Gender, populist attitudes, and voting: explaining the gender gap in voting for populist radical right and populist radical left parties

Abstract

Empirical studies have demonstrated that compared to almost all other parties, populist radical right (PRR) parties draw more votes from men than from women. However, the two dominant explanations that are generally advanced to explain this disparity – gender differences regarding socio-economic position and lower perceptions regarding the threat of immigrants – cannot fully explain the difference. The article contends that it might actually be gender differences regarding the conceptualisation of society and politics – populist attitudes – that explain the gender gap. Thus, the gap may be due, in part, to differences in socialisation. The article analyses EES 2014 data on voting for the populist radical right and the populist radical left in nine European countries. Across countries, the gender gap in voting for the PRR is indeed partly explained by populist attitudes. For populist radical left parties, the results are less clear, suggesting that populism has different meanings to voters on the left and on the right.

Keywords:

• Populism
• gender gap
• voting behaviour
• populist radical right
• left-wing populism
• Europe

The electoral success of populist parties does not seem to subside. In fact, it appears that the opposite is true: in addition to the continuing success of populist radical right parties (PRR), in recent years, there has also been a surge in populist (radical) left parties. Recent examples include Podemos in Spain, SYRIZA in Greece, Movimento 5 Stelle in Italy,¹ the Dutch Socialistische Partij, and Die Linke in Germany. The continued, and perhaps even growing, success of populist parties presents us with an interesting question: why do voters continue to support populist parties?

Research in the last 10 to 15 years has provided answers to this question. Reasons for the support for the populist radical right are found in opposition to immigration (e.g. Immerzeel 2015; Lubbers et al. 2002; Norris 2005; Spierings and Zaslove 2015b), lower levels of trust in political institutions (Norris 2005) and being less well educated, younger, and male (Harteveld et al. 2015; Immerzeel 2015; Spierings and Zaslove 2015b). In contrast, supporters of the populist radical left are marginally more supportive of economic equality than supporters of other parties (Visser et al. 2014). The socio-economic profile of (populist) radical left parties is, however, less clear. Visser et al. (2014) show that those with lower incomes and those who are unemployed are more likely to possess a radical left ideology. Ramiro (2016) finds that those who identify with the working class are more likely to support radical left parties. Education is, however, more complicated: both less well and more highly educated voters are attracted to the radical left (Ramiro 2016). Nevertheless, there is a quality that the populist radical right and populist left parties have in common: their populist attitudes (Akkerman and Zaslove 2014; Akkerman et al. 2014).

Despite our growing knowledge of why voters support the PRR, a recurring puzzle is the gendered nature of their support. In other words, the question is: why do PRR parties receive more male than female support? In seeking an answer to this question, we start with the abundant literature on the so-called PRR gender gap (see Spierings and Zaslove 2015b; Spierings et al. 2015). This literature suggests two main explanations for the gender gap: first, the difference is due to socio-economic position, and second, it is due to programmatic attitudes, most notably anti-immigration and law and order attitudes (Harteveld et al. 2015; Immerzeel et al. 2015; Spierings and Zaslove 2015b). However, most notably, neither of these explanations helps us fully understand why more men than women vote for populist (radical right) parties. The socio-economic explanation argues that women are more likely to be employed in the public sector and less likely to be employed in labour-intensive jobs and are therefore less likely to be threatened by deindustrialisation. These factors, in turn, make women less likely to be attracted to PRR parties (see Harteveld et al. 2015; Immerzeel et al. 2015; see also Ivarsflaten 2005, 2008; Spierings and Zaslove 2015b). However, these factors only partly explain the gap. A second possible explanation is that women feel less threatened by immigration – they are not in direct competition with migrants – and thus they may hold weaker anti-immigrant attitudes (Immerzeel et al. 2015; Spierings and Zaslove 2015b). Even if it were true that less competition exists between migrants and women, this does not translate into lower levels of anti-immigrant attitudes among women. It has been shown repeatedly that there is scarcely any difference between men and women regarding anti-immigration attitudes (Harteveld et al. 2015; Norris 2005; Spierings and Zaslove 2015b). In sum, these approaches cannot fully account for the gender gap.

Therefore, in this study, we investigate a new, third explanation: whether the populist nature of the parties in question explains why more men than women support PRR parties. In order to investigate this line of research, it is necessary to focus not only on the PRR but also on the populist left. This approach is in line with the party literature that considers populism to be a thin-centred ideology (Mudde 2007). Thin-centred ideologies are not full ideologies and must attach themselves to other ideologies ‒ for example, to the anti-immigration ideology of the radical right. It is our contention that the thin-centred ideology of populism appeals to voters with certain political attitudes, which are on average more prevalent among men than women due (ceteris paribus) to political socialisation processes. As noted, given that we focus on a broad definition of populism, it is important that we not only focus on the populist radical right but also on the populist radical left (PRL). As a result, if our explanation holds, it might also shed light on the recent increase in voting for populist radical left parties.

The article is organised as follows. We begin by briefly discussing the concept of populism, relating it to voter attitudes. We link this discussion to the gendered processes of socialisation, which leads to our contention regarding the gender gap in voting for populist parties. Next, we examine the extent to which the gender gap exists among populist radical right voters and focus on whether the gap found is partly due to a gender difference regarding populist attitudes. Because this analysis demonstrates that populism can indeed partly explain the gender gap, we then move to the populist radical left to further test whether a populist ideology and populist attitudes are gendered in the same way on the left as on the right. The analyses are based on the latest voter data from the European Election Survey 2014 (Schmitt et al. 2015) for nine European countries: Austria, Denmark, Finland, France, Germany, Greece, Italy, the Netherlands, and Spain.

Theoretical background

It is common to claim that there is a gender gap with regard to voting for populist radical right parties, and there is certainly ample evidence that this is the case. Since the early rise of populist radical right parties, research has identified a gender gap.² Betz (1994: 143) notes that in the late 1980s and the early 1990s, only 43% of the supporters of the French National Front and 40% of the supporters of the Austrian Freedom Party were women. Kitschelt and McGann (1997: 76) confirm these findings, noting that ‘men are overrepresented in the rightist electorate with a 60–40 to 70–30 margin’. They go on to say, ‘This pattern underlines the importance of the authoritarian antifeminist thrust in the contemporary extreme right’ (Kitschelt and McGann 1997: 76). Norris (2005: 145) again confirms these findings. She notes that men outnumbered women in Austria and Denmark by a ratio (compared to the national average) of 1.3 to 0.7 and in France by 1.3 to 0.8 in their support for the radical right. At the same time, however, there is also evidence that the gender gap is perhaps not as large as is often claimed. Spierings and Zaslove (2015b) note that we should be careful not to overstate the degree to which the populist radical right is gendered, especially in comparison to centre-right parties. Nevertheless, the authors conclude that there is no doubt that more men than women vote for populist radical right parties. As discussed above, significant attention has been paid to explaining this gender gap, and socio-economic explanations have proven fruitful, but the question of why women have lower likelihoods of voting for PRR parties than men remains partly unanswered.

