Welfare chauvinism in times of labour market segmentation: how different employment contracts moderate the impact of welfare chauvinism on support for radical right parties

I examine how different employment contracts moderate the impact of welfare chauvinist preferences on radical right support. Welfare chauvinism has become a cornerstone of radical right’s nativist electoral programme. Yet, there are mixed findings on how welfare chauvinist preferences affect radical right support. While some studies find a positive association, others find little correlation. One reason for such ambiguity is: voters who support welfare chauvinism may prioritise other competing issue preferences. They may thus vote for other parties, even if such parties do not offer welfare chauvinist programmes. From this perspective, the crucial question is: under what conditions do voters who support welfare chauvinism prefer the radical right? Among other reasons, I argue that they may do so when they experience economic risk from insecure employment contracts. Differences in employment protection legislation strictness for different employment contracts yield differences in employment security for these different contracts. Using cross-national data from the European Social Survey (Rounds 1 and 7), I find that employed workers, who support welfare chauvinism and have temporary contracts, vote most for radical right parties. I regard this finding as evidence that voters supporting welfare chauvinism prefer radical right parties under conditions of employment insecurity.


Introduction
Radical right parties are present in most West European countries today. Studies show that welfare chauvinism is now a cornerstone of these parties' electoral programmes. For Mudde (2007), welfare chauvinism is the radical right's nativist interpretation of the welfare state. Welfare chauvinism proposes that "fairly generous social benefits are to be guaranteed for the native needy (mainly pensioners and the sick), while 'aliens' are to be excluded" (132). Could the radical right's recent electoral success be tied to their nativist welfare chauvinist appeals?
Recent studies, however, offer mixed findings on the impact of welfare chauvinism against immigrants on radical right support. Goerres et al. (2018) found that voters who supported welfare chauvinist policies were consistently more likely to support the Alternative for Germany (AfD) (see also Häusermann et al. 2013). By contrast, de Koster et al. (2012) demonstrated that Dutch voters who preferred welfare chauvinist policies were not significantly more likely to vote for the Party for Freedom (PVV) in the Netherlands. In fact, the authors demonstrated that such voters were as likely to vote for radical right and leftist parties. Likewise, Rydgren (2008) found that welfare chauvinism did not consistently predict radical right support.
A plausible reason for such mixed results is: voters who prefer welfare chauvinist policies have other issue preferences. Such voters may not always prioritise their welfare chauvinist preferences. They may consider other political issues to be more salient. These voters' other issue preferences may thus influence their party choice, even if they do prefer welfare chauvinist policies. They may therefore vote for mainstream parties, even if such parties do not offer welfare chauvinist policies. Conversely, voters may choose parties based on their welfare chauvinist preferences, when welfare chauvinist issues are salient to their party choice.
From this standpoint, it is pertinent to ask: under what conditions do voters who support welfare chauvinist policies prefer radical right parties to other parties? Recent literature on economic risk shows that economically insecure individuals frequently favour radical right parties (Mayer 2015;Gidron and Hall 2020;Goerres et al. 2018). I build on these findings and posit that voters who support welfare chauvinism prefer radical right parties, when they face economic risk.
Economic risk is multifaceted and influenced by a range of factors. In this paper, I build on such studies and focus on one specific form of economic risk: employment insecurity arising from employment on insecure contracts. I argue that employment contracts are one relevant source of economic risk today, among others. This perspective ties in with recent research on labour market segmentation (Rueda 2005;Emmenegger 2009). Workers hired on different employment contracts experience diffused employment insecurity. Such differences in insecurity arise from contemporary firms' employment strategies. Contemporary firms adopt a mixed employment strategy and hire workers on permanent, temporary, and part-time contracts (Palier and Thelen 2010; Emmenegger et al. 2012). Firms do this for several reasons, of which one reason is to lower labour costs associated with dismissals. Temporary and part-time workers are cheaper and easier to dismiss than permanent workers because employment protection legislation is laxer for temporary and part-time workers. Rueda (2005) highlights succinctly that such workers tend to "act as a buffer" (61) for workers on permanent contracts by bearing the brunt of dismissals when firms seek to lower labour costs. Contemporary employment insecurity is therefore linked to employment on insecure contracts. Such developments mean that employment insecurity should no longer be solely conceptualised as being employed or unemployed.
Workers' on different employment contracts therefore face different levels of employment security (Rueda 2005). Unemployed workers clearly face the greatest employment insecurity. Among employed workers, however, workers on temporary contracts and part-time contracts are also more vulnerable to unemployment, in comparison with workers on permanent contracts. Such workers (and the unemployed) are therefore more likely to experience employment insecurity than workers on permanent contracts (Burgoon and Dekker 2010).
There is, however, mixed evidence in terms of how such different employment contracts affect party choice (Schwander 2018). King and Rueda (2008) argued that workers on insecure temporary and part-time contracts are theoretically more likely to vote for radical right parties. There is, however, limited empirical evidence which suggests that these workers support such parties (Marx and Picot 2013;Rovny and Rovny 2017). Limited empirical support for King and Rueda's (2008) proposition may arise from heterogeneity among workers on insecure contracts. As Häusermann et al. (2014) noted, the use of insecure contracts has proliferated in social groups which had hitherto enjoyed secure contracts. This means that the background of workers who have insecure contracts could vary substantially. With such heterogeneity in background, different groups of workers on insecure contracts could prefer different solutions to overcome their employment insecurity (Schwander 2018). To better assess the effects of employment contracts on party choice, it might thus be more useful to conduct comparisons within groups of workers who share similar background or policy preferences.
In this paper, I focus on a specific group of workers who share similar policy preferences in terms of welfare chauvinism. This approach builds on the study conducted by Emmenegger et al. (2015). I examine how different employment contracts affect support for radical right parties among workers who support welfare chauvinism. Specifically, I argue that voters who support welfare chauvinism are more likely to prioritise their welfare chauvinist preferences when they face substantial employment insecurity arising from their insecure contracts. This is because welfare chauvinist workers on insecure contracts face higher levels of employment insecurity, which leaves them even more concerned about welfare competition from immigrant (Van Oorschot and Uunk 2007). Such workers are thus more likely to pay attention to welfare competition and support welfare chauvinist solutions than workers who have welfare chauvinist preferences, but secure employment. They are hence more likely to vote for radical right parties. At a broader level, this paper engages both the policy preferences and labour market insider-outsider studies. It does so by studying the effect of welfare chauvinism and employment status on party choice. The rest of the paper proceeds as follows. I first review the literature on welfare chauvinism and its impact on radical right support. I next explore the relationship between economic risk and radical right support. Crucially, I argue that one form of contemporary economic risk today is employment insecurity which stems from workers' employment contracts. I then posit how different employment contracts condition welfare chauvinism's impact on radical right vote. Thereafter, I present the data and method. The final two sections discuss the results and conclude.

