Abstract

This study analyses the relationship between attitudes toward immigration and deteriorating economic conditions in times of crisis. We examine three questions: First, how are a vulnerable position in the labour market and recent changes to an individual’s economic situation related to perceived ethnic threat? Second, what is the role of the nation’s economic and immigration context? Last, are relationships at the individual level between economic conditions and perceived ethnic threat affected by contextual variables?

Data from 23 countries sampled in the fifth round of the European Social Survey (ESS-5, 2010) is used. At the micro level, unemployment, job insecurity and income deprivation during the three years prior to the survey affect perceived ethnic threat, as predicted by group conflict theory. These effects are, however, relatively small. Among the contextual variables, only growth in gross domestic product (GDP) shows an effect in the expected direction: perceived threat is higher in countries where GDP growth is lower. However, the study design does not allow the conclusion that changes in the economic context lead to changes in attitudes toward immigrants. The significant cross-level interaction for economic growth indicates that the threat-inducing effect of unemployment is stronger in contexts where the growth in GDP is high. This finding contradicts our hypothesis. One could explain this by the emergence of a generalized feeling of economic insecurity in countries severely hit by the economic crisis. In these countries, strong feelings of economic insecurity—and the resulting levels of perceived ethnic threat—might also be present among those who are employed, thereby diminishing the gap between them and the unemployed.

1. Introduction

The global financial meltdown—initiated by the collapse of Lehman Brothers in September 2008—has plunged Europe into its most severe economic crisis since the 1930s (Hemerijck, Knapen and Van Doorne 2010). Although the full impact of the crisis still has to become clear and varies considerably across countries, it is evident that it is severely affecting the social conditions of certain strata within European populations. Economic growth has declined substantially and was still negative in 2012 (e.g. a GDP of −0.6 per cent in the Eurozone and −0.5 per cent in the EU27),1 while in a number of countries (such as Greece, Cyprus, Portugal, Slovenia and Hungary), unemployment rates—especially among younger cohorts—have increased steeply (Eurostat 2013). In 2012, 10.5 per cent of the population of the EU27 countries was unemployed, which amounted to at least 16 million Europeans. The unemployment rate was even higher in the Eurozone, at 11.4 per cent—varying between 4.4 per cent in Austria and 25.0 per cent in Spain.2 At the same time, policymakers have been calling for austerity measures and welfare budgets are being cut. In sum, European populations are confronted with economic insecurity on a scale seldom witnessed before.

The current grim economic conditions also potentially affect intergroup relationships, and notably those between majority groups and immigrants. One of the most established approaches to explain attitudes toward immigration—namely group conflict theory (Blalock 1967; Olzak 1992)—contends that the roots of anti-immigration sentiments and ethnic prejudice should be sought in the realm of the economy. Group conflict theory predicts that economic insecurity induces ethnic competition for scarce material goods. As a result, negative attitudes toward immigration are expected to be more prevalent among individuals in vulnerable socio-economic positions, that is, those with lower educational status and income, the unemployed, and low-skilled workers (Lancee and Pardos-Prado 2013; Kunovich 2002; Fetzer 2000, 2012), and in countries with poor economic performance (Quillian 1995; Coenders 2001; Semyonov, Raijman and Gorodzeisky 2006).

In this study, we investigate empirically whether and how economic conditions—at the individual as well as the country level—influence perceived ethnic threat among the majority population. Against the background of an economic crisis, do indicators of individual economic insecurity affect threat perceptions? Does the severity of the crisis in a country have an impact on the levels of perceived threat among the population? Last, is the effect of individual economic insecurity different across contexts? To answer these questions, we analyse comparative survey data from round five of the European Social Survey (ESS-5 2010). This current study contributes to existing literature in several ways. First, the timing of the data collection (after the start of the economic crisis) allows us to test whether findings from previous research are still valid in times of economic crisis. Accordingly, this study attempts to enhance understanding of the consequences of the economic crisis. Second, we use several innovative measurement instruments to operationalize economic insecurity. Concretely, we analyse two scales gauging the perceived deterioration of income situation and job conditions by means of retrospective questions, as well as a one-item measurement of job (in)security. Third, we perform formal tests for measurement equivalence of the multiple indicator variables, thereby increasing the measurement validity of the study. Fourth, a comparative design using data from a wide variety of European countries allows us to study the extent to which the interplay between economic insecurity and perceived threat is context dependent.

