Populism

Although the concept of populism is a highly contested concept, scholars increasingly agree about an appropriate definition of the term. Mudde defines populism as ‘an ideology that 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’ (Reference Mudde2004: 543).Footnote ² Hawkins proposes a highly similar definition: populism is, according to him, ‘a Manichean discourse that identifies Good with a unified will of the people and Evil with a conspiring elite’ (Reference Hawkins2009: 1042). These definitions have in common that: (1) they focus on the antagonistic relationship between the good people and the evil elite; and (2) they conceive of populism as a set of ideas (Hawkins et al., Reference Hawkins, Riding and Mudde2012).

Populists conceive of the people as a homogeneous entity, devoid of divisions. They are ‘people-centrist’ because they argue that the will of the people should be the point of departure of all political decision-making (Taggart, Reference Taggart2000). The term ‘the people’ could refer to the nation, the electorate, peasants, or to the proletariat (Canovan, Reference Canovan1981). Populists are anti-elitist because they accuse the elite of being incompetent, arrogant, and/or selfish (Canovan, Reference Canovan2002; Laclau, Reference Laclau2005; Barr, Reference Barr2009). The critique could be directed toward political elites, economic elites, media elites, and/or cultural elites (Jagers and Walgrave, Reference Jagers and Walgrave2007).

The mentioned definitions do not only have in common that they emphasize people-centrism and anti-elitism. They also share with each other that they perceive populism as more than merely a particular rhetoric, style, or strategy. Populism is conceived of as being a substantive message – or a set of ideas (Hawkins et al., Reference Hawkins, Riding and Mudde2012). As a set of ideas, populism can be attached to different ideologies, ranging from left-wing to right-wing and from progressive to conservative. Populism has therefore been compared to a chameleon: it takes on the color of the environment in which it occurs (Taggart, Reference Taggart2000).

Most previous studies on populist voter bases focused on either the radical right or the radical left. If we aim at making inferences about what the electorates of populist parties in general have in common, however, relying on these studies will lead to flawed conclusions for two main reasons: (1) these analyses are not inclusive enough because many other (non-radical) parties that might also express a populist message are excluded from the analysis; and (2) these analyses are contaminated by the specific ideological positions of the voters for these particular party families (e.g. anti-immigrant attitudes when it comes to radical right parties and attitudes toward welfare redistribution when it comes to radical left parties).

Who votes for a populist?

Kriesi et al. (Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006; Reference Kriesi, Grande, Lachat, Dolezal and Bornschier2008) claim that the Western European societal landscape has changed profoundly. A new political cleavage is said to have developed, which distinguishes the winners from the losers of globalization. The winners are those who profit from international competition (e.g. highly educated entrepreneurs) and the losers are those who feel threatened by the opening of borders (e.g. lower educated employees in traditionally protected sectors). Kriesi et al. argue that the transformation of the Western European societal landscape is also represented in terms of political supply and electoral behavior: ‘In most countries, it is [the] parties of the populist right who have been able to formulate a highly attractive ideological package for the “losers” of economic transformations and cultural diversity’ (Reference Kriesi, Grande, Lachat, Dolezal and Bornschier2008: 19). As a result, it might be expected that those with lower socioeconomic positions are likely to vote for populist radical right parties. Various studies have suggested that this is indeed the case. To start with, it has been shown that those who vote for populist radical right parties tend to be lower educated (Lubbers et al., Reference Lubbers and Scheepers2002; Arzheimer, Reference Arzheimer2009; Bovens and Wille, Reference Bovens and Wille2010). It has also been shown that radical right voters tend to come from lower social classes (Van der Brug et al., Reference Van der Brug, Fennema and Tillie2000; Arzheimer and Carter, Reference Arzheimer and Carter2006), and are more often unemployed (Van der Brug et al., Reference Van der Brug, Fennema and Tillie2000; Lubbers et al., Reference Lubbers and Scheepers2002).

