Abstract
A considerable portion of European citizens are in favour of limited or conditional access for migrants to welfare provisions. Previous studies found that this welfare chauvinism is stronger among citizens with less favourable economic positions. This study seeks to explain the relationship between economic risk, both objective and subjective, and welfare chauvinism by looking at two distinct mechanisms: the traditional economic explanation of economic egalitarianism and the cultural explanation of ethnic threat. Given the lack of longitudinal studies, we also examine whether changes in economic risk, economic egalitarianism and threat can explain changes in welfare chauvinism over time. Using a four-wave panel-study (2013–2015) collected in Great Britain and the Netherlands, these relationships were studied both cross-sectionally and longitudinally. The longitudinal mediation model was tested by making use of parallel process latent growth curve modelling. In both Great Britain and the Netherlands, economic egalitarianism and ethnic threat explained the link between economic risk and welfare chauvinism. Furthermore, in both countries, an increase over time in perceptions of ethnic threat was found to be the driving force behind an increase in welfare chauvinism, irrespective of changes in economic egalitarianism.
Introduction
Even though a strong welfare system remains a defining characteristic of most European countries, there are growing concerns about the future of the welfare state (Taylor-Gooby, 2011). It is argued to be challenged by individualization (Giddens, 1994), declining economic growth (O’Connor, 2006), and increasing cultural heterogeneity (Alesina and Glaeser, 2004; Schmidt-Catran and Spies, 2016). Additionally, there are concerns that the welfare debate is becoming ‘racialized’ as welfare programmes are more and more associated with poor immigrants (Harell, Soroka and Ladner, 2014).
Arguably, these developments are reflected in the extent to which European citizens support welfare chauvinism and disagree with one another about the conditions under which immigrants should have access to welfare benefits (Van Oorschot, 2006). Immigration and the subsequent discussion about immigrants’ access to the welfare state have been suggested to pose a challenge to the social cohesion of contemporary European societies, by eating away at popular support for redistribution in general and therefore threaten the public legitimacy of the welfare state (Eger, 2010).
Traditionally, attitudes towards the welfare state are explained by economic self-interest and the ideology of economic egalitarianism (Pierson, 2001). In addition, the preference for an exclusive and chauvinistic welfare state is most often and most strongly expressed by people who find themselves in the least favourable economic situations (Achterberg and Houtman, 2006). This could be because they are more likely to feel threatened by immigrants (Coenders and Scheepers, 2008). Interestingly, these are in turn also the people that, from the perspective of economic self-interest, tend to be more supportive of a generous welfare state. In short, people who are economically vulnerable could be expected to be more welfare chauvinistic because they experience more threat from immigrants, but they could also be expected to be less chauvinistic because they are typically more egalitarian.
Despite this tension among a considerable part of the populace and the relevance of chauvinistic welfare attitudes for the legitimacy of the welfare state (Van der Waal et al., 2010), very little research has been devoted to the relative importance of economic and anti-immigrant incentives in explaining welfare chauvinism. The first aim of this study is therefore to examine how welfare chauvinism is rooted in individual’s subjective and objective economic risk, and whether this can be explained by the endorsement of economic egalitarianism and ethnic threat perceptions.
Theories on welfare chauvinism often have a dynamic character: changes in individuals’ economic positions, preferences and feelings of ethnic threat are assumed to be the drivers behind changes in welfare chauvinism. Yet, there is a lack of longitudinal studies. The second aim of this study is therefore to investigate changes in welfare chauvinism over time. While cross-sectional studies can explain why certain individuals are more chauvinistic than others, they don’t further our knowledge about why a given individual becomes more or less chauvinistic over time. This is unfortunate since scholars and politicians are particularly concerned about the possibility of increasing welfare chauvinism and declining support for and legitimacy of the welfare state. Using a comparative four-wave panel study in two countries, we therefore also examine whether changes in welfare chauvinism can be explained by changes in subjective and objective economic security, economic egalitarianism, and ethnic threat. This also allows us to control for unobserved confounders common in cross-sectional research and to investigate whether these relationships are merely correlational or causal (Finkel, 1995). All in all, we test whether the same theoretical mechanisms can explain differences in welfare chauvinism between individuals (contemporaneous) as well as within individuals over time (longitudinal).
Specifically, this study looks at changes in welfare chauvinism between 2013 and 2015 in Great Britain and the Netherlands. These two countries and the specific timespan are interesting for several reasons. For one, recent research suggests that welfare attitudes are especially likely to fluctuate over short time periods (Jeene, Van Oorschot and Uunk, 2014). Further, expressions of anti-immigrant attitudes and sentiments have been common in both countries. This is reflected in the rise in right-wing extremist voting. Additionally, during the period of data collection, the media in both countries critically evaluated Europe’s migration policy, spoke of a refugee crisis and meticulously covered its effects on both countries (De Volkskrant, 2015; The Guardian, 2015). Research has shown that both the frequency and the tone of media coverage of immigration influences changes in anti-immigrant attitudes (Boomgaarden and Vliegenthart, 2009). The media coverage reflects the increase in the influx of refugees and asylum seekers into Europe, which rose by a third during 2013 and by another 47% in 2014 (UNHCR, 2015). Especially such a sudden increase in immigration has been shown to be related to resistance to foreigners (Coenders and Scheepers, 2008), particularly when national rhetoric reinforces the idea that immigration is a threat (Hopkins, 2010).
Great Britain and the Netherlands are also similar in the sense that they have been recovering from the economic crisis of 2008. At the same time, the two countries differ in welfare regime type, with Great Britain being more liberal than the Netherlands (Esping-Andersen, 1990). It is therefore interesting to see whether similar economic and demographic developments have impacted welfare chauvinism in the same way in both countries.
Theory
Welfare Chauvinism
The welfare state came into being by virtue of the national community and the accompanying mutual obligation between people belonging to one nation. However, a homogenous national community is becoming less evident, given the growing diversity in many European societies. Immigrants are often considered ‘outsiders’ of the national community and are therefore more likely to be excluded from welfare support. Andersen and Bjørklund (1990, p. 212) coined the term welfare chauvinism to describe the idea that ‘welfare services should be restricted to our own [people]’. In other words, welfare chauvinism denotes an attitudinal preference for an exclusionist welfare state that is defined by ethnic group boundaries. In the current study, we look at welfare chauvinism in the soft rather than strict sense (Reeskens and Van Oorschot, 2012). That is, while a comparatively small portion of the European public would exclude immigrants from welfare benefits all together, and thus support strict welfare chauvinism, most people would allow immigrants conditional access to the welfare state, and support a soft form of welfare chauvinism. One of these conditions is, for example, that immigrants should first work and pay taxes in the host country. Finally, empirical research suggests that welfare chauvinism is most strongly expressed by people who are in a less economically secure position (Achterberg and Houtman, 2006).
Economic Risk and Ethnic Threat
Ethnic competition theory can help shed light on the relationship between objective and subjective economic risk and welfare chauvinism. This theoretical framework has often been applied to explain negative attitudes towards immigrants. The theory integrates realistic group conflict theory and social identity theory (Scheepers, Gijsberts and Coenders, 2002). The former argues that competition over scarce resources between social groups results in a conflict of interest that has a negative effect on intergroup relations (LeVine and Campbell, 1972). The latter seeks to explain negative outgroup attitudes by means of the social-psychological mechanisms of social identification and the general tendency to perceive of one’s ingroup in a favourable way (Tajfel and Turner, 1979). Ethnic competition theory postulates that the process of social identification is intensified under conditions of intergroup competition (Scheepers, Gijsberts and Coenders, 2002). This competition may revolve around either scarce material and immaterial resources, such as jobs, housing, political power, and cultural values (Olzak, 1992).