2. Economic conditions and attitudes toward immigration

2.1 Anti-immigration attitudes or perceived ethnic threat as a response to perceived economic threat

The various versions of group conflict theory share as a central premise that the negative attitudes of group members toward other social groups are essentially rooted in perceived intergroup competition for scarce goods (Jackson 1993). In this line of thinking, anti-immigration attitudes—a specific form of negative out-group attitudes—stem from the perception that immigrant groups pose a threat to certain interests of the own social group. Perceived threat has even been identified as being an intrinsic factor in ethnic prejudice (Blumer 1958). The general idea of group conflict theory is that hostile attitudes toward immigration can be regarded as a defensive reaction to perceived intergroup competition for scarce goods.

According to the nature of the interests or the ‘scarce goods’ that are felt to be threatened, various types of perceived threat can be distinguished (Stephan et al. 1998; Stephan, Ybarra and Bachman 1999). Perceived economic threat has probably received the widest scientific attention (Olzak 1992; Quillian 1995; Citrin et al. 1997; Dustmann and Preston 2004; Meuleman 2011). This type of threat refers to the perception that social groups have to compete for scarce economic benefits, such as well-paid jobs, jobs that are secure, affordable housing, or welfare-state resources. Perceived economic threat is thus related to the view that majority and minority groups are locked in a zero-sum game for economic resources (Blumer 1958; Blalock 1967; Quillian 1995). It should be noted, however, that although material goods are important, they are not the only interests at stake in intergroup competition. Cultural (Zarate et al. 2004) or symbolic threats (Stephan, Ybarra and Bachman 1999), for example, refer to the perception that out-groups adhering to different cultural traditions pose a threat to the own world view, which is believed to be morally right (Stephan et al. 1998).

2.2 Individual and contextual determinants of perceived ethnic threat

Group conflict theory not only claims that perceptions of threat are the driving force behind anti-immigration attitudes, but also elaborates on how such threat perceptions are rooted in economic conditions at the micro (individual) as well as at the macro (country) level. In the individual formulation of this theoretical framework, feelings of ethnic threat are expected to develop primarily among those who are—or at least perceive themselves to be—individually exposed to competition from immigrant groups. Concretely, the perception of threat will be most widespread among individuals who hold similar positions to immigrant groups and are therefore most vulnerable to competition (Mayda 2006; O’Rourke and Sinnott 2006; Scheve and Slaughter 2001; Citrin et al. 1997; Fetzer 2000; Dustmann and Preston 2004; Scheepers, Gijsberts and Coenders 2002). Previous studies have provided empirical evidence for the existence of such self-interest effects. For example, low-skilled, blue-collar workers have been found to perceive stronger ethnic threat than their high-skilled, white-collar counterparts, as minority groups are predominantly active in the lower segments of the labour market (Cummings 1980; Mayda 2006). In addition, those at the lower end of the income distribution scale have been reported to hold less favourable out-group attitudes, probably because they possess fewer financial means to protect themselves against competition from minority groups. For similar reasons, workers who feel insecure about their job, experience deterioration in job conditions, or experience income deprivation can develop negative out-group attitudes, as these experiences might strengthen the perception that out-groups pose a threat to their economic position (De Weerdt, Catellani, De Witte and Milesi 2007). Further, the lower educated can be expected to be more prone to perceived economic threat (Kunovich 2004; Hello, Gijsberts and Scheepers 2002), although this education effect will probably also operate indirectly via income and occupational position. However, the relevance of education as a predictor of threat perceptions does not only derive from self-interest notions (Hagendoorn and Nekuee 1999). Education increases cognitive capacities, stimulates critical thinking and increases knowledge about foreign cultures (De Witte 1999; Hagendoorn and Nekuee 1999; Hainmueller and Hiscox 2007).

Group conflict theory is also useful in conceptualizing contextual effects, because it looks beyond intra-individual processes and takes group processes into account (Bobo 1983). The development of perceived threat is conceived as a fundamentally collective process, through which a particular social group comes to define other groups (Blumer 1958). Consequently, group conflict theorists also look for the origins of perceived threat in the context in which intergroup relationships take place. It is not accidental that most research building on group conflict theory considers threat as a collective phenomenon (Meuleman, Davidov and Billiet 2009). The trigger for hostility toward immigrants is the threat to the group’s resources or status, rather than to those of individuals (Lancee and Pardos-Prado 2013: 108).