However, Kriesi et al. (Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006, Reference Kriesi, Grande, Lachat, Dolezal and Bornschier2008) pay almost no attention to populist parties that are not radical right. This is remarkable because it might be expected that the losers of globalization are not only inclined to vote for populist radical right parties, but also for populist parties with other ideologies. After all, populists from left, right, and center all claim that they stand up for ‘ordinary citizens’ who, according to them, are neglected and betrayed by a condescending political and/or economic elite. Those with lower socioeconomic positions will most likely identify with this category of ‘ordinary citizens’ or ‘the common man’, and therefore feel attracted to the message that they are victims of the behavior of the evil elite.

It has indeed been found that the unemployed and those with lower incomes are also likely to support radical left ideologies (Visser et al., Reference Visser, Lubbers, Kraaykamp and Jaspers2014). Moreover, Lubbers and Scheepers (Reference Lubbers, Gijsberts and Scheepers2007) have shown that those with lower class positions are more likely to vote for populist radical left parties than for other parties, and Ramiro (Reference Ramiro2016) has demonstrated that those who identify with the working class are likely to vote for the radical left. Interestingly, however, all of these studies show that the relationship between education and radical left support is more complicated. Often, education exerts a positive effect on radical left voting. Most likely because those with a higher education often subscribe to values emphasized by radical left parties (e.g. solidarity, equality). In line with this finding, Pauwels (Reference Pauwels2014) has shown that when it comes to voting for populist parties in Belgium, Germany, and the Netherlands, education does not exert an overall negative effect.

I therefore expect that someone’s socioeconomic position – but not education – has a negative effect on populist voting. This leads to the following hypothesis:

But why exactly would a low socioeconomic position lead to voting for a populist party? In general, I expect that individuals with lower socioeconomic positions will be more likely to hold certain attitudes, and will, for that reason, be inclined to vote populist. In other words, I expect the relationship between someone’s socioeconomic position and his or her inclination to vote populist to be mediated by various political attitudes.

First, the relationship between someone’s socioeconomic position and populist voting might be mediated by Euroscepticism. Kriesi et al. (Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006, Reference Kriesi, Grande, Lachat, Dolezal and Bornschier2008) argue that the losers of globalization tend to be more Eurosceptic than the winners, because they feel threatened by the opening of borders and the process of further European unification. They might, for instance, fear that immigrants from Eastern European countries ‘steal’ their jobs. Those who are Eurosceptic might in turn feel attracted to the populist message because populism and Euroscepticism are often closely related (Harmsen, Reference Harmsen2010). This might be the case for two reasons: (1) those who despise the political elite in Brussels will tend to agree with the populist message that ‘Brussels’ does not listen to ordinary people in the member states anymore; and (2) Eurosceptics might also vote for populists not because they criticize the European elite, but because they accuse the national political elite of transferring too much power to the EU.

Various studies have indeed found that populist parties positioned at the fringes of the political spectrum tend to express Eurosceptic attitudes (Hooghe et al., Reference Hooghe, Marks and Wilson2002; De Vries and Edwards, Reference De Vries and Edwards2009; Halikiopoulou et al., Reference Halikiopoulou, Nanou and Vasilopoulou2012), and that supporters of such parties (both left- and right-wing) tend to be rather Eurosceptic as well (Lubbers and Scheepers, Reference Lubbers, Gijsberts and Scheepers2007; Werts et al., Reference Werts, Scheepers and Lubbers2013; Visser et al., Reference Visser, Lubbers, Kraaykamp and Jaspers2014; Ramiro, Reference Ramiro2016). I therefore expect that:

Second, it might be expected that the relationship between someone’s socioeconomic position and populist voting is not only mediated by negative attitudes toward the EU, but also by distrust toward politics in general. Various studies have shown that those who are unemployed and those with lower incomes are more likely to be politically dissatisfied (e.g. Bernauer and Vatter, Reference Bergh2012), most probably because they tend to blame the political establishment for their economic position. Research has also shown that those voting for populist radical right parties tend to be distrustful toward politics (Betz, Reference Betz1994; Swyngedouw, Reference Swyngedouw2001; Bélanger and Aarts, Reference Bélanger and Aarts2006). It is likely that the politically discontented vote for populist radical right parties because the negative attitude toward politics of these parties is congruent with their own ideas. Bergh (Reference Bernauer and Vatter2004) has shown that it is not so much distrust toward the entire political system that predicts voting for the populist radical right, but distrust toward the political elites within that system. This makes sense as the populist message is not directed against the system, but against the corrupted elites.