One prominent hypothesis that is derived from this theory is that especially people in low economic standing are in competition with and feel threatened by immigrants, due to their similar social positions (Scheepers, Gijsberts and Coenders, 2002). Despite being put to the test by an ever growing body of research, this proposition has received mixed results. On the one hand, various studies found that ethnic threat perceptions are more common among people with lower income and lower occupational status, such as manual workers and unemployed (Savelkoul et al., 2011; Schneider, 2008). On the other hand, in a recent meta-analysis conducted by Hainmueller and Hopkins (2014), it was concluded that actual labour market competition is not related to anti-immigrant attitudes. Natives who are in competition with immigrants on the labour market, due to having the same skill-set or working in the same industry, are generally not more negative towards immigrants. In addition, Kuntz, Davidov and Semyonov (2017) showed that unemployed people do not perceive immigrants to be a threat, while people who perceive their economic position to be vulnerable do feel threatened by immigrants. Combined, these latter studies suggest it might be less relevant whether people are objectively at economic risk or in competition with immigrants, but rather that subjective perceptions of economical vulnerability induce perceptions of immigrant threat.
Following these lines of research, we test whether we can replicate the distinct findings on objective and subjective economic risk in relation to ethnic threat perceptions in the case of welfare chauvinism. Applying the distinction between objective and subjective economic risk to the issue of welfare chauvinism specifically, we focus on people who actually receive welfare benefits themselves (as an objective indicator of risk) and those who perceive income insecurity (as a subjective indicator), as these groups might feel they have more to lose from allowing immigrants wider access to the welfare state.
Subsequently, it can be argued that people who experience more threat from immigrants have stronger exclusionist attitudes (Schneider, 2008). This perceived threat can be both economic and cultural. In fact, sociotropic concerns about the impact of immigrants on the nation’s culture and economy from one of the strongest predictors of anti-immigrant attitudes (Hainmueller and Hopkins, 2014). People who perceive immigrants to be an economic threat may be more reluctant to allow immigrants to make use of welfare benefits in order to protect the economic longevity of the welfare state. Further, it has been argued that the welfare debate is becoming ‘racialized’ as welfare programmes are more and more associated with poor immigrants (Harell, Soroka and Ladner, 2014). Consequentially, concerns about cultural differences between natives and immigrants become more relevant in relation to welfare spending. People who perceive immigrants to be a cultural threat to their country may be more likely to want to exclude them from using ‘their’ welfare benefits. More generally, it can be argued that people who feel threatened by immigrants, economically and culturally, will be less willing to include them in the welfare system.
In sum, economic risk is expected to be positively related to perceptions of ethnic threat (Hypothesis 1a), which are in turn expected to be positively associated with welfare chauvinism (Hypothesis 1b).
Economic Risk and Economic Egalitarianism
Most research seeking to explain the connection between individuals’ economic position and welfare attitudes focuses on economic egalitarianism: the ideological belief concerning the preferred or tolerated degree of economic differences in society (Svallfors, 1991; Gilens, 1995).
One’s economic standing has often been found to be directly related to one’s ideological stance about economic egalitarianism (Van der Waal et al., 2010). Based on the principle of economic self-interest, it can be argued that especially people who are economically vulnerable tend to be more egalitarian, because it holds the promise of becoming economically better off (Lipset, 1963). People who are worried about their own economic position will be more likely to want to off-set the economic risk they face by supporting a more egalitarian society (Naumann, Buss and Bähr, 2015). Conversely, when people’s economic situation improves, they have less to fear from income inequalities. Thus people with higher economic standing are typically more supportive of market liberalism. That said, it must also be noted that people’s economic ideology can operate independent of their own economic position (Andreß and Heien, 2001), and that, for example, high-earners can also support a more generous form of redistribution (Fong, 2001).
Subsequently, people who are more egalitarian can also be argued to be less chauvinistic. Egalitarianism, following Frankfurt (1987, p. 21), can be defined as ‘the doctrine that it is desirable for everyone [emphasis added] to have the same amounts of income and of wealth’. Thus, a priori, egalitarianism doesn’t distinguish between natives and immigrants. Welfare chauvinism, on the other hand, explicitly puts immigrants apart and restricts their access to welfare benefits (Gelissen, 2000; Mewes and Mau, 2013). From an ideological standpoint, such an exclusionist welfare state would go against the very nature of economic egalitarianism. For egalitarian people, a fair society is exemplified by small differences in standards of living and income, irrespective of the ethnicity of its citizens (Pierson, 2001).
To be clear, we do not argue that welfare chauvinists cannot also prefer a strong, generous welfare state. We merely argue that egalitarian people are less likely to want to exclude immigrants from access to the welfare state, irrespective of how generous the welfare system is. Put differently, welfare chauvinists can be in favour of a big pie—but one that is mostly shared by natives. There is evidence that the willingness to exclude immigrants gets larger as the pool of welfare resources becomes scarcer (Reeskens and Van Oorschot, 2012).
All in all, economic risk is expected to be positively related to the endorsement of economic egalitarianism (Hypothesis 2a), which is in turn expected to be negatively related to welfare chauvinism (Hypothesis 2b).
Taken together, the hypotheses put forward in the theory section reflect two distinct and opposing links between economic risk and welfare chauvinism. On the one hand, people who are economically vulnerable might feel threatened by immigrants, which would lead to a more exclusive stance towards granting welfare rights to immigrants. On the other hand, based on self-interest, people who are economically insecure might support economic egalitarianism. Based on this ideological position, they would be less chauvinistic in their welfare attitudes.
To investigate these two opposing mediating mechanisms, we first test a contemporaneous mediation model to analyse differences between individuals and to examine whether the relations between subjective and objective economic risk and welfare chauvinism can be explained by ethnic threat and economic egalitarianism. Second, taking advantage of our longitudinal panel data, we analyse changes over time by means of latent growth curve models to test to what extent we find evidence of causal mechanisms.
Methods
Sample
The study made use of the ‘Welfare State Under Strain’ (WESTUS) data, a four-wave panel-study collected between 2013 and 2015 in Great Britain and the Netherlands (Ford et al., 2015).1 The time periods between the four waves were 4, 8, and 4 months, respectively. The samples were drawn by YouGov (Great Britain) and TNS-NIPO (the Netherlands). There was no self-selectivity as the participants could not sign up for the survey themselves and were randomly drawn. Participants completed the surveys online. The obtained samples in the first wave were representative of the adult Dutch and British populations in terms of gender, education, and the region of residence.2 However, there was non-random attrition across the waves, as indicated by the results of a set of logistical regressions, predicting the likelihood of dropping out between two consecutive waves with the main and the control variables (Supplementary Table S1). This is accounted for by making use of full information maximum likelihood (FIML) estimations (Asendorpf et al., 2014).
Two sample selections were made. First, non-natives were excluded (N = 440 in Great Britain and 169 in the Netherlands). Second, given the explicit interest in changes over time in the current study, 1398 participants from Great Britain were excluded because they were newly added to the sample in the last wave. The final sample in Great Britain consisted of 5052 participants at wave 1, 4212 at wave 2, 3571 at wave 3, and 3460 at wave 4. In the Netherlands, the final sample consisted of 3922 participants at wave 1, 3081 at wave 2, 2030 at wave 3, and 1449 at wave 4.
Measures
Dependent variable
Welfare chauvinism was measured with four items. Respondents were asked to indicate how long they thought immigrants should work and pay taxes before they are entitled to four welfare benefits: disability benefits, housing benefits, unemployment support, and income support. These items were measured in the last three waves on 12-point scales ranging from 0 ‘immediately’ to 11 ‘never’. The intermediate options specify a number of years (1–10).3 Previous research also tried to capture the conditionality of immigrant access to welfare benefits (Reeskens and Van Oorschot, 2012), but used a nominal measure and did not distinguish different types of welfare benefits. The composite reliability value was satisfactory in Great Britain (ρ = 0.92) and the Netherlands (ρ = 0.90), respectively (Raykov, 1997).