Welfare chauvinism and radical right support
In the context of fiscal austerity, governments are under pressure to find cost savings to ease their fiscal burden (Häusermann and Kriesi 2015). West European governments have sought to reduce social expenditure or tighten conditions attached to benefit recipiency. However, this restructuring of welfare has led to "widespread anxiety among voters who (expect to) depend on social services and transfers (the sick, the old, the unemployed) and… [have] a strong demand for political protection of social security" (Schumacher and Van Kersbergen 2016: 302). Such cutbacks have generated demands for alternative proposals on the future of the welfare state, especially as centre-left and centre-right parties seem to converge on welfare state retrenchment. In contrast to mainstream parties, radical right parties propose that "the benefits of the welfare state should be limited to the [countries'] 'own people'" (Mudde 2007: 131). By following a nativist logic, radical right parties propose to minimise the burden of welfare state retrenchment on the native population. They instead seek to restrict benefit access to non-natives, arguing that "immigrants make excessive use of the welfare state, which makes it unaffordable" (Schumacher and Van Kersbergen 2016: 300).
Some members of the public have such welfare chauvinist preferences. Studies on welfare deservingness show that the public generally considers the elderly and the sick as most deserving of welfare (van Oorschot 2006;van Oorschot and Uunk 2007). By contrast, immigrants are frequently viewed as least deserving of welfare, even less deserving than the unemployed. Van Oorschot (2006) showed that lowereducated individuals with lower socioeconomic status, and more rightist political views support welfare chauvinism more. Individuals with welfare chauvinist opinions support the notion that "welfare services should be restricted to 'our own'" (Andersen and Bjørklund 1990: 212), and consider immigrants as a potential source of competition over scarce welfare resources (Goerres et al. 2018). They prefer welfare cuts to be directed at 'undeserving immigrants', should cuts be made. From this perspective, public sentiments on welfare chauvinism relate to welfare distribution. Namely, to whom should we redistribute, and how much (Häusermann and Kriesi 2015)?
It would be logical to assume that voters with welfare chauvinist preferences are more likely to support radical right parties. Yet, recent studies yield mixed findings. Häusermann et al. (2013) suggested that pro-redistribution voters with conservative positions on social issues were more likely to support radical right parties. Similarly, Bornschier and Kriesi (2012) found that welfare chauvinism was highly correlated with voters' opinions on other cultural issues. They then showed that cultural opinions, and by extension welfare chauvinism, significantly influenced radical right vote. Furthermore, Goerres et al. (2018) demonstrated that welfare chauvinist attitudes were consistently correlated with AfD support.
By contrast, Rydgren (2008) discovered that welfare chauvinism was an inconsistent predictor of radical right support. It was also a weaker predictor than other immigrant attitudes. In addition, de Koster et al. (2012) found no significant association between welfare chauvinist opinions and radical right support. In fact, they highlighted that voters supporting welfare chauvinism did not differ significantly in their likelihood of supporting radical right, centre-left, or centre-right parties. These contrasting results prompt an examination of the direct relationship between welfare chauvinist preferences and radical right support.

