of statements). The respondents could place themselves on a scale ranging from 1 (‘completely disagree’) to 4 (‘completely agree’). In the scope of the Voting Aid Application (‘De Stemtest/Test électoral’) that was designed by the University of Antwerp and the UCLouvain for the 2019 elections, the same statements were presented to all (mainstream and populist radical) Flemish and francophone political parties. These parties determined their official position on the same scales as the respondents in the RepResent survey did. We obtained access to these data and constructed congruence measures for each respondent in the survey with the populist radical parties VB and PVDA-PTB (for a similar approach, see Lesschaeve et al., 2018). The ideological congruence scale runs from 0 (disagreement on all 18 statements) to 17 (agreement on all 18 statements). In sum, we have three ideological congruence measures, i.e. one that indicates the ideological proximity between each Flemish participant and the VB, one between each Flemish participant and the PVDA and one between each Walloon participant and the PTB.
  To analyse this, we first conduct descriptive analyses. We present mean scores for protest attitudes and ideological considerations for each party’s voter base in Flanders and Wallonia. In particular, we compare the mean scores of voters of radical left and radical right populist parties to the scores of voters of mainstream parties and to each other. These analyses give first insights into the protest attitudes and policy preferences of each party’s voter base. Second, in order to test our hypotheses, we conduct multivariate logistic regressions (see Appendix B for descriptive statistics for all variables in the logistic regressions and Appendix C for correlation matrices). We conduct these analyses for the VB, the PVDA and the PTB to investigate which determinants were important in explaining these radical populist votes. For this analysis, we proceed in two steps.
  In the first step, we compare the radical populist parties with the mainstream parties. For these analyses, we construct a dichotomous variable for vote choice, with respondents scoring 1 if they responded voting for the radical populist party, and 0 if they responded voting for a mainstream party. In particular, to explain the VB vote in Flanders, we include respondents who voted for VB (=1) and all mainstream parties (=0) and exclude respondents who voted for the PVDA. To explain the PVDA vote, we include PVDA voters (=1) and all mainstream party voters in Flanders (=0) and exclude VB voters. To explain the PTB vote in Wallonia, we include PTB voters (=1) and all mainstream party voters in Wallonia (= 0). While this method allows us to describe more general patterns by drawing conclusions about the comparison between populist radical party voters on the one hand and mainstream party voters on the other hand, it also has its limitations. As most voters have choice sets out of which they select one party to vote for, clustering all mainstream parties in the same reference category may affect the informativeness of the results. Take, for instance, a person who votes for the VB. The migration issue might explain why the person did not vote for Groen or the sp.a, but it might not explain why the person did not vote for the N-VA.
  Therefore, in a second step, we take a closer look at the difference between the radical populist party voters and the voters of the one mainstream party that comes closest as an alternative for them. In particular, we construct a variable that compares radical populist voters with the electorates of the party that represent their second-best vote choice. In order to identify this second-best vote choice, we rely on the following question included in wave 2 of the RepResent panel survey:

  Table 1 Average propensity to vote for one of the following parties for populist voters

  For VB voters		For PVDA voters		For PTB voters
  (N = 393)		(N = 133)		(N = 189)
  CD&V	2.08	CD&V	2.73	cdH	1.58
  Groen	1.35	Groen	4.58	DéFI	2.83
  N-VA	4.33	N-VA	1.09	Ecolo	2.29
  Open Vld	2.05	Open Vld	1.80	MR	1.67
  PVDA	1.98	PVDA	8.67	PS	3.00
  sp.a	1.72	sp.a	5.19	PTB	8.52
  VlaamsBelang	8.85	VlaamsBelang	1.26

  Note: The second choice in boldface.

  There are several political parties that are active in Flanders/Wallonia. Could you indicate the extent to which it is likely that you will one day vote for each of the following parties in upcoming elections?

  Respondents answered by giving each party a score on an 11-point scale ranging from 0 (‘very unlikely’) to 10 (‘very likely’). As Table 1 shows, for VB voters, the second option is clearly the N-VA. For the PVDA voters, it is clearly sp.a. And for PTB voters, it is the PS.⁶x DéFI comes close but the N for this party is rather small. Furthermore, the party has a very different status in Brussels (mainstream, established) and in Wallonia (new, alternative). We hence constructed three dummy variables that compare the electorates of VB (=1) with N-VA (=0), PVDA (=1) with sp.a (=0) and PTB (=1) with PS (=0). As we compare parties that are ideologically closer to each other (at least compared with the analysis in the first step), this analysis allows us to conduct a tougher test for ideological considerations and investigate whether – irrespective of protest attitudes – there are still ideological reasons to vote for radical populist parties.
  In the multivariate analyses, we control for a number of socio-demographic variables, i.e. sex, age, level of education and for political interest as well as populist attitudes, as these have been shown to play an important role as well in explaining radical populist voting behaviour (Ramiro, 2016; Van Hauwaert & Van Kessel, 2018). They were measured in the pre-electoral wave of the survey (see Appendix D for more information on the operationalisation of these variables).⁷x We present the analyses without weights. Analyses applying weights for age, sex, education level and vote choice are available on request.

• 4 Results

  4.1 Descriptive Results

  Table 2 Protest attitudes – mean scores

  Flanders			Wallonia
  Party	Political trust	Satisfaction with democracy	Party	Political trust	Satisfaction with democracy
  CD&V	4.87 (1.98)	3.39 (0.86)	cdH	4.11 (2.03)	3.03 (1.02)
  Groen	4.57 (2.22)	3.20 (0.99)	DéFI	3.21 (1.91)	2.59 (0.95)
  N-VA	4.57 (2.01)	3.05 (0.95)	Ecolo	3.57 (2.09)	2.84 (0.98)
  Open VLD	4.73 (2.00)	3.17 (0.90)	MR	4.40 (2.19)	3.17 (0.95)
  sp.a	3.76 (2.29)	2.95 (1.00)	PS	3.62 (2.19)	2.76 (1.04)
  PVDA	2.85 (2.12)	2.50 (0.96)	PTB	1.88 (2.08)	1.96 (0.85)
  VB	2.92 (2.44)	2.24 (0.96)
  Blank/invalid	2.83 (2.58)	2.44 (1.05)	Blank/ invalid	2.24 (2.40)	2.35 (1.03)

  Note: Standard deviations between parentheses.

  Table 3 Ideological considerations mean scores – Flanders

  Party	Migration	Economic redistribution	Ideological congruence with VB	Ideological congruence with PVDA
  CD&V	7.09 (2.07)	3.35 (0.69)	9.06 (2.24)	9.73 (1.98)
  Groen	5.98 (2.04)	3.38 (0.68)	7.65 (1.91)	11.76 (2.30)
  N-VA	7.89 (2.09)	3.04 (0.90)	10.10 (2.20)	8.47 (2.33)
  Open VLD	7.12 (2.03)	2.82 (0.96)	9.32 (2.05)	9.64 (2.07)
  sp.a	6.94 (2.28)	3.48 (0.74)	8.65 (2.33)	11.26 (2.11)
  PVDA	5.95 (2.31)	3.66 (0.64)	8.74 (2.44)	11.94 (2.43)
  VB	7.72 (2.78)	3.21 (0.92)	10.69 (2.34)	9.06 (1.92)

  Note: Standard deviations between brackets.