Mediator variables
Ethnic threat was measured with two questions that were asked in all four waves and read as follows: ‘Would you say it is generally good or bad for Britain’s/the Netherlands’ economy that people come to live here from other countries?’ and ‘Would you say Britain’s/the Netherlands’ cultural life is generally undermined or enriched by people who come to live here from other countries?’. Answers were given on 11-point Likert scales ranging from 0 ‘good for the economy’ to 10 ‘bad for the economy’ and from 0 ‘enriched’ to 10 ‘undermined’, respectively. These items clearly refer to immigrants posing socio-economic or socio-cultural threats to Great Britain and the Netherlands. Similar items have been used in previous studies (Scheepers, Gijsberts and Coenders, 2002). The composite reliability value was satisfactory in Great Britain (ρ = 0.87) and the Netherlands (ρ = 0.79).
Economic egalitarianism was measured by asking the respondents to indicate the extent to which they agreed with the following statement: ‘For a society to be fair, differences in people’s standard of living should be small’. This item was measured in all four waves on five-point Likert scales, ranging from 1 ‘strongly disagree’ to 5 ‘strongly agree’. This question taps into the fairness of income differences, which has previously been used as an indicator of economic egalitarianism (Van Der Waal et al., 2010).
Independent variables
Subjective economic risk was measured by asking the respondents the following question: ‘How likely or unlikely is it that during the next 12 months there will be some periods when you don’t have enough money to cover your household necessities?’ This question was measured in all four waves on five-point Likert scales ranging from 1 ‘very unlikely’ to 5 ‘very likely’.
Objective economic risk was assessed by having the respondents specify whether or not they currently claimed any of four welfare benefits: disability benefits, housing benefits, unemployment support, and income support. These four dichotomous items were measured in all four waves with separate yes/no questions and were transformed into one count variable that indicated the number of welfare benefits respondents received, ranging from 0 to 4.
Control variables
Males, older people and lower educated people have been shown to be more likely to hold negative perceptions of immigrants (Coenders and Scheepers, 2008). Three variables were therefore controlled for in relation to all hypothesized paths: gender (1 = male), age (in years), and education. The latter was measured with seven categories in accordance with the ISEC-97 scale and was treated as quasi-metric in the analyses.
Analysis
Mplus was used to derive the descriptive results, test the measurement models, fit the latent growth curve models and estimate the structural models. Table 1 shows the ranges, means, standard deviations, and composite reliability statistics of the main variables, in all four waves separately and for both Great Britain and the Netherlands. All main continuous variables (welfare chauvinism, ethnic threat, economic egalitarianism, and subjective economic risk) can be considered normally distributed, given that kurtosis and skewness values are within the acceptable range of −1 to +1. Objective economic risk, the other main predictor, is specified as a count variable, and estimated with a Poisson model.
Ranges, means, standard deviations, and composite reliabilities in all four waves for both countries
| Range | Wave 1 | Wave 2 | Wave 3 | Wave 4 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | ||
| Great Britain | |||||||||||||
| Welfare chauvinism | 0–11 | 6.10 | 3.16 | 0.91 | 6.18 | 3.05 | 0.92 | 6.10 | 3.07 | 0.92 | |||
| Ethnic threat | 0–10 | 5.00 | 2.80 | 0.87 | 5.30 | 2.81 | 0.87 | 4.94 | 2.68 | 0.87 | 4.71 | 2.59 | 0.87 |
| Economic egalitarianism | 1–5 | 3.41 | 1.01 | 3.42 | 1.05 | 3.46 | 1.03 | 3.57 | 1.00 | ||||
| Subjective risk | 1–5 | 2.69 | 1.35 | 2.53 | 1.32 | 2.44 | 1.32 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.26 | 0.86 | 0.23 | 0.80 | 0.07 | 0.39 | 0.07 | 0.41 | ||||
| Male | 0–1 | 0.49 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | ||||
| Age | 18–85 | 50.9 | 15.8 | 52.3 | 14.7 | 52.6 | 14.6 | 53.8 | 14.5 | ||||
| Education | 1–7 | 4.84 | 1.96 | 4.81 | 1.94 | 4.84 | 1.97 | 4.86 | 1.95 | ||||
| The Netherlands | |||||||||||||
| Welfare chauvinism | 0–11 | 5.34 | 3.02 | 0.89 | 5.25 | 3.03 | 0.90 | 5.45 | 2.90 | 0.90 | |||
| Ethnic threat | 0–10 | 5.34 | 2.36 | 0.79 | 5.32 | 2.43 | 0.79 | 4.96 | 2.07 | 0.79 | 5.09 | 2.13 | 0.79 |
| Economic egalitarianism | 1–5 | 3.28 | 0.96 | 3.41 | 1.04 | 3.40 | 1.03 | 3.48 | 0.98 | ||||
| Subjective risk | 1–5 | 2.52 | 1.30 | 2.37 | 1.26 | 2.41 | 1.28 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.21 | 0.76 | 0.26 | 0.82 | 0.08 | 0.40 | 0.05 | 0.37 | ||||
| Male | 0–1 | 0.51 | 0.50 | 0.51 | 0.50 | 0.51 | 0.50 | 0.53 | 0.50 | ||||
| Age | 18–94 | 50.3 | 17.3 | 52.5 | 17.1 | 54.3 | 16.6 | 56.0 | 16.3 | ||||
| Education | 1–7 | 4.25 | 1.81 | 4.26 | 1.81 | 4.28 | 1.82 | 4.30 | 1.84 | ||||
| Range | Wave 1 | Wave 2 | Wave 3 | Wave 4 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | ||
| Great Britain | |||||||||||||
| Welfare chauvinism | 0–11 | 6.10 | 3.16 | 0.91 | 6.18 | 3.05 | 0.92 | 6.10 | 3.07 | 0.92 | |||
| Ethnic threat | 0–10 | 5.00 | 2.80 | 0.87 | 5.30 | 2.81 | 0.87 | 4.94 | 2.68 | 0.87 | 4.71 | 2.59 | 0.87 |
| Economic egalitarianism | 1–5 | 3.41 | 1.01 | 3.42 | 1.05 | 3.46 | 1.03 | 3.57 | 1.00 | ||||
| Subjective risk | 1–5 | 2.69 | 1.35 | 2.53 | 1.32 | 2.44 | 1.32 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.26 | 0.86 | 0.23 | 0.80 | 0.07 | 0.39 | 0.07 | 0.41 | ||||
| Male | 0–1 | 0.49 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | ||||
| Age | 18–85 | 50.9 | 15.8 | 52.3 | 14.7 | 52.6 | 14.6 | 53.8 | 14.5 | ||||
| Education | 1–7 | 4.84 | 1.96 | 4.81 | 1.94 | 4.84 | 1.97 | 4.86 | 1.95 | ||||
| The Netherlands | |||||||||||||