H1(a)
Voters who prefer welfare chauvinist policies are significantly more likely to vote for radical right parties.

H1(b)
Voters who prefer welfare chauvinist policies are not significantly more likely to vote for radical right parties.
These contrasting results may seem puzzling because parties are unlikely to adopt programmatic stances which do not appeal to segments of their electorate. The same logic applies to radical right parties. Scholars have shown that radical right parties campaign on welfare chauvinism to attract voters who have welfare chauvinist opinions (Andersen and Bjørklund 1990;Schumacher and Van Kersbergen 2016). They adopt welfare chauvinist positions to fill a gap in political representation-voters who support redistribution, but are conservative on social issues such as immigration (Lefkofridi et al. 2014).
There is one plausible way to square this circle. Voters who support welfare chauvinism could be more concerned about other issues when they vote. Put differently, they could find other issues more salient, even if they prefer welfare chauvinist policies. If other issues are more salient, they could vote for other parties such as centreleft and centre-right parties which better address these issues (de Koster et al. 2012). They may do so, even if other parties do not have policies which meet their welfare chauvinist preferences. Conversely, if welfare chauvinist issues are more salient, voters who support welfare chauvinism are more likely to vote for radical right parties. They could prioritise their welfare chauvinist preferences over other competing preferences. This conjecture therefore raises an important question: under what conditions do voters who support welfare chauvinism choose radical right parties over other parties? I posit that voters who support welfare chauvinism are more likely to choose radical right parties, when they face substantial economic risk arising from insecure employment.

Economic risk and radical right vote
A burgeoning number of studies provide evidence that economic risk is linked to contemporary radical right support (see Mayer 2015;Goerres et al. 2018). They find that economically insecure individuals vote most for radical right parties (Gidron and Hall 2017;Goerres et al. 2018). For instance, Im et al. (2019) showed that economically insecure workers who are vulnerable to automation are significantly more likely to vote for radical right parties in Europe.
Such voters could support radical right parties more because they face material deprivation or status decline (Engler and Weisstanner 2019). Studies on material deprivation emphasise the impact of economic risk on individuals' economic wellbeing (Colantone and Stanig 2018). When individuals face economic risk, they are more likely be unemployed. Unemployment frequently entails a substantial loss of income which renders it difficult to make ends meet. Put differently, economically insecure individuals vote for radical right parties more because they fear the material effects of unemployment.
On the other hand, other studies focus on the impact of economic risk on status decline (see Mayer 2015;Im et al. 2019). They highlight that unemployment results in status decline. Ezzy (1993) noted that workers undergo a 'status passage' when they become unemployed. They go from being 'valued workers' to 'unvalued unemployed'. Unemployment also leads to income loss which could cause status decline. Economically insecure individuals therefore vote for radical right parties because they fear status decline arising from unemployment. Put differently, economically insecure individuals vote for radical right parties because they fear being left behind (Gidron and Hall 2017: 61).
Regardless of the mechanism, there seems to be ample theoretical support for the core argument of this paper: voters who support welfare chauvinism are more likely to support radical right parties, when they face economic risk. But in the context of economic transformations and labour market disruptions, what types of economic risk do such voters respond to?