Blalock’s (1967) distinction between actual and perceived competition is crucial to understanding the role played by contexts in shaping perceived threat (see also Semyonov et al. 2004). Actual competition is used to denote objective conditions of competition between members of different groups. Perceived competition, on the other hand, refers to the interpretations of this objective situation made by individual group members. It is the subjective perception that out-group members pose a threat to in-group interests (Bobo 1983). Blalock (1967) proposed a connection between actual and perceived competition: the degree of actual competition can determine the extent to which individuals experience ethnic rivalry. Thus, actual competition has an indirect rather than a direct impact on anti-immigration attitudes, via perceived competition and threat.

The concept of actual competition at the contextual level is usually operationalized by means of two sets of contextual variables, namely economic conditions and immigrant group size. In times of a downward economic trend (as has been the case since 2008) economic competition intensifies, because the material goods that are the object of competition become scarcer, consequently leading to higher levels of perceived threat. By contrast, in more prosperous times competition becomes less intense and the perception that majority and minority groups are locked in a zero-sum game is reduced (Blalock 1967). In addition to economic conditions, the immigrant group size (i.e. the proportion of immigrants in the population) is hypothesized as having a crucial impact on perceptions of threat (Blalock 1967). However, empirical evidence for the impact of national-level economic indicators on anti-immigration attitudes is mixed. Some studies have reported a benign influence of economic prosperity on out-group attitudes (Quillian 1995; Coenders 2001; Semyonov, Raijman and Gorodzeisky 2006), whereas others have not been able to replicate these results (Strabac and Listhaug 2008; Davidov and Meuleman 2012).3

There are good theoretical reasons to assume that individual and national economic conditions do not act independently of each other, but instead interact. The precise indicators of this cross-level interaction, however, are not immediately clear. Because perceptions of a declining national economy increase xenophobia (Jones and Lambert 1965; Kunovich 2002; Fetzer 2000, 2012), it seems logical to expect that the relationship between individual economic positions and ethnic threat will be more pronounced in poorer countries. However, empirical studies by Kunovich (2004) and Fachini and Mayda (2009) reported the opposite, namely that the effects on xenophobia of individual occupation and income are stronger in richer countries. Kunovich (2004: 25–6) suggested several explanations for this finding. These refer to changes in the relative position of disadvantaged groups and in the perception of collective threat in a context of economic downturn. It is not possible to single out a decisive explanation, but this current research clearly supports the idea that the effects of labour market position, education, and income on ethnic prejudice differ across countries, and that this is (partly) due to variation in the economic conditions of the countries (Kunovich 2004: 40).

2.3 Group conflict theory in times of economic crisis

In the past, the theoretical propositions set out above have been approached mostly from a static perspective, considering a single point in time (for an exception, see Lancee and Pardos-Prado 2013). Meuleman, Davidov and Billiet (2009), however, formulated a dynamic version of group conflict theory, arguing that changes in economic conditions might be more relevant than absolute levels of economic success with regard to our understanding of perceived ethnic threat. Following this logic, actual competition could remain constant at a high level without changing attitudes and it is only if sudden changes in economic conditions occur that out-group attitudes change. This dynamic formulation of group conflict theory is particularly useful in times of economic crisis, when severe shocks can suddenly affect individual as well as national economic security.

A strict test of this dynamic argument would ideally require analysing longitudinal cross-national survey data. Unfortunately, such data sources are non-existent at present. In practice, several second-best strategies can be developed as alternatives to assess the impact of economic shocks. First, it is possible to look at repeated cross-sectional data. Previous research has used the measurements of anti-immigration attitudes included in the ESS for this purpose. This approach has revealed that, on average, anti-immigration attitudes remained quite stable in Europe during the period from 2002 to 2006 (Meuleman, Davidov and Billiet 2009), while public opinion in European countries has predominantly moved toward a more restrictive approach to immigration since the beginning of the crisis in 2008 (Meuleman, Davidov and Billiet 2013). These findings therefore seem to be in line with the group conflict framework. A drawback, however, is that a repeated cross-sectional approach does not allow us to test whether the observed rise in anti-immigration attitudes since 2008 can be attributed to the economic crisis.