In this section, we have two aims. First, we briefly address the concept of populism from a party ideology perspective as well as from an individual-level voter perspective. Second, we theorise how populist attitudes at the individual level might be gendered, which might help to explain the gender gap in populist voting. In doing so, we show that this approach offers a new, additional explanation to our understanding of the gender gap while at the same time offering an additional explanation for how structural factors and socio-economic position might be linked to it ‒ i.e. populist attitudes may function as a mediator.

Populism, a thin-centred ideology – parties and attitudes

The literature on populism makes an important distinction regarding the nature of populism: it argues that populism is a thin-centred ideology (Mudde 2004; see also Akkerman et al. 2014). The thin-centred nature of populism implies that it is necessary to distinguish between populism and what is referred to as the attaching ideology. Populism is a thin-centred ideology because it is not able to stand on its own. The argument goes as follows: a full ideology is an ideology that has a position (policy position or programme) that covers the full range of policy domains (Akkerman et al. 2014; Mudde 2004; Stanley 2008). Thin-centred ideologies do not have these qualities: they are ideologies insofar as it is possible to identify the core characteristics of populism (a minimal definition); however, they must rely on other ideologies for the full content of their ideology, which accounts for the left- and the right-wing nature of populist parties (Akkerman et al. 2014; Mudde 2004, 2007; Mudde and Kaltwasser 2013).

This analysis begs the question: what is populism? The thin-centred ideological approach points to several core characteristics (Akkerman et al. 2014; Mudde 2004; Mudde and Kaltwasser 2013). First, populism is people-centred. Populists argue that the people are the basis of the democratic decision-making process: democratic representation begins with the people. Second, there is an emphasis on the people versus the elites. It is not enough for populists to be simply people-centred; they must also juxtapose the people with the elites, who are usually identified as the corrupt elites. Third, this distinction must be viewed as antagonistic, or Manichean, such that the good people are juxtaposed with the bad (corrupt) elites. The fourth element is the notion of the general will. Populist notions of representation focus on the idea that the party or movement in question represents the general will of the people (Akkerman et al. 2014; Mudde 2004; Mudde and Kaltwasser 2013; Rooduijn 2013; Zaslove 2008).

This notion of populism is what unifies populist radical left and populist radical right parties both on the supply and demand sides (Akkerman et al. 2014; Akkerman and Zaslove 2014; Mudde and Kaltwasser 2013). Populist parties of both the left and the right focus on the distinction between the elite and the people. It is certainly quite possible that their conception of the people may be different: radical left populists may be less exclusionary – in part due to a focus on economic issues and not culture – than the populist radical right (Mudde and Kaltwasser 2013).

The populist nature of populist parties is often identified with the supply side ‒ i.e. with the programmes of the parties and the discourses of the leaders (see Hawkins 2009; Rooduijn 2013). However, recent research has also identified populist attitudes among voters (Akkerman et al. 2014; Hawkins et al. 2012; Spruyt et al. 2016). The distinction between elites and the people and the notion that sovereignty should lie not with the elites but with the people are also found among the supporters of populist parties. These studies have found that populist voters can be distinguished from non-populist voters ‒ i.e. those who have more pluralist attitudes (favouring compromise and listening to other opinions) ‒ and from voters who have more elitist attitudes ‒ i.e. arguing that politicians should lead and not follow the people, or that successful individuals from business or independent experts should govern (Akkerman et al. 2014; Hawkins et al. 2012).

Gendered political socialisation

Given that populism is a thin-centred ideology, it is possible that the gendered nature of support for populist radical right and populist radical left parties may be due to populism and that it may have less to do with the attaching ideology ‒ i.e. the radical right or the radical left nature of the parties in question. In other words, it may be the populist attitudes of the supporters that lie at the heart of why women do or do not support populist parties. In exploring this argument, we first discuss how these attitudes might be gendered, and we then move on to how these gendered attitudes might explain the gender gap in voting for PRR and PRL parties.

We start by recalling the insight that women and men tend to be socialised differently (Sapiro 1983; Trevor 1999). Parents and society reward different behaviours by girls and boys, and different examples are set about ‘correct behaviour’. These differences in socialisation evidently include political behaviour; part of this process can for instance be found in the way media depict female and male politicians differently (e.g. Gidengil and Everitt 2003; Ross and Sreberny 2000). The differences in socialisation include elements that relate to populism as an attitude. For instance, women are depicted as less aggressive, more consensus-seeking, and friendlier (e.g. Ross and Sreberny 2000; Sawer 2002; Van Zoonen 2001), and these differences are not without consequences, as prior studies suggest that gender differences in average political attitudes explain certain political behaviours.

For instance, Golebiowska (1999) shows that women are more politically tolerant than men. Verba et al. (1997) attempt to explain why men are more engaged in politics than women and note that considerations that are normally associated with political engagement (such as education or income) do not explain the gender gap. Rather, the authors claim that, on average, it appears that men and woman may have different ‘tastes for politics’. Fish (2002) makes the differences even more tangible, arguing that men are more prone to conflictual politics and women prefer consensus-seeking.

Populist attitudes and the gender gap

Mudde (2007) makes a similar argument regarding populist radical right parties, taking a first step in translating this general theorisation to the question at hand here. He refers to the argument as the ‘low efficacy theory’ (Mudde 2007: 116), and argues that the gendered nature of support for the PRR may have less to do with the ideas of the parties in question and more to do with the image of the parties as radical or even extreme parties. Because women often have lower levels of political efficacy and lower levels of political interest, they are more likely to vote for ‘established parties’ (Mudde 2007: 115–16). In other words, women may disagree less with the message and more with the image of the parties in question. Mudde thus identifies the genderedness of political socialisation but does not fully develop this argument with regard to populist attitudes, given his focus on radical parties. Our focus is on populist parties and not on the radical or even the extreme nature of the parties per se.

In sum, given the above insights, there is reason to assume that there may also be something about the thin-centred ideology of populism (again, its qualities), such as its people-centredness or perhaps its antagonistic (Manichean) qualities, that is less attractive for women voters. Consequently, women might vote less for populist parties than men even though they may agree with the programmatic views of the parties.

For the PRR, the gender gap in voting has been robustly demonstrated. Our explanation, however, is initially not conditional on the populist radical right: Populism could well be expected to unify both the left and the right on this front. The foregoing raises the question of whether there is a gender gap among the populist radical left. One of the first large voter studies of the attitudes of those with a left-wing ideology (Visser et al. 2014) does not show such a gap. In fact, it might also be counterintuitive to assume a gender gap on the left, given that it has been shown that women vote more for the left for socio-economic and ideological (attitude) reasons. Thus, given that the radical left is more inclined to support redistribution and income equality (e.g. Bowyer and Vail 2011; Inglehart and Norris 2003; March and Rommerskirchen 2015), we may even expect more women to support radical left parties. In other words, if our notion of gendered populist attitudes holds, the gender dynamics in voting for the populist radical left would counter the normal trend for women to be more supportive of left-wing parties. To test the general validity of our argument, both the populist radical right and left will be studied.