Different employment contracts as a potential source of economic risk
Economic risk is multifaceted and is influenced by a range of factors (Im et al. 2019). In this paper, I focus on one specific type of economic risk: employment insecurity arising from employment on insecure contracts. This form of economic risk begins with firms' employment strategies.
Firms today adopt a mixed employment strategy. They hire workers (a) on permanent contracts, (b) temporary contracts, and (c) part-time contracts (Rueda 2005;Emmenegger 2009; Palier and Thelen 2010). Firms utilise this mixed employment strategy to lower overall labour costs. Labour costs vary across employment contracts for several reasons. One reason pertains to differences in employment protection legislation (EPL). EPL strictness varies across different employment contracts. EPL is generally stricter for workers on temporary and part-time contracts than for workers on permanent contracts (Palier and Thelen 2010). These differences mean that it is easier and cheaper to dismiss temporary and part-time workers than permanent workers. In other words, such differences lower the dismissal costs of workers on temporary and part-time contracts, relative to workers on permanent contracts. Whenever firms seek to lower labour costs, firms would probably be more inclined to dismiss temporary and part-time workers because they are cheaper and easier to fire. Rueda (2005) aptly summarises that such workers tend to "act as a buffer" (61) for workers on permanent contracts by bearing the brunt of dismissals when firms seek to lower labour costs. In this way, different employment contracts give rise to varying levels of employment insecurity and is hence one type of economic risk today (Rueda 2014).
Such differences in dismissal costs may persist, even after accounting for other contributing factors such as the economic performance of workers' employment sectors. This is because EPL frequently favours the employment interests of permanent workers, even when economic conditions are poor (Rueda 2014). Consequently, dismissal costs for permanent workers generally remain higher than dismissal costs for temporary and part-time workers, even if economic conditions are poor. As a result, workers on insecure temporary and part-time contracts are arguably more likely to be fearful that unemployment could always be just around the corner, because they are easier and cheaper to fire. They may also be more likely to face employment insecurity, even if they are employed in booming sectors. Put simply, they face a greater and more salient threat of dismissal than permanent workers. Temporary and part-time workers hence face greater and more consistent insecurity than permanent workers. Burgoon and Dekker (2010) stressed this point when they showed that temporary and part-time workers feel a strong sense of economic insecurity. 1 There is, however, mixed evidence in terms of how different employment contracts affect party choice, especially for radical right parties (Schwander 2018). King and Rueda (2008) argued that workers on insecure contracts are more likely to vote for protest parties, including radical right ones. They suggested that workers on insecure contracts have different economic interests from workers on secure contracts. When centre-left parties cater to the interests of workers on secure contracts, workers on insecure contracts may turn away from such parties to support protest parties such as radical right ones. There is, however, limited evidence supporting King and Rueda's claims (Marx and Picot 2013). For instance, Rovny and Rovny (2017) found that workers on insecure contracts are less likely to support centre-right parties, but are more likely to support radical left parties. They, however, find that workers on temporary and part-time contracts do not vote for radical right parties significantly more than workers on permanent contracts. Based on these studies, I express a second set of hypotheses relating employment insecurity to radical right support.

H2(a)
Workers on insecure temporary and part-time contracts are more likely to vote for radical right parties.

H2(b)
Workers on insecure temporary and part-time contracts are not more likely to vote for radical right parties.

