Working class economic insecurity and voting for radical right and radical left parties

The radical right ’ s rise is widely assumed to go hand-in-hand with increasing economic insecu- rity, as manual workers are considered typical radical right voters. We question whether economic insecurity actually drives working class members to vote for radical parties, both right and left. Using European Social Survey data from 12 Western European countries (2002 – 2014), we tested whether less secure employment links to vote for such parties. We did this by dis- tinguishing people in permanent employment from those in the same social class but not in permanent employment. Our outcome was surprising: whereas perceived job insecurity corre- lated with radical right voting, actual economic insecurity in terms of temporary employment was not associated with greater likelihood of voting for a radical right party among the working class. Instead, it was the radical left for which we found indications that it appealed more to groups of people in such an insecure economic position.


Introduction
The rise of radical right parties has often been attributed to their appeal to the manual working class (Betz, 1994;Morgan and Lee, 2018;Oesch, 2008;Oesch and Rennwald, 2018;Rydgren, 2012). The argument goes that the manual working class perceives their labour market position as insecure, due to perceptions of a threat to their economic position emanating from the arrival of immigrants and from ongoing globalization processes. The manual working class is therefore thought to be more susceptible than other voters to the radical right's anti-immigrant and anti-elitist rhetoric (Betz, 1994;Kriesi et al., 2008). While ample studies have found empirical support for this radical right working class thesis, debate continues about whether the relationship can be understood by the economic insecurity attributed to the working class (Bornschier and Kriesi, 2012;Carella and Ford, 2020;Gidron and Hall, 2017;Gidron and Mijs, 2019;Kurer, 2020;Oesch, 2008;Sipma and Lubbers, 2020;Vlandas and Halikiopoulou, 2016).
Previous studies suggest that cultural attitudes, such as anti-immigrant sentiment, may be more strongly related to voting for the radical right than economic concerns (Lucassen and Lubbers, 2012;Oesch and Rennwald, 2018). They conclude that within the manual working class in particular, the perception of a cultural threat posed by immigrants, more than the perception of an economic threat, is relevant to understand why they vote for radical right parties (Bornschier and Kriesi, 2012). Theoretically, however, perceptions of a cultural threat within the manual working class might be explained by the insecure economic position of working class voters (Halikiopoulou and Vlandas, 2020;Lucassen and Lubbers, 2012). Such economic insecurity could give rise to perceptions of threat, both cultural and economic, implying that economic insecurity is at the basis of support for the radical right. As this seems theoretically sound, we propose that a better empirical test is needed to investigate whether economic insecurity within the manual working class can explain support for the radical right. We construct this test using an important temporary manifestation of economic insecurity: type of employment (i.e., permanent, temporary and self-employment). These employment types provide an objective indicator of actual job insecurity, while possibly also reflecting workers' perceptions of job insecurity. Nowadays, more and more Europeans find themselves in an insecure economic position due to the increasing share of temporary employment contracts, particularly within the manual working class (OECD, 2021). It is therefore relevant to assess whether this manifestation of economic insecurity affects manual working class voting for radical right parties.
Investigating the role of economic insecurity in the voting behaviour of the manual working class, however, requires us to go beyond the radical right literature, as working class people are also overrepresented among the radical left's electorate Visser et al., 2014). Radical left parties champion social justice and wealth redistribution (Ramiro, 2016), therefore appealing to voters in economic insecure positions. Studies show that those in more insecure employment types, such as temporary employment, are more likely to vote for radical left parties (Marx, 2016). We theorize here, however, that within the working class those who are more economically insecure vote either for the radical left or for the radical right.
The main contribution of the current study is to clarify the role of economic insecurity within social classes in voting for radical right and radical left parties, gaining greater insight into the influence of labour market position on voting behaviour. We set out to answer the following research question: to what extent does economic insecurity affect voting for the radical right or radical left within social classes, particularly within the manual working class?

Economic insecurity and radical right voting
Whereas classical studies assume that economic grievances among the lowest social classes explain why they vote for the radical right, later empirical research indicates that economic grievances play only a modest role (Bornschier and Kriesi, 2012;Ivarsflaten, 2005;Lucassen and Lubbers, 2012;Oesch and Rennwald, 2018). For example, Gidron and Mijs (2019) found reduced income to be unrelated to support for a radical right party, even in times of economic recession. Instead, cultural or identity-related explanations have been found to be more influential in radical right voting (Lucassen and Lubbers, 2012). Such cultural explanations usually entail national identity threats and often attitudes towards immigrants (Gidron and Mijs, 2019). For radical left voting, economic considerations have been found to play a leading role, especially support for a more equal income distribution Visser et al., 2014). The conclusion is thus easily drawn that cultural concerns outweigh economic ones for radical right voting, with the reverse being true for radical left voting. But why then would a lower class position be associated with cultural concerns and consequently with a vote for the radical right?