I label those who vote populist out of political distrust ‘protest voters’. According to Bergh, ‘protest voting is the act of voting for a political party as a means of expressing political distrust. The extent to which supporters of a particular party are protest voters is measured by the effect of political distrust on support for that party’ (Reference Bernauer and Vatter2004: 385). Although the literature focuses on voting for the radical right, it could be expected that protest voting also holds for other parties that express a populist message (see Schumacher and Rooduijn, Reference Schumacher and Rooduijn2013). After all, that support for the radical right can be explained by protest voting is not due to the radical right ideology of the party, but to the anti-elitist message in which the party attacks the ‘established political order’. Support for a left-wing populist party or any other party that expresses such a message might just as well be explained by protest voting (Ramiro, Reference Ramiro2016). Some studies have indeed shown that political distrust exerts a positive effect on voting for populists (Doyle, Reference Doyle2011; Hooghe et al., Reference Hooghe, Marien and Pauwels2013; Pauwels, Reference Pauwels2014). This leads to the third hypothesis:

A third political attitude that might positively affect populist voting is someone’s ideas about direct democracy.Footnote ³ Because populists regard politicians and representative institutions with suspicion, they often demand institutional innovations (Webb, Reference Webb2013). ‘They argue that there is a growing “silent majority” which pays taxes, produces welfare and obeys the law but is politically quiescent’ (Pauwels, Reference Pauwels2014: 59). It has therefore been claimed that populists want to restore the primacy of the people and are often in favor of introducing more tools of direct democracy, such as popular initiatives and referendums. The underlying idea is that the political system should resemble as closely as possible ‘a pure democratic regime where the people are given the first and final word’ (Mény and Surel, Reference Mény and Surel2002b: 13). Populism has been associated with direct democracy by various scholars (see Canovan, Reference Canovan1981; Taggart, Reference Taggart2000), and it might be expected that those who vote for a party that is in favor of more direct democracy, share this viewpoint. I therefore expect that:

Case selection, measurements, and method

To assess what the electorates of populist parties have in common, I focus on the voter bases of 15 prototypical populist parties in 11 Western European countries.Footnote ⁴ I included the following parties: the Freiheitliche Partei Österreichs (Freedom Party of Austria) in Austria (see Heinisch, Reference Heinisch2008); the Vlaams Blok/Vlaams Belang (Flemish Block/Flemish Interest) (see Pauwels, Reference Pauwels2011a) and Lijst Dedecker/Libertair, Direct, Democratisch (List Dedecker/Libertarian, Direct, Democratic, LDD) (see Pauwels, Reference Pauwels2010) in Belgium; the Schweizerische Volkspartei (Swiss People’s Party) in Switzerland (see Albertazzi, Reference Albertazzi2008); Die Linke in Germany (see March, Reference March2007); the Dansk Folkeparti (Danish People’s Party) in Denmark (see Rydgren, Reference Rydgren2004); Perussuomalaiset (the Finns Party) in Finland (see Arter, Reference Arter2010); the FN in France (see Surel, Reference Surel2002); Forza Italia (Go Italy, FI) (see Tarchi, Reference Tarchi2008) and M5S (see Fella and Ruzza, Reference Fella and Ruzza2013) in Italy; the Lijst Pim Fortuyn (List Pim Fortuyn) (see Van Kessel, Reference Van Kessel2011), the Socialistische Partij (Socialist Party, SP) (see March, Reference March2011) and the PVV (Vossen, Reference Vossen2011) in the Netherlands; the Fremskrittspartiet (Progress Party) in Norway (see Jupskas, Reference Jupskas2015); and Sverigedemokraterna (Sweden Democrats) in Sweden (see Jungar, Reference Jungar2015).Footnote ⁵ A large majority of these parties can be categorized as populist radical right parties. Only Die Linke and the SP are populist radical left parties.Footnote ⁶ LDD, FI, and M5S are neither radical right nor radical left.Footnote ⁷

Because I am interested in assessing what the voter bases of different kinds of populist parties have in common, the most essential criteria for my case selection are: (1) that the included parties are prototypically populist (meaning that most country experts agree that they can be categorized as populist); and (2) that there is sufficient variation between the selected parties in terms of their ideologies. Both criteria are fulfilled.

