Support for restrictive immigration policies in the European Union 2002–2013: the impact of economic strain and ethnic threat for vulnerable economic groups

ABSTRACT

Testing propositions from ethnic competition theory, we examine contextual and individual determinants of support for restrictive immigration policies in 26 European Union member states between 2002 and 2013, a period characterized by enduring economic downturn. We hypothesize that natives in vulnerable economic positions, similar to many migrants, are more restrictive toward immigration, because they perceive more economic strain and more ethnic threat. We expect that natives are more restrictive in times of economic decline – when national unemployment rates and debts increase – especially those who hold similar economic positions as many migrants. We enriched European Social Survey data (2002–2013, containing more than 210,000 respondents) with cross-national data on the economic situation. We indeed find that support for immigration restrictiveness is higher among natives in more vulnerable socio-economic positions. They perceive more economic strain, which is directly related to restrictiveness. But economic strain also increases perceptions of ethnic threat, which is strongly related to restrictiveness. We do, however, not find strong empirical evidence that economic decline more strongly affects support for restrictive immigration policies among vulnerable economic groups. Stronger changes in national debts induce more restrictiveness among the full population, but especially among those who perceive more economic strain and more ethnic threat.

KEYWORDS:

• Anti-immigration attitudes
• ethnic competition theory
• perceived ethnic threat
• perceived economic strain
• European Social Survey

Introduction and research questions

The global financial-economic crisis that started in 2008 has affected national economies and individual households throughout the European Union (EU). Member states were confronted with higher levels of unemployment and national debt, and individual households experienced a steadily decline of purchasing power (Eurostat 2014a,d,f). Europe has been confronted with severe economic insecurity, which potentially affects intergroup relations (Billiet et al. 2014). Using individual-level data from the European Social Survey (ESS), enriched with country level data from Eurostat covering the long period from 2002 to 2013, we examine to what extent recent changes in economic conditions affect natives’ support for restrictive immigration policies, that is, individuals and collectivities loudly sharing their views that immigration should be put on hold. These data provide possibilities to test hypotheses rigorously regarding the impact of severe changes in economic conditions on immigration restrictiveness among vulnerable natives from many EU member states. The data cover a longer period of time than previous contributions examined.

Previous research has convincingly shown that especially vulnerable economic groups hold restrictive views toward immigrants (Scheepers et al. 2002; Kunovich 2004). Ethnic competition theory predicts that natives who hold similar social positions as the majority of the migrant population – such as manual workers, lower educated, unemployed, and self-employed natives – compete more with migrants on the labor market, and therefore perceive more ethnic threat. These threat perceptions induce antagonism toward immigrants (Scheepers et al. 2002).¹

In this contribution, we fill two important lacunas in previous cross-national research on anti-immigration attitudes. The first lacuna is on the extent to which economic downturn fuels anti-immigrant sentiments in Europe, especially among natives in vulnerable socio-economic positions. We therefore examine the impact of recent increases in unemployment rates and national government debts while controlling for a more elaborate set of macro-economic and demographic indicators than previous research on anti-immigration attitudes has done. We assess the impact of the level and changes in economic conditions simultaneously.

The second lacuna we aim to fill is to theorize and empirically examine the mediating role of perceived economic strain in relation to the previously empirically evidenced mediating role of perceived ethnic threat. Ethnic competition theory traditionally predicts that vulnerable economic groups perceive more ethnic threat, which translates into more antagonism toward immigrants. Many researchers have used this theoretical reasoning (e.g. Kunovich 2004; Schneider 2008; Schlueter and Scheepers 2010). We take it one step further and address the role of perceived economic strain in relation to anti-immigration sentiments and perceived ethnic threat. Perceptions of economic strain concern perceptions of economic vulnerability, that is, perceptions of failing to make ends meet. Gesthuizen and Scheepers (2010) found that especially people in vulnerable economic positions – such as unemployed, inactive, and part-time working people and people with a lower socio-economic status and/or lower level of educational attainment – perceive more economic strain. We will test to what extent and how perceived economic strain is related to anti-immigration attitudes.

In this contribution, we answer four research questions. First, to what extent can individual differences in support for restrictive immigration policies among natives in the EU between 2002 and 2013 be explained in terms of differences in individuals’ (vulnerable) socio-economic positions? Second, to what extent can socio-economic differences in support for restrictive immigration policies be explained by perceived economic strain and perceived ethnic threat? Third, to what extent can cross-national differences in support for restrictive immigration policies be explained in terms of recent changes in unemployment and national government debt? Fourth, to what extent does the relation between, on the one hand, individual socio-economic position and perceptions of economic strain and ethnic threat and, on the other hand, support for restrictive immigration policies vary with recent changes in unemployment and national government debt?

We innovate and improve upon previous research on anti-immigration attitudes in several ways. First, we compiled a comprehensive longitudinal database, containing information of more than 210,000 individuals from 26 EU member states, covering a period of more than a decade, which is longer than previous studies. Second, we conduct more elaborate analyses: we examine the impact of a more elaborate set of macro-economic and demographic indicators and examine the impact of the level of and changes in economic conditions simultaneously. Third, we disentangle the theoretical mechanisms that explain the negative relation between vulnerable socio-economic positions and anti-immigration attitudes. Fourth, we examine a more elaborate set of cross-level interactions between changes in national economic conditions and individual’s socio-economic position as well as their perceptions of economic strain and ethnic threat.

Theoretical framework and hypotheses

We derive hypotheses from ethnic competition theory (Scheepers et al. 2002; Schneider 2008) to explain individual and cross-national differences in support for restrictive immigration policies. This theory posits that ethnic competition, at the individual level as well as at the contextual level, increases perceptions of ethnic out-group threat, which in turn increase antagonism toward immigrants among ethnic majority members (Blalock 1967; Scheepers et al. 2002). The competition may be objectively assessable or only subjectively perceived (Blalock 1967), but what matters is that individuals who experience ethnic competition are expected to perceive more ethnic threat (McLaren 2003; Schlueter and Scheepers 2010). Ethnic competition theory predicts that these perceptions are especially present among vulnerable economic groups who hold similar positions to ethnic minorities in general and even more so in times of economic decline (Quillian 1995; Scheepers et al. 2002; Billiet et al. 2014).