Theories propose that due to the inherent economic insecurity of manual working class employment, cultural concerns, and especially anti-immigration attitudes, develop (Meuleman, 2011;Scheepers et al., 2002). Economic insecurity drives 'natives' to be protective of what they have, in both economic and cultural terms. Because immigrants are more often employed in manual jobs, 'natives' in the manual working class may perceive competition with immigrants for resources more acutely than people in other social classes. This stirs unfavourable attitudes towards immigration and dissatisfaction with existing political elites (as they have not prevented this competition from arising) (Olzak, 1992;Scheepers et al., 2002), translating into votes for the radical right (Rydgren, 2008;Savelkoul and Scheepers, 2017). If these theoretical premises hold, people with economic insecurity in the manual working class would feel most threatened and therefore adhere most to traditional and nationalist values and the idea of a safe, ethnically homogeneous communitythe same values promoted by radical right parties (Betz, 1994;Kriesi et al., 2008;Zaslove, 2008).
Despite ample empirical support for the relation between social class and cultural and/or economic preferences in radical right voting, it remains unclear whether this relationship is actually driven by economic insecurity (Bornschier and Kriesi, 2012;Gidron and Hall, 2017;Lucassen and Lubbers, 2012;Oesch, 2008). Members of the manual working class are more likely to have flexible employment relations and low-wage jobs, and they experience economic hardship more often than people in higher social classes. Moreover, it is often assumed that belonging to the manual working class equates with economic insecurity. To test whether it is actually the economic position that fosters manual working class voting for the radical right, we used specific indicators to distinguish economic insecurity within the working class. Particularly, we zoomed in on the temporality of the employment type, as this is a main case of insecurity and because temporary employment is on the rise in Western Europe.
Temporary employment relations are a manifestation of economic insecurity, because people with a temporary job have more uncertainty about their future (financial) prospects and a greater anticipation that they will have to compete in the labour market in due time. Since perceptions of insecurity correlate with perceiving immigrants as more of a threat to the own economic position (Billiet et al., 2014), it is to be expected that also and particularly within the working class those in temporary employment will have the highest levels of cultural and political resentment. Therefore they will be more likely to vote for a radical right party compared to other members of the working class.
Economic insecurity in terms of temporary employment might be expected to affect voting behaviour in higher classes too, but we anticipate that, due to lower levels of ethnic competition in jobs found among higher social classes, temporary employment in these classes will bring less insecurity and be less associated with a perception of immigrant threat. Consequently, having temporary employment is less likely to be associated with voting for the radical right in the higher classes.
Summarizing, in this study we expect to reproduce earlier findings that manual workers are more likely to vote for the radical right, and we also expect people in temporary employment to be more likely to vote for the radical right compared to people in permanent employment. Thus, we hypothesize that within the manual working class, those in temporary employment are more likely to vote for the radical right than those in permanent employment, and the difference is greater than in other social classes (H1a), because members of the manual working class in temporary employment perceive greater job insecurity (H1b) and have stronger anti-immigrant attitudes and more political distrust (H1c).

Economic insecurity and radical left voting
Overrepresentation within the radical left's electorate of members of the manual working class and people with an insecure employment type can be explained by voters' economic concerns. Thus, lower wage and less secure economic positions translate into perceptions of economic unfairness and inequality, aligning with the views of radical left parties, which strive for greater social justice and wealth redistribution (Ramiro, 2016). Indeed, the radical left does particularly well in times of economic decline (March and Rommerskirchen, 2015;Vail, 2009) and among people with economically left-wing policy preferences, such as support for income redistribution (Akkerman et al., 2017;Bowyer and Vail, 2011;Visser et al., 2014). As people in temporary employment are at greater risk of losing their job, they are expected to be more in favour of social policies that protect their position (Burgoon and Dekker, 2010;Cusack et al., 2006;Marx, 2014;Rehm, 2009). Policies championed by the radical left may be more directly in line with the interests of people in an insecure economic position than radical right policies. Hence, members of the manual working class in temporary employment might be more likely to vote for the radical left than their working class counterparts in permanent employment.