In this study, we follow a different approach and confine ourselves to the analysis of data from a single time point (namely the 2010 round of the ESS), but we use retrospective measurements that record changes in economic conditions both at the individual and at the country level.

2.4 Hypotheses

The hypothesized relationships between our theoretical concepts are shown in Fig. 1. In line with dynamic conflict group theory, three kinds of expected relationships (hypotheses) are distinguished: (1) relationships between attitudes toward immigration (perceived threat) and economic predictors at the micro level; (2) effects of context (country) level variables on the perception of ethnic threat; and (3) interaction effects between micro and macro-level predictors.

Figure 1.

Expected relationships between factors at the micro level and expected between level interactions.

Figure 1.

Expected relationships between factors at the micro level and expected between level interactions.

The general hypothesis at the micro level (H1) states that worse individual economic conditions and economic insecurity will increase perceived ethnic threat. One may expect that a vulnerable position in the job market is related to perceived ethnic threat (H1.1). More specifically, being unemployed or being in job categories that are most vulnerable to competition from unskilled immigrants (blue-collar jobs), are expected to foster feelings of threat (De Witte and Meuleman 2007). In line with the dynamic formulation of group conflict theory, we also expect that a deterioration in income (H1.2) and employment (H1.3) in the recent past will reinforce the perception of threat from immigrants. In addition to past experiences, poor future economic prospects may also affect the attitude toward immigrants. It could be expected that in the light of competition for jobs in a context where they are scarce, feeling insecure about employment will also affect the perception of threat (H1.4).

With regard to the direct effects of contextual factors on perceived threat, two general hypotheses are distinguished: one concerning the economic conditions and the other concerning the size of the immigrant population. The general hypotheses at the country (context) level predict higher perceived threat in countries with poor economic performance and a diverse population (H2). More specifically, increased perceived ethnic threat is expected in countries where economic growth is low (H2.1a) or has decreased (H2.1b), and in countries where the unemployment level is high (H2.2a) or has recently increased (H2.2b). This is because a marked deterioration of the economy in a country is likely to affect individuals’ perceptions. Furthermore, we expect that the size of the immigrant population (H2.3a) or a sudden increase (H2.3b) in the inflow of immigrants will affect the perception of ethnic threat.

Finally, the cross-level interaction hypothesis states that societal economic conditions and the size of the immigrant population affect the relationships between economic conditions and perceived threat as specified in Hypothesis H1. We expect that the micro-level relationships will become stronger when the economic situation in a country deteriorates because of the economic crisis (H3.1). Stronger relationships at the micro level are also expected if the size of the immigrant population has recently increased (H3.2). The dotted lines in Fig. 1 mean that we do not focus on these relationships in this study. The direct relationships between the control variables and perceived threat have been well documented in previous research and we do not specify interaction hypotheses with control variables.

3. Data and measurements

3.1 Data and samples

To study the impact of the economic crisis on attitudes toward immigration, data from the ESS round 5 (2010) is used. This round contains the rotating module ‘work and family’. Change is not studied within individual cases, since ESS data is not panel data and consists of repeated cross-sections rather than repeated measurements of the same individuals. The lack of ‘real’ longitudinal data is substituted by retrospective questions, used to measure the (subjective) perception of change in the period from 2008 to 2010. European countries have been confronted by a financial crisis since 2008 and the consequences of this were still not fully manifest during the data collection (autumn 2010 to spring 2011). We analyse whether changes in the economic conditions vary not only over time, but also between countries. The following 23 EU4 countries participated in the ESS-5: Belgium (BE – 744), Bulgaria (BG – 988), Switzerland (CH – 604), Cyprus (CY – 524), the Czech Republic (CZ – 1,277), Germany (DE – 1,433), Denmark (DK – 798), Estonia (EE – 733), Spain (ES – 959), Finland (FI – 925), France (FR – 841), the United Kingdom (GB – 1,044), Greece (GR – 1,114), Hungary (HU – 769), Ireland (IE – 1,012), Lithuania (LT – 729), the Netherlands (NL – 870), Norway (NO – 819), Poland (PL – 939), Portugal (PT – 872), Sweden (SE – 745), Slovenia (SI – 607), and the Slovak Republic (SK – 842).5