| Welfare chauvinism | 0–11 | 5.34 | 3.02 | 0.89 | 5.25 | 3.03 | 0.90 | 5.45 | 2.90 | 0.90 | |||
| Ethnic threat | 0–10 | 5.34 | 2.36 | 0.79 | 5.32 | 2.43 | 0.79 | 4.96 | 2.07 | 0.79 | 5.09 | 2.13 | 0.79 |
| Economic egalitarianism | 1–5 | 3.28 | 0.96 | 3.41 | 1.04 | 3.40 | 1.03 | 3.48 | 0.98 | ||||
| Subjective risk | 1–5 | 2.52 | 1.30 | 2.37 | 1.26 | 2.41 | 1.28 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.21 | 0.76 | 0.26 | 0.82 | 0.08 | 0.40 | 0.05 | 0.37 | ||||
| Male | 0–1 | 0.51 | 0.50 | 0.51 | 0.50 | 0.51 | 0.50 | 0.53 | 0.50 | ||||
| Age | 18–94 | 50.3 | 17.3 | 52.5 | 17.1 | 54.3 | 16.6 | 56.0 | 16.3 | ||||
| Education | 1–7 | 4.25 | 1.81 | 4.26 | 1.81 | 4.28 | 1.82 | 4.30 | 1.84 | ||||
Ranges, means, standard deviations, and composite reliabilities in all four waves for both countries
| Range | Wave 1 | Wave 2 | Wave 3 | Wave 4 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | ||
| Great Britain | |||||||||||||
| Welfare chauvinism | 0–11 | 6.10 | 3.16 | 0.91 | 6.18 | 3.05 | 0.92 | 6.10 | 3.07 | 0.92 | |||
| Ethnic threat | 0–10 | 5.00 | 2.80 | 0.87 | 5.30 | 2.81 | 0.87 | 4.94 | 2.68 | 0.87 | 4.71 | 2.59 | 0.87 |
| Economic egalitarianism | 1–5 | 3.41 | 1.01 | 3.42 | 1.05 | 3.46 | 1.03 | 3.57 | 1.00 | ||||
| Subjective risk | 1–5 | 2.69 | 1.35 | 2.53 | 1.32 | 2.44 | 1.32 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.26 | 0.86 | 0.23 | 0.80 | 0.07 | 0.39 | 0.07 | 0.41 | ||||
| Male | 0–1 | 0.49 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | ||||
| Age | 18–85 | 50.9 | 15.8 | 52.3 | 14.7 | 52.6 | 14.6 | 53.8 | 14.5 | ||||
| Education | 1–7 | 4.84 | 1.96 | 4.81 | 1.94 | 4.84 | 1.97 | 4.86 | 1.95 | ||||
| The Netherlands | |||||||||||||
| Welfare chauvinism | 0–11 | 5.34 | 3.02 | 0.89 | 5.25 | 3.03 | 0.90 | 5.45 | 2.90 | 0.90 | |||
| Ethnic threat | 0–10 | 5.34 | 2.36 | 0.79 | 5.32 | 2.43 | 0.79 | 4.96 | 2.07 | 0.79 | 5.09 | 2.13 | 0.79 |
| Economic egalitarianism | 1–5 | 3.28 | 0.96 | 3.41 | 1.04 | 3.40 | 1.03 | 3.48 | 0.98 | ||||
| Subjective risk | 1–5 | 2.52 | 1.30 | 2.37 | 1.26 | 2.41 | 1.28 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.21 | 0.76 | 0.26 | 0.82 | 0.08 | 0.40 | 0.05 | 0.37 | ||||
| Male | 0–1 | 0.51 | 0.50 | 0.51 | 0.50 | 0.51 | 0.50 | 0.53 | 0.50 | ||||
| Age | 18–94 | 50.3 | 17.3 | 52.5 | 17.1 | 54.3 | 16.6 | 56.0 | 16.3 | ||||
| Education | 1–7 | 4.25 | 1.81 | 4.26 | 1.81 | 4.28 | 1.82 | 4.30 | 1.84 | ||||
| Range | Wave 1 | Wave 2 | Wave 3 | Wave 4 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | Mean | SD | ρ | ||
| Great Britain | |||||||||||||
| Welfare chauvinism | 0–11 | 6.10 | 3.16 | 0.91 | 6.18 | 3.05 | 0.92 | 6.10 | 3.07 | 0.92 | |||
| Ethnic threat | 0–10 | 5.00 | 2.80 | 0.87 | 5.30 | 2.81 | 0.87 | 4.94 | 2.68 | 0.87 | 4.71 | 2.59 | 0.87 |
| Economic egalitarianism | 1–5 | 3.41 | 1.01 | 3.42 | 1.05 | 3.46 | 1.03 | 3.57 | 1.00 | ||||
| Subjective risk | 1–5 | 2.69 | 1.35 | 2.53 | 1.32 | 2.44 | 1.32 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.26 | 0.86 | 0.23 | 0.80 | 0.07 | 0.39 | 0.07 | 0.41 | ||||
| Male | 0–1 | 0.49 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | ||||
| Age | 18–85 | 50.9 | 15.8 | 52.3 | 14.7 | 52.6 | 14.6 | 53.8 | 14.5 | ||||
| Education | 1–7 | 4.84 | 1.96 | 4.81 | 1.94 | 4.84 | 1.97 | 4.86 | 1.95 | ||||
| The Netherlands | |||||||||||||
| Welfare chauvinism | 0–11 | 5.34 | 3.02 | 0.89 | 5.25 | 3.03 | 0.90 | 5.45 | 2.90 | 0.90 | |||
| Ethnic threat | 0–10 | 5.34 | 2.36 | 0.79 | 5.32 | 2.43 | 0.79 | 4.96 | 2.07 | 0.79 | 5.09 | 2.13 | 0.79 |
| Economic egalitarianism | 1–5 | 3.28 | 0.96 | 3.41 | 1.04 | 3.40 | 1.03 | 3.48 | 0.98 | ||||
| Subjective risk | 1–5 | 2.52 | 1.30 | 2.37 | 1.26 | 2.41 | 1.28 | 2.44 | 1.26 | ||||
| Objective risk | 0–4 | 0.21 | 0.76 | 0.26 | 0.82 | 0.08 | 0.40 | 0.05 | 0.37 | ||||
| Male | 0–1 | 0.51 | 0.50 | 0.51 | 0.50 | 0.51 | 0.50 | 0.53 | 0.50 | ||||
| Age | 18–94 | 50.3 | 17.3 | 52.5 | 17.1 | 54.3 | 16.6 | 56.0 | 16.3 | ||||
| Education | 1–7 | 4.25 | 1.81 | 4.26 | 1.81 | 4.28 | 1.82 | 4.30 | 1.84 | ||||
Measurement invariance of the two latent variables, welfare chauvinism and ethnic threat, was assessed in two ways. First, invariance was tested across waves within each country separately. In both countries, partial scalar invariance across waves could be concluded for welfare chauvinism and ethnic threat. Second, invariance was tested across countries within each wave separately. In all four waves, partial metric invariance between the two countries was concluded (see Supplementary Tables S2–S4 for a thorough discussion). Substantively, these tests of measurement invariance show that it is warranted to compare covariances, regression coefficients, means and intercepts across the four waves, within each country separately. Moreover, the covariances and regression coefficients can also be compared across the countries, within each wave (Kline, 2010). Note that our hypotheses refer to these regression coefficients. However, as no full scalar invariance between the countries could by achieved, one should be careful in interpreting the differences in means and intercepts between Great Britain and the Netherlands. The analyses were therefore performed in the two countries separately.
To prevent estimation problems in our complex longitudinal models, we used mean-composite scales instead of latent factors for both ethnic threat and welfare chauvinism. That is, we approximated latent measurement models of welfare chauvinism and ethnic threat by using the loadings, residual variances, covariances and factor variances to calculate specific measurement error corrections (Wang and Wang, 2012). To use the correction method for measurement error in the two single item measures for economic egalitarianism and economic security, a reliability of 0.8 was assumed. Objective economic risk was not corrected for measurement error, as the questions used to measure it are more factual, referring to actual welfare dependency, and do not necessarily relate to an underlying construct.