Furthermore, people in an insecure economic position might be particularly attracted to a radical left party rather than more moderate left-wing parties. Mainstream left parties, such as social democrats, were in the past popular among the working class, as they were thought to defend working class economic interests (Jansen et al., 2013). However, recent research on the voting behaviour of people in temporary employment indicates that social-democratic parties have lost the trust of many of these voters, as they are perceived as mainly defending the interests of people with a permanent job (Emmenegger, 2009;Lindvall and Rueda, 2014). Those in temporary employment might then opt for a variety of other parties, including the Greens and the radical left (Lindvall and Rueda, 2014;Marx, 2014Marx, , 2016Marx and Picot, 2013). We expect the radical left to be particularly popular among people with temporary employment in a manual working class job. Indeed, the radical left presents itself as not only serving the interests of people with insecure employment, but also as explicitly serving the manual working class. This leads to the expectation that people in temporary employment will be more likely to vote for the radical left than people with a permanent job. We hypothesize that within the manual working class, those in temporary employment are more likely to vote for the radical left than those in permanent employment, and the difference is greater than in other social classes (H2a), because members of the manual working class in temporary employment perceive greater job insecurity (H2b), and are more in favour of income redistribution and have more political distrust (H2c).

Data source and variables
To test our hypotheses we used data from 12 Western European countries collected in seven waves of the European Social Survey (ESS). This cross-country dataset is a biennial survey of the population of a large set of European countries, and provides information on both political behaviour and work status. To have sufficient numbers of radical right voters in the dataset, we pooled all the surveys from 2002 to 2014. This time span includes the period before the start of the economic recession in 2008, the period of the economic recession itself and its aftermath. It is relevant to test whether our expectations hold in these different periods, because the recession increased people's perceived job insecurity (Chung and Van Oorschot, 2011). That might be especially true among people in an insecure economic position. Each wave includes information on working status, allowing us to study the role of economic insecurity in voting behaviour. The complementary modules ESS2 (2004) and ESS5 (2010) further allowed us to test perceived job insecurity as a possible explanation for why people in insecure economic positions vote as they do.
All of the included 12 Western European countries 1 had both a radical right and a radical left party. We excluded respondents aged younger than 18 years and first-generation non-Western migrants. After listwise deleting respondents with missing values, we were left with a main dataset consisting of 62 country-year combinations and 80,895 respondents. Listwise deleting respondents barely affected the distribution on our main variables. In the full sample, consisting of 100,671 respondents, 9.4 percent of the respondents prefer a radical right party and 7.8 percent prefer a radical left party. In the smaller sample used for our analyses, this is respectively 9.6 and 7.9 percent (for more details see Table A.2 in appendix). The distribution of labour market position within the manual working class (13.4 percent in full sample; 13.6 percent in used sample) is also rather similar in both samples: respectively 72.5 and 73.0 percent with permanent employment and 8.8 and 7.8 percent in temporary employment (see same appendix). A subset of this dataset, consisting of surveys from ESS rounds 2 and 5 (2004 and 2010), includes a measure of perceived job insecurity and encompasses 16 country-year combinations for 11 countries (excluding the United Kingdom) and 9083 respondents.
We ran our models both weighted (using the post-stratification weights in ESS) and unweighted, noting that core elements of the weighting are accounted for in our models. 2 Below we present the unweighted results, but where relevant different are discussed in the text (section 4.1) and considered in terms of robustness, being included among a large set of robustness checks (section 4.2).

Dependent variable
The dependent variable in our study, voting behaviour, was based on a combination of two questions on political preferences. First, we measured what respondents voted during the last parliamentary elections. Second, as there could be a time gap between the last election and the moment of data collection, we used the question of what party respondents currently felt closest to, if that differed from their previous voting behaviour. Solely using party closeness was not feasible, as 43% of the people who voted in the last election reported not feeling close to any particular party. Based on these questions, we created a three-category dependent variable reflecting respondents' party preference: a non-radical party (0), a radical right party (1) and a radical left party (2). Respondents with a missing value on our dependent variable where either not eligible to vote, didn't know if they had voted and/or what party they voted for, or refused to answer this question. We deleted 12,892 respondents with a missing value (12.8%) and 12,761 non-voters (11.2%).
We based our categorization of parties on previous literature. The group of radical right parties consists of those characterized by an anti-immigration stance, nationalism, authoritarianism and populism (Immerzeel, 2015;Mudde, 2007;Norris, 2005). In the group of radical left parties, we included communist, democratic socialist and populist socialist parties (March, 2008;March and Rommerskirchen, 2015;Ramiro, 2016). Table 1 provides an overview of the included parties.

Independent variables: labour market insecurity
We expected manual workers in temporary employment to be the most insecure in the labour market. We therefore created a classification combining employment type and social class based on established occupational schemes. Regarding measurement of employment type, we used the question asking respondents whether they were in paid work or not. If respondents were not in paid work, we coded them as unemployed (actively looking and not actively looking for a job) or other, with the latter category made up of respondents in education, permanently sick or disabled, retired, in community or military service or a homemaker. Among those in paid work we distinguished workers in permanent employment from those in temporary employment and the self-employed. 3 Based on the question of how many employees a respondent had, we categorized self-employed respondents as self-employed without personnel (solo self-employed) or self-employed with personnel. This produced a classification of employment types consisting of six categories: (1) permanent contract, (2) temporary contract, (3) solo self-employed, (4) self-employed with personnel, (5) unemployed and (6) other.