A shortcoming of this selection of cases is that a large majority of parties (12 out of 15) is not only populist but also radical. This makes it rather difficult to disentangle populism and radicalism. The reason that so many of the selected parties are radical parties is that most populist parties in Western Europe are also radical parties (Rooduijn and Akkerman, Reference Rooduijn and Akkerman2017). However, that does not mean that all populist parties are radical. Some of the most successful ones are neither radical left nor radical right (e.g. FI and M5S in Italy). Excluding such non-radical populist parties from our case selection would be problematic. Not only because we would then miss out on some highly successful populist parties, but also – and this is probably even more important – because the results could then be driven by these parties’ radicalness instead of their populism.

I have dealt with this large number of selected radical parties in two ways. First, I have assessed the electorates of all parties separately. In other words, I have not run a pooled analysis, but separate regressions for each selected party. The reason is that the found effects in a pooled analysis could be due to some specific features of certain types of populist parties and not to populism as such. For instance, a pooled analysis shows that those who hold radical attitudes are more likely to vote for populists. This effect is probably due to the radical nature of a majority of the parties under analysis. However, those who vote for other types of populists might not be radical at all. Hence, a pooled analysis does not provide us with an answer to the question what unites voter bases of populist parties across cases. Hence, in order to answer the question what unites the voter bases of populist parties in general, we should investigate the voter bases of all the populist parties separately. Second, in order to disentangle populism from radicalism, I have also focused more specifically on the three selected populist parties that are not radical. See the ‘Findings’ section.

The individual-level data used in this study come from the ESS (2002–12). The dependent variable is based on which party a respondent voted for during the last national election. I recoded this variable so that a respondent scores a ‘1’ if he or she voted for a populist party and a ‘0’ if he or she voted for one of the mainstream parties. Parties have been categorized as mainstream when they are liberal, social democratic, Christian democratic, or conservative. If a respondent voted for another party, the variable has been set to missing.Footnote ⁸

The socioeconomic variables have been measured as follows: class is measured with the Erikson–Goldthorpe–Portocarero classification scheme (Erikson et al., Reference Erikson, Goldthorpe and Portocarero1979),Footnote ⁹ income is measured in deciles, and employment status is measured with a dummy variable in which 0=employed and 1=unemployed. Level of education is measured by means of five categories: (1) less than lower secondary education; (2) lower secondary education completed; (3) upper secondary education completed; (4) post-secondary non-tertiary education completed; (5) tertiary education completed.

Euroscepticism is measured on an 11-points scale with the question whether European unification should go further (0) or whether it has already gone too far (10).Footnote ¹⁰ Political distrustFootnote ¹¹ is measured by means of a scale including two variables: trust in parliament and trust in politicians.Footnote ¹² The two items are summed up and recoded so that they range from 0 (complete trust) to 10 (no trust at all). The reliability of this scale is satisfactory (Cronbach’s α=0.84). Someone’s attitude toward direct democracy is measured with a statement about the importance of referendums. On a scale from 0 (not at all important) to 10 (extremely important) respondents could indicate how important it is for a democracy that ‘citizens have the final say on political issues by voting directly in a referendum’.Footnote ¹³

I control for several attitudinal variables, such as an individual’s general left-right radicalism (‘where would you place yourself on [a scale] where 0 means the left and 10 means the right?’; the variable was recoded so that −5=‘left’ and ‘5’=‘right’, and then squared); satisfaction with the economy (‘how satisfied are you with the present state of the economy?’; 0=‘extremely dissatisfied’ and 10=‘extremely satisfied’), satisfaction with the government (‘how satisfied are you with the way [the government] is doing its job?’; 0=‘extremely dissatisfied’ and 10=‘extremely satisfied’), someone’s attitude toward income redistribution (‘the government should take measures to reduce differences in income levels; 1=‘disagree strongly’ and 4=‘agree strongly’), someone’s attitude toward immigrants (1=‘many should be allowed to come and live here’ and 4 ‘none should be allowed to come and live here’) and someone’s political interest (1=‘not at all interested’ and 4=‘very interested’). I also control for age, gender (0=male; 1=female) and religiosity (ranging from ‘not at all religious’ [0] to ‘very religious’ [10]).