Individual differences in support for restrictive immigration policies

Previous research has supported the hypothesis that individuals in vulnerable economic positions – the lower educated, manual workers, the unemployed, and the self-employed – are more exclusionist toward immigrants (Scheepers et al. 2002; Schneider 2008; Mewes and Mau 2013; Billiet et al. 2014). In terms of ethnic competition theory, indicators of ethnic competition are expected to increase perceived ethnic threat among natives in similar economic positions as migrants, which in turn increases their restrictiveness toward immigration (Schlueter and Scheepers 2010). In line with previous research, we hypothesize that support for restrictive immigration policies is stronger among (a) natives with a lower level of educational attainment, (b) native manual workers, (c) unemployed natives, and (d) self-employed natives (H1a), because they perceive more ethnic threat (H1b).

Ethnic competition theory strongly emphasizes the role of ethnic competition and the consequent perceptions of ethnic threat that drive anti-immigrant sentiments. We argue that vulnerable socio-economic groups not only perceive more ethnic threat, but also more economic strain (Gesthuizen and Scheepers 2010), which in turn affects their restrictiveness toward immigration. In a context of ethnic competition over scarce jobs and welfare resources, socio-economic differences in restrictive immigration stances may be explained by perceptions of economic strain in two distinct ways. First, economically strained natives could have more or less rational reasons to be restrictive toward immigrants. Natives who face economic hardship will try to keep jobs or look for better jobs; or, in cases of unemployment, will claim welfare benefits. It is in their personal interest to limit the number of competitors in the labor market and the welfare state (Malchow-Moller et al. 2008). Migrants compete particularly with economically vulnerable natives in the labor market as well as in the realm of the welfare state, because compared to natives they have lower labor market resources and a stronger dependency on welfare benefits (Brücker et al. 2002; Barret and McCarthy 2008). Hence, economically vulnerable natives may support restrictive immigration policies because they think that immigration deteriorates their economic opportunities and entitlements. We hypothesize that support for restrictive immigration policies is stronger among (a) natives with a lower level of educational attainment, (b) native manual workers, (c) unemployed natives, and (d) self-employed natives, because they perceive more economic strain (H1c).

Second, to take the argument one step further building on classic theoretical mechanisms (Berkowitz 1988): natives who face economic hardship inducing frustration may also respond by blaming immigrants for their possible or future misfortune (Rothschild et al. 2012). They may ‘scapegoat’ ethnic out-groups, because they are currently highly salient groups in political and societal debates.² Hence economically strained natives may consider ethnic out-groups a threat to the economic interests of the ethnic in-group. We therefore expect that natives who perceive economic strain consequentially also perceive more ethnic threat, and for that reason they would be more restrictive toward immigration. We hypothesize that support for restrictive immigration policies is stronger among (a) natives with a lower level of educational attainment, (b) native manual workers, (c) unemployed natives, and (d) self-employed natives, because they perceive more economic strain and therefore more ethnic threat (H1d).

Changes in unemployment and national debt

Ethnic competition theory is especially suitable to explain antagonism toward immigrants in times of competitive economic circumstances. Economic conditions affect the degree of perceived ethnic competition over scarce economic resources such as employment and a share of the national welfare (Quillian 1995; Semyonov et al. 2008). The unemployment rate reflects scarcity of employment, which indicates ethnic labor market competition. The national government debt indicates the long-term and short-term sustainability of the financial position of national governments and thereby the economic and political relevance of austerity. In recent years, EU member states with sharp increases of national debt – such as Cyprus, Greece, Ireland, Portugal, and Spain – had to be financially supported by other EU member states and financial institutions such as the International Monetary Fund, in exchange for austerity measures to decrease the annual national debt and hereby improve the sustainability of the government finances (European Central Bank 2015). Within countries, people in vulnerable socio-economic positions are strongly affected by both the unemployment rate and national government’s austerity measures than people in higher positions. The former are, on average, more at risk of being or becoming unemployed (Eurostat 2015) and budget cuts have a relative stronger effect on their income position because they receive more (frequently) support from the government (Eurostat 2014e). Previous research on the relation between macro-economic conditions and anti-immigrant attitudes used a limited set of macro-economic indicators, namely (changes in) national welfare and unemployment rate. Some studies found less antagonism toward immigrants in economically more affluent countries (Schneider 2008; Semyonov et al. 2008; Billiet et al. 2014), although other studies did not establish a significant association (Semyonov et al. 2006; Gorodzeisky and Semyonov 2009; Meuleman et al. 2009). With regard to the unemployment rate, many studies did not find a significant association with anti-immigrant attitudes (Scheepers et al. 2002; Sides and Citrin 2007; Meuleman et al. 2009; Billiet et al. 2014). However, Coenders and Scheepers (2008) established a positive association between change in unemployment in the past five years and antagonism toward immigrants.

Following Olzak (1994) and Meuleman et al. (2009), we argue that sudden economic changes may shock the native population. A constant high level of competition may not affect antagonism toward immigrants, but if economic conditions suddenly deteriorate, people may become more antagonistic toward ethnic out-groups (Meuleman et al. 2009; Turner and Cross 2015). People might be overwhelmed – shocked – by these sudden economic changes and its (actual and perceived) consequences for the labor market and welfare state, because there was little time to absorb the changes (Olzak 1994). People may be exposed directly and indirectly and, moreover, alarmingly via media messages (Hopkins 2007). In this vein, Schlueter and Davidov (2013) showed that negative immigration-related news reports increased perceived group threat, over and beyond immigrant group sizes. In times of higher levels of unemployment and national government debt, employment and government resources generally become scarcer.³ We expect that short-term economic deterioration increases perceptions of economic strain and ethnic threat, which is in turn related to more restrictiveness toward immigrants. We hypothesize that support for restrictive immigration policies is stronger in countries that experienced a stronger increase of (a) unemployment and (b) national government debt (H2).