Of these employment types, permanent employment was considered the most secure. Temporary employment is less secure, because the employment relation could end when the contract ends, and this employment type is often involuntary (Marx, 2014). The solo self-employed, whose voting behaviour is very heterogeneous (Jansen, 2017), can be divided into two groups (Van Stel and De Vries, 2015). The first consists of successful, highly skilled solo self-employed who perceive their freedom and autonomy as advantageous. The second, less successful group is made up mostly of low-skilled workers who became solo self-employed because of a lack of alternative employment. Solo self-employment within the manual working class was regarded mainly as the latter, and we considered it an insecure position. This distinction enabled us to also investigate the effect of solo self-employment within the classes separately.
We measured social class based on the EGP class scheme (Erikson and Goldthorpe, 2002), which is predominantly derived from the International Standard Classification of Occupations (ISCO) and thus hinges on differences in occupation and employment types and not, for instance, in capital. While the latter is rightfully considered part of class distinctions in the broader sociological sense, we depart from the literature on radical voting behaviour and expand on that. The occupation-based social class categorization has undergone many revisions of late, with these revisions generally concerning the recategorization of occupations of the higher classes. In Europe, these occupations have rapidly expanded in recent decades, necessitating for instance, differentiation between technical and sociocultural professionals (see, e.g., Güveli and De Graaf, 2007;Oesch, 2006). However, the manual working classwhich is our main interest hereremained present in these different class schemes. We applied Ganzeboom's program to construct a variable with an 2 Age, gender and education (used in ESS post-stratification weights) are already included as control variables in our models. By including random intercepts at the country/round (level 2) and country level (level 3), we control for differences between these higher-level units which also includes differences in designs and data collection between round and countries. For more information on ESS weights, see: https://www. europeansocialsurvey.org/methodology/ess_methodology/data_processing_archiving/weighting.html. 3 The small third category 'no contract' (1.1%) was considered missing.
EGP score for ESS data (Leiulfsrud et al., 2005). 4 The original EGP class scheme consists of 11 categories, which we combined into three categories. 5 In line with our focus, we excluded the self-employed (IVa, IVb and IVc) from the class grouping. 6 Our social class variable thus consists of four categories: (1) manual working class, (2) middle class, (3) high class and (4) respondents without an occupational status (unemployed, retired, etc.). Subsequently, we constructed the variable labour market position by combining employment type with EGP-based social class (see descriptive statistics in Table A.1, in appendix). For each class, we grouped respondents according to their employment type: permanent employment, temporary employment, solo self-employed or self-employed. This created 12 groups, to which we additionally included the unemployed and "other" as separate categories. Our final variable thus consists of 14 categories.

Mediating variables
Two of the seven ESS waves provided information on perceived job insecurity, in the module "Family, Work and Well-being" administered in 2004 (ESS2) and 2010 (ESS5). In an additional model using only these two waves, we assessed the role of job insecurity perception, measured by answers to the statement "my job is secure". Response categories were (0) very true, (1) quite true, (2) a little true and (3) not at all true. Job insecurity was only applicable to employees, and therefore could only be used to test for mediation of potential differences between respondents in permanent and in temporary employment.
For all surveys, we included anti-immigration attitudes, political distrust and support for income redistribution as mediating variables. Anti-immigration attitudes, derived from six items, 7 ranged from (0) weak to (3) strong (Cronbach's α = 0.88). Political distrust was based on three items asking respondents about their trust in politicians, parliament and political parties (Cronbach's α = 0.89).
The scale ranged from (0) complete trust to (10) no trust at all. Leftist economic attitudes were measured by support for income redistribution.

Control variables
As socio-demographic control variables, we included those found to be most influential in studies on voting for radical parties and which could also relate to economic insecurity (Immerzeel, 2015;Spierings and Zaslove, 2017). 8 Education is the most relevant control variable, as people in the manual working class tend to be lower educated, and the lower educated are more likely to vote for the radical right and radical left. Moreover, education captures socio-economic, cognitive and cultural aspects Savelkoul and Scheepers, 2017;Visser et al., 2014). In other words, education entails more than an economic explanation of voting behaviour. By including it in our model on the role of economic insecurity, we control for the confounding effects of the different aspects education captures. Education was measured using the ISCED classification and consists of the following categories: (0) less than lower secondary education; (1) lower secondary education; (2) upper secondary education; (3) post-secondary, non-tertiary, education; and (4) tertiary education. Moreover we included gender, coded (0) male and (1) female. Age was included as an interval variable and ranges from 18 to 102. Religion was measured by a question asking respondents whether they considered themselves as (1) religious or (0) not. Ethnicity consists of three categories: (0) those born in the country of residence, (1) first-generation Western migrants and (2) second-generation Western and non-Western migrants. As noted, we excluded first-generation non-Western migrants from our analyses.