Returning to the core of our argument, we focus on the idea that populist attitudes translate into ‘populist behaviour’: voting for populist parties. As argued above, these parties, either on the left or the right, are not only populist but also connect populism with other ideologies. In other words, there are many reasons why voters might support populist parties. However, only by including attitudes that tap into these different components of the ideologies can we examine which factors are most important. From this perspective, and focusing on the PRR, the distinction between populist attitudes and anti-immigration attitudes might also shed new light on the linkage between the socio-economic positions of men and women and voting for the PRR. The structural socio-economic argument claims that women are less likely to be employed in labour-intensive jobs, such as factory jobs and jobs that are challenged by immigrants ‒ i.e. workers who are willing to receive less pay (see Harteveld et al. 2015; Immerzeel et al. 2015; Ivarsflaten 2005, 2008). Here it might not just be the economic structure and the perceived threat from labour immigration that lead to the higher likelihood among men to vote for the PRR but also the idea that politicians have sold them – the low-skilled workers – out, which feeds into a populist mind-set: perceiving society to be divided between the good people and the misrepresenting political elite.

In sum, we expect that populist attitudes partly explain the gender gap (ceteris paribus) in voting for populist parties. However, the effect might be partly mediated by the relationship among gender, socio-economic position, and voting populist, in particular on the right side of the political spectrum.

Methods and data

Survey and case selection

We use the European Elections Surveys (EES) 2014 (Schmitt et al. 2015) as our data source. These data are the only cross-national data that include voting behaviour as well as items related to our core explanatory concepts, including indicators that are needed to construct indexes on populism and attitudes regarding anti-immigration, law and order, and economic policies. The sample size and elections surveyed ensure that we have a sufficient number of respondents who indicate having voted for the populist radical right or populist radical left. We focus on West and South European countries and, of those, we include all countries present in the EES data that have a populist party, based on the criterion that at least 5% of the respondents who voted indicated that they voted for that party during the last national elections.³ This approach makes it possible to include multiple countries while also having group sizes that potentially allow for disaggregating the analyses by country, at least for some countries. Thus, for instance, Belgium and Sweden have not been included because only approximately 3% to 4% of the (voting) respondents indicated Flemish Interest and the Sweden Democrats as their party of choice. If a country had both a populist radical right and a populist radical left party, we included the country in both analyses if for at least one of these choices the 5% threshold was met. These criteria resulted in the inclusion of eight countries: seven with a PRR party and five with a PRL party. In three countries, both were present (Table 1). The classification of parties as populist follows Van Kessel (2015), with three exceptions. Based on other studies, we also include the Dutch Socialist Party (Akkerman et al. 2014; Lucardie and Voerman 2012; Otjes and Louwerse 2015), Podemos (Gómez-Reino and Iván Llamazares 2015) and Golden Dawn (Ellinas 2013).⁴

Table 1. Parties’ classification.

	Populist radical right	Populist radical left
Austria	Austrian Freedom Party	‒
	Alliance for the Future of Austria
Denmark	Danish People’s Party	‒
Finland	True Finns	‒
France	National Front	‒
Germany	‒	The Left
Greece	Popular Orthodox Rally	Coalition of the Radical Left
	Golden Dawn
Italy	Northern League	Five Star Movement
Netherlands	Party for Freedom	Socialist Party
Spain	‒	We Can^1

Notes: All other parties present in the data were considered not populist. Party classification is based on Van Kessel (2015) with the exception of three parties. These are included according to the following studies: the Dutch Socialist Party (Lucardie and Voerman 2012; Akkerman et al. 2014; Otjes and Louwerse 2015), Podemos (Gómez-Reino and Iván Llamazares 2015), and Golden Dawn (Ellinas 2013). Golden Dawn is a borderline case, it is clearly right-wing and anti-migration, but not per se populist. We have rerun several analyses removing Golden Dawn voters from the sample, which gave similar results. See note 5 regarding the classification of the Five Star Movement.

¹ Podemos was only included in two of the sampled dependent variables: the last European elections and ‘what would you vote today’, and not for the last national elections because their rise took place afterwards.

Dependent variable and models

The EES includes three different vote choice variables that allow us to construct a robust indicator of ‘voting populist’ (which is at the core of this study) and to test the mediating impact of populist attitudes on the relationship between gender and voting populist. Thus, by combining the different party-choice variables, we can create an interval variable and conduct an ordinary least-squares regression (OLS). This approach is crucial for the comparison of coefficients across models. Regular vote models often use logistic regressions but, as Mood (2010) clearly argues – unfortunately, only recently has this insight become more widespread – the transformation of the dependent variable in logistic regressions has the effect that the coefficients cannot simply be compared across models. In other words, testing mediating effects using logistic regression models is problematic, as it requires the inclusion of an additional variable. This approach can lead to a change in the coefficient of the other variables for technical rather than substantive reasons. In other words, a change in the coefficient cannot be ascribed with certainty to the mediating effect. We therefore revert to OLS models. The downside of using OLS models, however, is that the stand-alone coefficients must be interpreted carefully given the dependent variable’s scale; for comparing the coefficients across models, this issue is not a problem.

Still, one could make the argument that logistic models do provide some insight into changing coefficients: if large decreases are found, it is highly unlikely that they are a result of including only one additional variable, as is the case here. Acknowledging that there is a trade-off between the different particularities of the models, we perform and report the results of multiple robustness tests, including running the logistic models for the different 0–1 vote variables. Another possible alternative is the use of propensity-to-vote scores, but this approach is not satisfactory for two reasons. First, propensity scores for populist radical right parties are highly skewed towards the extremes, which does not improve on OLS regression models.⁵ Second, there is a gendered problem regarding concept measurement and discrepancy: the gender gap in propensity to vote shows a different dynamic than the gender gap in actual party choice. In other words, the difference between propensity scores and party choice is gendered, which makes it an unreliable proxy for understanding voting behaviour (see Spierings and Zaslove 2015a, 2015b). In one of the robustness tests, however, we did include the propensity score as an additional item in our dependent variable. In summary, we report the results of multiple robustness tests to determine whether they confirm or disconfirm the core results.