The conditional impact of employment contracts on welfare chauvinist preferences and radical right support
Studies on the political consequences of labour market segmentation may have found mixed results, because workers on insecure temporary and part-time contracts are heterogeneous. As Häusermann et al. (2014) noted, non-standard employment (temporary or part-time contracts) is increasingly prevalent in social groups which had hitherto enjoyed secure standard employment (permanent contracts). The authors demonstrated that employment on such insecure contracts has increased among higher-educated and higher-skilled workers. In other words, it is likely that temporary and part-time workers have diverse backgrounds. They may thus have diverse interests and preferences. Even if they have a shared interest in alleviating their employment insecurity, they may be divided in terms of their preferred solutions to overcome their common predicament (Schwander 2018). Different groups of workers on insecure contracts may prefer different political solutions to their employment insecurity. They may hence vote for different parties.
Conversely, this conjecture means that we should not rule out the effect of employment contracts on support for radical right parties. Instead, it suggests that comparing the effects of different employment contracts within similar groups is more appropriate, such as among workers who have similar political preferences. For example, workers' support for radical right parties could vary according to their employment contracts, even if they share similar welfare chauvinist preferences. In other words, the effect of employment status on radical right support may only manifest within groups with similar political preferences.
In an earlier section, I posited that voters who support welfare chauvinism may vote for radical right parties when they prioritise their welfare chauvinist preferences over other issue preferences. This may occur when such voters experience elevated levels of economic risk. Among workers, elevated levels of economic risk may arise from insecure employment contracts, all things equal. Put differently, workers may prioritise their welfare chauvinist preferences, when they face employment insecurity arising from their temporary and part-time contracts.
Existing studies show that individuals in socially and economically risky positions may prefer more restrictive welfare allocation (Van Oorschot 2006). Van Oorschot and Uunk (2007) demonstrated that they are more likely to prefer "stricter conditionality in welfare rationing to prevent social protection from being used by competing groups" (65). They may prefer restrictive welfare allocations because they need social protection when they become unemployed. Social protection alleviates income loss during unemployment. Because immigrants are more frequently non-employed and more reliant on social welfare than the indigenous population, individuals in economically risky positions may be more concerned about welfare competition from immigrants (van Oorschot and Uunk 2007). They could therefore prefer more restrictive allocation against immigrants.
In a similar vein, employment insecurity arising from different employment contracts may have comparable effects on workers who prefer welfare chauvinism. Put differently, employment insecurity could magnify such workers' existing support for welfare chauvinism. To begin, workers supporting welfare chauvinism are already more convinced of welfare competition from immigrants. Employment-insecure welfare chauvinists could, however, be even more concerned about welfare competition from immigrants. They could be more concerned, because they would be more reliant on welfare to mitigate the effects of income loss during unemployment. By contrast, workers who support welfare chauvinism may place less importance on their welfare chauvinist preferences when they face less employment insecurity. They are less likely to rely on welfare because they are less likely to experience unemployment. Employment-insecure workers supporting welfare chauvinism could therefore be more willing and ready to support welfare chauvinist solutions for their economic malaise.
In terms of party choice, workers who support welfare chauvinism, and experience employment insecurity, could therefore be more attracted to the radical right's appeals. This is because workers on insecure temporary and part-time contracts with welfare chauvinist preferences could find welfare issues more salient than workers on secure permanent contracts with welfare chauvinist preferences. The former is hence more likely to base their party choice on their welfare chauvinist opinions. Economically insecure workers supporting welfare chauvinism could thus support radical right parties' welfare chauvinist policies more because they have a stronger and more pressing desire to reduce welfare competition.
H3 Workers with welfare chauvinist preferences are more likely to vote for radical right parties, when they are on insecure temporary and part-time contracts.

Data and method
Data I used cross-sectional data from the European Social Survey (ESS) Rounds 1 and 7. 2 It is a biennial cross-national survey, and it maintains permanent modules on respondents' sociodemographic background and party choice. It also contains rotating modules on different policy issues. I chose Rounds 1 and 7 because these are the only rounds which contain a variable asking respondents' opinions on welfare chauvinism in the form of welfare competition. I chose the following countries: Austria, Belgium, Switzerland, Germany, Denmark, Finland, France, Netherlands, and Norway. I chose these countries on two conditions: (a) having electorally successful radical right party(ies), (b) and are represented in both Rounds 1 and 7. The final sample consists of 8740 observations, after omitting observations with missing values on the dependent variable and covariates.