At the contextual level, we included unemployment rate as a percentage of the active population, change in the unemployment rate compared to the previous year (Eurostat, 2018b), GDP per capita at current market prices, GDP per capita growth compared to the previous year (Eurostat, 2018a) and foreign-born population as a percentage of the total population (OECD, 2018).

Analytical strategy
Given the structure of our data, we employed multilevel multinomial logistic regression analyses. Respondents were considered as first-level units, with country-year combinations considered as second-level units and countries as third-level units. We opted for multinomial regression, as our dependent variable consists of three categories. We conducted our analyses in three steps. We started, first, with the descriptive analysis (Fig. 1). Second, we tested whether radical right and radical left voting were affected by respondents' labour market position, to determine whether the effect of the employment type differed across social classes (Table 2, H1a and H2a). Third, we tested whether the effect of labour market position on radical voting was mediated by subjective job insecurity (H2b and H2b) and anti-immigration attitudes, political distrust and support for income redistribution (H1c and H2c). Since subjective job insecurity was only available for a subset of our data, we started by testing H1c and H2c on the full sample (Table 2), and thereafter tested H1b, H1c, H2b and H2c on the smaller subset of data (Table 3). We present the relevant estimates of our multinomial regression models in Tables 2 and 3. Full models, including the estimates of the control variables, are presented in Table B.1 and B.2 in appendix. Fig. 1 shows the percentage of respondents who voted for radical right and radical left parties in each labour market position. Respondents in the manual working class (together with the unemployed) were most likely to vote for a radical right party. Further, manual working class respondents voted for a radical right party more often than a radical left party. However, those in temporary employment in the manual working class (i.e., those with an insecure economic position) were not more likely to vote for the radical right compared to those in permanent employment within the manual working class. We did find differences here for the radical left. Insecurity was linked to a higher probability of voting for the radical left within the manual working class. It remains to be seen if this result holds in the multilevel multinomial logistic regression analysis.

Main results
Before elaborating on our results regarding the influence of insecurity on radical right and radical left voting within the manual working class, we first discuss the direct effects of social class and employment type on voting behaviour (see Table B.3 in appendix). Overall, the higher respondents' social class, the less likely they were to vote for a radical right party (Model B.3a), in line with previous studies. Congruently, high social class respondents were less likely to vote for a radical left party, compared to manual working class respondents, and middle class respondents were slightly more likely to vote for a radical left party than manual working class respondents (Model B.3b). We found no significant effect of employment type on radical right voting (Model B.3a), except for a negative effect of "other" compared to permanent employment. However, employment type was significantly related to radical left voting (Model B.3b). Respondents in temporary employment were significantly more likely to vote for a radical left party compared to respondents in permanent employment (b = 0.25). The self-employed, both solo and those with personnel, were significantly less likely to vote for a radical left party compared to the permanent employed (b = − 0.24; b = − 0.81).
Regarding the extent that an insecure employment type affected voting behaviour within the working class, we expected manual working class respondents in an insecure position, that is, in temporary employment and possibly also the self-employed, to be more likely to vote for a radical right party (H1a) or a radical left party (H2a) compared to people in a more secure position, namely members of the manual working class in permanent employment.
In Model 1a of Table 2, we put this expectation to test. To ease interpretation of the interaction effect of employment type with social class, we combined the two variables into one labour market position variable. The table presents the same model with different reference categories, also to ease interpretation. Model 1a shows a surprising absence of differences in radical right voting between the temporary and permanent employed across all social classes. The coefficients are both small and statistically insignificant at the conventional levels. We therefore cannot conclude that people in an insecure employment type in the manual working class are more likely to vote for a radical right party than people in more secure employment. We therefore cannot support H1a. We did find differences between the self-employed and the others that varied across social classes, but not in the expected direction. The selfemployed in the higher classes were more likely to vote for a radical right party than the permanent employed higher class respondents (b = 0.24). Both the self-employed with personnel (b = − 0.22) and the solo self-employed (b = − 0.24) in the manual working class were less likely to vote for a radical right party than the permanent employed in the manual working class.