Following the reasoning of Scheepers et al. (2002), we propose that natives differ in the degree to which economic conditions affect their support for restrictive immigration policies. In times of stronger economic decline, natives who are generally more vulnerable and therefore more likely to perceive economic strain and ethnic threat are expected to support restrictive immigration policies more strongly. Hence, we expect that differences in support between economic groups are larger in times of stronger economic decline. We hypothesize that support for restrictive immigration policies is stronger among (a) natives with a lower level of educational attainment, (b) native manual workers, (c) unemployed natives, and (d) self-employed natives, especially in countries that experienced a stronger increase of (i) unemployment and (ii) national government debt (H3).

We further expect that natives who perceive more economic strain and/or ethnic threat are more restrictive in times of recent economic decline. In times of scarcity those who feel threatened and/or economically strained may feel even more strained or threatened, and therefore have a more urgent need to limit the number of immigrants. We therefore hypothesize that support for restrictive immigration policies is stronger among (a) natives who perceive more economic strain and (b) natives who perceive more ethnic threat, especially in countries that experienced a stronger increase of (i) unemployment and (ii) national government debt (H4). Our hypotheses are summarized in Figure 1.

Figure 1. Overview of hypotheses.

Data, measurements, and analytical strategy

Data description

Data on support for restrictive immigration policies and individuals’ economic position were derived from the ESS. The ESS is a biennial multi-country survey. What makes the ESS unique in cross-national research is its aim to meet the highest methodological standards. In order for the information gathered to be truly comparable across all countries involved, the survey employs the highest standards in its approach to sample design, response rates, questionnaire design, and fieldwork procedures across all participating countries. The target group encompassed all persons over 15 years of age who live in a private household in one of the participating countries. The individual chance of selection is independent of nationality, legal status, citizenship, or language. Respondents were randomly drawn from the population of individuals within each country, in order to reach representativeness of each sample. We use the most recent data of six ESS waves (2002, 2004, 2006, 2008, 2010, 2012). Our data are structured in three levels: individuals are nested in country-year combinations, which are in turn nested in countries. We selected all countries that were EU member states during at least half of the data collection period (2002–2013), hence all countries that were a member on 1 January 2007. From these 27 countries, 26 countries participated in at least 1 wave of the ESS. Only Malta did not participate in any wave. We selected all country-year combinations that were available (n_cy = 119). From each country-year combination, we only selected native-born respondents (n_i = 210,247), because the sampling procedure was not designed to optimize population representativeness among migrants. In addition, migrants may have different interpretations or interests with regard to support for restrictive immigration policies.⁴ The country-year variables were derived from Eurostat (2014a,b,c,f). The indicators were standardized across countries and over time, which makes them highly comparable and equivalent. In our analyses, we control for the levels of immigration, unemployment, national government debt, and for recent changes in immigration.

Individual-level variables

The dependent variable support for restrictive immigration policies is measured by three items: ‘to what extent do you think [country] should allow a) people of the same race or ethnic group as most [country]’s people to come and live here?; b) people of a different race or ethnic group from most [country] people?; and c) people from the poorer countries outside Europe?’. After recoding, response categories ranged from (0) ‘allow many to come and live here’ to (3) ‘allow none’. Perceived ethnic threat is, just like Coenders et al. (2013, 2014), measured by three items: (1) ‘Would you say it is generally bad or good for [country]’s economy that people come to live here from other countries?’; (2) ‘Would you say that [country]’s cultural life is generally undermined or enriched by people coming to live here from other countries?’; (3) ‘Is [country] made a worse or a better place to live by people coming to live here from other countries?’. For these items, the response scale was reversed, so 0 indicated no perceived ethnic threat and 10 indicated maximum perceived ethnic threat.⁵ Perceived economic strain is, just like Visser et al. (2013), measured with a single item: ‘Which of the descriptions on this card comes closest to how you feel about your household’s income nowadays?’ The response categories were (1) ‘living comfortably on present income’, (2) ‘coping on present income’, (3) ‘finding it difficult on present income’, and (4) ‘finding it very difficult on present income’.

The level of education is measured as the highest level of education respondents achieved, coded in the following ordered categories: 0, ‘Less than lower secondary education (ISCED 0–1)’; 1, ‘Lower secondary education completed (ISCED 2)’; 2, ‘Upper secondary education completed (ISCED 3)’; 3, ‘Post-secondary non-tertiary education completed (ISCED 4)’; 4, ‘Tertiary education completed (ISCED 5–6)’. This variable is included in a linear fashion. Employment status is measured nominally, with the categories ‘non-manual worker’, ‘manual worker’, ‘self-employed’, ‘unemployed’, ‘retired’, and ‘not in labor market for other reasons’. Non-manual workers are the reference category. We focus on differences in support with manual workers, the self-employed and unemployed people. At the individual level, we control for variables that were previously found to be related to antagonism toward immigrants: religiosity (indicated by frequency of church attendance), age, and gender (cf. McFarland 1989; Billiet 1995; Coenders and Scheepers 2008; Semyonov et al. 2008).