  Results show that, in line with expectations, the electorates of parties differ when it comes to protest attitudes (see Table 2). In particular, electorates of the populist radical right and populist radical left have higher levels of political distrust and dissatisfaction than the electorates of the mainstream parties.
  First, in Flanders, we observe that VB and PVDA voters display the lowest levels of political trust and satisfaction with democracy, compared to all mainstream parties. Similar conclusions can be drawn for Wallonia, where PTB voters have the lowest trust in politics and are most dissatisfied, compared to the mainstream party voters. Interestingly, while radical populist party voters display the strongest protest attitudes, the electorates of other parties also score rather low, indicating that all citizens are generally distrustful and rather dissatisfied with politics. Other interesting points to note are the positions of sp.a voters in Flanders, DéFI voters in Wallonia, and blank/invalid voters. First, sp.a voters and DéFI voters are situated between the voters of radical populist parties and voters of the mainstream parties in terms of their levels of protest attitudes. Moreover, in line with Hooghe et al. (2011), we find that voters who cast a blank/invalid vote display comparatively high levels of political distrust and dissatisfaction, close to those of radical populist voters. This indicates that voting invalid or blank is also used as a way to express dissatisfaction with politics.
  Regarding ideological considerations, results are also in line with expectations. In Flanders (see Table 3), VB voters display more radical positions related to migration than almost all other voters, except for N-VA voters. Regarding economic redistribution, PVDA voters have the most radical position on this issue, although the difference from sp.a voters is small. On a broader range of topics too, we see that radical populist party voters are most in line with ‘their’ party. Although differences are not huge, VB voters are most congruent with the VB: on average, VB voters agree with 11 out of 18 statements with the VB, while all other voters agree with the VB on 10 or less statements. Although differences are again rather small, we also see that PVDA voters tend to agree most with the positions of the PVDA, but it is also the case for the electorate of other parties on the left end of the spectrum. In particular, not only PVDA voters but also Groen voters agree on average with 12 statements with the PVDA and sp.a voters on average agree with 11 statements. All other parties agree with 10 or less statements.
  In Wallonia too (see Table 4), PTB voters are the most in favour of taxing great fortunes more, although – just as in Flanders – the difference with the social democrats and also Ecolo is small. Regarding ideological congruence for the PTB in Wallonia, differences are less pronounced. Especially, differences between the PTB and Ecolo are small, with mean scores indicating that Ecolo voters are the most in line with the positions of the PTB as a party, followed by PTB voters. They only clearly distinguish themselves from voters of the centre/centre-right parties, cdH and MR.

  Table 4 Ideological considerations mean scores – Wallonia

  Party	Migration	Economic redistribution	Ideological congruence with PTB
  cdH	6.63 (2.33)	3.24 (0.80)	10.69 (2.21)
  DéFI	7.48 (2.24)	3.34 (0.78)	11.08 (2.36)
  Ecolo	6.73 (2.04)	3.42 (0.71)	12.14 (2.25)
  MR	7.61 (2.12)	2.95 (0.81)	9.62 (2.33)
  PS	6.85 (2.26)	3.43 (0.74)	11.06 (2.50)
  PTB	6.69 (2.80)	3.59 (0.70)	11.72 (2.49)

  Note: Standard deviations between parentheses.

  Table 5 Drivers of VB vote – Flanders

  	Probability to vote for VB
  	Model I	Model II	Model III	Model IV
  Intercept	0.191	3.904***	3.643***	1.575*
  	(0.405)	(0.511)	(0.533)	(0.618)
  Female	-0.357**	-0.298*	-0.320*	-0.219
  	(0.130)	(0.141)	(0.142)	(0.146)
  Age	-0.021***	-0.028***	-0.032***	-0.032***
  	(0.004)	(0.005)	(0.005)	(0.005)
  Level of education (ref. none or primary education)
  Secondary education	-0.335	-0.323	-0.312	-0.393
  	(0.179)	(0.196)	(0.198)	(0.201)
  Tertiary education	-1.066***	-1.120***	-1.119***	-1.183***
  	(0.187)	(0.205)	(0.207)	(0.210)
  Political interest	-0.072**	-0.008	-0.016	-0.011
  	(0.024)	(0.026)	(0.026)	(0.027)
  Populist attitudes	0.248**	-0.087	-0.173	-0.222*
  	(0.082)	(0.090)	(0.096)	(0.097)
  Political trust		-0.047***	-0.049***	-0.045***
  		(0.012)	(0.012)	(0.013)
  Satisfaction with democracy		-0.755***	-0.745***	-0.697***
  		(0.082)	(0.083)	(0.084)
  Cultural assimilation of immigrants			0.119***	0.083**
  			(0.030)	(0.030)
  Taxing great fortunes more			-0.017	0.083
  			(0.078)	(0.080)
  Congruence with VB				0.202***
  				(0.030)
  McFadden’s adj. R2	0.048	0.168	0.176	0.200
  Correctly classified in %	76.36	79.13	79.37	79.73
  N	1,658	1,658	1,658	1,658

  Note: The dependent variable is the probability to vote for VB as compared to Flemish mainstream parties. Entries are logit coefficients and standard errors (in parentheses) of a logistic regression. Sign.: *<0.05; **<0.01; ***<0.001.

  4.2 Drivers of the Radical Populist Vote

  For the analysis of the determinants of the populist radical vote, we proceed in two steps. First, we compare the voters of the three different radical populist parties with the voters of the mainstream parties. Second, we compare those radical populist voters to the voters of parties that appear to be radical populist voters’ second-best choice.
  First, in Table 5, we present the results of the analysis where VB voters are compared with the voters of the mainstream parties in Flanders (i.e. the voters of CD&V, Groen, N-VA, Open Vld and sp.a). Model I contains control variables only. Model II adds the protest attitude indicators, Model III adds the two core issues that we have identified for the radical populist parties, and Model IV, the final model, adds ideological congruence with VB. All in all, these models show that the VB vote is driven by both protest attitudes and ideological considerations. We also observe that McFadden’s adjusted R² increases when adding protest attitudes and ideological considerations to our model. The percentage of correctly classified cases is, at 80%, the highest in the final model. Our interpretation below is hence based on Model IV, unless indicated differently.
  Regarding protest attitudes, we observe that the lower voters’ level of political trust, the more likely they are to vote for the VB. The same holds for voters’ satisfaction with democracy. For both indicators, the results are clear and robust across models. These results hence clearly support H1 for VB, stating that the stronger the voters’ protest attitudes, the more likely they are to vote for the VB, compared to the mainstream parties.
  Regarding ideological considerations, the results are also in line with the expectations. The more citizens are in favour of cultural assimilation of immigrants, the more likely they are to vote for the VB. It confirms H2a on radicalness of positions towards migration. Moreover, the more the voters’ positions are congruent with the positions of the VB, the more likely they are to vote for the VB, compared to the mainstream parties. This confirms H3a. These effects, as well as the effects of political trust and satisfaction with democracy, are illustrated in Figure 1 that plots the substantive effects of political trust (a), satisfaction with democracy (b), favouring cultural assimilation (c) and ideological congruence (d) on the probability to vote for the VB. The plotted points show the predicted probability of voting for the VB across the full range of values of these four variables. The figure clearly shows that the probability to vote for VB decreases as political trust increases (from 0.44 to 0.17). The same holds for satisfaction with democracy, with increasing satisfaction reducing the probability to vote for VB (from 0.63 to 0.09). On the other hand, the predicted probability to vote for the VB increases with the request that immigrants should adapt to European culture (from 0.19 to 0.35) as well as with ideological proximity with VB (from 0.06 to 0.66).
  Last, our logistic analyses also give insight into the role played by socio-demographic characteristics, political interest and populist attitudes. We observe that younger people and the lower educated are more likely to vote for the VB. The findings for populist attitudes are mixed. Model I shows that, in line with what one could expect, the stronger citizens’ populist attitudes, the more likely they are to vote for the VB. Yet, this effect disappears in Models II and III and becomes negative in Model IV when protest attitudes and ideological considerations are added.