The multi-item indicator that we use for our main OLS party choice models is based on the respondents’ vote choice in the last general election; their vote choice if the general elections were held today; and their vote choice in the last elections for the European Parliament. Combining these, we can include those who never intended to vote for a PRR or PRL party, those who voted for a PRR or PRL party, those who have only voted once or twice, and those who do so consistently. Creating a multi-item indicator in this manner also makes the results less sensitive to the particularities of each of these variables or elections, thus increasing the reliability of measuring who votes for populists and who does not. That the three vote items can be combined and tap into the same underlying political behaviour is confirmed empirically: the three dichotomous indicators for voting PRR correlate between 0.70 and 0.85; a factor analysis shows that the three items load on a single factor;⁶ and Cronbach’s alpha is very high (0.92).

Based on the conceptual literature (see above), we classify parties into three categories: populist radical left parties, populist radical right parties, and not populist parties (NP) (see Table 1). Respondents who indicated that they did not vote or did not want to answer the question were not included in the analysis. The survey answers included all major and minor parties; those who said ‘other party’ nevertheless were considered to have voted for a non-populist party.⁷ In the present analysis, we study voting for the populist radical right and voting for the populist radical left separately. The dependent variables indicate the number of times that respondents indicate that they have voted or would vote for the PRR/PRL versus NP. This scale runs from 0 to 3 for the PRR. For the PRL, it runs from 0 to 2 because in Spain only two dependent variables were available: Podemos did not run in the first national election. We coded those voting for the PRL in all elections as ‘2’, those who never did so as ‘0’, and those who did so at least once but not always as ‘1’. To control for different levels of support for populist parties between countries, fixed-effects models are used.

Independent variables

To measure the so-called ‘gender gap’ or differences between the average among women and the average among men, we use self-reported sex. In the regression analysis, men are the reference category. A negative coefficient gives the extent to which women have lower chances of voting for a populist party. If populist attitudes partly explain the gender gap, the negative coefficient should decrease after the inclusion of the populist attitudes variable. The latter is based on six items, which were selected based on the theoretical concept of populism (Akkerman et al. 2014; Mudde 2004) and a factor analysis (see Appendix A1). In the factor analysis, we included a multitude of related items that are not per se related to populism, including political attitudes focusing on internal political efficacy, political interest, closeness to a political party, political trust, and the idea that politics and people constitute opposites. The latter two tap into the theoretical concept of a thin-centred populist ideology (Akkerman et al. 2014; Mudde 2004; Spierings et al. 2015), and indeed the six items⁸ that relate to ‘trust’ and ‘political opposites’ all load on one and the same factor (factor loadings > 0.670). The fact that the other items do not load on the same factor indicates that these six items measure populist attitudes – attitudes that focus on the tension between politicians and political institutions on the one hand and ordinary citizens on the other hand. It also indicates that this concept is distinct from other political attitudes. As such, our empirical factor analysis confirms that there is a cluster of items that load on one factor, and, conceptually, these are the items that relate to the core notion of populism as a Manichean ideology distinguishing the good people from the evil political elite.⁹ Each of the six items has four answer categories, and these are recoded so that a higher score indicates higher populist attitudes. These scores are summed and rescaled to a 0–10 scale.

In the literature, other variables have been used to explain the gender gap in PRR voting, and these variables are also included. The most important ideological variables are anti-immigration, law and order, and anti-EU attitudes (Immerzeel 2015; Spierings and Zaslove 2015b). Given that this study is the first, to our knowledge, that examines PRL parties and gender, we also include economic attitudes – the classic left–right divide. Anti-immigration attitudes are measured on a scale from 0 (fully opposed to a restrictive policy) to 10 (fully in favour of a restrictive policy). Taking a strong law and order stance is measured on a similar scale: 0 (fully support privacy rights even if they hinder combating crime) to 10 (fully in favour of restricting privacy rights in order to combat crime). The EES included multiple items measuring attitudes towards the EU. The three that were theoretically closest to general attitudes towards EU integration and the EU loaded on a single factor.¹⁰ Each of the three items was rescaled to a variable from 0 to 10, with a higher score indicating more anti-EU attitudes, and the average score of these three was used in the analysis. The last ideological attitude measured was an economic left–right scale. Two specific items measure the economic dimension well and have better response rates than a general left–right item: ‘Are you in favour of or opposed to the redistribution of wealth from the rich to the poor in your country?’ and ‘Are you in favour of or opposed to state intervention in the economy?’ These two items load on a single factor, and we calculated the average score from 0 to 10, where 0 is economically extreme left-wing and 10 extremely right-wing.

In addition, we include occupational status as an explanation for voting for the PRR. This variable is also an indicator that is generally used to assess whether the gender gap in voting for the PRR can be explained by structural socio-economic differences between men and women. Including this variable also allows us to examine whether populist attitudes are part of the mechanism that links socio-economic position to the gender gap. If the effect found for employment decreases after the inclusion of populist attitudes, then it does (see Harteveld et al. 2015; Immerzeel 2015; Ivarsflaten 2008). We distinguish among owners of small businesses, manual labourers, and the unemployed, given that research demonstrates that these groups are prone to vote for the radical right (e.g. Ivarsflaten 2008). We also include the non-employed separately; this category includes pensioners, those who are unfit for work, and stay-at-home parents. The reference category is ‘other employed’, which refers to those employed in white-collar jobs. In addition, we control for generational and age effects by including age. Those aged 18 and over were included; the age of 18 is used as the origin.

The purpose is to make valid comparisons across the models. As a result, all models are estimated on the sample of respondents with valid scores on all variables, including the most complete model. In this manner, we can ensure that changes between the models are not due to differences in sample size and the inclusion of different respondents.

Results

Are women less populist?

Before addressing the gender gap in voting for the PRR and PRL, we establish whether and to what extent a face-value gender gap in fact exists. Figure 1 provides the (weighted) percentage of votes for the party families for women. Comparing the PRR parties to the non-populist parties demonstrates a clear difference, which is the case for each of the vote choice variables: populist radical right parties draw relatively more votes from men than do non-populist parties. The difference is between 7 and 9 percentage points, and if we disaggregate the data by country, this pattern is found in 16 of the 21 cases. In the three Italian cases, the opposite is found; however, the subsample of Italian PRR voters is the smallest of all countries, making this result less reliable. The two other exceptions are found for one of the three Dutch party choice variables and one of the three Austrian variables. In both cases, no clear difference between men and women (either positive or negative) was found. Overall, the picture is quite clear: on average, PRR parties draw more votes from men than other parties. At the same time, as has been argued previously (Spierings and Zaslove 2015b), we should not exaggerate this gap, as all bars in Figure 1 are between 43% and 51%.

Figure 1. Gender differences in PRR and PRL voting behaviour.