Contextual independent variables

Unemployment rate is measured as the natural logarithm of country’s annually average percentage of people who were unemployed in the year of survey. Unemployed persons are all persons 15–74 years of age who were not employed during the reference week, had actively sought work during the past 4 weeks, and were ready to begin working immediately or within 2 weeks (Eurostat 2014f). National debt is measured as the natural logarithm of the general government gross debt expressed in percentages of GDP in the year of survey. Government debt is defined in the Maastricht Treaty as consolidated general government gross debt of the whole general government sector at nominal value, outstanding at the end of the year. The general government sector comprises central government, state government, local government, and social security funds. The relevant definitions are provided in Council Regulation 479/2009, as amended by Council Regulation 679/2010. Data for the general government sector are consolidated between sub-sectors at the national level. The series are measured in percentage of GDP (Eurostat 2014a). The level of immigration is measured as the natural logarithm of the number of immigrants per 1000 inhabitants aged 15–64 (the working population) (Eurostat 2014b,c). ‘Immigration’ denotes the action by which a person establishes his or her usual residence in the territory of a member state (of the EU) for a period that is, or is expected to be, of at least 12 months, having previously been usually resident in another member state (of the EU) or a third country. Change in unemployment, national debt, and immigration is measured as the absolute difference between the year of survey and the preceding year. The differences are calculated with the operationalizations described above.⁶

Analytical strategy

Individuals (n_i = 210,247) are nested in country-year combinations (n_cy = 119), which are in turn nested in countries (n_c = 26). We therefore conducted two-level hierarchical ordinary least squares (OLS) structural equation modeling (level 1: native individuals; level 2: country-years) in Mplus 7.31 (Muthén and Muthén 1998–2015). We computed robust standard errors for the country-year level parameters, in order to take into account the stratification and non-independence of country-year observations. We dealt with missing values on individual-level variables by means of multiple imputation and created 100 new datasets with imputed data (Graham et al. 2007). For each respondent we calculated the mean score support for restrictive immigration policies and perceived ethnic threat.⁷ All independent variables were grand-mean centered. We used the maximum likelihood estimation with robust standard errors and the expectation–maximization algorithm. In the models with cross-level interactions between changes in unemployment, respectively, national debt and individuals’ economic position, perceived economic strain, and perceived ethnic threat, the effects of the relevant individual-level variables were made random across country-years. In the mediation models, we specified perceived economic strain and perceived ethnic threat as mediating variables that were predicted by education and employment status (the categories manual worker, unemployed, and self-employed), that in turn predict support for restrictive immigration policies. Perceived economic strain is also modeled to have an effect on perceived ethnic threat.

Results

Country differences in support for restrictive immigration policies

Member states of the EU substantially differ in the degree of public support for restrictive immigration policies. On a scale from 0 to 3, the average value of support is 1.470, which means that, in the period 2002–2013, EU natives wished to allow ‘a few’ to ‘some’ people from other countries (see Table 1). There is substantial between-country variation, and some overtime variation within and between countries. Public support for restrictive immigration policies is, on average between 2002 and 2013, lowest in Sweden (m = 0.828), Poland (m = 1.158), Bulgaria (m = 1.258), Germany (m = 1.296), and Denmark (m = 1.301), and it is highest in Greece (m = 2.007), Cyprus (m = 1.944), Hungary (m = 1.881), Latvia (m = 1.872), and Portugal (m = 1.860). Across all years of participation in the ESS, the most restrictive country, Greece, experienced stronger increases of unemployment and national debt compared to the least restrictive country, Sweden. In terms of bivariate correlations, changes in unemployment (r = 0.173) and national debt (r = 0.263) are stronger related to support for restrictive immigration policies than changes in immigration, and the levels of unemployment, national debt, and immigration (see Appendix 3).

Table 1. Mean support for restrictive immigration policies per country-year (N_i = 210,247; N_cy = 119; N_c = 26; average results of 100 multiple imputed datasets).

Country/year	2002/2003	2004/2005	2006/2007	2008/2009	2010/2011	2012/2013	(Country)
Austria	1.640 (0.715)	1.427 (0.774)	1.507 (0.737)	1.444 (0.770)			1.505 (0.754)
Belgium	1.429 (0.742)	1.475 (0.795)	1.380 (0.746)	1.325 (0.756)	1.459 (0.774)	1.383 (0.707)	1.408 (0.755)
Bulgaria				1.197 (0.929)		1.318 (0.939)	1.258 (0.936)
Cyprus			1.954 (0.590)	1.734 (0.481)	1.943 (0.676)	2.175 (0.613)	1.944 (0.612)
Czech Republic	1.543 (0.727)	1.681 (0.795)		1.739 (0.740)	1.773 (0.771)	1.798 (0.772)	1.717 (0.771)
Denmark	1.360 (0.668)	1.379 (0.682)	1.292 (0.650)	1.285 (0.673)	1.236 (0.662)	1.261 (0.673)	1.301 (0.670)
Estonia		1.776 (0.764)	1.715 (0.750)	1.646 (0.774)	1.552 (0.731)	1.508 (0.727)	1.630 (0.755)
Finland	1.539 (0.679)	1.564 (0.670)	1.547 (0.658)	1.461 (0.677)	1.630 (0.646)	1.463 (0.650)	1.531 (0.666)
France			1.503 (0.725)	1.436 (0.703)	1.459 (0.717)	1.383 (0.728)	1.445 (0.719)
Germany	1.320 (0.696)	1.493 (0.786)	1.466 (0.769)	1.198 (0.754)	1.249 (0.741)	1.045 (0.669)	1.296 (0.753)
Greece					2.007 (0.753)		2.007 (0.753)
Hungary	1.854 (0.625)	1.945 (0.756)	1.981 (0.720)	1.837 (0.703)	1.811 (0.727)	1.868 (0.751)	1.881 (0.718)
Ireland	1.243 (0.670)	1.239 (0.753)	1.231 (0.737)	1.405 (0.779)	1.587 (0.870)	1.513 (0.845)	1.379 (0.797)
Italy	1.258 (0.801)	1.398 (0.849)				1.234 (0.860)	1.311 (0.840)
Latvia			1.879 (0.923)	1.865 (0.884)			1.872 (0.904)
Lithuania				1.376 (0.817)	1.201 (0.868)	1.309 (0.816)	1.301 (0.834)
Luxembourg	1.556 (0.791)	1.500 (0.784)					1.527 (0.788)
The Netherlands	1.425 (0.668)	1.478 (0.729)	1.509 (0.754)	1.332 (0.705)	1.373 (0.728)	1.370 (0.750)	1.416 (0.723)
Poland	1.361 (0.690)	1.261 (0.775)	1.075 (0.776)	1.047 (0.738)	1.062 (0.741)	1.095 (0.766)	1.158 (0.757)
Portugal	1.775 (0.803)	1.894 (0.775)	1.904 (0.848)	1.838 (0.847)	1.805 (0.803)	1.919 (0.861)	1.860 (0.827)
Romania				1.439 (0.944)			1.439 (0.944)
Slovakia		1.264 (0.827)	1.266 (0.847)	1.387 (0.867)	1.552 (0.861)	1.606 (0.864)	1.423 (0.866)
Slovenia	1.434 (0.701)	1.457 (0.779)	1.461 (0.755)	1.389 (0.739)	1.373 (0.732)	1.357 (0.798)	1.414 (0.751)
Spain	1.420 (0.814)	1.393 (0.837)	1.540 (0.854)	1.665 (0.830)	1.509 (0.863)	1.340 (0.890)	1.491 (0.855)
Sweden	0.889 (0.645)	0.893 (0.702)	0.822 (0.667)	0.782 (0.655)	0.737 (0.622)	0.818 (0.654)	0.828 (0.661)
United Kingdom	1.531 (0.746)	1.495 (0.755)	1.589 (0.755)	1.556 (0.743)	1.631 (0.793)	1.665 (0.781)	1.581 (0.765)
(Wave)	1.437 (0.743)	1.479 (0.803)	1.499 (0.811)	1.451 (0.817)	1.514 (0.817)	1.442 (0.828)	1.470 (0.807)