  Predicted probability to vote for VB by levels of political trust, satisfaction with democracy, cultural assimilation of immigrants and ideological congruence with VB
  Note: Predicted probability of voting for the VB (as compared to Flemish mainstream parties) over the range of values of political trust (a), satisfaction with democracy (b), cultural assimilation of immigrants (c) and congruence with VB (d), while all other covariates are held at their means (the categorical variable level of education is held at the reference category). The points represent the predicted probability of voting for the VB, and the vertical lines are the 95% confidence intervals. Predicted probabilities are generated from the estimates shown in Model IV (Table 5).

  Second, we turn to the PVDA vote and proceed in the same manner (see Table 6). First, we introduce the control variables in Model I and then we add – step by step – the protest attitude indicators (Model II), the core issues (Model III) and ideological congruence with PVDA (Model IV). We can see that our model appears to predict the PVDA vote better than the VB vote, as we reach more than 90% of correctly classified observations in our fourth and final model. Furthermore, Table 6 shows that the PVDA vote is also driven by protest attitudes and ideological considerations.
  Regarding protest attitudes, and just as in the case of the VB, political trust is driving the vote for the PVDA as compared to voting for the mainstream parties, supporting H1a. Moreover, Model IV shows that the more dissatisfied voters are with the way democracy works, the more likely they are to vote for the PVDA, confirming H1b. Overall, results for protest attitudes point towards a confirmation of H1, indicating that the stronger the voters’ protest attitudes, the more likely they are to vote for the PVDA, compared to the mainstream parties.
  Regarding ideological considerations, we observe that radical positions of voters on economic redistribution does indeed affect the probability to vote for the PVDA, supporting H2b (see Model IV). Interestingly, however, migration is also a decisive topic for the radical left populist PVDA voters. As the analysis shows, the less voters demand cultural assimilation of immigrants, the more likely they are to vote for the PVDA. Ultimately, with respect to ideological congruence, our expectations that the more voters’ positions are congruent with the positions of the PVDA, the more likely they are to vote for the party compared to the mainstream parties, are confirmed (H3b).
  To get a better sense of the magnitude of these effects, Figure 2 presents the predicted probability of voting for the PVDA across the full range of values of (a) political trust (b) satisfaction with democracy, (c) cultural assimilation of immigrants (d) taxing great fortunes more and (e) ideological congruence with the PVDA. It shows that the probability to vote for the PVDA decreases as political trust increases (from 0.14 to 0.03), satisfaction with democracy increases (from 0.10 to 0.04), demands for cultural assimilation of immigrants increase (from 0.20 to 0.05), and that the probability to vote for the PVDA increases with the request for taxing great fortunes more (from 0.03 to 0.09) and as ideological proximity with PVDA increases (from 0.01 to 0.34).
  Finally, it can be noted that, just like for VB voters, PVDA voters tend to be younger than mainstream party voters. The role played by populist attitudes is clearer for the PVDA than for the VB voters: the stronger the populist attitudes, the higher the likelihood to vote for the PVDA.

  Table 6 Drivers of PVDA vote – Flanders

  	Probability to vote for PVDA
  	Model I	Model II	Model III	Model IV
  Intercept	-5.019	-1.923*	-2.881**	-5.380***
  	(0.740)	(0.862)	(0.968)	(1.114)
  Female	-0.113	-0.112	-0.082	-0.133
  	(0.204)	(0.209)	(0.216)	(0.220)
  Age	-0.020**	-0.026***	-0.019**	-0.016*
  	(0.006)	(0.007)	(0.007)	(0.007)
  Level of education (ref. none or primary education)
  Secondary education	-0.373	-0.382	-0.554	-0.472
  	(0.298)	(0.307)	(0.316)	(0.321)
  Tertiary education	-0.513	-0.478	-0.542	-0.533
  	(0.297)	(0.306)	(0.313)	(0.320)
  Political interest	0.045	0.100*	0.083*	0.074
  	(0.039)	(0.040)	(0.042)	(0.043)
  Populist attitudes	1.055***	0.651***	0.654***	0.640***
  	(0.144)	(0.154)	(0.164)	(0.170)
  Political trust		-0.077***	-0.063**	-0.058**
  		(0.019)	(0.019)	(0.020)
  Satisfaction with democracy		-0.270*	-0.277*	-0.316**
  		(0.114)	(0.117)	(0.120)
  Cultural assimilation of immigrants			-0.241***	-0.169***
  			(0.043)	(0.046)
  Taxing great fortunes more			0.656***	0.379*
  			(0.163)	(0.170)
  Congruence with PVDA				0.268***
  				(0.048)
  McFadden’s adj. R2	0.062	0.108	0.165	0.201
  Correctly classified in %	90.55	90.48	90.91	91.12
  N	1397	1397	1397	1397

  Note: The dependent variable is the probability to vote for PVDA as compared to Flemish mainstream parties. Entries are logit coefficients and standard errors (in parentheses) of a logistic regression. Sign.: *<0.05; **<0.01; ***<0.001.

  Predicted probability to vote for PVDA by levels of political trust, satisfaction with democracy, cultural assimilation of immigrants, taxing great fortunes more and ideological congruence with the PVDA
  Note: Predicted probability of voting for PVDA (as compared to Flemish mainstream parties) over the range of values of political trust (a), satisfaction with democracy (b), cultural assimilation of immigrants (c), taxing great fortunes more (d) and congruence with PVDA (e), while all other covariates are held at their means (the categorical variable level of education is held at the reference category). The points represent the predicted probability of voting for PVDA, and the vertical lines are the 95% confidence intervals. Predicted probabilities are generated from the estimates shown in Model IV (Table 6).

  Finally, we focus on the PTB vote for which we find a different pattern (see Table 7). While we also observe that protest attitudes play an important role in explaining the vote, ideological considerations appear to be less decisive. Consequently, the proportion of correctly classified observation is very close between Models II and III, showing that adding ideological variables does not add much to our models.
  As Model II indicates, the PTB vote is strongly driven by protest attitudes. Both indicators of protest attitudes are statistically significant. This also holds for the models in which we control for ideological considerations. The lower the voters’ levels of political trust (H1a), and the less satisfied they are with democracy (H1b), the more likely they are to vote for the PTB compared to mainstream parties. Our first hypothesis receives full support for the PTB vote.