Are women also less attracted to populist radical left parties? Figure 1 shows that the distribution of women and men among PRL parties is more skewed towards men than for the non-populist parties, with a gap of 2 to 4 percentage points. This difference is thus considerably smaller than is found for the PRR parties; the PRL sits consistently between the PRR and non-populist parties. Again, disaggregating this result by country demonstrates that the pattern is repeated for 10 out of 14 cases.¹¹

As noted, these results are not controlled for explanations for voting for the PRR and PRL. For example, women are on average somewhat less anti-immigrant in our sample (see Table 2; see Harteveld et al. 2015), which might imply that the gender gap for voting for the PRR is somewhat larger than indicated. Similarly, since the 1990s, women have become relatively more left-wing economically (2003) – as is also the case in our sample (see Table 2) – thus, the net gap of voting for the PRL between women and men might be larger than suggested by the figures presented above. Therefore, we need to control for these factors in order to establish the net gender gap in voting for the PRR (and PRL) and subsequently to analyse to what extent populist attitudes might explain the residual gap.

Table 2. Descriptive statistics on explanatory variables for men and women separately.

	Mean men	s.d. men	Mean women	s.d. women
Populist attitudes (0–10)	4.21	2.29	4.13	2.20
Strongly populist (proportion 8–10)	0.08	0.28	0.06	0.24
Economic left‒right position (0–10)	4.43	2.14	4.34	1.96
Anti-migration attitudes (0–10)	5.65	2.92	5.52	2.86
Law and order attitudes (0–10)	5.00	2.84	5.38	2.72
Anti-EU attitudes (0–10)	4.67	2.36	4.75	2.16
Occupational status
Not employed: unemployed	0.05	0.22	0.06	0.23
Not employed: other	0.31	0.46	0.42	0.49
Employed: manual labour	0.09	0.29	0.02	0.16
Self-employed: small business	0.04	0.19	0.03	0.18
Employed: other	0.51	0.50	0.47	0.50
Age (18 = 0)	32.7	15.9	33.7	16.2

Note: Descriptives are based on cases that had valid scores on all variables, including the votes variables; n = 3360 (1683 women and 1519 men); weighted (w1).

Source: EES 2014.

In this respect, it is noteworthy that Table 2 shows that the average populist attitudes among men and women differ only slightly.¹² A closer look at the distribution shows that the difference between women and men is also found for higher populist attitudes (scores of 8 to 10), but the difference remains small. The within-group variation among women and men is much greater than the between-group variation. However, the degree of populism can also be expected to be related to the other explanatory variables. Only by assessing the variables together and including them simultaneously in a multiple regression model can we account for this difference, and only in this manner can we assess whether the degree of populism might explain the gender gap in voting for the PRR or PRL parties. An OLS model with populist attitudes as the dependent variable and the other variables from Table 2 as independent variables (and country dummies) demonstrates that women are slightly less populist on average (ceteris paribus): 0.076 less populist. This gap is relatively small, but with a certainty of 95% we can expect that in the population at large, women are indeed less populist.

The gender gap in voting for the populist radical right

In Table 3, we present the multivariate analysis. Before turning to the mediating influence of populist attitudes, we need to establish the overall validity of the model and the extent to which a gender gap exists, as these two elements are fundamental to the further analysis. Because the control variables act in accordance with the literature, it is safe to assume that our data and model are generally valid.

Table 3. Regression models explaining voting for a populist party.

	Populist radical right (0–3)			Populist radical left (0–2)
	1	2	3	4	5	6
Sex (1 = woman)	−0.153***	−0.124***	−0.113**	−0.016	−0.016	−0.010
	(0.04)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)
Populism			0.08***			0.09***
			(0.01)			(0.01)
Economic left‒right		0.01	0.01		−0.06***	−0.05***
		(0.01)	(0.01)		(0.01)	(0.01)
Anti-migration		0.05***	0.05***		−0.00	0.00
		(0.01)	(0.01)		(0.02)	(0.01)
Law and order		0.01	0.01		−0.02***	−0.02**
		(0.01)	(0.01)		(0.01)	(0.01)
Anti EU integration		0.15***	0.11***		0.06***	0.01
		(0.01)	(0.01)		(0.01)	(0.01)
Occupation (ref = employed other)
Not employed: unemployed		−0.01	−0.04		0.15*	0.11
		(0.08)	(0.08)		(0.07)	(0.06)
Not employed: other		0.05	0.04		0.06	0.04
		(0.04)	(0.04)		(0.04)	(0.04)
Employed: manual labour		0.47***	0.45***		0.15*	0.13
		(0.07)	(0.07)		(0.08)	(0.07)
Self-employed: small business		0.29**	0.28**		0.13	0.14
		(0.10)	(0.10)		(0.08)	(0.08)
Age		−0.004**	−0.003**		−0.004***	−0.003**
		(0.00)	(0.00)		(0.00)	(0.00)
Intercept	0.60***	−0.55***	−0.84***	0.30***	0.55***	0.30***
	(0.06)	(0.09)	(0.09)	(0.04)	(0.08)	(0.08)
Adjusted R square	0.04	0.24	0.25	0.06	0.12	0.17
N	2443			1908

Note: All models are fixed effect models. The countries include are: Austria, Denmark, Finland, France, Greece, Italy, and the Netherlands (PRR), and Germany, Greece, Italy, the Netherlands, and Spain (PRL). The intercept thus only holds for one country, when respondents score 0 on all variables; it says very little and has not been included either.

^* p < 0.05

^** p < 0.01

^*** p < 0.001.

Source: EES voter survey 2014.

Regarding the gender gap in voting for the PRR compared to voting for non-populist parties, Model 1 shows that a statistically significant (p < 0.001) gender gap exists in voting for the PRR, but it should not be overestimated, as it is 0.15 on a three-point scale. Moreover, Model 2 shows that the regular explanatory variables explain 19% of this gap, but it remains substantively and statistically significant. In other words, after taking into account that, for instance, women’s and men’s averages differ with regard to anti-immigration attitudes and socio-economic positions, women are (still) less likely to vote for the PRR compared to men. This result justifies a further analysis of the gender gap by adding our new explanation of populist attitudes to the model (Model 3) to test whether it can explain part of the remaining gap.

In Model 3, we do indeed find a positive effect of holding more populist attitudes on voting for the PRR. Together with the findings above, there is a small gender gap in holding populist attitudes (ceteris paribus). This result suggests a link between sex and populist attitudes and voting behaviour, and indeed, we find that when comparing the coefficient of the dummy variable for respondents’ sex in Models 2 and 3, it decreases. This decrease is a modest 9%. Considering, however, that we only included one additional variable and that the modest explanatory power we find here is equal to or slightly stronger than the effect found in other studies attempting to explain the gender gap in voting for the PRR (e.g. Harteveld et al. 2015; Immerzeel et al. 2015; Spierings and Zaslove 2015b), the result should not be ignored. Examining the control variables, we also see that most of the other variables – anti-EU attitudes being the exception – are not affected considerably by including populist attitudes, which is another indication that we are not tapping into another concept or presenting a spurious relationship. In light of these findings, it should also be noted that the variables measuring socio-economic position, such as unemployment or holding a manual labour position, are only very modestly influenced by the inclusion of populist attitudes. This result suggests that populist attitudes do not mediate the linkage between socio-economic position and voting for the PRR and thus that the structural socio-economic gender gap explanation (e.g. Harteveld et al. 2015) is separate from the one we present here.