Source: ESS pooled dataset wave 1 (2002/2003); wave 2 (2004/2005), wave 3 (2006/2007), wave 4 (2008/2009), wave 5 (2010/2011), wave 6 (2012/2013). Author’s calculations.

Note: Standard deviations are given in brackets.

We highlight some remarkable changes in support within and between countries. In Estonia, Germany, and Poland, support gradually declined over time. The economies of these countries generally grew in that period. Public support for restrictive immigration policies increased in Cyprus and Ireland. Just like Greece, these countries experienced economic difficulties and needed substantial support from the International Monetary Fund and the European Central Bank. In other countries, there are less pronounced changes in mean support over time.

Individual differences in support for restrictive immigration policies

The differences in support for restrictive immigration policies between native individuals from different economic groups are presented in the upper half of Model 1 in Table 2. In accordance with H1, we find that the lower educated are more restrictive toward immigration than the higher educated (γ = −0.096, p < .001). Next, native manual workers (γ = 0.144, p < .001), unemployed natives (γ = 0.184, p < .001), and self-employed natives (γ = 0.073, p < .001) are more restrictive than native non-manual workers. Furthermore, retired natives (γ = 0.102, p < .001) and natives who are not in the labor market (γ = 0.116, p < .001) are more restrictive than native non-manual workers. These findings are in accordance with propositions from ethnic competition theory. We further find that older natives (γ = 0.005, p < .001) are more restrictive than younger natives, while we do not find statistically significant effects of church attendance and gender.

Table 2. Hierarchical OLS regression estimates of individual support for restrictive immigration policies on individual and country-year characteristics (N_i = 210,247; N_cy = 119; N_c = 26; average results of 100 multiple imputed datasets).

	γ	SE	γ	SE	γ	SE
	Model 1		Model 2		Model 3
Intercept	1.473	(0.046)***	1.473	(0.045)***	1.473	(0.045)***
Individual-level variables
Education	−0.096	(0.006)***	−0.156	(0.055)**	−0.157	(0.055)**
Non-manual worker (ref.)
Manual worker	0.144	(0.010)***	0.364	(0.644)	0.364	(0.648)
Unemployed	0.184	(0.023)***	0.931	(0.915)	0.927	(0.921)
Self-employed	0.073	(0.009)***	−0.523	(1.902)	0.553	(1.940)
Retired	0.102	(0.011)***	0.104	(0.011)***	0.104	(0.011)***
Not in labor market	0.116	(0.009)***	0.118	(0.009)***	0.119	(0.009)***
Church attendance	−0.007	(0.005)	−0.007	(0.005)	−0.007	(0.005)
Age	0.005	(0.001)***	0.005	(0.001)***	0.005	(0.001)***
Female	0.007	(0.012)	0.005	(0.012)	0.005	(0.012)
Country-year-level variables
Δ Unemployment rate	0.016	(0.021)	0.018	(0.020)	0.023	(0.019)
*Education			−0.002	(0.002)
*Manual worker			0.001	(0.007)
*Unemployed			0.014	(0.010)∼
*Self-employed			0.005	(0.005)
Δ National debt	0.011	(0.007)∼	0.007	(0.007)	0.008	(0.007)
*Education					−0.001	(0.000)
*Manual worker					0.000	(0.001)
*Unemployed					0.002	(0.002)
*Self-employed					−0.002	(0.001)∼
Δ Immigration	−0.012	(0.006)*	−0.010	(0.004)*	−0.010	(0.004)*
Unemployment rate	−0.012	(0.006)*	−0.017	(0.005)**	−0.017	(0.005)**
National debt	0.000	(0.001)	0.001	(0.001)	0.001	(0.001)
Immigration	−0.003	(0.005)	−0.002	(0.004)	−0.002	(0.004)
−2 Log likelihood	468,340		1,242,502		1,242,502
BIC	468,560		1,242,919		1,242,918
	Model 4		Model 5		Model 6
Intercept	0.354	(0.042)***	−0.140	(0.331)	−0.141	(0.333)
Individual-level variables
Education	−0.030	(0.006)***	−0.030	(0.003)***	−0.030	(0.003)
Non-manual worker (ref.)
Manual worker	0.054	(0.006)***	0.054	(0.006)***	0.054	(0.006)***
Unemployed	0.050	(0.010)***	0.048	(0.009)***	0.048	(0.009)***
Self-employed	0.045	(0.006)***	0.045	(0.006)***	0.045	(0.006)***
Retired	0.035	(0.006)***	0.037	(0.007)***	0.037	(0.007)***
Not in labor market	0.041	(0.007)***	0.044	(0.007)***	0.044	(0.007)***
Church attendance	−0.001	(0.003)	−0.001	(0.003)	−0.001	(0.003)
Age	0.004	(0.000)***	0.004	(0.000)***	0.004	(0.000)***
Female	−0.014	(0.008)*	−0.014	(0.008)*	−0.014	(0.008)*
Perceived economic strain	0.022	(0.005)***	0.149	(0.102)∼	0.149	(0.102)∼
Perceived ethnic threat	0.213	(0.004)***	0.040	(0.069)	0.040	(0.069)
Country-year-level variables
Δ Unemployment rate	0.005	(0.015)	0.000	(0.019)	0.003	(0.014)
*Economic strain			0.002	(0.002)
*Ethnic threat			−0.001	(0.002)
Δ National debt	0.006	(0.004)∼	0.004	(0.003)∼	−0.004	(0.004)
*Economic strain					0.001	(0.001)∼
*Ethnic threat					0.001	(0.000)**
	Model 4		Model 5		Model 6
Δ Immigration	−0.007	(0.003)*	−0.006	(0.004)∼	−0.006	(0.004)∼
Unemployment rate	−0.008	(0.004)*	−0.011	(0.007)*	−0.011	(0.007)*
National debt	0.000	(0.001)	0.000	(0.001)	0.000	(0.001)
Immigration	0.000	(0.004)	0.002	(0.003)	0.002	(0.003)
−2 Log likelihood	391,353		1,717,941		1,717,935
BIC	391,598		1,718,296		1,718,290