  Second, regarding ideological considerations, our expectations about the PTB’s core issue of economic redistribution (H2b) and about ideological congruence on a broader range of topics (H3b) are not confirmed. As discussed in our descriptive analyses, this can potentially be explained by the fact that differences in ideological congruence are small in Wallonia, especially with mainstream parties on the left. This is in line with our theoretical expectations. We had identified a unique selling point of radical right populist parties, whereas radical left populist parties were mainly adopting more radical positions on existing socio-economic issues, making them less distinctive from other left-wing parties, as our analyses confirm. Interestingly, however, ideological considerations play a role in an unexpected way. As for the PVDA, we observe for the PTB that the less voters are in favour of cultural assimilation, the more likely they are to vote for the PTB. Again, we illustrate the effects in Figure 3. As this figure shows, the predicted probability to vote for PTB, as compared to mainstream parties in Wallonia, declines with (a) increasing levels of political trust (from 0.40 to 0.05) and (b) increasing satisfaction with democracy (from 0.44 to 0.07). It also decreases with (c) growing demands for the cultural assimilation of immigrants (from 0.35 to 0.17).
  Last, similar to the results for the PVDA, younger people and people with strong populist attitudes are more likely to vote for the PTB. But similar to the results for the VB, the lower educated are more likely to vote for the PTB.

  Table 7 Drivers of PTB vote – Wallonia

  	Probability to vote for PTB
  	Model I	Model II	Model III	Model IV
  Intercept	-3.292***	0.863	0.635	0.177
  	(0.597)	(0.706)	(0.787)	(0.869)
  Female	0.131	0.048	0.015	-0.007
  	(0.172)	(0.184)	(0.186)	(0.187)
  Age	-0.017**	-0.025***	-0.024***	-0.023***
  	(0.006)	(0.007)	(0.007)	(0.007)
  Level of education (ref. none or primary education)
  Secondary education	-0.457	-0.637*	-0.660*	-0.653*
  	(0.248)	(0.269)	(0.272)	(0.272)
  Tertiary education	-0.984***	-1.005***	-1.021***	-1.030***
  	(0.248)	(0.269)	(0.272)	(0.272)
  Political interest	-0.002	0.064	0.071*	0.070*
  	(0.030)	(0.033)	(0.033)	(0.033)
  Populist attitudes	0.831***	0.330**	0.350**	0.337*
  	(0.124)	(0.127)	(0.133)	(0.133)
  Political trust		-0.085***	-0.082***	-0.083***
  		(0.019)	(0.019)	(0.019)
  Satisfaction with democracy		-0.595***	-0.595***	-0.591***
  		(0.117)	(0.118)	(0.118)
  Cultural assimilation of immigrants			-0.109**	-0.098**
  			(0.037)	(0.038)
  Taxing great fortunes more			0.244	0.199
  			(0.132)	(0.136)
  Congruence with PTB				0.050
  				(0.040)
  McFadden’s adj. R2	0.061	0.172	0.181	0.180
  Correctly classified in %	83.42	84.23	84.32	83.96
  N	1,110	1,110	1,110	1,110

  Note: The dependent variable is the probability to vote for PTB as compared to Walloon mainstream parties. Entries are logit coefficients and standard errors (in parentheses) of a logistic regression. Sign.: *<0.05; **<0.01; ***<0.001.

  Predicted probability to vote for PTB by levels of political trust, satisfaction with democracy and cultural assimilation of immigrants
  Note: Predicted probability of voting for PTB (as compared to Walloon mainstream parties) over the range of values of political trust (a), satisfaction with democracy (b) and cultural assimilation of immigrants (c), while all other covariates are held at their means (the categorical variable level of education is held at the reference category). The points represent the predicted probability of voting for PTB, and the vertical lines are the 95% confidence intervals. Predicted probabilities are generated from the estimates shown in Model IV (Table 7).

  In this first step of the analysis, we focused on the differences between the electorate of the three radical populist parties and the electorate of the mainstream parties as a whole. While this approach allows us to describe general patterns, it also implies that our comparison includes voters who are ideologically rather distant from each other. In order to reflect the voting process better, our additional analyses focus on the comparison of parties that are closer to each other in the radical populist voter’s choice set. More specifically, we compare voters of radical populist parties with voters of the one mainstream party that represents (on average) the second-best choice for the radical populist voters. In this second step, we repeat the analyses above, but instead of comparing the radical populist party voters to those voting for the mainstream parties, we compare them with those who voted for their second-best choice (for the analyses see Appendix F).
  The comparison of VB voters with N-VA voters shows that these voters differ significantly in their protest attitudes, both in terms of political trust and satisfaction with democracy (see Table F.1 and Figure F.1). The difference regarding the ideological variables is, however, not found between the electorates of these two parties. Therefore the protest attitudes are the ones that specifically drive radical right-wing populist voters to the VB. Ideological considerations do not matter anymore when VB voters are compared to the voters of their second-best option N-VA.
  Results differ when looking at the left side of the ideological spectrum. When comparing PVDA to sp.a voters, we find differences regarding satisfaction with democracy but not political trust. PVDA voters appear to be less satisfied with democracy as compared to sp.a voters, but this finding does not replicate for political trust (see Figure F.2). As already hinted at earlier when discussing the descriptive analyses, this can be explained by the fact that sp.a voters are situated somewhere in between the voters of radical populist parties and the voters of the mainstream parties regarding their protest attitudes. Interestingly, in the case of PVDA, we do find ideological differences (see also Figure F.3). The probability to vote for the PVDA decreases as demands for cultural assimilation of immigrants increase (from 0.65 to 0.26).
  For the PTB, we find a similar pattern as we have seen for VB: PTB voters differ from PS voters in terms of protest attitudes, as they also display significantly lower levels of political trust and satisfaction with democracy (see Table F.3). They do not differ, however, with respect to the ideological variables, meaning that not only socio-economic positions do not allow differentiating between PTB and PS, but also that cultural assimilation of immigrants is not a differentiating factor, contrary to what is found in Flanders for PVDA-sp.a, and contrary to what had been emphasised for the choice for PTB against all mainstream parties in Wallonia.