Variables
I used the variable imbleco to construct respondents' opinions on welfare chauvinism. It is a continuous variable ranging from 0 to 10. Higher values indicate greater support for the notion that immigrants consume welfare more than they contribute to it. I consider respondents reporting higher values on this variable to prefer welfare chauvinist policies and are more concerned about welfare competition from immigrants.
To measure employment insecurity arising from insecure contracts, I adapted Emmenegger's (2009) classification system. Emmenegger distinguishes between upscales, workers on permanent contracts, temporary contracts, part-time work, and the unemployed. Likewise, I followed his classification of employment contracts. Upscales are workers who are in "privileged positions in the labour market" (133) and need not worried about unemployment. Upscales are coded as employees belonging to European Socio-economic Classification (ESeC) category 1: large employers, higher managers, and professionals with permanent contracts. 3 Workers on permanent contracts are employees drawn from all other ESeC categories who are not on fixed-term contracts, or in part-time work. Workers on temporary contracts are employees on fixed-term contracts. Part-time workers are employees working for less than 30 hours a week (Rovny and Rovny 2017). The unemployed are individuals who are actively or inactively looking for work in the past 7 days. I excluded the self-employed from this variable because I am primarily interested in labour market participants who are in paid employment. Finally, I included a category of non-employed. They are non-labour market participants who are either studying, retired, disabled or sick, or doing housework. This category merely serves as a benchmark to compare results of labour market participants with non-labour market participants. In sum, this variable consists of-(a) upscales (reference category), (b) workers on permanent contracts, (c) workers on temporary contracts, (d) part-time workers, (e) the unemployed, (f) the non-employed. Respondents are assigned exclusively to one of these categories.
In the labour market segmentation literature, there are four commonly used measures of employment insecurity arising from differences in employment. According to Rovny and Rovny (2017), these measures are based on two conceptualisations: one set directly measures current employment contracts and statuses (Rueda 2005;Emmenegger 2009), the other measures individuals' occupational risk of falling into non-standard employment (Rehm 2009;Schwander and Häusermann 2013). The focus of this paper is workers' employment insecurity arising from their employment on insecure contracts. The core argument is: workers' current employment contracts influence how their welfare chauvinist preferences affect radical right support. Conversely, when their contracts change, their employment insecurity may also change. The theoretical approach of this paper means that a snapshot measure of employment insecurity is more appropriate than a risk-based measure, especially when the latter calculates workers to have similar levels of employment insecurity regardless of their actual current employment status and contracts. I therefore operationalised my variable to measure current employment contracts and statuses based on Emmenegger's (2009) classification. I preferred Emmenegger's because it is more nuanced than Rueda's (2005).
I included sociodemographic covariates commonly used in similar studies. Individual-level controls include age, gender (reference: male), education in years, religiosity, union membership (reference: not union member), if one belongs to a minority group (reference: belong to minority group), self-reported position on the left-right scale. Household-level controls include household income, respondents' marital status (reference: not married), if they have children at home (reference: no child at home), and if their partner is in paid employment (reference: partner not in paid work). I controlled for respondents' partners being in paid employment because employment-insecure workers could be less insecure in dual-earner households than in single-earner households.
My dependent variable is respondents' vote in the last national elections. I recoded parties into three party families based on Rovny and Rovny's (2017) study: centre-left, centre-right, and radical right (reference outcome). I only included these party families because they are generally the most electorally successful today. I excluded respondents who did not vote because studies show that vote abstention is largely determined by factors unrelated to issue opinions (Brady et al. 1995).

Method
I conducted the analyses using multinomial probit regression. I used a pooled model with country-clustered standard errors and country fixed effects because there are only 9 countries. I preferred a pooled model because Bryan and Jenkins (2015) highlighted that a minimum of 27 level-2 cases is needed to reliably estimate the standard errors of level-2 parameters in a multilevel model. I also included yeardummies to account for differences between Rounds 1 and 7, and design weights.
I performed stepwise analyses using 4 models. The first model consist of controls to estimate their effects on party choice. The second model includes respondents' support for welfare chauvinism. The third model adds different employment contracts and statuses. The final model includes an interaction term composed of respondents' support for welfare chauvinism and their employment contracts and statuses. The original terms remained in the model together with their interaction term (Golder 2003). Table 1 presents the mean and standard deviations for welfare chauvinist preferences for different parties' electorate. It shows that voters of radical right parties generally prefer welfare chauvinist policies. Interestingly, voters of the centre-left and centre-right also prefer welfare chauvinist policies, on average. The standard deviations of centre-left and centre-right parties nevertheless reveal that these parties' electorates are divided between voters who prefer welfare chauvinist policies, and those who do not. Descriptively, it would seem that supporters of welfare chauvinism are as likely to vote for centre-left, centre-right, and radical right parties (de Koster et al. 2012). Put simply, voters supporting welfare chauvinism may not automatically prefer radical right parties. Rather, they do so when they prioritise their welfare chauvinist preferences. They may do so when they face economic risk in the form of employment insecurity arising from insecure contracts.  Table 2 presents results from the regression analyses. I turn first to Model 1 which consists only of control variables. Younger, less-educated, less-religious, nonethnic minority individuals with rightist political views are significantly more likely to vote for radical right than centre-left parties. Likewise, younger, less-educated, less-religious, unmarried men in lower income households whose partners are not in paid work are significantly more likely to vote for radical right than centre-right parties. Most of these associations remain significant after adding welfare chauvinist preferences, employment insecurity, and their interaction term.