Source: ESS’s wave 1 (2002/2003), wave 2 (2004/2005); wave 3 (2006/2007), wave 4 (2008/2009), wave 5 (2010/2011), and wave 6 (2012/2013). Authors’ calculations.

Notes: Unstandardized effects are presented, standard errors are given in brackets. All exogenous variables are grand-mean centered. The effects of education, manual worker, unemployed, and self-employed are random across country-years in Models 2 and 3. The effects of perceived economic strain and perceived ethnic threat are random across country-years in Models 5 and 6. ∼p < .10; *p < .05; **p < .01; ***p < .001.

Mediating effects of perceived economic strain and perceived ethnic threat

In Model 4 of Table 2, we have a first look at the parameters of the mediators’ perceived economic strain and perceived ethnic threat that actually turn out to be significant (0.022, respectively, 0.213). Moreover, in Table 3, we present the results of the mediation analyses.⁸

Table 3. Decomposition of direct, indirect, and total individual effects of individual socio-economic position on restrictiveness toward immigrants (N_i = 210,247; N_cy = 119; N_c = 26; average results of 100 multiple imputed datasets).

Dependent variable:	Restrictiveness toward immigrants
Independent variable:	Education	Manual worker	Unemployed	Self-employed	Economic strain
Direct effect	−0.030 (0.003)***	0.054 (0.006)***	0.050 (0.010)***	0.045 (0.006)***	0.022 (0.005)***
Indirect effects (total)	−0.078 (0.006)***	0.035 (0.009)***	0.098 (0.018)***	−0.020 (0.010)*	0.087 (0.012)***
Perceived economic strain	−0.003 (0.001)***	0.002 (0.001)**	0.016 (0.004)***	−0.003 (0.001)**
Strain via ethnic threat	−0.012 (0.002)***	0.009 (0.002)***	0.063 (0.010)***	−0.011 (0.004)**
Perceived ethnic threat	−0.063 (0.005)***	0.024 (0.007)***	0.019 (0.012)	−0.007 (0.009)	0.087 (0.012)***
Total effect	−0.108 (0.007)***	0.089 (0.011)***	0.148 (0.026)***	0.025 (0.012)*	0.109 (0.012)***

Source: ESS’s wave 1 (2002/2003), wave 2 (2004/2005); wave 3 (2006/2007), wave 4 (2008/2009), wave 5 (2010/2011), and wave 6 (2012/2013). Authors’ calculations.

Notes: Unstandardized effects are presented (standard errors are given in brackets). All exogenous variables are grand-mean centered. ∼p < .10; *p < .05; **p < .01; ***p < .001.

There are three mediating mechanisms (see Appendix 4). First, we find that the lower educated perceive more ethnic threat than the higher educated (γ = −0.297, p < .001), and that manual workers perceive more ethnic threat than non-manual workers (γ = 0.113. p < .001). The unemployed and the self-employed do not perceive more or less ethnic threat than non-manual workers. In turn, we find in Table 3 that those, who perceive more ethnic threat, are more supportive of restrictive immigration policies (indirect effects are γ = −0.063, respectively, 0.024, p < .05). Our findings support H1b regarding the lower educated and manual workers, but not for the unemployed and the self-employed. Second, we find in Appendix 4 that the lower educated (γ = −0.137, p < .001), manual workers (γ = 0.104, p < .001), and the unemployed (γ = 0.723, p < .001) perceive more economic strain, while the self-employed perceive less strain (γ = −0.126, p < .001). In turn, we find in Table 3 that those who perceive more economic strain are more supportive of restrictive immigration policies (indirect effects are −0.003 (p < .001), 0.002 (p < .01), 0.016 (p < .001), and −0.003 (p < .01) for, respectively, education, manual workers, the unemployed, and the self-employed. These findings confirm H1c. Third, we find in Appendix 4 that those who perceive more economic strain actually perceive more ethnic threat (β = 0.407, p < .001), which in turn increases support for restrictive immigration policies. So we find in Table 3 that lower educated (−0.012, p < .001), manual workers (0.009, p < .001), and the unemployed (0.063, p < .001) are more supportive of restrictive immigration policies, because they perceive more ethnic threat, that is fueled by perceptions of economic strain. The self-employed (−0.011, p < .01) are less supportive because they perceive less ethnic threat fueled by economic strain. We find that about 80% of the effect of perceived economic strain (β = 0.109, p < .001) is mediated by perceived ethnic threat (β = 0.087, p < .001). These findings corroborate H1d.