From the very outset, the ESS was designed as a research instrument aimed at making cross-cultural comparisons. Therefore, elaborate attention has been paid to ensure the methodological quality of the survey.6 Translation of the questionnaire into each native language, for example, follows the rigorous procedures for cross-cultural surveys set out in Harkness, Van de Vijver and Mohler (2003: 35–56). Respondents were selected by means of strict probability samples of the resident populations aged 15 and above. Although many countries were not able to meet the target response of 70 per cent, response rates are reasonably high for most countries.7 The total number of valid cases for all European countries in the ESS-5 is 50,781. Only part of the country samples are used in this study. Because our study deals with attitudes toward immigration, we exclude all foreign-born respondents, non-citizens, and people who stated they belong to an ethnic minority group. The samples are also restricted to residents who were engaged in paid work during the seven days preceding the survey or were unemployed. Because of these restrictions, the number of cases dropped to 25,201.

3.2 Measurements8

3.2.1 Micro-level variables: perceived ethnic threat

The ESS-5 contains three items measuring the concept ‘perceived ethnic threat’ (questions B38, B39 and B40). Respondents were asked to assess the consequences of immigration for the country’s economy, cultural life and living conditions in general (see Table 1 for the exact wordings). Answers were recorded on 11-point scales with specific end labels for each item. Confirmatory factor analysis shows that the three indicators adequately capture one single factor. The parameters of the selected measurement model for ‘perceived ethnic threat’ and the two other latent variables used in this study are reported in Appendix 1 (Tables A.1 and A.2). Based on these three items, we create a sum scale, THREAT, ranging from 0 to 10, where higher scores indicate a higher level of perceived threat. Fig. 2 presents country averages for the THREAT scale. There is considerable variation between countries, as the country averages range from 3 to 7.

Figure 2.

Country means for the THREAT scale.

The acronyms (country codes) are detailed in Section 3.1 where the countries in the sample are listed).

Figure 2.

Country means for the THREAT scale.

The acronyms (country codes) are detailed in Section 3.1 where the countries in the sample are listed).

Table 1.

Question wording and answer scales of the micro-level variables

Item Wording Response categories 
THREAT   
B23 Would you say it is generally bad or good for [country]’s economy that people come to live here from other countries? 0 (bad for the economy) – 10 (good for the economy) 
B39 Would you say that [country]’s cultural life is generally undermined or enriched by people coming to live here from other countries? 0 (cultural life undermined) – 10 (cultural life enriched) 
B40 Is [country] made a worse or a better place to live by people coming to live here from other countries? 0 (worse place to live) – 10 (better place to live) 

 
INCOME DEPRIVATION Please tell me to what extent each of the following has applied to you in the last three years.  
G8 I have had to manage on a lower household income. 0 (not at all) – 6 (a great deal) 
G9 I have had to draw on my savings or get into debt to cover ordinary living expenses. 
G10 I have had to cut back on holidays or new household equipment. 

 
 JOB DETERIORATION Please tell me whether or not each of the following has happened to you in the last three years. Have you … 
G58 … had to do less interesting work? 1 (yes) – 2 (no) 
G59 … had to take a reduction in pay? 
G60 … had to work shorter hours? 
Item Wording Response categories 
THREAT   
B23 Would you say it is generally bad or good for [country]’s economy that people come to live here from other countries? 0 (bad for the economy) – 10 (good for the economy) 
B39 Would you say that [country]’s cultural life is generally undermined or enriched by people coming to live here from other countries? 0 (cultural life undermined) – 10 (cultural life enriched) 
B40 Is [country] made a worse or a better place to live by people coming to live here from other countries? 0 (worse place to live) – 10 (better place to live) 

 
INCOME DEPRIVATION Please tell me to what extent each of the following has applied to you in the last three years.  
G8 I have had to manage on a lower household income. 0 (not at all) – 6 (a great deal) 
G9 I have had to draw on my savings or get into debt to cover ordinary living expenses. 
G10 I have had to cut back on holidays or new household equipment. 