In Model 2a (Table 2), we included anti-immigration attitudes, political distrust and support for income redistribution, to test H1c. In line with previous research, we found anti-immigration attitudes (b = 1.61) and political distrust (b = 0.20) to be relevant indicators of radical right support. However, there was no evidence of the expected mediation effect, as labour market insecurity had no direct effect in the first place, refuting H1c. The significant difference between the self-employed and permanent employed in the high class decreased upon inclusion of the mediating variables (from 0.24 to 0.06); these factors thus partially mediated the effect. 9 Other coefficients for the effect of labour market position on radical right voting did not change substantially after including anti-immigration attitudes and political distrust. An interesting finding is that people who favoured income redistribution appeared to be less likely to  vote for the radical right. We also tested whether the effect we found of temporary employment on radical left voting (see Table B.3) was strongest in the manual working class. Model 1b of Table 2 shows that the difference between the temporary and permanent employed in their likelihood of voting for a radical left party was significant in the manual working class (b = 0.35) and middle class (b = 0.28), but the latter is less robust across alternative modelling strategies. 10 In the highest class is it absent altogether (b = 0.17). This lends support to H2a, as respondents in an insecure employment type were more likely to vote for a radical left party than people in more secure employment, especially among members of the manual working class.
Again, we added anti-immigration attitudes, political distrust and support for income redistribution to the model (Model 2b, Table 2). In contrast to the results for the radical right, these showed that the radical left was relatively more popular among people who supported income redistribution (b = 0.64) and favoured immigration (b = − 0.55), compared to all other parties. Political distrust had, similar to the radical right, a positive effect on radical left voting (b = 0.17). A comparison between Model 1b and Model 2b shows no clear change in the coefficients for temporary employment versus permanent employment within the manual working class (b = 0.35; b = 0.33). 11 Although we did find an effect of labour market insecurity, it was not mediated by the political attitudes, therefore providing no support for H2c.
Regarding the potential mediators, the difference between the permanent employed and the solo self-employed in the high class was stronger after including political distrust and anti-immigration attitudes (b = − 0.14; b = − 0.24). The high class solo self-employed were generally less likely to vote for a radical left party than their permanent employed counterparts, but they also expressed higher levels of political distrust; the effect presented in Model 2b was suppressed by these attitudes.
Next to anti-immigration attitudes, political distrust and support for income redistribution, we expected perceived job insecurity to mediate the effect of economic insecurity on radical right (H1b) and radical left (H2b) voting. Because questions regarding perceived job insecurity were asked only to employees and were included only in 2004 and 2010, we used a smaller sample in these analyses.
Model 3, firstly, shows that our previous results were partly replicated in this smaller sample. As in the larger sample, temporary employed manual working class respondents were not more likely to vote for the radical right compared to the permanent employed in the manual working class (b = 0.13; Model 3a, Table 3). Neither were they significantly more likely to vote for the radical left than the permanent employed in the manual working class, due to the smaller sample size (b = 0.28; Model 3b, Table 3). Though the effect size was similar, the standard error was larger in this much smaller sample.
In Table 3 the focus is on perceived insecurity, included in Models 4a and 4b. Perceived job insecurity led to an increased likelihood of radical right voting (b = 0.10). Thus, while the objective situation was found to be not relevant in radical right voting, the perceived situation was. Still, H1b must be rejected because we found no direct effect of the objective situation in the first place. Moreover, the differences between social classes in likelihood of voting for the radical right remained after including perceived job insecurity (cf .  Table B.4), with working class respondents perceiving higher levels of job insecurity than respondents in the higher classes (Table B.5). To summarize, the effect of social class on radical right voting seems to be based neither on objective nor subjective job insecurity. 10 When including ESS post-stratification weights, the b-coefficient for middle class is 0.16 (standard error = 0.10) and (thus) not significant (plevel = 0.107); for the manual working class the b-coefficient decreases to 0.24 with a standard error of 0.12 and a p-level of 0.050. 11 A similar comparison for het middle class shows a decline from 0.28 to 0.21, and from 0.16 to 0.09 in the weighted model. Thus the biggest decline was found in the model without a significant effect to start with. This further underscores that for the middle class the effect is less robust (but might deserve further attention in future work).
Also in these models we included anti-immigration attitudes, political distrust and support for income redistribution. The effect of job insecurity on radical right voting dropped by 60%, losing its significance in Model 5a (b = 0.04) compared to Model 4a (b = 0.10) ( Table 3). 12 Anti-immigration attitudes (b = 0.10) and political distrust (b = 0.03) were both significantly correlated with perceived job insecurity (Table B.4). The effect of perceived job insecurity on radical right voting was thus mediated by political distrust and antiimmigration attitudes.
Surprisingly, job insecurity had no significant effect on radical left voting (b = 0.04; Model 5b, Table 3), and we therefore found no support for H2c. This is noteworthy, as we did find evidence of a correlation between temporary employment and stronger perceptions of job insecurity.