• 5 Conclusion

  Previous research has shown that protest attitudes and ideological considerations are among the two main drivers of radical populist voting behaviour (e.g. Akkerman et al., 2017; Oesch, 2008; Van der Brug et al., 2000; Van Hauwaert & Van Kessel, 2018). This article investigated how these two drivers affected the radical populist vote, both on the left and on the right side of the ideological spectrum in one particular election, namely the 2019 Belgian elections. In particular, we investigated how the populist radical right VB vote and the populist radical left PVDA and PTB vote can be explained by political trust, satisfaction with the functioning of democracy, policy preferences regarding migration and economic redistribution, and ideological congruence.
  In a first descriptive analysis, we showed that voters with the strongest protest attitudes, i.e. politically distrustful and dissatisfied voters, are more inclined to turn to these radical populist parties, as compared to mainstream parties. However, voters who cast blank/invalid votes display similar levels of protest attitudes, and all voters in general display low levels of trust and satisfaction. Besides, when it comes to ideological considerations, we showed that voters of the VB (and the N-VA) adopt more radical policy positions on migration, and that voters of the PVDA-PTB adopt more radical positions on economic redistribution (and the sp.a in Flanders, and Ecolo and the PS in Wallonia). We also investigated voters’ positions on a broader range of topics via a measure of ideological congruence. We highlighted that VB voters were indeed most ideologically congruent with the VB. Congruence is somewhat lower for the PVDA-PTB, as their voters are also congruent with other left-wing parties (Groen and sp.a in Flanders, Ecolo in Wallonia).
  Second, we included these determinants in two multivariate analyses. Overall, we show that both protest attitudes and ideological considerations drive the vote for radical populist parties, but we also show that it depends on that particular comparison. When comparing radical populist party voters to the voters of all mainstream parties, our results confirm that both protest attitudes and ideological considerations play a role. This is the case for the VB vote that is explained by lower levels of political trust and satisfaction with democracy, and by radical positions on migration and an overall congruence with the policy positions of the party. Similar results are found for the PVDA: lower political trust, lower satisfaction with democracy and higher ideological congruence explain the PVDA vote. Moreover, radical positions regarding economic redistribution, and surprisingly also regarding migration (antagonist position to the VB), drive the PVDA vote. Finally, the PTB vote shares some features of the VB vote, and some of the PVDA. Overall, protest attitudes play a strong role, but ideological considerations are less prominent. As for the PVDA, it is radical positions on migration that drive the vote more than a radical position on socio-economic issues.
  Insights into the general pattern of comparing radical populist voters to mainstream voters were gained with this first multivariate analysis; yet, citizens generally have choice sets out of which they select a party to vote for and will be closer to some parties than to others. Therefore, in a second analysis, we compared radical party voters to the voters the party of their second-best choice (VB vs. N-VA, PVDA vs. sp.a, PTB vs. PS). It is a tougher test of the role of ideological considerations. Overall, we show that protest attitudes matter more when comparing radical populist voters to the voters of their second-best party. Only for PVDA voters, radical positions on migration drove the vote compared to sp.a. VB voters or PTB voters were not distinct from N-VA or PS voters regarding any of the ideological considerations. Rather, the difference was guided strongly by protest attitudes.
  This study was not without limitations and hence opens avenues for future research. First, we did not study the role of issue salience. Some issues might be more important for the electorate of certain parties. It would be interesting to study if, and if so, how, issue salience affects the vote for populist radical parties. Second, while our panel data allow us to make careful predictions about the direction of the effects, it does not allow us to fully solve the issue of endogeneity and to fully disentangle the causal structure or interactive nature of protest attitudes and ideological considerations in driving the radical populist vote. For instance, the fact that the VB has been excluded from every coalition on all levels of government for the past decades may have strongly aggravated political distrust and dissatisfaction with democracy, meaning that it is not distrust and dissatisfaction leading to the populist vote choice, but populist vote choice leading to distrust and dissatisfaction. Similarly, voting for a certain party and partisanship may affect loyalty to that party, thereby further increasing ideological congruence and proximity between the voter’s policy positions and the policy positions of the party voted for. Ideally, future research would take these different causal mechanisms into account and give more insights into the different directions of these relationships. Third, we studied the role of economic redistribution for the PVDA-PTB, yet other socio-economic issues might also be interesting to take into account for explaining the radical left populist vote.
  Finally, these three parties managed to attract younger voters (Van der Brug, 2010). It remains to be seen if they are going to be able to keep them in the long run. In order to keep their voters, these parties have an incentive to maintain their protest component prominent, and therefore to remain in the opposition. Being able to portray itself as an alternative to mainstream is electorally important for a party, but also very difficult to maintain in the long run, as the case of the N-VA shows. These parties also have an incentive to maintain their core issues at the centre of the political agenda. Here, the social and political contexts might change and become less favourable to them. But mainstream parties also have a role to play by shifting the agenda back to their own core issues.

References

• Akkerman, T., de Lange, S. & Rooduijn, M. (Eds.). (2016). Radical parties in Western Europe: Into the mainstream? Routledge.

• Akkerman, A., Zaslove, A. & Spruyt, B. (2017). ‘We the People’ or ‘We the Peoples’? A comparison of support for the populist radical right and populist radical left in the Netherlands. Swiss Political Science Review, 23(4), 377-403.

• Bélanger, E. & Aarts, K. (2006). Explaining the rise of the LPF: Issues, discontent, and the 2002 Dutch election. Acta Politica, 41(1), 4-20.

• Bergh, J. (2004). Protest voting in Austria, Denmark, and Norway. Scandinavian Political Studies, 27(4), 245-273.

• Betz, H. G. (1994). Radical right-wing populism in Western Europe. Macmillan.

• Bull, M. J. & Heywood, P. (Eds.). (1994). West European communist parties after the revolutions of 1989. Palgrave.

• De la Torre, C. (Ed.). (2019). Routledge handbook of global populism. Routledge.

• Delwit, P. (2014). Nouvelle gauche, vieille recette. Pire.

• Delwit, P., Pilet, J.-B. & van Haute, E. (Eds.). (2011). Les partis politiques en Belgique. Éditions de l’Université de Bruxelles.

• Ford, R. & Goodwin, M. J. (2010). Angry white men: Individual and contextual predictors of support for the British National Party. Political Studies, 58, 1-25.

• Ford, R., Goodwin, M. & Cutts, D. (2012). Strategic Eurosceptics and polite xenophobes: Support for the United Kingdom Independence Party (UKIP) in the 2009 European Parliament elections. European Journal of Political Research, 51(2), 204-234.

• Hayward, J. (Ed.). (1996). Elitism, populism, and European politics. Oxford University Press.

• Henry, L., van Haute, E. & Hooghe, M. (2015). Confiance, satisfaction et comportement électoral dans un Etat fédéral. In K. Deschouwer, P. Delwit, M. Hooghe, P. Baudewyns & S. Walgrave (Eds.), Décrypter l’électeur. Le Comportement Électoral et Les Motivations de Vote (pp. 213-234). Lannoo.

• Hooghe, M., Marien, S. & Pauwels, T. (2011). Where do distrusting voters turn if there is no viable exit or voice option? The impact of political trust on electoral behaviour in the Belgian regional elections of June 2009. Government and Opposition, 46(2), 245-273.

• Inglehart, R. F. & Norris, P. (2016). Trump, Brexit, and the rise of populism: Economic have-nots and cultural Backlash. HKS Working Paper No. RWP16-026. http://ssrn.com/abstract=2818659 or doi: 10.2139/ssrn.2818659

• Ivarsflaten, E. (2008). What unites right-wing populists in Western Europe?: Re-examining grievance mobilization models in seven successful cases. Comparative Political Studies, 41(1), 3-23.

• Lefevere, J. & Van Aelst, P. (2019). Towards a first-order arena? Domestic vs European attention in Belgian Party Campaigns, 2014-2019. Presented at ECPR General Conference, Wroclaw, Poland.

• Lesschaeve, C., Van Erkel, P. & Meulewaeter, C. (2018). Thinking alike: Two pathways to leadership-candidate opinion congruence. Journal of Elections, Public Opinion and Parties, 28(4), 488-515.

• Meguid, B. M. (2005). Competition between unequals: The role of mainstream party strategy in Niche Party success. American Political Science Review, 99(3), 347-359.

• Mudde, C. (2004). The populist Zeitgeist. Government and Opposition, 39, 541-563.

• Oesch, D. (2008). Explaining workers’ support for right-wing populist parties in Western Europe: Evidence from Austria, Belgium, France, Norway, and Switzerland. International Political Science Review, 29(3), 349-373.

• Pauwels, T. (2014). Populism in Western Europe. Comparing Belgium, Germany and the Netherlands. Routledge.

• Ramiro, L. (2016). Support for radical left parties in Western Europe: Social background, ideology and political orientations. European Political Science Review, 8(1), 1-23.

• Ramiro, L. & Gomez, R. (2017). Radical-left populism during the Great Recession: Podemos and its competition with the established radical left. Political Studies, 65(1S), 108-126.

• Rooduijn, M. & Akkerman, T. (2017). Flank attacks: Populism and left-right radicalism in Western Europe. Party Politics, 23(3), 193-204.