The results of two types of analyses are presented. First, a contemporaneous mediation model was fitted to investigate differences between individuals and to examine whether the effects of subjective and objective economic risk on welfare chauvinism could be explained by ethnic threat and economic egalitarianism. To account for the occasions-within-respondent data structure, a complex survey method was used that takes the clustering of the data into account (Asparouhov, 2005). Second, latent growth curve models were estimated in order to explicitly model changes in variables over time, individual differences in these changes, and relations between rates of change in different variables (Selig and Preacher, 2009, p. 152). Parallel process latent growth curve modelling makes it possible to examine whether changes over time in welfare chauvinism can be explained by changes over time in economic risk, ethnic threat, and economic egalitarianism. Again, the single indicator method was used to correct for measurement error. Finally, a structural model was fitted that imposed a mediation structure on the slopes of the main variables (Bollen and Curran, 2006).
Results
Contemporaneous Mediation Analysis
A structural model was fitted that used the single indicator method. This model included welfare chauvinism as the dependent variable, subjective and objective economic risk as the independent variables and ethnic threat and economic egalitarianism as the mediator variables. The residuals of the mediators were allowed to be correlated. Gender, age, and education were controlled for in relation to all estimated paths. Figure 1 shows the unstandardized coefficients of the contemporaneous mediation model and Table 2 gives the total effects, the indirect effects and the remaining direct effects.
Contemporaneous mediation model
Notes: The nesting of the data was accounted for by the clustering adjustment of the sandwich estimator. Gender, age, and education were controlled for (see Supplementary Table S5). Unstandardized coefficients are presented as Great Britain/the Netherlands. Two-sided P-values are denoted as *P < 0.05, **P < 0.01, ***P < 0.001. The latent variables ethnic threat and welfare chauvinism are measured by single indicators with fixed residual variances representing assumed composite reliabilities of welfare chauvinism (0.92 in Great Britain, 0.90 in the Netherlands) and ethnic threat (0.87 in Great Britain, 0.79 in the Netherlands).
Contemporaneous mediation model: the total, direct, and indirect effects of subjective and objective economic risk on welfare chauvinism, via ethnic threat and economic egalitarianism
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| Ethnic threat | Economic egalitarianism | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| Subjective risk | 0.148*** (0.042) | 0.166*** (0.024) | −0.140*** (0.038) | 0.122* (0.057) |
| Objective risk | −0.119 (0.086) | 0.064 (0.041) | −0.151*** (0.046) | −0.032 (0.080) |
| The Netherlands | ||||
| Subjective risk | 0.258*** (0.052) | 0.165*** (0.024) | −0.041*** (0.008) | 0.134** (0.047) |
| Objective risk | −0.353** (0.110) | −0.017 (0.048) | −0.059*** (0.013) | −0.276** (0.096) |
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| Ethnic threat | Economic egalitarianism | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| Subjective risk | 0.148*** (0.042) | 0.166*** (0.024) | −0.140*** (0.038) | 0.122* (0.057) |
| Objective risk | −0.119 (0.086) | 0.064 (0.041) | −0.151*** (0.046) | −0.032 (0.080) |
| The Netherlands | ||||
| Subjective risk | 0.258*** (0.052) | 0.165*** (0.024) | −0.041*** (0.008) | 0.134** (0.047) |
| Objective risk | −0.353** (0.110) | −0.017 (0.048) | −0.059*** (0.013) | −0.276** (0.096) |
Notes: Unstandardized coefficients presented. Two-sided P-values denoted as *P < 0.05, **P < 0.01, ***P < 0.001.
Contemporaneous mediation model: the total, direct, and indirect effects of subjective and objective economic risk on welfare chauvinism, via ethnic threat and economic egalitarianism
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| Ethnic threat | Economic egalitarianism | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| Subjective risk | 0.148*** (0.042) | 0.166*** (0.024) | −0.140*** (0.038) | 0.122* (0.057) |
| Objective risk | −0.119 (0.086) | 0.064 (0.041) | −0.151*** (0.046) | −0.032 (0.080) |
| The Netherlands | ||||
| Subjective risk | 0.258*** (0.052) | 0.165*** (0.024) | −0.041*** (0.008) | 0.134** (0.047) |
| Objective risk | −0.353** (0.110) | −0.017 (0.048) | −0.059*** (0.013) | −0.276** (0.096) |
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| Ethnic threat | Economic egalitarianism | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| Subjective risk | 0.148*** (0.042) | 0.166*** (0.024) | −0.140*** (0.038) | 0.122* (0.057) |
| Objective risk | −0.119 (0.086) | 0.064 (0.041) | −0.151*** (0.046) | −0.032 (0.080) |
| The Netherlands | ||||
| Subjective risk | 0.258*** (0.052) | 0.165*** (0.024) | −0.041*** (0.008) | 0.134** (0.047) |
| Objective risk | −0.353** (0.110) | −0.017 (0.048) | −0.059*** (0.013) | −0.276** (0.096) |
Notes: Unstandardized coefficients presented. Two-sided P-values denoted as *P < 0.05, **P < 0.01, ***P < 0.001.
First, in both Great Britain and the Netherlands, people who perceived themselves to be at economic risk perceived more ethnic threat and were therefore more chauvinistic. The same cannot be said for people who were at objective economic risk (those who received welfare benefits), because they did not perceive more ethnic threat than people who were not at objective economic risk.
Second, in Great Britain and the Netherlands, both people who were at subjective and objective economic risk were found to be more egalitarian and were therefore less chauvinistic in their welfare attitudes.
These results are in line with our hypotheses. Whereas welfare chauvinism is stronger among those who feel threatened by immigrants (H1b), it is lower among those in favour of egalitarianism (H2b). The results of this contemporaneous mediation analysis also confirm two opposing mediating mechanism. On the one hand, people who are economically vulnerable (objectively or subjectively) support economic egalitarianism and are thus less chauvinistic. On the other hand, those who perceive less income security feel more threatened by immigrants and are thus more chauvinistic.
An additional comparison of the relative sizes of the indirect effects indicated that the effect of subjective risk on welfare chauvinism was stronger via ethnic threat perceptions than via economic egalitarianism.4 In contrast, economic egalitarianism was more important in explaining the relationship between objective risk and welfare chauvinism, as we found no association between objective risk and ethnic threat. This is congruent with previous research that found objective economic circumstances to be irrelevant for anti-immigrant attitudes (Hainmueller and Hopkins, 2014).
Next to these mediating links, there was a remaining direct effect of subjective economic risk on welfare chauvinism in Great Britain, whereas in the Netherlands both subjective and objective economic risk still had direct effects on welfare chauvinism. These results suggest partial mediation. In Great Britain, the contemporaneous mediation model explained 45% of the variance of welfare chauvinism. In the Netherlands, the model explained 37% of this variance. The results for the control variables are shown in Supplementary Table S5.
Longitudinal Mediation Analysis
Latent growth curves were modelled for all the main variables using the corrected single items and mean-composite scales as observed variables. For each variable, two factors were specified: the random intercept, representing individuals’ scores at the first wave, and the random slope, indicating how much individuals change, on average, between two time points. The parameterization of the slope factors was coded in accordance with the number of months passed since the first wave (T1 = 0; T2 = 4; T3 = 12; T4 = 16), reflecting the uneven time intervals between waves (Von Soest and Hagtvet, 2011).
In order to investigate the changes over time for the main variables, a model was fitted that included the growth models of welfare chauvinism, ethnic threat, economic egalitarianism, subjective and objective economic risk. The model assumes that an individual is described by five intercepts and five slopes, and that these ten properties follow a ten-dimensional multivariate normal distribution. The means and variances of the five estimated slopes are reported in Table 3. Supplementary Table S6 shows the correlations between the intercepts and slopes.