Robustness checks
The results show, in contrast to our expectations based on key theoretical frameworks, no effect of objective job insecurity on radical right voting. To increase confidence in this finding, we employed several robustness checks. The results did not differ when we used the original nine EGP categories for social class instead of our three categories (Table B.6). Nor did distinguishing between men and women or 'natives' and migrants yield different results for radical right voting. This implies that the null finding is rather consistent, with no effect present even among "likely" cases, such as men and 'natives' only. This is in line with findings in the literature of a persisting gender gap in radical right voting, which is partly explained by gender differences in economic insecurity, though the influence of socio-economic factors on voting for the radical right do not appear to differ between men and women (Givens, 2004;Mayer, 2015;Spierings and Zaslove, 2017).
The effect of an insecure labour market position on radical left voting is found, but our results indicate that it is mostly present among 'natives' and men. Women were more likely than men to vote for a radical left party (Table B.3), though additional analysis showed an effect of labour market insecurity on radical left voting only among men, and also in the weighted models the effect is stronger for men, albeit insignificant for both women and men. A possible explanation is that men generally contribute more to household income than women, making labour market position more important in their voting behaviour (Strøm, 2014). Or, the greater effect of economic insecurity on men's left-wing voting could be because male-dominated occupations have been particularly impacted by the flexibilization of the labour market, or because women, including those in secure jobs, are more likely to vote left anyway (Inglehart and Norris, 2003;Mayer, 2015;Spierings and Zaslove, 2017).
We also estimated the effects within each country and each wave. The difference in radical right voting between the temporary employed and the permanent employed in the manual working class was not significant in any country or wave, nor was a direct effect of temporary employment found in any of the countries studied. Interestingly, though, a direct effect of temporary employment was present in two waves. Respondents in temporary employment were more likely than the permanent employed to vote for a radical right party in the 2002 wave (b = 0.38), and less likely to vote radical right in 2008, when the economic crisis hit Europe (b = − 0.42). All significant differences between the permanent and temporary employed in likelihood of voting for a radical left party were in the expected direction, and were significant in 2002, 2012 and 2014 and in Belgium, Germany, Finland and Norway.

Conclusion and discussion
An increasing number of Europeans are experiencing economic insecurity (Blossfeld et al., 2011;Chung and Van Oorschot, 2011). More are working in temporary employment (OECD, 2021) and more are solo self-employed (Van Stel and De Vries, 2015). This is said to have contributed to the rise of the radical right. The aim of our study was to examine whether these insecure job positions have indeed affected radical right voting, as well as radical left voting. We conceptualized the idea of an insecure labour market position by combining employment type with social class, and formulated the following research question: to what extent does economic insecurity affect voting for the radical right or radical left within social classes, particularly within the manual working class?
We expected people in an objectively insecure labour market position to be more likely to vote for a radical right party compared to counterparts in more secure employment. Contradicting this popular idea, we found no significant differences between the permanent, temporary and self-employed, not even within the manual working class. In fact, the solo self-employed in the manual working class, which we also classified as an insecure economic position (Van Stel and De Vries, 2015), were significantly less likely to vote for a radical right party than the permanent employed in the manual working class. We tested our expectations in countries and periods separately, as country-specific labour market regulations might affect the extent that a particular employment type correlates with voting behaviour. However, we must conclude that among manual workers, being in a less secure employment type apparently does not provide fertile ground for radical right parties. Whereas previous research suggests that the relation between social class and radical right voting is mainly explained by cultural attitudes, such as anti-immigrant sentiment, and less so by economic concerns (Gidron and Mijs, 2019;Lucassen and Lubbers, 2012;Oesch and Rennwald, 2018), we expected that economic insecurity could still give rise to anti-immigrant sentiments and therewith explain voting behaviour. Our results suggest, however, that objective economic insecurity does not drive the manual working class's vote for the radical right. This implies that cultural threat perceptions that lead to voting radical right are not rooted in objective insecurity in terms of employment type as measured here.
In contrast to the objective economic situation, we found that perceived job insecurity did foster radical right voting; albeit the effect was small. If economic insecurity affects radical right voting, it is through subjective insecurity rather than more objective insecurity. This is line with Kurer (2020), who found that the perception or fear for income decline relates to radical right voting, whereas actual impoverishment has no effect. Furthermore, this effect was mediated by anti-immigration attitudes and political distrust, implying that cultural threat may originate from perception of economic insecurity or that both are part from a larger world view or predisposition such as societal pessimism (more below). People who perceived their job as insecure considered immigrants as a possible threat to their employment and blamed traditional parties for their inability to guarantee job security, making them more likely to vote for a radical right party. This influence of perceived job insecurity contrasted with the lack of an effect of the objective economic position on radical right voting. People with a permanent manual working class job may have reasons to be insecure about their position as well, as they may question what their permanent contract is worth when companies outsource more of their production to low-wage countries or workers are replaced by robotization or automation. However, our analyses indicate that the higher likelihood of voting for a radical right party among the manual working class, regardless of employment type, is not explained by a higher level of perceived economic insecurity.