• Rooduijn, M., Van Kessel, S., Froio, C., Pirro, A., De Lange, S., Halikiopoulou, D., Lewis, P., Mudde, C. & Taggart, P. (2019). The PopuList: An overview of populist, far right, far left and Eurosceptic parties in Europe. http://www.popu-list.org.

• Schumacher, G. & Rooduijn, M. (2013). Sympathy for the ‘devil’? Voting for populists in the 2006 and 2010 Dutch general elections. Electoral Studies, 32(1), 124-133.

• Swyngedouw, M. (2001). The subjective cognitive and affective map of extreme right voters: Using open-ended questions in exit polls. Electoral Studies, 20(2), 217-241.

• Van der Brug, W. (2003). How the LPF fuelled discontent: Empirical tests of explanations of LPF support. Acta Politica, 38(1), 89-106.

• Van der Brug, W. (2010). Structural and ideological voting in age cohorts. West European Politics, 33(3), 586-607.

• Van der Brug, W., Fennema, M. & Tillie, J. (2000). Anti-immigrant parties in Europe: Ideological or protest vote? European Journal of Political Research, 37(1), 77-102.

• Van der Brug, W., Fennema, M. & Tillie, J. (2005). Why some anti-immigrant parties fail and others succeed: A two-step model of aggregate electoral support. Comparative Political Studies, 38(5), 537-573.

• Van Erkel, P., Lefevere, J., Walgrave, S., Jennart, I., Kern, A., Marien, S. & Baudewyns, P. (forthcoming). Des Transferts de Voix Avant ou Pendant la Campagne ? Une Étude sur les Transferts Électoraux lors des Élections de 2019 en Belgique. In J.-B. Pilet, K. Deschouwer, A. Kern & J. Lefevere (Eds.), Les Élections Belges du 26 mai 2019. Une Analyse des Comportements Électoraux. Presses universitaires de Louvain .

• Van Hauwaert, S. M. & Van Kessel, S. (2018). Beyond protest and discontent: A cross-national analysis of the effect of populist attitudes and issue positions on populist party support. European Journal of Political Research, 57, 68-92.

• Wauters, B. & Pittoors, G. (2019). Populist party leaders in Belgium: An analysis of VB and PVDA-PTB. Polish Political Science Review, 7(1), 1-23.

Appendix A List of Statements

1. Hosting transit migrants must be a punishable offence.

2. Situation tests must be put in place to detect discrimination in employment.

3. There must be a test on European values in order to obtain the Belgian nationality.

4. If the request for asylum of families with children is rejected, these families can be placed in detention pending their repatriation.

5. By 2024, company cars that run with petrol or diesel must be banned.

6. The VAT on electricity must be reduced from 21% to 6%.

7. There must be a tax on plane tickets in order to raise their price.

8. Nuclear power plants must remain operational after 2025.

9. One cannot drink alcohol at all while driving (zero tolerance).

10. Abortion must be allowed beyond the 12th week of a pregnancy.

11. Sperm donation must no longer be anonymous.

12. Big fortunes must be taxed more heavily.

13. Wages must no longer be automatically indexed.

14. The fingerprints of all citizens must be kept in a central database.

15. Shops must be able to choose when to do sales.

16. A retirement pension of at least 1,500 € per month must be put in place.

17. The government should be composed of an equal number of men and women.

18. Important political decisions must be handled by citizens via a referendum.

Appendix B Descriptive Statistics of Variables in the Logistic Regression Analyses
Table B.1 Descriptive statistics of variables in the logistic regression analyses – Flanders

	Mean	Standard deviation	Minimum	Maximum
Female	0.42	0.49	0	1
Age	52.15	15.98	18	91
Level of education			1	5
No or primary education (ref.)	0.12	0.33	0	1
Secondary education	0.40	0.49	0	1
Higher education	0.48	0.50	0	1
Political interest	5.96	2.72	0	10
Populist attitudes	3.59	0.75	1	5
Political trust	11.84	6.97	0	30
Satisfaction with democracy	2.85	1.03	1	5
Migration issue	7.25	2.38	0	10
Economic redistribution issue	3.22	0.86	1	4
Ideological congruence with VB	9.54	2.42	0	17
Ideological congruence with PVDA	9.78	2.48	0	17

Table B.2 Descriptive statistics of variables in the logistic regression analyses – Wallonia

	Mean	Standard deviation	Minimum	Maximum
Female	0.52	0.50	0	1
Age	46.59	14.23	18	88
Level of education			1	5
No or primary education (ref.)	0.12	0.33	0	1
Secondary education	0.37	0.48	0	1
Higher education	0.51	0.50	0	1
Political interest	5.15	2.93	0	10
Populist attitudes	3.61	0.74	1	5
Political trust	9.75	6.90	0	30
Satisfaction with democracy	2.64	1.05	1	5
Migration issue	7.09	2.37	0	10
Economic redistribution issue	3.33	0.90	1	4
Ideological congruence with PTB	10.92	2.53	0	17

Appendix C Correlation Matrices
Table C.1 Matrix of correlations of independent variables for the Flemish models (VB & PVDA)

	Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
(1)	Female	1.000
(2)	Age	-0.237	1.000
(3)	Secondary education (ref. primary education)	-0.043	-0.025	1.000
(4)	Higher education (ref. primary education)	0.072	-0.066	-0.781	1.000
(5)	Political interest	-0.272	0.241	-0.084	0.134	1.000
(6)	Populist attitudes	-0.002	0.135	0.110	-0.158	-0.042	1.000
(7)	Satisfaction with democracy	-0.010	-0.036	-0.077	0.102	0.135	-0.249	1.000
(8)	Political trust	-0.005	-0.072	-0.042	0.094	0.270	-0.325	0.605	1.000
(9)	Cultural assimilation of immigrants	-0.044	0.269	0.006	-0.033	0.107	0.240	-0.062	-0.036	1.000
(10)	Taxing big fortunes more	-0.007	0.178	0.078	-0.113	0.003	0.313	-0.111	-0.161	0.051	1.000
(11)	Ideological congruence with VB	-0.103	0.012	0.065	-0.059	-0.019	0.097	-0.196	-0.162	0.244	-0.161	1.000
(12)	Ideological congruence with PVDA	0.149	-0.112	0.030	-0.036	-0.114	0.131	0.015	-0.076	-0.235	0.325	-0.362	1.000

Table C.2 Matrix of correlations of independent variables for the Walloon models (PTB)

	Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
(1)	Female	1.000
(2)	Age	-0.166	1.000
(3)	Secondary education (ref. primary education)	-0.021	-0.070	1.000
(4)	Higher education (ref. primary education)	0.025	-0.001	-0.775	1.000
(5)	Political interest	-0.216	0.139	-0.100	0.154	1.000
(6)	Populist attitudes	-0.026	0.124	0.049	-0.066	-0.070	1.000
(7)	Satisfaction with democracy	-0.070	-0.052	-0.107	0.127	0.177	-0.307	1.000
(8)	Political trust	-0.094	-0.059	-0.094	0.105	0.358	-0.332	0.605	1.000
(9)	Cultural assimilation of immigrants	-0.057	0.207	0.025	-0.036	0.130	0.218	-0.084	-0.051	1.000
(10)	Taxing big fortunes more	0.044	0.146	0.060	-0.069	-0.039	0.273	-0.201	-0.214	0.044	1.000
(11)	Ideological congruence with PTB	0.133	-0.034	-0.028	0.042	-0.064	0.159	-0.111	-0.114	-0.178	0.307	1.000