For matters of clarity, Figure 2 shows the mediating mechanisms of the relation between educational attainment and support for restrictive immigration policies. Natives with a lower level of education are more supportive of restrictive immigration policies (γ = −0.030, p < .001). They perceive more economic strain (γ = −0.137, p < .001) and more ethnic threat (γ = −0.297, p < .001). Economic strain directly induces support (β = 0.022, p < .001), but also induces ethnic threat (β = 0.407, p < .001). In turn, ethnic threat induces support for restrictive immigration policies (β = 0.213, p < .001).

Figure 2. Relation between educational attainment and anti-immigration attitudes.

In terms of to what extent the effects of socio-economic position are mediated by perceived economic strain and ethnic threat, we find that, respectively, 72%, 39%, and 66% of the effects of education, manual work, and unemployment are explained by economic strain and ethnic threat. Interestingly, the indirect effect of education (γ = −0.078, p < .001) and manual work (γ = 0.035, p < .001) runs mainly via perceived ethnic threat (γ = −.063, p < .001, resp., γ = 0.024, p < .001), while the indirect effect of unemployment (γ = 0.098, p < .001) mainly runs via economic strain through economic threat (γ = 0.063, p < .001). Thus, unemployed natives perceive more economic strain, which induces perceived ethnic threat, which in turn induces their support of restrictive immigration policies. We further find a negative indirect effect of self-employment. The main reason for this effect is that self-employed natives perceive less economic strain (γ = 0.126, p < .001), which also decreases their perceived ethnic threat.

Changes in unemployment and national debt

The effects of national economic conditions on support for restrictive immigration policies are presented in the bottom half of Model 1 in Table 2. In accordance with H2b, we find that in countries that experienced a stronger recent increase of national government debt (γ = 0.011, p < .10), natives more strongly support restrictive immigration policies. The effect of the increase of unemployment is not statistically significant (γ = 0.016, p > .10), refuting H2a. Regarding the macro-level control variables, we find that the effects of immigration change and unemployment rate (both γ = −0.012, p < .05) are contrary to what we would expect. The effects of national debt level and immigration level are not statistically significant.

The cross-level interactions between individual’s socio-economic position and changes in unemployment and national debt are presented in Model 2 and Model 3 of Table 2. We find that unemployed natives are somewhat more supportive of restrictive immigration policies in times of a stronger increase of unemployment (γ = 0.014, p < .10). This finding is in accordance with H3c. Furthermore, self-employed natives are less restrictive in times of a stronger increase of national debt (γ = −0.002, p < .10). This finding refutes H3d. More in general, there is little empirical support for our hypothesis that specific economic groups are more restrictive in times of stronger economic decline.

In Models 4–6 of Table 2, we present the unconditional and conditional effects of perceived economic strain and perceived ethnic threat. We already saw that natives who perceive more economic strain (γ = 0.022, p < .001) and more ethnic threat (γ = 0.213, p < .001) are more supportive of restrictive immigration policies. Tests of cross-level interactions presented in Model 6 of Table 2 show that in times of a stronger increase of national debt, the effects of economic strain (γ = 0.001, p < .10) and ethnic threat (γ = 0.001, p < .01) are stronger. This finding corroborates H4a/b with regard to change in national debt. We did not find corresponding conditional effects with regard to changes in the unemployment rate (γ = 0.002, respectively, γ = −0.001, p > .10).

Conclusion and discussion

Driven by theoretical notions of ethnic competition theory, we examined to what extent strong changes in national economic conditions affected natives’ support for restrictive immigration policies in 26 EU member states between 2002 and 2013. This theory proposes that antagonism toward immigrants is especially prevalent among vulnerable economic groups who hold positions similar to migrants – and even more so under competitive economic conditions. We argued that especially sudden changes in economic conditions fuel natives’ restrictiveness. In the long run, people may get used to the certain levels of unemployment and welfare. However, sudden economic changes such as during the recent financial-economic crisis may have ignited perceptions of economic strain and ethnic threat, which in turn fuel antagonism toward immigrants (Olzak 1994; cf. Meuleman et al. 2009), and consequently support for restrictive immigration policies.

Using large-scale data covering a more extensive time period than previous studies and virtually all member states of the EU, we found empirical support for our hypothesis, derived from ethnic competition theory (Blalock 1967; Scheepers et al. 2002), that restrictiveness toward immigrants is especially strong among vulnerable natives who are lower educated, manual workers, or unemployed. They perceive more economic strain and more ethnic threat. Natives who perceive more economic strain support more restrictive immigration policies. Much of the relation between economic strain and restrictiveness is explained by perceived ethnic threat. Natives who experience economic difficulties view ethnic out-group members as a threat more strongly. In turn, these perceptions of threat are related to support for restrictive immigration policies. These findings underline the theoretical relevance of the mechanisms we deduced, which is a rather innovative addition to this line of theorizing. Ethnic competition theory traditionally emphasizes the mediating role of perceived ethnic threat, perceptions that are especially prevalent among vulnerable economic groups. We argued and showed that perceptions of economic strain in turn explain (part of) the relation between individual natives’ socio-economic position and their perceptions of ethnic threat. Moreover, economic strain directly affects the degree of restrictiveness. To some extent natives perceive more threat from ethnic out-groups because they face economic hardship themselves. This finding especially holds for unemployed natives, whose restrictiveness can be largely attributed to perceived economic strain and not solely to perceived ethnic threat. That said, perceived ethnic threat remains the strongest predictor of support for restrictiveness toward immigration (cf. Schneider 2008).

Recent increases of government debts are moderately related to stronger public support for restrictive immigration, while recent increases of unemployment did not affect restrictiveness. These economic changes affected all socio-economic groups’ restrictiveness equally. However, we found that in times of stronger national debts increases, natives who perceive more economic strain and who perceive more ethnic threat actually turn out to support more restrictive immigration policies. These are interesting findings, because it shows that natives’ perceptions matter more for anti-immigration attitudes than their actual socio-economic position. Furthermore, national debt increases are likely to result in austerity measures or at least to expectations of austerity measures – which generally lead to budget cuts in the domain of social protection and employment. Under these economic conditions, natives who are economically vulnerable or perceive threat from ethnic out-groups do feel even more vulnerable or threatened – feelings that actually ignite anti-immigration sentiments.