Means and variances of the slopes of the five main variables
| Great Britain | The Netherlands | |||
|---|---|---|---|---|
| Mean | Variance | Mean | Variance | |
| b (SE) | b (SE) | b (SE) | b (SE) | |
| S(Welfare chauvinism) | ?>0.026* (0.014) | 0.069 (0.055) | 0.011 (0.020) | 0.122* (0.075) |
| S(Ethnic threat) | −0.082**** (0.007) | 0.031**** (0.008) | −0.173**** (0.009) | 0.037**** (0.010) |
| S(Economic egalitarianism) | 0.033**** (0.004) | ?>0.006*** (0.002) | 0.033**** (0.005) | 0.004* (0.003) |
| S(Subjective risk) | −0.050**** (0.004) | 0.013**** (0.002) | −0.013** (0.006) | 0.018**** (0.004) |
| S(Objective risk) | −0.614**** (0.056) | 0.158**** (0.020) | −0.414**** (0.039) | 0.180**** (0.018) |
| Great Britain | The Netherlands | |||
|---|---|---|---|---|
| Mean | Variance | Mean | Variance | |
| b (SE) | b (SE) | b (SE) | b (SE) | |
| S(Welfare chauvinism) | ?>0.026* (0.014) | 0.069 (0.055) | 0.011 (0.020) | 0.122* (0.075) |
| S(Ethnic threat) | −0.082**** (0.007) | 0.031**** (0.008) | −0.173**** (0.009) | 0.037**** (0.010) |
| S(Economic egalitarianism) | 0.033**** (0.004) | ?>0.006*** (0.002) | 0.033**** (0.005) | 0.004* (0.003) |
| S(Subjective risk) | −0.050**** (0.004) | 0.013**** (0.002) | −0.013** (0.006) | 0.018**** (0.004) |
| S(Objective risk) | −0.614**** (0.056) | 0.158**** (0.020) | −0.414**** (0.039) | 0.180**** (0.018) |
Notes: Unstandardized coefficients are presented.
P < 0.10,
P < 0.05,
P < 0.01,
P < 0.001.
Slopes denoted as S(variable). For the slope variances, the reported P-values are one-sided, as is appropriate when using Wald tests (Hox, 2010).
Means and variances of the slopes of the five main variables
| Great Britain | The Netherlands | |||
|---|---|---|---|---|
| Mean | Variance | Mean | Variance | |
| b (SE) | b (SE) | b (SE) | b (SE) | |
| S(Welfare chauvinism) | ?>0.026* (0.014) | 0.069 (0.055) | 0.011 (0.020) | 0.122* (0.075) |
| S(Ethnic threat) | −0.082**** (0.007) | 0.031**** (0.008) | −0.173**** (0.009) | 0.037**** (0.010) |
| S(Economic egalitarianism) | 0.033**** (0.004) | ?>0.006*** (0.002) | 0.033**** (0.005) | 0.004* (0.003) |
| S(Subjective risk) | −0.050**** (0.004) | 0.013**** (0.002) | −0.013** (0.006) | 0.018**** (0.004) |
| S(Objective risk) | −0.614**** (0.056) | 0.158**** (0.020) | −0.414**** (0.039) | 0.180**** (0.018) |
| Great Britain | The Netherlands | |||
|---|---|---|---|---|
| Mean | Variance | Mean | Variance | |
| b (SE) | b (SE) | b (SE) | b (SE) | |
| S(Welfare chauvinism) | ?>0.026* (0.014) | 0.069 (0.055) | 0.011 (0.020) | 0.122* (0.075) |
| S(Ethnic threat) | −0.082**** (0.007) | 0.031**** (0.008) | −0.173**** (0.009) | 0.037**** (0.010) |
| S(Economic egalitarianism) | 0.033**** (0.004) | ?>0.006*** (0.002) | 0.033**** (0.005) | 0.004* (0.003) |
| S(Subjective risk) | −0.050**** (0.004) | 0.013**** (0.002) | −0.013** (0.006) | 0.018**** (0.004) |
| S(Objective risk) | −0.614**** (0.056) | 0.158**** (0.020) | −0.414**** (0.039) | 0.180**** (0.018) |
Notes: Unstandardized coefficients are presented.
P < 0.10,
P < 0.05,
P < 0.01,
P < 0.001.
Slopes denoted as S(variable). For the slope variances, the reported P-values are one-sided, as is appropriate when using Wald tests (Hox, 2010).
In both Great Britain and the Netherlands, on average, people’s ethnic threat perceptions have decreased over time, they have become more egalitarian, and they are less at an economic risk, both subjectively and objectively. The latter reflects the economic recovery in both countries. Furthermore, these four slopes varied significantly between individuals in both countries. Lastly, people have become slightly, and only marginally significantly, more chauvinistic in their welfare attitudes in Great Britain, but this was not significantly so in the Netherlands. Moreover, the slope of welfare chauvinism only varied between individuals in the Netherlands but not in Great Britain. Despite the fact that, on average, there was little to no change in welfare chauvinism over time or little to no variation in this change between individuals, and the caution this warrants, it is still possible to regress the slope of welfare chauvinism on the slopes of ethnic threat, economic egalitarianism, subjective and objective economic risk (Snijders and Bosker, 1999, Chapter 12). Generally, there is little power to detect slope variance and change over time without covariates (Bryk, 1992), so change in welfare chauvinism can still occur as a function of change in the other variables. Moreover, it is important to stress that the slopes of objective economic risk, subjective economic risk, and ethnic threat were negative, as this influences the interpretation of the regression coefficients. They now describe the effect of a decrease in objective and subjective economic risk and ethnic threat, which is opposite to the way regression coefficients are typically read in cross-sectional analyses.
Subsequently, a model was fitted imposing a structure on the ten latent effects of the growth models in order to examine the longitudinal mediation paths, and investigate whether changes over time in economic risk were associated with changes over time in welfare chauvinism, and whether this can be explained by changes over time in ethnic threat and economic egalitarianism. This structural model represents a longitudinal mediation process via the slopes of the mediator variables (Selig and Preacher, 2009). This model included the slope of welfare chauvinism as the dependent variable, the slopes of subjective and objective economic risk as the independent variables, and the slopes of ethnic threat and economic egalitarianism as the mediator variables. Furthermore, the slope of welfare chauvinism was regressed on the intercepts of the independent and the mediator variables, and the slopes of the mediator variables were regressed on the intercepts of the independent variables. This was done to isolate the influence of the changes over time, by controlling for the effects of the scores at the first wave. Lastly, the covariance between the intercepts was accounted for. Similar to the contemporaneous mediation model, the residuals of the intercepts and the slopes of the mediators were allowed to be correlated. Gender, age, and education were controlled for in all estimated paths. Figure 2 presents the unstandardized coefficients, and Table 4 shows the total effects, the indirect effects and the remaining direct effects. The full covariance structure between the intercepts and slopes is shown in Supplementary Table S7. And the results for age, gender, education, and intercepts are shown in Supplementary Table S8.
Longitudinal mediation model
Notes: The effects of the slopes of objective and subjective economic risk on the slope of welfare chauvinism, via the slopes of ethnic threat and economic egalitarianism. Gender, age, and education were controlled for (see Supplementary Table S8). For the sake of clarity, the covariances between slopes and intercepts and the relationships between the intercepts are not shown (see Supplementary Table S7). Unstandardized coefficients are presented as Great Britain/the Netherlands. Two-sided P-values are denoted as *P < 0.10, **P < 0.05, ****P < 0.001. The latent variables ethnic threat and welfare chauvinism are measured by single indicators with known residual variances estimated from the reliabilities in Table 1.