Robustness

To validate whether the mediating role of populist attitudes is robust, we have run several variations on the models presented above, adjusting both the populist attitudes and party choice variables. Appendix B2 provides a full overview of the extent to which the coefficient for the gender dummy decreases.

Varying the indicators used in the populist attitudes variable or changing it to a dummy (strongly populist vs. not strongly populist) does show similar results (a 7‒9% decrease). Next, we estimated party-choice models per election type using logistic regression models and a multinomial model using the dependent variable: voted for the PRR at least once (versus never having voted for a populist party or having voted for the PRL at least once). As noted, the results of these models should be interpreted carefully, as the changes in coefficients are modest, and the fact that the models are estimated using logistic regression impacts on the changes in the coefficients for technical reasons as well. Nevertheless, all these models still show a decrease in the gender gap indicator after the inclusion of populist attitudes, although the decrease is somewhat smaller than in the main OLS models. This result is likely to be (partly) caused by the choice of model. Finally, we varied the indicators included in the party-choice variable by only examining national elections (not the European elections) and also by adding propensity to vote for the PRR as an additional item. Both of these models show effects similar to the models presented in Table 2.

In sum, for the overall pooled model, the general direction of our findings is very robust with regard to changes in the core explanatory variables, the dependent variables and the estimation technique. The substantive size of the gender gap explained by populist attitudes varies across the models. Based on the results from the different models, the decrease seems to be between 5% and 10%, which, as noted, is a modest increase, but it is similar to what has been found for other explanations of the gender gap in the literature.

Country-specific models

For four countries, the subsample sizes (>40 PRR voters) also allow us to repeat the analysis on the disaggregated sample by country. This approach allows us to examine whether the results might be context- and party-specific. The countries are Austria, Denmark, Finland, and Greece. The core coefficients of these analyses are provided in Table 4. These results confirm the overall picture: women vote for the PRR slightly less often, and this disparity is partly explained by a difference in populist attitudes: in all cases, the gender gap coefficient decreases after including the populism variable as a mediating factor. It should be noted that both the gender gap and the mediating role of populist attitudes vary considerably across countries. Denmark and Greece stand out as countries where populism explains a considerable degree of the gender gap.

Table 4. The impact of populism on the gender gap per country.

	Populist radical right			Populist radical left
	Before entering populism	After inclusion populism	Difference	Before entering populism	After inclusion populism	Difference
Austria	−0.180^**	−0.170	−6%	n.a.	n.a.
Denmark	−0.029	−0.022	−24%	n.a.	n.a.
Finland	−0.216*	−0.205	−5%	n.a.	n.a.
France	n.a.	n.a.		n.a.	n.a.
Germany	n.a.	n.a.		−0.131*	−0.109	−17%
Greece	−0.176	−0.129	−27%	0.127	0.112	−12%
Italy	n.a.	n.a.		0.070	0.063	−10%
Netherlands	n.a.	n.a.		−0.118*	−0.116*	−2%
Spain	n.a.	n.a.		−0.114	−0.086	−25%

Notes: Only de B coefficients are given here; the underlying models are the same as models 2, 3, 5 and 6 in Table 2, but then estimated per country.

^* p < 0.05

^** p < 0.10

n.a. indicates there was no considerable PRR or PRL in that country, or in the cases of France, Italy, and The Netherlands that they did not meet the criteria of number of PRR voters to allow separate estimates (>40).

To the extent that the sample sizes permitted, we reran all the robustness variations for these countries separately (Appendix B2). The analyses largely confirm our general conclusion with regard to the PRR gender gap being partly explained by populist attitudes, but they show some variations as well. Whereas Greece shows the strongest decreases (up to 41%), the models for Austria include a considerable increase in the gender gap after the inclusion of populist attitudes. Moreover, the logistic regression models show more changes hovering around the 0% mark. Due to the technicalities involved with logistic models, it is safer to only focus on the changes in the dummy coefficient that are, for instance, at least 4%. Taking this criterion into account, 21 of the 22 country-specific models confirm a decreasing gender gap after controlling for populist attitudes.

In sum, we find a robust pattern of a modest sex gap in voting for the PRR. This pattern decreases after taking into account populist attitudes among women and men. It seems that the populism in PRR parties does attract men somewhat more than women. Or, phrased differently, populism seems to deter women slightly more than men from voting for PRR parties. The country variation within the overall pattern should be further examined in future research. For example, in Austria, a similar gender gap is found as in Greece (see Table 4), but populism is less of a mediating factor, which might relate to a somewhat different degree of masculinity in the style and rhetoric of the parties.

The gender gap in voting for the populist radical left

The question is, are the above results due to how populism manifests itself in combination with the attaching ideology – the radical right (e.g. anti-immigration positions) – or does it only have something to do with the thin-centred ideology of populism? In order to further investigate this question, we turn to the populist radical left.

Table 3 includes results from models similar to those we used for the PRR parties, but in this case explaining PRL voting based on the data for Germany, Greece, Italy, the Netherlands, and Spain. In general, it appears that economic attitudes are much more important for voting for the PRL than for the PRR, but both party types attract more voters with strong anti-EU attitudes than non-populist parties. Similarly to the case of the PRR, the overall models seem to fit the results found in the literature, thus confirming their validity.

Regarding gender and populism, first of all, Models 4‒6 to some extent confirm what we observed in Figure 1: there is a gender gap for PRL voting, but it is considerably smaller than it was for the PRR. In fact, the gap is not statistically significant at conventional levels. A closer look at Table 4 demonstrates that the gender gap for the PRL is also more varied across countries. The largest negative gender gaps are found for Germany, the Netherlands, and Spain. For the German sample, approximately one-sixth of the gender gap appears to be explained by gender differences in holding populist attitudes; for Spain, this value is actually 25%, but the original gap is not statistically significant. For the Netherlands, only a very mediating effect is found. For the other two countries, no mediating effect is found that explains why women vote less for populist parties because no such gender gap was found in the model without the populist attitudes to start with.

Similar robustness tests as for voting for the PRR were conducted (Appendix 2). Given the mixed results over the countries, the pooled model’s results cannot be given much weight: the decrease in the gender gap after including populist attitudes might seem to be substantial, but this result must be viewed relative to the hardly existing gap to start with, which distorts the picture. The country-specific models for the three countries that registered a negative gender gap in Table 4 further show rather unstable results for the Dutch case, which might be related to the presence of a PRR party as well ‒ i.e. multiple populist options were offered to voters. For the German Die Linke and the Spanish Podemos, the role of populist attitudes seems to be rather robust and in line with the expectations.