Our research shows that recent increases of immigration decrease support for restrictive immigration policies among natives. This is an odd finding from the perspective of ethnic competition theory, because the theory would expect that increased immigration indicates ethnic competition, inducing perceptions of threat which in turn would ignite out-group hostility. Pettigrew et al. (2010) showed that the presence of foreigners is also related to more positive interethnic contact, which in turn decreases hostility toward foreigners, partially via decreased individual and collective threat. The question remains whether the same mechanism is at work in times of sudden increases of immigration: basically, there is very little time to build up interethnic contact. It could be that the negative correlation between recent increases of immigration and restrictiveness is actually a time-lagged derivative of previous waves of immigration. Another explanation of the negative effect of changes in immigration on support for restrictiveness regards the operationalization of immigration. We treated immigrants as a general group, not distinguishing between Western and non-Western immigrants. Making this distinction may be useful in future research. Furthermore, in their meta-study of the effect of out-group size on anti-immigrant sentiment, Pottie-Sherman and Wilkes (2015) found that perceived out-group size is more likely to have a positive effect on anti-immigrant sentiments than actual immigration. Hence, future research should consider including both actual and perceived (changes in) immigration in their models.

Europe currently recovers from an economic crisis, not seen for many decades. Following the logic of ethnic competition theory, one would now expect a decline of antagonism toward immigrants. Our findings show, however, that economic conditions are not one-to-one related to migration attitudes. The study suggests that economic recovery will be associated to a decline in antagonism only if three conditions are met. First, natives’ perceptions of economic competition should decrease to the same extent as the objective decrease of economic competition. This is not necessarily the case (Blalock 1967). Second, the economic recovery should benefit all economic groups equally. If this is not the case, anti-immigrant sentiments may persist or even increase in specific vulnerable economic strata. Third, perceptions of economic ethnic threat should not be ‘replaced’ by other forms of ethnic threat. If in times of economic prosperity, economic ethnic threats may become less salient but at the same time perceived cultural and safety threats from particular ethnic out-groups might become more salient. Hence, the net result may be the same degree of antagonism toward immigrants. Should exclusionist attitudes toward immigrants indeed persist, then social cohesion and solidarity between ethnic groups may be impeded in Europe’s increasingly diverse societies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Menno van Setten is a Ph.D. candidate at the European Research Center on Migration and Ethnic Relations at Utrecht University. His primary research interests concern cross-national and longitudinal comparisons of ethnic exclusionism, in particular in relation to the welfare state. He is a member of a British–Dutch research team which collected panel data on political, economic, ethnic, and welfare attitudes in Britain and the Netherlands.

Peer Scheepers is a full professor Social Science Research Methodology and vice-dean in the Faculty of Social Sciences at Radboud University Nijmegen, served as a national coordinator of the ESS, and is a Member of the Royal Dutch Academy of Arts and Sciences (K.N.A.W.). His primary research interests concern (cross-national and longitudinal) comparisons regarding political attitudes and behavior, more particularly related to ethnic exclusionism and social capital. He has published numerous contributions in international scientific journals.

Marcel Lubbers is a professor in political sociology. He has published internationally on understanding radical right voting behavior and euroscepticism. For that research he obtained a VENI grant. Extending his work on nationalism and cultural consumption, he was recipient of a VIDI grant. He also collaborated in the international data collection program Social and Cultural Integration Processes of new immigrants (NORFACE), which is continued in the NWO-middelgroot program (2013–2018) ‘New Immigrant Survey in the Netherlands’.

Additional information

Funding

This work was supported by the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO) [grant number 464-11-060].

Notes

1 These groups are more likely to perceive more ethnic out-group threat, even if there is no actual competition.

2 For instance, ethnicity is salient in debates about immigration, welfare entitlement, and criminality. The scapegoating implies that the target group that is scapegoated can be changed. Should other groups become more salient in one way or the other, the scapegoating may be directed at these groups.

3 Government resources become scarcer if national debt increases, because EU member states are obliged to strive to a yearly budget deficit of less than 3% (Council Regulation 679/2010).

4 The number of respondents for each country-year combination is presented in Appendix 1.

5 Before researchers can adequately and validly examine causal interrelations between latent constructs, its level of measurement equivalence should be established (Van de Vijver and Leung 1997). This rule applies to all types of analyses where groups of respondents from different countries and at different points in time are examined. Unfortunately, the number of groups specified in a multiple group measurement model is limited. Therefore it is not possible to directly assess the level of measurement equivalence across all country-years in one multiple group model. Using χ²-based fit indices, we established partial metric invariance across countries within each wave of the ESS for both support for restrictive immigration policies and perceived ethnic threat (see Appendix 2). We calculated Likert mean scores for support for restrictive immigration policies and perceived ethnic threat to decrease the number of parameters that need to be estimated.

6 The mean and standard deviation of support for each country-year combination are presented in Table 1.

7 The individual-level and contextual-level correlations can be found in Appendix 3.

8 According to Preacher and Hayes (2004), a variable is considered a mediator if (1) X significantly predicts Y, (2) X significantly predicts M, and (3) M significantly predicts Y controlling for X. Furthermore, the mediator should be measured without error, and Y should not cause M (non-recursive effect). We can only assume that perceived economic strain is measured without error, because it is measured by only one item. Given our theoretical model, we think it is not plausible that support for restrictive immigration policies would induce more perceived ethnic threat or more perceived economic strain.). All effects of the Xs of interest on Y, all but two effects of X on M, and both effects of M on Y are statistically significant (see Appendix 4). Hence, our mediation model largely meets the criteria of Preacher and Hayes (2004). The full model can be found in Appendix 4.