I conducted separate logistic regression analyses per populist party (DV=voted for either a populist party [1] or a non-populist party [0]), and included year-fixed effects to take into account that respondents are nested within years. The observations are weighted using the ESS design weights, which correct for the fact that due to the sampling design used, respondents in different countries have different probabilities of being sampled. Observations with missing values were deleted listwise. The independent variables class and education have been modeled as continuous variables to ease interpretation. Including them as dummy variables does not lead to substantially different results (these analyses are available upon request). I corrected for conducting a large number of multiple tests simultaneously by applying Bonferroni’s adjustments.

Findings

Table 1 gives an overview of the direction of the regression coefficients. The table shows whether the effects are positive or negative. Statistically insignificant effects have not been displayed (see the Online Appendix Tables 1–4 for the full models). In a first step, I have only included the sociodemographic variables in my models. The displayed effects are the directions of the regression coefficients in models in which the attitudinal variables have not been included yet. Social class exerts a negative effect in only 5 out of 15 cases. Income is negatively related to populist voting in 3 out of the 15 cases, and in none of the cases is employment status significantly related to populist voting. This all indicates that Hypothesis 1 should be rejected: the voter bases of different kinds of populist parties have not in common that they consist of individuals who are more likely to hold lower socioeconomic positions.

Table 1 Logistic regression analyses explaining voting for a populist party compared to voting for a mainstream party (with Bonferroni adjustments)

FPÖ=Freiheitliche Partei Österreichs (Freedom Party of Austria); VB=Vlaams Blok/Vlaams Belang (Flemish Block/Flemish Interest); SVP=Schweizerische Volkspartei (Swiss People’s Party); DF=Dansk Folkeparti (Danish People’s Party); PS=Perussuomalaiset (the Finns Party); FN=Front National (National Front); LPF=Lijst Pim Fortuyn (List Pim Fortuyn); PVV=Partij voor de Vrijheid (Party for Freedom); FrP=Fremskrittspartiet (Progress Party); SD=Sverigedemokraterna (Sweden Democrats); SP=Socialistische Partij (Socialist Party); LDD=Lijst Dedecker/Libertair, Direct, Democratisch (List Dedecker/Libertarian, Direct, Democratic); FI=Forza Italia (Go Italy); M5S=Movimento Cinque Stelle (Five Star Movement); X=no observations.

In a first step I estimated the effects of the sociodemographic variables. In a second step I included the effects of the attitudinal variables. The effects of the variables ‘Euroscepticism’ and ‘Referendum’ have been estimated separately because of the large number of missing cases.

Gray shaded areas are significant at P<0.05 (the Bonferroni adjustment for multiple comparisons was utilized to most conservatively estimate statistical significance).

Education exerts a negative and statistically significant effect in 6 out of the 10 radical right cases. When it comes to voting for other populist parties, however, education does not exert a statistically significant effect. Age is negatively related to populist voting in 6 out of the 15 cases under analysis. In only three cases gender is negatively related to populist voting. In 5 out of the 15 cases under study religiosity exerts a negative effect on populist voting.

In a second step, I have added the attitudinal variables.Footnote ¹⁴ Political distrust turns out to exert a positive and statistically significant effect on populist voting in 8 out of the 15 cases under analysis. This suggests that we should reject Hypothesis 3. The findings regarding radicalism, satisfaction with the economy, satisfaction with the government and income distribution are mixed. Those who hold negative attitudes toward immigrants are more likely to vote populist only when it comes to populist radical right parties. Political interest is negatively related to populist voting in none out of the 15 cases under analysis.

In a third and fourth step, I have also added the variables ‘Euroscepticism’ and ‘Referendum’ separately. I included these variables separately because of the large number of missing cases regarding these variables (see above). ‘Euroscepticism’ exerts a positive and significant effect in 2 out of the 15 cases. We therefore have to reject Hypothesis 2. ‘Referendum’ exerts a significant positive effect in only 1 out of the 12 cases under study. In all other cases the effects are not statistically significant. This means that we also have to reject Hypothesis 4.