Longitudinal mediation model: the total, direct, and indirect effects of the slopes of subjective and objective economic risk on the slope welfare chauvinism, via the slopes of ethnic threat and economic egalitarianism
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| S(ethnic threat) | S(Economic egalitarianism) | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| S(Subjective risk) | −0.013 (0.241) | 0.086 (0.066) | 0.057 (0.116) | −0.156 (0.287) |
| S(Objective risk) | −0.157 (0.092) | −0.011 (0.021) | −0.023 (0.045) | −0.124 (0.111) |
| The Netherlands | ||||
| S(Subjective risk) | 0.186 (0.236) | −0.056 (0.094) | −0.027 (0.088) | 0.269 (0.237) |
| S(Objective risk) | 0.011 (0.101) | −0.025 (0.038) | 0.021 (0.044) | 0.014 (0.107) |
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| S(ethnic threat) | S(Economic egalitarianism) | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| S(Subjective risk) | −0.013 (0.241) | 0.086 (0.066) | 0.057 (0.116) | −0.156 (0.287) |
| S(Objective risk) | −0.157 (0.092) | −0.011 (0.021) | −0.023 (0.045) | −0.124 (0.111) |
| The Netherlands | ||||
| S(Subjective risk) | 0.186 (0.236) | −0.056 (0.094) | −0.027 (0.088) | 0.269 (0.237) |
| S(Objective risk) | 0.011 (0.101) | −0.025 (0.038) | 0.021 (0.044) | 0.014 (0.107) |
Notes: Unstandardized coefficients are presented. Slopes are represented as S(variable).
Longitudinal mediation model: the total, direct, and indirect effects of the slopes of subjective and objective economic risk on the slope welfare chauvinism, via the slopes of ethnic threat and economic egalitarianism
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| S(ethnic threat) | S(Economic egalitarianism) | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| S(Subjective risk) | −0.013 (0.241) | 0.086 (0.066) | 0.057 (0.116) | −0.156 (0.287) |
| S(Objective risk) | −0.157 (0.092) | −0.011 (0.021) | −0.023 (0.045) | −0.124 (0.111) |
| The Netherlands | ||||
| S(Subjective risk) | 0.186 (0.236) | −0.056 (0.094) | −0.027 (0.088) | 0.269 (0.237) |
| S(Objective risk) | 0.011 (0.101) | −0.025 (0.038) | 0.021 (0.044) | 0.014 (0.107) |
| Total effect | Indirect effects via two mediators | Direct effect | ||
|---|---|---|---|---|
| S(ethnic threat) | S(Economic egalitarianism) | |||
| b (SE) | b (SE) | b (SE) | b (SE) | |
| Great Britain | ||||
| S(Subjective risk) | −0.013 (0.241) | 0.086 (0.066) | 0.057 (0.116) | −0.156 (0.287) |
| S(Objective risk) | −0.157 (0.092) | −0.011 (0.021) | −0.023 (0.045) | −0.124 (0.111) |
| The Netherlands | ||||
| S(Subjective risk) | 0.186 (0.236) | −0.056 (0.094) | −0.027 (0.088) | 0.269 (0.237) |
| S(Objective risk) | 0.011 (0.101) | −0.025 (0.038) | 0.021 (0.044) | 0.014 (0.107) |
Notes: Unstandardized coefficients are presented. Slopes are represented as S(variable).
First of all, in both Great Britain and the Netherlands, if people’s subjective or objective economic risk decreased over time, they did not feel less threatened by immigrants; refuting Hypothesis 1a. However, supporting Hypothesis 1b, both British and Dutch people who felt less threatened by immigrants than they did before, also became less chauvinistic, as can be seen from the significant relationship between the slope of ethnic threat and the slope of welfare chauvinism.
In the case of the Netherlands, a decrease over time in people’s economic vulnerability, either subjective or objective, did not result in less egalitarianism; refuting Hypothesis 2a. Further, in neither Great Britain nor the Netherlands did people who became more egalitarian over time also become less chauvinistic, refuting Hypothesis 2b.
Only in Great Britain did we find partial support for Hypothesis 2a: the slopes of subjective and objective economic risk were found to be positively related to the slope of economic egalitarianism. That is, British people who were less at a subjective or objective economic risk than before became less egalitarian, supporting Hypothesis 2a.
Discussion
The present study had two aims. First, it set out to examine whether welfare chauvinism is rooted in subjective and objective economic risk and whether this can be explained by two distinct mechanisms: the endorsement of economic egalitarianism and perceptions of ethnic threat. Furthermore, given the lack of longitudinal studies that allow for causal inferences, our second aim was to study changes in welfare chauvinism over time by examining whether they are related to changes in economic risk, economic egalitarianism and ethnic threat.
Concerning the contemporaneous part, in both Great Britain and the Netherlands, people who felt more economically secure perceived less threat from immigrants and were therefore less supportive of welfare chauvinism. Importantly, we did not find that people who received more welfare benefits and were thus at objective economic risk, perceived more ethnic threat or were more chauvinistic. This corroborates previous research that shows that one’s actual economic position has little bearing on one’s attitude towards immigrants (Hainmueller and Hopkins, 2014; Kuntz, Davidov and Semyonov, 2017). We extend on this research by showing that this not only holds for people’s labour market position but also when looking at welfare benefits and welfare chauvinism specifically. We note that our measurement of ethnic threat refers to sociotropic concerns about the impact of immigrants on the nation’s culture and economy. Research on anti-immigrant attitudes showed that these sociotropic threats are more relevant than threats to one’s personal situation (Hainmueller and Hopkins, 2014). Yet, the question to what extent perceived personal threats from migrants are relevant to explain welfare chauvinism remains open for future research.
Furthermore, in both countries it was found that people who felt more economically secure and depended less on welfare benefits were less egalitarian which meant that, in turn, they were more supportive of welfare chauvinism. These findings are in line with our expectations and previous research (Achterberg and Houtman, 2006).
Taken together these findings support two opposing links between economic risk and welfare chauvinism. On the one hand, people at either subjective or objective economic risk are more supportive of economic egalitarianism, and because of this ideological self-interest they are less chauvinistic. On the other hand, people at subjective but not objective economic risk also perceive more threat from immigrants and are therefore less in favour of granting welfare rights to immigrants.
However, the results from our longitudinal mediation analyses also suggest that all but two of these relationships are correlational rather than causal. First, in both Great Britain and the Netherlands, people who perceived less ethnic threat than they did before also supported welfare chauvinism less than before. But within-individual changes over time in objective or subjective economic risk, or economic egalitarianism did not result in changes over time in welfare chauvinism. Second, and only in Great Britain, people who became less economically vulnerable, subjective or objective, also became less egalitarian.
As such, the results of our contemporaneous models and longitudinal models are to quite some extent incongruent. Our findings thus raise concerns about the validity of cross-sectional studies on the relationship between economic risk and welfare chauvinism. While subjective and economic risk and economic egalitarianism are helpful in understanding between individual differences in welfare chauvinism, they do not explain within individual variation over time. In fact, our study suggests ethnic threat to be the only causal predictor of welfare chauvinism. We encourage other researchers to put other common correlates of welfare chauvinism to a similar longitudinal test, as doing so may prohibit scholars from drawing invalid conclusions based purely on cross-sectional research.
In addition, it would be interesting to consider third variables that may explain why some relationships are found in the contemporaneous but not the longitudinal analyses. For example, perhaps we only find a correlational, perhaps even spurious, relationship between subjective risk and ethnic threat because they are both positively related to another factor, like authoritarian personality traits (Gallego and Pardos-Prado, 2014).
Furthermore, our longitudinal analyses show that, in general, people in both Great Britain and the Netherlands hardly changed in their endorsement of welfare chauvinism between 2013 and 2015. In Great Britain, there was a slight and only marginally significant increase in welfare chauvinism, and in the Netherlands, there was no significant change over time in welfare chauvinism. Our results therefore do not support the idea of eroding solidarity in Europe. The lack of change in welfare chauvinism is remarkable given the prominent position of immigration and the refugee crises in the public and political debate during this timeframe. One reason for this stability in public support for chauvinism could be that the actual percentage of immigrants are less consequential for people’s anti-immigrant and welfare attitudes than is often assumed (Hjerm and Schnabel, 2012). Instead, overestimating the total number of immigrants and misjudging the type of immigrants, by for instance focusing disproportionally on asylum seekers and ‘welfare tourists’, might be particularly relevant for anti-immigrant attitudes and welfare chauvinism (Blinder, 2013; Hjorth, 2016).