Results
In Model 2, I added individuals' support for welfare chauvinism to the model. I find that voters who support welfare chauvinism are significantly more likely to vote for radical right than centre-left and centre-right parties. These associations remain significant even in Model 3 which includes differences in employment contracts.
Model 3 shows the direct effects of different employment contracts. Workers on permanent and temporary contracts, part-time workers, the unemployed, and the non-employed are all significantly more likely to vote for radical right than centreleft parties in comparison with upscales. Similarly, workers on permanent, or temporary contracts, in part-time work, or unemployed are all significantly more likely to vote for the radical right than the centre-right, in comparison with upscales. Among the employed, workers on temporary contracts have the biggest coefficient, which could indicate that individuals facing higher employment insecurity are more likely to vote for radical right than centre-left and centre-right parties.
Model 4 includes an interaction term composed of welfare chauvinist preferences and employment insecurity. As interaction terms are challenging to interpret, Table 2 Regression results  Table 2 (continued) Fig. 1 illustrates the results. It is based on a modified version of Model 4 (see Online Appendix Table A3). I recoded welfare chauvinist preferences into a binary variable for easier interpretation. Presenting partial effects of employment insecurity at different levels of welfare chauvinist support (as a continuous variable) would be graphically overwhelming and difficult to interpret. I recoded welfare chauvinist preferences such that respondents with values above 5 are considered supporters of welfare chauvinism, and vice versa. Figure 1 presents discrete effects for a change in type of employment contract, conditional on individuals' support for welfare chauvinism. In other words, it presents the level of change in vote probability between upscales and other employment contracts. All other covariates are held at mean values. Figure 1 shows that differences in employment insecurity generally do not have significant impact on centre-left and centre-right support, regardless of welfare chauvinist support. By contrast, employment insecurity significantly moderates the impact of welfare chauvinist support on radical right vote. Unemployed voters who support welfare chauvinism are more likely to support radical right parties than upscale ones by 14.8% points. Temporary workers who support welfare chauvinism are more likely to vote for radical right parties than upscales who support welfare chauvinism by 9.2% points. By contrast, the difference in support is smallest between upscale and part-time workers who support welfare chauvinism at 5.5% points.
It is also interesting to examine how non-employed voters who support welfare chauvinism vote for radical right parties relative to other groups. Non-employed voters who support welfare chauvinism are 7.9% points more likely to vote for radical Fig. 1 Conditional effects of employment contracts and employment status on party choice right parties than upscale workers who support welfare chauvinism. Non-employed is a crude category consisting of all non-labour market participants. Nevertheless, it serves as a benchmark to examine if certain groups of voters are more likely to vote for radical right parties than the rest of society. Temporary workers and the unemployed who support welfare chauvinism have higher propensities of voting for radical right parties than similar non-labour market participants. By contrast, permanent, part-time, and upscale workers who support welfare chauvinism have lower propensities of voting for radical right parties than similar non-labour market participants. I checked if these results are sensitive to attitudes towards immigration, satisfaction with democracy, unemployment rate, and migration inflows (Golder 2003). There are no substantive differences (see Online Appendix Tables A5 and A6). Figure 2 presents predicted probabilities of voting for radical right parties across different levels of support for welfare chauvinism and for different employment contracts. It is based on Model 4 and uses the original welfare chauvinism variable as a continuous variable. Figure 2 reiterates results in Fig. 1. Among voters who support welfare chauvinism, the unemployed and temporary workers have highest propensities of voting for radical right parties. Among voters who strongly support welfare chauvinism (welfare chauvinism = 10), temporary workers are 3.1% points more likely to vote for radical right parties than the unemployed.
By contrast, upscales who support welfare chauvinism are least likely to vote for radical right parties. Permanent and part-time workers who support welfare chauvinism are more likely to vote for radical right parties than their respective counterparts who do not support welfare chauvinism. But they are substantially less likely to vote for radical right parties in comparison with temporary workers and the unemployed who support welfare chauvinism. Permanent and part-time workers who strongly support welfare chauvinism are 11.3 and 12.1% points less likely to vote for radical right parties than temporary workers who have similar support for welfare chauvinism. They are also 8.2 and 9.0% points less likely to vote for radical right parties than unemployed supporters of welfare chauvinism, respectively. Figure 3 demonstrates that supporters of welfare chauvinism generally turn away from centre-left parties in support for radical right parties. Crucially however, temporary workers who support welfare chauvinism are least likely to support centre-left parties, and most likely to vote for radical right parties. Figures 1, 2, and 3 cumulatively show that employment insecurity, in the form of employment contracts, conditions the impact of welfare chauvinist preferences on radical right support. Namely, workers who support welfare chauvinism and face greater employment insecurity are more likely to vote for radical right parties.