This comparative analysis is one of the first assessments of voting for the PRL and shows that gender and political attitudes do play a role in such voting preferences. For two countries, the pattern found for the PRL seems to be confirmed, but for the three others, it is not. These latter three are the countries where a strong PRR party is also present. This complex relationship deserves more attention in further research.

Conclusion

A recurring question within the field of populism is, why do more men than women vote for populist parties? As noted, research over the last few decades implies that we know more about the reasons that voters support populist parties, in particular populist radical right parties (e.g. Betz 1994; Immerzeel 2015; Ivarsflaten 2008; Lubbers et al. 2002; Norris 2005). We also know more about why more men vote for populist radical right parties than do women. However, also as noted, we are still not able to fully explain the gender gap (Harteveld et al. 2015; Immerzeel et al. 2015; Spierings and Zaslove 2015b). Spierings and Zaslove (2015b) conclude that perhaps the best manner in which to understand the gender gap is to consider PRR parties as a more radical version of the mainstream right. If so, we should also take the populist left into account.

In this study, we thus examine whether it was populism rather than right-wing (or left-wing) ideology that explains the so-called gender gap. In the first step of our analysis, we assessed whether women do indeed vote less for populist parties, and whether this tendency indeed holds for both the populist radical right and the populist radical left. It does seem to hold for both the PRR and PRL that men tend to vote more for them, although the tendency is considerably stronger and more consistent for the radical right than for the radical left. However, even for the PRR, we should be careful in labelling these parties as Männerparteien. These parties draw a large minority of their votes from women even though they are relatively traditional regarding gender issues (see Akkerman 2015; Spierings and Zaslove 2015a). In other words, these parties represent women’s concerns, albeit gender-traditional and anti-feminist positions (see Celis and Childs 2012).

Next, our analysis showed that for the PRR parties, the populism variables do indeed partly explain why more men vote for the PRR than do women. In the case of the radical left, this pattern is only found in two cases. In these two cases, no PRR party was present, while in the other three cases, a PRR party was present. In other words, it is possible that the gender gap in voting for the PRR and to a lesser extent the PRL in part occurs because populism turns women away from voting for populist radical right parties. Until now, the structural socio-economic approach has been the main explanation for the gender gap in voting for the PRR. The mediating impact of populist attitudes on voting behaviour employed in this study adds to the existing explanations of the gender gap in voting for populists, and in particular the populist radical right. In other words, the relationship demonstrated in this study provides an additional explanation for the gender gap that is not accounted for by the existing explanations. Thus, for example, populist attitudes do not mediate the socio-economic factors but constitute an independent and alternative explanation for part of the gender gap. In fact, we only found one other clear mediation effect of populist attitudes for both the PRL and the PRR. Populist attitudes were connected with opposition to European integration, suggesting that opposition to European integration may be about the divide between the ‘evil political elite’ and the ‘good people’ and not so much about the EU itself. From a gender perspective, this finding might be particularly relevant, as the EU has been a driving force for women’s emancipation and the emancipation of lesbian, gay, bisexual and transgender (LGBT) people across Europe.

That populism does not play the same role in attracting or repelling women and men from the PRL suggests that the thin-centred populist ideology works differently for the left than for the right. It would be interesting to compare our results with supply-side analyses within the context of party systems that have different combinations of left- and right-wing populist parties. In other words, in systems where there are several populist parties, and in particular populist radical right and populist radical left parties, the question is whether this presence of both party types affects whether women vote for (certain) populist parties.

Our results for the PRR and the PRL in the cases of Germany and Spain were robust across countries and estimation methods: they all point towards populism as possessing a modest mediating impact on women voting less for the populist radical left. At the same time, the gender gap differs considerably across countries, as we found before using the ESS instead of EES data (Spierings and Zaslove 2015b), and the mediating role of populism also varied considerably. The particular country, election or party seems to play a major role in shaping to what extent and why women vote less for PRR parties. Further analyses that also take into account the ideological programmes and leadership of these parties might explain these differences. This context dependency deserves to be theorised and studied in more detail by bringing together the demand and supply literature in a systematic way (see Rydgren 2007; Spierings and Zaslove 2015a). In this article, we hope to have made a first step in theorising and presenting how populist attitudes as a gendered concept help to further explain who votes for the PRR and who does not.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Niels Spierings is Assistant Professor of Political and Gender Sociology at Radboud University. His work focuses on issues of inequality and participation, in particular, gender, politics, and Islam. Recent key publications include Social Media, Parties, and Political Inequalities, with K. Jacobs, Springer (2016) and ‘Gender Equality Attitudes among Turks in Western Europe and Turkey: The Interrelated Impact of Migration and Parents’ Attitudes’, Journal of Ethnic and Migration Studies (2015). [[email protected]]

Andrej Zaslove is Assistant Professor of Comparative Politics in the Department of Political Science at Radboud University. His recent publications include ‘Italy. A Case of Mutating Populism?’, with B. Verbeek, Democratization (2016) and ‘The Impact of Populist Radical Right Parties on Foreign Policy: The Northern League as a Junior Coalition Partner in the Berlusconi Governments’, with B. Verbeek, European Political Science Review (2015). [[email protected] ]

Acknowledgements

We would like to thank the anonymous reviewers, the editors of the special section, and the participants of the panel ‘Still Men’s Parties? Gender and the (Radical) Right in Comparative Perspective’ at the 4th European Conference on Politics and Gender, Uppsala University, Sweden in 2015 for their helpful comments.

Notes

1. It is more difficult to classify Movimento 5 Stelle. Early on in its political development, it was more clearly a left-wing party (both regarding demand and supply). More recently, it has received support from voters from both the left and the right, while its political message has even on occasion included more right-wing issues (Bordignon and Ceccarini 2013, 2015; Natale 2014).

2. For a similar discussion regarding this section please see Spierings and Zaslove (2015b).

3. Spain is the exception here because Podemos did not participate in this national election but, based on the party’s other election results, we consider it to be currently a major populist party in Europe, so we included Spain.

4. Considering the Golden Dawn as a PRR party is debatable, as it is clearly anti-immigration and right-wing but not unambiguously populist. There is no ideal solution here because the party is likely a communicating vessel with the Populist Orthodox Rally (POR), and placing it in the non-populist group or removing it from the sample would distort the analysis. Moreover, not including it would remove a considerable number of voters. We carried out a robustness check by rerunning some models with only the POR. The results still showed a decrease in the coefficient found for the female dummy after the inclusion of populist attitudes.

As noted above, the classification of the Movimento 5 Stelle is also somewhat problematic. There is consensus that it is a populist party, however, as noted, it also attracts voters from the centre-right, while some statements, especially regarding immigration, make classification more difficult. However, classifying it as populist radical right party would be incorrect. Thus, for substantive and methodological reasons (i.e. similar to issues mentioned above with Golden Dawn) we have chosen to classify it as a populist radical left party.