What drives members of the manual working class to vote for the radical right, if not economic insecuritywhich is central to theoretical arguments in the literature? Though that question remains unanswered here, our results suggest that it are cultural concerns, not rooted in economic insecurity, that explain the working class vote for radical right parties. Our study, nonetheless, advances previous research by addressing differences in employment type and perceived job insecurity within the manual working class. To understand what differentiates people within the manual working class, study of other manifestations of economic insecurity may be relevant as well, such as low wages, experiences of (long-term) unemployment and economic hardship. Or, perhaps the literature on voting behaviour focuses too much on employment when considering class, bypassing the role of capital as a buffer. In that case it would be worthwhile to consider the concept-measurement discrepancy between class and class-based occupational schemes. Besides, people in the manual working class could feel more economically deprived in general, compared to higher social classes, and might perceive less opportunity to improve their economic position, leading to political discontent and other forms of societal pessimism, explaining voting for the radical right. In this regard, other authors have addressed non-economic (or cultural) explanations of radical right voting and found that economic insecurity indeed plays only a limited role (Bornschier and Kriesi, 2012;Lucassen and Lubbers, 2012;Oesch, 2008;Oesch and Rennwald, 2018). However, this issue is not settled (Halikiopoulou and Vlandas, 2020). We recommend that more manifestations of economic insecurity (and perceptions thereof) be analysed conditionally, to determine whether economic situations do play a role after all in voting for the radical right.
With regard to cultural explanations of radical right voting, perhaps objective and subjective economic insecurity only play a role when people already perceive immigrants as a threat. This implies that immigration attitudes would not only mediate economic insecurities, but would moderate their effect too.
We did find that those with an insecure employment type (i.e., temporary employment) were more likely to vote for a radical left party than those in permanent employment, and this effect was most clearly evident among the manual working class men. Even though the losers-of-globalization thesis (Kriesi et al., 2008) has thus far been applied mainly to radical right voting, it may actually explain radical left voting, as these parties give voice to the political discontent of voters facing the most acute negative economic consequences of globalization. Whereas radical right parties predominantly focus on anti-immigration policies, radical left parties decry the economic consequences of globalization (Ramiro, 2016). People in an insecure labour market position might therefore perceive the radical left as more capable of rectifying their economic insecurity than the radical right, and therefore be more likely to vote for the radical left (Sipma and Berning, 2021). However, the interpretation we tested for here, the mediation by political distrust and support for income redistribution, did not explain the found effect of economic insecurity on radical left voting. It would be interesting to study what political attitudes and policy preferences do explain this effect, for instance, by including more specific attitudes towards social policies regarding job insecurity. Surprisingly, we found no effect of perceived job insecurity on radical left voting, even though people with temporary contracts are more likely to perceive their jobs as insecure. This result raises new questions, requiring further study. For instance, could manual workers in temporary positions be more likely to vote radical left for reasons of solidarity, and less because of their own perceived insecurity. With the current data this is impossible to assess.
More generally, our study lays bare an interesting puzzle: why does the radical right attract people in subjectively insecure jobs, whereas the radical left seems to attract groups of those in more objectively insecure employment positions? One possible answer lies in the issues on which these parties take a dominant stance. When people are in temporary employment, they may vote for parties that propose policies which could directly improve their own situation, such as radical left parties. When people are not in a particularly insecure position objectively, but mainly perceive their situation as such, they may be less supportive of targeted policies to improve the situation of people in weak economic positions, since these policies would not directly serve their own economic interests. Rather they may then translate their sense of insecurity to other feelings of societal pessimism resulting in a vote for the radical right (Steenvoorden and Harteveld, 2017).
Finally, our contribution suggests that to better understand the relationship between social class, economic insecurity and voting behaviour, a gender or intersectional lens is needed. But, other than perhaps expected, given the persisting gender gap in radical right voting, our study demonstrates that gender differences in the role of economic insecurity are more prominent in predicting radical left voting. Indeed, we found that economic insecurity only affected voting for the radical left, and only among men. We provided some potential explanations for these results. Future work with new data might delve into these. This touches upon a limitation of our study; that is, we focused on the manual working class and compared it to the middle and higher classes. Given the strong labour market gender segregation, in particular in manual work, our focus on the rather traditional categorization of working class versus other classes, may mask gender differences.
In sum, besides a small effect of perceived job insecurity on radical right voting, this study found that an objectively more insecure labour market position, in terms of one's employment type, did not increase the likelihood of voting for a radical right party, including among working class voters. Instead, it was the radical left for which we found indications that it appealed more to groups of people in such an insecure economic position. This indicates that care should be taken in using economic insecurity to explain why people from the manual working class are more likely to vote for the radical right.