Discussion and conclusion
While voters who support welfare chauvinism may find radical right parties' welfare chauvinist programmes appealing, they may not automatically vote more for such parties. This is because they may prioritise their other issues preferences over their welfare chauvinist ones. They may therefore vote for other parties which appeal to their other issue preferences, even if such parties do not have welfare chauvinist programmes. This shift could explain why recent studies yield mixed findings on the impact of welfare chauvinism on radical right support (de Koster et al. 2012).
I thus argue that voters who prefer welfare chauvinism may support radical right parties more, when they consider their welfare chauvinist preferences to be salient. Based on existing studies, they may do so when they face economic risks (Van Oorschot 2006; Engler and Weisstanner 2019). Economic risks are however multifaceted and arise from different sources (Im et al. 2019). It therefore begs the question: what types of risks do voters respond to? In this paper I focus on one form of economic risk: employment insecurity arising from insecure employment contracts. This is a relevant and pertinent form of economic risk today because firms use mixed employment strategies to lower labour costs, including those arising from dismissals (Rueda 2005;Palier and Thelen 2010;Schwander 2018). Different types of contracts therefore yield different levels of employment security (Burgoon and Dekker 2010).
I therefore hypothesised that workers who support welfare chauvinism are more likely to consider their welfare chauvinist preferences salient, when they are employed on insecure temporary and part-time contracts. They are thus more likely to support radical right parties over other parties, such as centre-left and centre-right parties. The results here partially support this hypothesis. Among voters who support welfare chauvinist policies, the unemployed and temporary workers have the greatest propensities of voting for radical right parties. Among employed workers, temporary workers who support welfare chauvinism have the highest propensity of voting for radical right parties. Their higher propensity could reflect their elevated concerns about welfare competition, especially when they face the threat of unemployment (Van Oorschot 2006).
By contrast, permanent and upscale workers who support welfare chauvinist policies are less likely to vote for radical right parties. Surprisingly however, part-time workers who support welfare chauvinist policies are even less likely to vote for radical right parties than similar permanent and upscale workers who support welfare chauvinist policies. Since part-time workers are theoretically more economically vulnerable than permanent workers, one would expect part-time workers who support welfare chauvinist policies to vote for radical right parties more than permanent workers who similarly support welfare chauvinist policies. One plausible reason for the findings here is that part-time workers are heterogeneous. In fact, Emmenegger (2009) highlights that some workers actively choose to downshift into flexible parttime jobs. Some part-time workers may therefore be more aware of employment insecurities associated with part-time work than other part-time workers. Part-time workers who downshifted willingly into such jobs and support welfare chauvinism could therefore be more accepting of employment insecurity arising from such contracts than temporary workers who support welfare chauvinism. They could hence be less concerned about welfare competition from immigrants, even if they have "insecure" part-time jobs, and thus vote less for radical right parties. Conversely, part-time workers who downshifted involuntarily into such jobs and support welfare chauvinism could be as unaccepting of employment insecurity arising from such contracts as temporary workers who support chauvinism. They could hence be as concerned about welfare competition from immigrants and have similar propensities of voting for radical right parties as temporary workers who also support welfare chauvinist policies. Likewise, this group of part-time workers may also support radical right parties more than permanent workers who also support welfare chauvinist policies. This dataset unfortunately does not allow one to distinguish between these two groups of part-time workers. The results here thus reflect the net voting behaviour of these two groups of part-time workers. That is to say, these two groups' propensities of voting for radical right parties may cancel each other out. It is thus important to acknowledge that one cannot decisively conclude whether parttime workers who support welfare chauvinist policies vote for radical right parties as much as similar temporary workers, or less in this study. It may hence be fruitful for future research to distinguish various types of part-time workers and explore their voting behaviour.
Overall, these findings lend support to the main hypothesis of this paper, which stresses the conditional impact of employment insecurity (as employment contracts) on welfare chauvinism and radical right support. The findings here partially contrast with studies which find limited support for King and Rueda's (2008) proposition that workers on insecure contracts vote more for radical right parties (see for example Rovny and Rovny 2017;Schwander 2018). I find that workers on insecure temporary contracts vote more for radical right parties, when they support welfare chauvinist policies. This paper therefore brings together the effect of policy preferences (welfare chauvinism) and employment status.
At a broader level, these findings highlight that we should not only focus on the direct effects of economic transformations on party choice (Colantone and Stanig 2018;Im et al. 2019). Owing to firms' employment strategies, the disruptive effects of automation and job offshoring may vary for workers on different contracts. For instance, it may be cheaper to replace temporary workers than permanent ones. Such economic transformations may thus have varying effects on party choice for workers who are employed on different contracts. These effects could also be magnified or minimised depending on how mainstream parties adapt to these growing concerns about welfare competition and employment insecurity (Schumacher and Van Kersbergen 2016;Schwander 2018). This paper therefore reinserts the importance of understanding how labour market segmentation is the interface through which economic transformations may influence party choice. Future studies could thus focus on how economic transformations affect workers who have similar backgrounds and political preferences, but differ in their employment contracts.