An important drawback of this analysis is that the applied Bonferroni adjustment for multiple testing is highly conservative, and might therefore lead to an underestimation of the number of statistically significant effects. However, subsequent analyses indicate that even if we neglect the correction for multiple testing, we reach the same substantive conclusion that the voter bases of populist parties do not have much in common. Table 2 shows the results of regression analyses without Bonferroni corrections. Not surprisingly, many more effects turn out to be statistically significant. However, even without corrections for multiple testing, none of the variables exerts a statistically significant effect in all cases under analysis. There is only one variable that comes close: political distrust. This variable has a significant effect in 14 out of the 15 cases. However, this pattern is much less clear-cut if we look at the effects in greater detail. Table 3 shows that the effects of political distrust strongly differ across ESS waves. Although the regression coefficients of political distrust always have a positive sign, they fail to reach statistical significance in 28 out of the 61 separate regression analyses (the full analyses can be found in the Online Appendix Tables 5–10).

Table 2 Logistic regression analyses explaining voting for a populist party compared to voting for a mainstream party (without Bonferroni adjustments)

FPÖ=Freiheitliche Partei Österreichs (Freedom Party of Austria); VB=Vlaams Blok/Vlaams Belang (Flemish Block/Flemish Interest); SVP=Schweizerische Volkspartei (Swiss People’s Party); DF=Dansk Folkeparti (Danish People’s Party); PS=Perussuomalaiset (the Finns Party); FN=Front National (National Front); LPF=Lijst Pim Fortuyn (List Pim Fortuyn); PVV=Partij voor de Vrijheid (Party for Freedom); FrP=Fremskrittspartiet (Progress Party); SD=Sverigedemokraterna (Sweden Democrats); SP=Socialistische Partij (Socialist Party); LDD=Lijst Dedecker/Libertair, Direct, Democratisch (List Dedecker/Libertarian, Direct, Democratic); FI=Forza Italia (Go Italy); M5S=Movimento Cinque Stelle (Five Star Movement); X=no observations.

In a first step I estimated the effects of the sociodemographic variables. In a second step I included the effects of the attitudinal variables. The effects of the variables ‘Euroscepticism’ and ‘Referendum’ have been estimated separately because of the large number of missing cases.

Gray shaded areas are significant at P<0.01.

Table 3 The effects of political distrust on voting for a populist party for each wave separately (sociodemographic and attitudinal variables controlled for but not displayed)

FPÖ=Freiheitliche Partei Österreichs (Freedom Party of Austria); VB=Vlaams Blok/Vlaams Belang (Flemish Block/Flemish Interest); SVP=Schweizerische Volkspartei (Swiss People’s Party); DF=Dansk Folkeparti (Danish People’s Party); PS=Perussuomalaiset (the Finns Party); FN=Front National (National Front); LPF=Lijst Pim Fortuyn (List Pim Fortuyn); PVV=Partij voor de Vrijheid (Party for Freedom); FrP=Fremskrittspartiet (Progress Party); SD=Sverigedemokraterna (Sweden Democrats); SP=Socialistische Partij (Socialist Party); LDD=Lijst Dedecker/Libertair, Direct, Democratisch (List Dedecker/Libertarian, Direct, Democratic); FI=Forza Italia (Go Italy); M5S=Movimento Cinque Stelle (Five Star Movement); X=no observations.

Displayed are only the effects of political distrust on populist voting. I controlled for all socioeconomic and attitudinal variables except for ‘Euroscepticism’ and ‘Referendum’.

Gray shaded areas are significant at P<0.01.

Let us now focus on the three non-radical populist parties only. Is there something that unites the voter bases of these particular parties? Table 1 shows that none of the included variables exerts a statistically significant effect in all three cases. And although Table 2 shows that if we neglect the Bonferroni adjustments, political distrust becomes statistically significant in all three of them, Table 3 demonstrates that if we look at the different ESS waves, not even the effect of this variable is statistically significant in all five cases under analysis.

In the main analyses those who vote for a populist party are compared to mainstream parties. As a robustness check I have assessed to what extent a comparison with a different reference category leads to different results. Table 11 in the Online Appendix displays the results of an analysis based on a comparison of those who vote for populist parties with those who vote for non-populist parties – that is, not only mainstream parties, but also other non-populist parties (like, for instance, some single issue, green, or regional parties). This does not lead to substantively different results.