Further, it must be noted that our data covers a relatively short time-span, which might confound our ability to detect changes in welfare attitudes. However, recent research shows that welfare attitudes are particularly likely to fluctuate over short periods of time, but remain stable in the long run (Jeene, Van Oorschot and Uunk, 2014). It might also be that we found no change in welfare chauvinism due to the phrasing of the items used to measure it. The items refer to conditional access to welfare provisions, based on the number of years immigrants have worked and payed taxes, two factors that make people more deserving of welfare benefits (Kootstra, 2016). Concerns about immigrants also revolve around portrayals of immigrants as lazy welfare tourists that benefit from the welfare state without contributing to it (Van Oorschot, 2006). Unfortunately, these sentiments could not be assessed by our measurements.
We measured welfare chauvinism with questions regarding conditional access for migrants to welfare benefits. These items specifically refer to migrants, yet could also partially capture a general dislike for welfare spending. Participants who prefer strict conditional access for migrants could also prefer strict conditions for natives. Although we did not assess the impact of a general dislike for welfare spending, it is again interesting that we find that welfare chauvinism generally does not change over time, given the claim that rising ethnic diversity could undermine support for a strong welfare state in general (Alesina and Glaeser, 2004).
Moreover, it was found that between 2013 and 2015 and in both countries, subjective and objective economic risk decreased. This reflects the economic crisis of 2008 slowly coming to an end. In both Great Britain and the Netherlands, GDP has steadily increased over the last few years whereas, conversely, unemployment rates have slowly decreased (World Bank, 2015).
Furthermore, ethnic threat perceptions decreased between 2013 and 2015 in both countries. This might seem surprising given the heightened public and political debate on immigration. It also contrasts research suggesting that ethnic threat increases when there is an influx in immigration (Coenders and Scheepers, 2008), as this was the case in both Great Britain and the Netherlands during data collection (UNHCR, 2015). This suggests that increasing levels of immigration did not signify cultural or economic competition. Perhaps this is because public debates about immigration between 2013 and 2015 often revolved around issues of terrorism and criminality. Previous research found exposure to stories about ethnic crime to increase ethnic threat (Lubbers, Scheepers and Vergeer, 2010). It could be that the increase in immigration spurred on ethnic threat that was security-related, rather than economic or cultural. Findings from our own data offer some tentative support for this interpretation. The last two waves included two additional items, measured on 11-point scales, indicating how much participants agreed that terrorist threats and criminality increased because of immigrants. In Great Britain, ethnic terrorist threats increased, albeit insignificantly, from mean = 7.07 (SD = 2.23) to mean = 7.12 (SD = 2.16), Wald χ2(1) = 1.77, P > 0.05, and ethnic crime threats increased significantly from mean = 6.64 (SD = 1.95) to mean = 6.88 (2.00), Wald χ2(1) = 50.46, P < 0.001. In the Netherlands, perceptions of ethnic terrorist and crime threats increased significantly from mean = 6.68 (SD = 1.78) to mean = 6.79 (SD = 1.83), Wald χ2(1) = 4.71, P < 0.05, and from mean = 6.85 (SD = 1.72) to mean = 7.09 (SD = 1.84), Wald χ2(1) = 25.55, P < 0.001, respectively. Cross-sectional analyses showed that perceptions of ethnic terrorist and crime threats are positively related to welfare chauvinism in Great Britain and the Netherlands.
While, to a large extent, we found similar results in Great Britain and the Netherlands, both in the contemporaneous and longitudinal analyses, there were also some notable differences between the two countries. In Great Britain, the decrease in subjective and objective economic risk was associated with a decrease in the endorsement of economic egalitarianism. These findings confirm our expectations. However, in the Netherlands, there were no significant effects of changes in subjective and objective economic risk. These unexpected findings could perhaps be due to differences in the countries’ economic situations (Semyonov, Raijman and Gorodzeisky, 2006). It was argued that as people’s economic situation deteriorates, they would feel more threatened by immigrants and become more egalitarian. Perhaps this was not found in the Netherlands because, compared with Great Britain, there is less economic inequality and unemployment (World Bank, 2015).
Given the differences that were found between Great Britain and the Netherlands, future research could consider larger comparative studies that are able to test differences between countries in the mechanisms and processes underlying welfare chauvinism, particularly when it comes to explaining changes over time. Unfortunately, this was impossible in the current study because the data only included two countries.
Notwithstanding these limitations, the current study provides crucial contributions to the literature. Between-individual differences in welfare chauvinism were rooted in between-individual differences in economic risk, economic egalitarianism and ethnic threat. Furthermore, we are, to our knowledge, the first to examine these processes longitudinally with high-quality panel-data in two countries. This revealed that the explanatory mechanisms common in cross-sectional research are not as apt in explaining within-individual differences. Most notably, changes in economic risk were unrelated to changes in ethnic threat and welfare chauvinism, and changes in economic egalitarianism were unrelated to changes in welfare chauvinism. Future studies should be aware of this and could consider other explanations for within-individual differences. We took an important step by testing the applicability of common explanations for between-individual differences in predicting within-individual differences. We found that decreases in ethnic threat can drive decreases in welfare chauvinism. This result is important given the concerns about declining public support for the welfare state in times of increasing immigration.
Mathijs Kros is a PhD candidate at Utrecht University, the Netherlands. His current research interests include welfare chauvinism, prejudice, positive and negative interethnic contact, and hate crimes. His work has been published in the journals European Journal of Social Psychology and Ethnic and Racial Studies.
Marcel Coenders is Senior Researcher at the European Research Centre on Migration and Ethnic Relations (ERCOMER). His primary research concerns intergroup relations. His research interests include attitudes towards migrants and ethnic minorities, nationalistic attitudes and discrimination. His work has been published in the journals European Sociological Review, International Journal of Comparative Sociology, Journal of Social Issues, Journal of Ethnic and Migration Studies, International Migration Review, Social Science Research, Social Psychology, Ethnic and Racial Studies, Social Psychology Quarterly, Social Forces, European Journal of Personality, and International Journal of Intercultural Relations.
Footnotes
One more wave was collected in Great Britain, in between the third and fourth waves in the Netherlands. This wave was not used because it did not include the variables crucial to this study.
See Van Ossenbruggen, Vonk and Willems (2006) for an assessment of the representativeness of TNS-NIPO’s panel, and Supplementary Tables S9 and S10 for a demographic comparison between our samples and the British and Dutch populations.
Welfare chauvinism was also measured in the first wave but with a different question that asked about immigrants coming from outside of the EU, not immigrants in general. Research has shown that anti-immigrant attitudes are influenced by the immigrants’ region of origin (Schneider, 2008). For the contemporaneous analyses, first wave scores for welfare chauvinism were therefore estimated using FIML. Analyses were also performed without estimating missing values for the first wave. Results of these analyses are the same and are available upon request.
The effect sizes of the indirect paths were compared by standardizing the four continuous main variables before running the analyses. Only objective risk was not standardized, because it was treated as a count variable. The results from these additional analyses can be found in the Supplementary Table S11.
Funding
This work was supported by the Open Research Area in Europe for the Social Sciences (ORA) joint funding scheme and financed by the Netherlands Organisation for Scientific Research (NWO) and the Economic and Social Research Council (ESRC).
References
The Guardian (
United Nations High Commissioner for Refugees (UNHCR) (