Appendix D Measurement of Control Variables

Sex: Sex of the respondent
Question: ‘What is your gender?’
Scale: 1 = Man, 2 = Woman, 3 = Other

Age: Age of the respondent
Question: ‘What is your age?’
Respondents answered this question by filling in their age (e.g. 46)

Level of Education: Level of education of the respondent
Question: ‘What is the highest level of education you have achieved?’
Scale: 1 = None or elementary; 2 = Secondary, incomplete (lower ASO, BSO or TSO); 3 = Secondary, complete (ASO, BSO or TSO); 4 = Higher non-university; 5 = University

Political Interest: Level of political interest of the respondent
Question: ‘To what extent are you interested in politics in general?’
Scale: 0 to 10: 0 = Not interested at all, 10 = Extremely interested

Populist attitudes: Level of populist attitudes of the respondent
Three items to measure populist attitudes:
“Politicians must follow the people’s opinion.”
“Political opposition is more present between citizens and the elite than between citizens themselves.”
“I prefer being represented by an ordinary citizen rather than by a professional politician.”
Scale: All measures on 5-point scales ranging from 1 ‘totally disagree’ to 5 ‘totally agree’. A sum index was created (Cronbach’s alpha: 0.678).

Appendix E PP Results

Results for PP are presented. These results should be interpreted with caution, as the number of respondents for PP is low (N = 59).

Table E.1 Protest attitudes mean scores

Flanders			Wallonia
Party	Political trust	Satisfaction with democracy	Party	Political trust	Satisfaction with democracy
CD&V	4.87 (1.98)	3.39 (0.86)	cdH	4.11 (2.03)	3.03 (1.02)
Groen	4.57 (2.22)	3.20 (0.99)	DéFI	3.21 (1.91)	2.59 (0.95)
N-VA	4.57 (2.01)	3.05 (0.95)	Ecolo	3.57 (2.09)	2.84 (0.98)
Open VLD	4.73 (2.00)	3.17 (0.90)	MR	4.40 (2.19)	3.17 (0.95)
sp.a	3.76 (2.29)	2.95 (1.00)	PS	3.62 (2.19)	2.76 (1.04)
PVDA	2.85 (2.12)	2.50 (0.96)	PP	1.94 (1.94)	1.98 (0.94)
VB	2.92 (2.44)	2.24 (0.96)	PTB	1.88 (2.08)	1.96 (0.85)
Blank/invalid	2.83 (2.58)	2.44 (1.05)	Blank/ invalid	2.24 (2.40)	2.35 (1.03)

Note. Standard deviations between brackets.

Table E.2 Ideological considerations mean scores – Wallonia

Party	Migration	Economic redistribution	Ideological congruence with PTB	Ideological congruence with PP
cdH	6.63 (2.33)	3.24 (0.80)	10.69 (2.21)	9.04 (1.68)
DéFI	7.48 (2.24)	3.34 (0.78)	11.08 (2.36)	8.98 (2.02)
Ecolo	6.73 (2.04)	3.42 (0.71)	12.14 (2.25)	7.97 (2.13)
MR	7.61 (2.12)	2.95 (0.81)	9.62 (2.33)	9.77 (2.23)
PP	8.32 (2.33)	3.12 (0.91)	9.75 (2.20)	11.00 (2.01)
PS	6.85 (2.26)	3.43 (0.74)	11.06 (2.50)	8.89 (2.02)
PTB	6.69 (2.80)	3.59 (0.70)	11.72 (2.49)	9.28 (1.98)

Note. Standard deviations between parentheses.

Table E.3 Drivers of PP vote – Wallonia

	Probability to vote for PP
	Model I	Model II	Model III	Model IV
Intercept	-3.514	0.702	0.636	-2.855*
	(0.970)	(1.094)	(1.197)	(1.425)
Female	-0.138	-0.251	-0.183	0.009
	(0.285)	(0.297)	(0.303)	(0.317)
Age	-0.017	-0.026*	-0.028*	-0.029*
	(0.010)	(0.010)	(0.011)	(0.011)
Level of education (ref. none or primary education)
Secondary education	-0.330	-0.451	-0.482	-0.446
	(0.399)	(0.417)	(0.424)	(0.433)
Higher education	-0.829*	-0.807	-0.853*	-0.919*
	(0.401)	(0.420)	(0.425)	(0.441)
Political interest	-0.026	0.066	0.026	-0.020
	(0.050)	(0.053)	(0.054)	(0.057)
Populist attitudes	0.625**	0.087	-0.023	-0.117
	(0.204)	(0.198)	(0.212)	(0.217)
Political trust		-0.096**	-0.092**	-0.086**
		(0.031)	(0.031)	(0.031)
Satisfaction with democracy		-0.586**	-0.557**	-0.539**
		(0.185)	(0.188)	(0.187)
Cultural assimilation of immigrants			0.240**	0.150*
			(0.075)	(0.072)
Taxing big fortunes more			-0.361*	-0.169
			(0.161)	(0.168)
Congruence with PP				0.398***
				(0.080)
McFadden’s adj. R2	0.010	0.107	0.136	0.192
Correctly classified in %	93.97	93.97	94.08	93.97
N	979	979	979	979

Significance: * < 0.05; ** < 0.01; *** < 0.001.

Appendix F Comparison with Second-Best Choice
Table F.1 Drivers of VB vote as compared to the N-VA vote – Flanders

	Probability to vote for VB
	Model I	Model II	Model III	Model IV
Intercept	1.573	5.147	5.059	4.811
	(0.482)	(0.642)	(0.647)	(0.751)
Female	-0.312*	-0.207	-0.219	-0.208
	(0.155)	(0.170)	(0.171)	(0.172)
Age	-0.034***	-0.042***	-0.044***	-0.044***
	(0.005)	(0.006)	(0.006)	(0.006)
Level of education (ref. none or primary education)
Secondary education	-0.362	-0.327	-0.328	-0.345
	(0.221)	(0.245)	(0.246)	(0.246)
Higher education	-1.061***	-1.189***	-1.173***	-1.181***
	(0.228)	(0.252)	(0.254)	(0.254)
Political interest	-0.089**	-0.020	-0.019	-0.020
	(0.028)	(0.033)	(0.033)	(0.033)
Populist attitudes	0.348***	0.040	-0.019	-0.027
	(0.097)	(0.111)	(0.116)	(0.117)
Political trust		-0.069***	-0.069***	-0.068***
		(0.015)	(0.015)	(0.015)
Satisfaction with democracy		-0.645***	-0.639***	-0.630***
		(0.098)	(0.098)	(0.099)
Cultural assimilation of immigrants			0.026	0.023
			(0.035)	(0.036)
Taxing big fortunes more			0.138	0.148
			(0.092)	(0.093)
Congruence with VB				0.028
				(0.036)
McFadden’s adj. R2	0.078	0.198	0.197	0.196
Correctly classified in %	67.03	72.75	73.30	73.19
N	910	910	910	910

Note: The dependent variable is the probability to vote for VB as compared to voting for the N-VA. Entries are logit coefficients and standard errors (in parentheses) of a logistic regression.
Significance: * < 0.05; ** < 0.01; *** < 0.001.