5. In the EES data on the countries included here, almost two-thirds of all voters score either 1 or 11 on a propensity scale that ranges from 1 to 11.

6. Principal Axis Factoring; Eigenvalue: 2.60; KMO 0.74; Bartlett’s Test (p < 0.001); factor loadings all above 0.84.

7. Only a handful of people are included here; excluding them does not influence the conclusions.

8. These items are the following: ‘You trust the national parliament’; ‘The national parliament takes the concerns of citizens into consideration’; ‘My voice counts in the EU’; ‘My voice counts in my country’; ‘You trust the institutions of the EU’; and ‘The European Parliament takes into consideration the concerns of European citizens’.

9. It should be noted that these items are not the same ones as used by other studies of populist attitudes (Akkerman et al. 2014; Hawkins et al. 2012; Spruyt et al. 2016). The items from the other studies are relatively new and have not been included in larger comparative surveys, as yet.

10. These items include the following: ‘Whether the EU is generally a good thing or not’; ‘Whether European integration has gone too far or not’; and ‘Whether the EU or member states should have the most authority over citizens’.

11. Disaggregating the non-populist parties and focusing only on the main left-wing parties shows that the gender balance is generally more in favour of women among the mainstream left-wing parties. Differences in the gender balance in voting in favour of women are found in 10 of 14 cases for the mainstream left-wing party compared to the PRL party. The major exception is Greece, which shows a different pattern throughout, as does the Dutch case for the European elections.

12. If we do not focus on the overall scale but on the six indicators separately, we find no strong differences. The differences between men and women are not above 0.1 on scales ranging from 1 to 4.

Appendix 1.

Exploratory factor analysis populist attitudes

Table

Total variance explained
Factor	Initial Eigenvalues			Extraction sums of squared loadings			Rotation sums of squared loadings^a
	Total	% of variance	Cumulative %	Total	% of variance	Cumulative %	Total
1	3.970	30.542	30.542	3.599	27.685	27.685	3.054
2	2.405	18.500	49.042	1.919	14.764	42.448	2.091
3	1.057	8.128	57.170
4	0.940	7.228	64.398
5	0.923	7.102	71.500
6	0.811	6.236	77.736
7	0.654	5.033	82.768
8	0.552	4.243	87.011
9	0.539	4.149	91.160
10	0.372	2.860	94.020
11	0.330	2.538	96.558
12	0.251	1.934	98.492
13	0.196	1.508	100.000
Extraction Method: Principal Axis Factoring.
Note: ^aWhen factors are correlated, sums of squared loadings cannot be added to obtain a total variance.

Table

KMO and Bartlett’s test
Kaiser‒Meyer‒Olkin Measure of Sampling Adequacy		0.797
Bartlett’s Test of Sphericity	Approx. Chi-Square	11992.403
	df	78
	Sig.	0.000

Table

Structure matrix
	Factor
	1	2
qpp9_1 You trust the (NATIONALITY PARLIAMENT) / QPP9 For each of the following statements, please tell me to what extent it corresponds or not to your attitude or opinion.	0.776
qpp9_2 The (NATIONALITY PARLIAMENT) takes the concerns of (NATIONALITY) citizens into consideration / QPP9 For each of the following statements, please tell me to what extent it corresponds or not to your attitude or opinion.	0.785
qpp9_3 Sometimes politics and government seem so complicated that a person like you can’t really understand what’s going on / QPP9 For each of the following statements, please tell me to what extent it corresponds or not to your attitude or opinion.
qpp21 dichotomised: Do you consider yourself to be close to any particular political party? [yes/no]
d71_1 National political matters / D71 When you get together with friends or relatives, would you say you discuss frequently, occasionally or never about...?		0.800
d71_2 European political matters / D71 When you get together with friends or relatives, would you say you discuss frequently, occasionally or never about...?		0.785
d71_3 Local political matters / D71 When you get together with friends or relatives, would you say you discuss frequently, occasionally or never about...?		0.596
d72_1 My voice counts in the EU / D72 Please tell me to what extent you agree or disagree with each of the following statements.	0.751
d72_2 My voice counts in (OUR COUNTRY) / D72 Please tell me to what extent you agree or disagree with each of the following statements.	0.732
qp6_2 You trust the institutions of the EU / QP6 For each of the following statements, please tell me to what extent it corresponds or not to your attitude or opinion.	0.645
qp6_9 You are very interested in politics / QP6 For each of the following statements, please tell me to what extent it corresponds or not to your attitude or opinion.		0.506
qp6_4 The European Parliament takes into consideration the concerns of European citizens / QP6 For each of the following statements, please tell me to what extent it corresponds or not to your attitude or opinion.	0.720
qp6_7 It is very important for you which particular candidates have been elected as MEPs in the European Parliament / QP6 For each of the following statements, please tell me to what extent it corresponds or not to your attitude or opinion.
Extraction Method: Principal Axis Factoring. Rotation Method: Oblimin with Kaiser Normalisation.

Appendix 2.

Changes in coefficient for gender dummy after including populist attitudes in alternative models as robustness tests

Table

	Populist radical right					Populist radical left
Model	Pooled	AU	DK	FI	GR	Pooled	GE	NL	SP
Presented model	−9%	−6%	−24%	−5%	−27%	−38%^3	−17%	−2%	−25%
Variations on populism variable
Alternative 4-item populist attitudes scale with only divide indicators	−7%	−2%	−31%	−3%	−32%	−19%^3	−11%	−3%	−15%
Populism as a dummy (8 or higher = 1)	−9%	+11%	−10%	−11%	−41%	−44%^3	−5%	−3%	−7%
Variations on voting variable and estimation technique
Multinomial^1: PRR/PRL at least ones vs not populist	−8%	−^2	−^2	−^2	−11%	−^3	−^2	+4%	−^2
Logistic regression^1: if national election were held today	−3%	−14%	−4%	+2%	−17%	−^3	−4%	+6%	−21%
Logistic regression^1: last national election	−1%	−6%	−2%	+1%	−1%	−^3	+4%	−1%	−^4
Logistic regression^1: last European vote	−6%	−5%	+1%	−2%	−4%	−^3	−6%	+3%	−21%
OLS on vote choices including propensity	−7%	−2%	−8%	−6%	−30%	−^3	−18%	−3%	−29%
OLS on vote choices only included national election indicators	−10%	−4%	−^3	−7%	−28%	−10%^34	−20%	+1%	−^4

*p < 0.05; **p < 0.01; ***p<0.001.

Notes: 1 In logistic regression models changes in coefficients are partly due to the computing method, therefore the given changes should be treated carefully, especially the ones very close to zero; 2 no PRL and PRR present and thus no multinomial model has been run; 3 the gender gap was below 0.05, inflating every change to extreme percentage; 4 Spain not included because Podemos was no present in the data on the last national election; (5) there was no initial negative gender gap.