Predicted probability to vote for the VB as compared to voting for N-VA by levels of political trust and satisfaction with democracy
Note: Predicted probability of voting for the VB (as compared to voting for N-VA) over the range of values of political trust (a) and satisfaction with democracy (b), while all other covariates are held at their means (the categorical variable level of education is held at the reference category). The points represent the predicted probability of voting for the VB, and the vertical lines are the 95% confidence intervals. Predicted probabilities are generated from the estimates shown in Model IV (Table F.1.).
Predicted probability to vote for the PVDA as compared to voting for sp.a by levels of satisfaction with democracy and cultural assimilation of immigrants
Note: Predicted probability of voting for PVDA (as compared to voting for sp.a) over the range of values of satisfaction with democracy (a) and cultural assimilation of immigrants (b), while all other covariates are held at their means (the categorical variable level of education is held at the reference category). The points represent the predicted probability of voting for the PVDA, and the vertical lines are the 95% confidence intervals. Predicted probabilities are generated from the estimates shown in Model IV (Table F.2.).
Table F.2 Drivers of PVDA vote as compared to the sp.a vote – Flanders

	Probability to vote for PVDA
	Model I	Model II	Model III	Model IV
Intercept	-2.195	-0.102	-0.029	-0.858
	(0.890)	(1.049)	(1.199)	(1.362)
Female	-0.493	-0.568*	-0.534	-0.535
	(0.268)	(0.274)	(0.279)	(0.280)
Age	-0.028**	-0.030***	-0.025**	-0.025**
	(0.009)	(0.009)	(0.009)	(0.009)
Level of education (ref. none or primary education)
Secondary education	-0.235	-0.289	-0.369	-0.344
	(0.352)	(0.362)	(0.368)	(0.370)
Higher education	0.268	0.316	0.236	0.253
	(0.361)	(0.371)	(0.376)	(0.378)
Political interest	0.075	0.092	0.076	0.072
	(0.048)	(0.049)	(0.050)	(0.050)
Populist attitudes	0.770***	0.567**	0.656**	0.652**
	(0.178)	(0.190)	(0.202)	(0.202)
Political trust		-0.024	-0.023	-0.022
		(0.022)	(0.023)	(0.023)
Satisfaction with democracy		-0.372**	-0.331*	-0.341*
		(0.141)	(0.142)	(0.143)
Cultural assimilation of immigrants			-0.187***	-0.166**
			(0.056)	(0.059)
Taxing big fortunes more			0.155	0.114
			(0.199)	(0.200)
Congruence with PVDA				0.074
				(0.058)
McFadden’s adj. R2	0.039	0.061	0.080	0.080
Correctly classified in %	65.38	68.93	70.12	71.01
N	338	338	338	338

Note: The dependent variable is the probability to vote for PVDA as compared to voting for the sp.a. Entries are logit coefficients and standard errors (in parentheses) of a logistic regression.
Significance: * < 0.05; ** < 0.01; *** < 0.001.

Table F.3 Drivers of PTB vote as compared to the PS vote – Wallonia

	Probability to vote for PTB
	Model I	Model II	Model III	Model IV
Intercept	-2.139	1.856	2.207*	1.911
	(0.684)	(0.862)	(0.953)	(1.030)
Female	0.208	0.0511	0.039	0.022
	(0.201)	(0.219)	(0.221)	(0.222)
Age	-0.022**	-0.034***	-0.032***	-0.032***
	(0.008)	(0.009)	(0.009)	(0.009)
Level of education (ref. none or primary education)
Secondary education	-0.152	-0.522	-0.580	-0.576
	(0.283)	(0.314)	(0.318)	(0.318)
Higher education	-0.175	-0.414	-0.476	-0.485
	(0.286)	(0.316)	(0.319)	(0.320)
Political interest	0.002	0.062	0.062	0.059
	(0.035)	(0.040)	(0.040)	(0.041)
Populist attitudes	0.750***	0.357*	0.432**	0.421**
	(0.142)	(0.149)	(0.160)	(0.160)
Political trust		-0.086***	-0.085***	-0.085***
		(0.022)	(0.022)	(0.022)
Satisfaction with democracy		-0.524***	-0.553***	-0.551***
		(0.138)	(0.140)	(0.140)
Cultural assimilation of immigrants			-0.098*	-0.091
			(0.047)	(0.047)
Taxing big fortunes more			0.018	-0.015
			(0.166)	(0.172)
Congruence with PTB				0.036
				(0.047)
McFadden’s adj. R2	0.036	0.149	0.150	0.147
Correctly classified in %	64.10	70.94	72.01	73.08
N	468	468	468	468

Note: The dependent variable is the probability to vote for PTB as compared to voting for the PS. Entries are logit coefficients and standard errors (in parentheses) of a logistic regression.
Significance: * < 0.05; ** < 0.01; *** < 0.001.

Predicted probability to vote for the PTB as compared to voting for the PS by levels of political trust and satisfaction with democracy
Note: Predicted probability of voting for PTB (as compared to voting for the PS) over the range of values of political trust (a) and satisfaction with democracy (b), while all other covariates are held at their means (the categorical variable level of education is held at the reference category). The points represent the predicted probability of voting for the PTB, and the vertical lines are the 95% confidence intervals. Predicted probabilities are generated from the estimates shown in Model IV (Table F.3.).

Noten

• * This article extends a chapter that will be published in Dutch and in French in a forthcoming book. While the book chapter compares populist party voters to mainstream party voters regarding their protest attitudes and ideological considerations, this article investigates how both protest attitudes and ideological considerations explain the 2019 vote for the populist radical left and the populist radical right. The hypotheses formulated and analyses made in this article consequently differ from the book chapter, yet the theoretical framework in this article is only a slightly adapted version of the one in the book chapter.

• 1 Among mainstream parties (i.e. CD&V, Groen, Open VLD, N-VA, sp.a), 0 to 8% of the slogans contained a reference to ‘the people’ while anti-elitist messages were even less prevalent (Lefevere & Van Aelst, 2019). Among Vlaams Belang, 71% contained a reference to ‘the people’ and 17% were anti-elitist.

• 2 University of Antwerp, Vrije Universiteit Brussel, KU Leuven, Université libre de Bruxelles and UCLouvain.

• 3 A detailed description of the sample can be found in the introduction of this special issue.

• 4 We present analyses for PP in Appendix E. This study explains the drivers behind the vote of the successful populist radical parties during the 2019 Belgian elections. The reason to include the PP analyses in appendix is that PP in Wallonia is, like VB in Flanders, a populist radical right party (Rooduijn et al., 2019). Although unsuccessful during the 2019 elections – the party lost the one seat they had in the federal parliament and was dissolved by unanimous vote of the party members in June 2019 – we expect the same reasoning to apply for PP voters in Wallonia as for VB voters in Flanders.

• 5 As H2a and H2b are about being radical on ideological issues, we repeated our analyses and introduced the items with squared effects. For the measure of migration, we kept the item on cultural assimilation measured on an 11-point scale. As regards economic redistribution, our original item was a 4-point scale on taxing great fortunes. We re-ran the analyses with an alternative item measured on a similar 11-point scale to be able to add square effects (item on state intervention in the economy). The results of these analyses are very similar to the ones presented in the article and are available on request.

• 6 DéFI comes close but the N for this party is rather small. Furthermore, the party has a very different status in Brussels (mainstream, established) and in Wallonia (new, alternative).

• 7 We present the analyses without weights. Analyses applying weights for age, sex, education level and vote choice are available on request.