 
 JOB DETERIORATION Please tell me whether or not each of the following has happened to you in the last three years. Have you … 
G58 … had to do less interesting work? 1 (yes) – 2 (no) 
G59 … had to take a reduction in pay? 
G60 … had to work shorter hours? 

Formal tests on cross-country equivalence in previous studies have shown that the latent factor THREAT can be used for comparative analyses, since at least partial invariance has been obtained (Meuleman and Billiet 2011). The formal equivalence test is repeated for the ESS 2010 data in an integrated measurement model of THREAT and the two other multiple indicator variables. The test information is reported in Table 1 and in Appendix 1. As in previous ESS rounds, perceived ethnic threat is quasi-scalar (and metric) invariant for the 23 countries used in this study, which is a prerequisite for making valid cross-national comparisons.

3.2.2 Individual level indicators of economic conditions

In the analysis, we use four indicators of individual-level economic conditions and the perceptions of it.

(1) Work situation is a nominal variable combining the main activity (working vs. unemployed) with occupational information for the employed respondents based on the Erikson–Goldthorpe–Portocarero classification scheme (see Ganzeboom and Treiman 1996). Five categories are distinguished: unemployed, self-employed, higher service class, white-collar workers, and blue-collar workers. Effect coding is used, so that the estimated parameters refer to comparisons with the grand mean over the five categories. This coding facilitates cross-national comparisons of the impact of unemployment on perceived threat. Note that one of the categories (here, white-collar workers) is excluded from the analysis to avoid perfect collinearity between the predictors.

(2) Job security is measured with a single item indicating how secure the respondent felt about her or his job: ‘My job is secure’ (1 = ‘not at all true’, to 4 = ‘very true’).

(3) Income deprivation is a retrospective scale, indicating deterioration in the financial situation of the respondents compared with the period before the financial crisis. The scale consists of three items (questions G8, G9 and G1) reflecting the extent to which the respondents had to manage on a lower household income, to draw on their savings, and to cut back on holidays or new household items during the preceding three years (see Table 1). Each of these items is measured on a 6-point scale (0 = ‘not at all’, to 6 = ‘a great deal’). In addition here, a sum scale ranging from 0 (no deterioration in income situation) to 10 (great deterioration in income situation), is constructed.

(4) Job deterioration is a retrospective scale capturing whether the job conditions of the respondents had deteriorated since the beginning of the financial crisis in 2008 (questions G58, G59 and G60). Respondents were asked to indicate whether, compared with three years previously, they did less interesting work, received a lower salary, or worked shorter hours (see Table 1). Factor analysis shows that these three indicators measure a single dimension. In the analysis, we use a sum scale of the three items, rescaled from 0 (no deterioration in job conditions) to 10 (maximum deterioration in job conditions).

The measurement model for the three multiple indicator (latent) variables (THREAT, income deprivation, and job deterioration) was subjected to a formal equivalence test. We tested whether scalar invariance (and consequently metric invariance) applied to the 23 country samples. In addition to the standard specifications in a measurement model for related latent variables measured by multiple indicators, the test specified that the corresponding intercept and slope parameters of the relationships between factors and indicators are invariant over countries (Billiet 2003). After a stepwise test, an acceptable measurement model characterized by partial scalar and metric equivalence was obtained (Byrne, Shavelson and Muthén 1989; Steenkamp and Baumgartner 1998) (see Appendix 1). This means that the three constructs are useful for the comparison of means and relationships between the latent variables over countries.

3.2.3 Individual-level control variables

Our analysis controls for several potentially confounding variables. Gender is a dummy variable with the value 0 for women and 1 for men. Age is operationalized as dummy variables indicating cohort membership: 16–25, 26–35, 36–45, 46–55, and 56–65. The middle category (36–45) is used as the reference category. Education is recoded into a variable with four categories, based on the international classification scheme ISCED (OECD 1999): no or only primary education, lower-secondary education, higher-secondary education, and tertiary education (used as the reference category). Finally, urbanization is measured by the respondent being asked to specify the area in which he/she lived (a large city, a suburb of a large city, a country village, or the countryside). This variable is recoded to range from 0 (rural area) to 10 (urban area).

3.2.4 Country-level variables

Our analysis includes various country variables measuring actual threat at the contextual level. For each of the variables, we include the level of these variables (averaged over the years 2009 and 2010 to increase the reliability of the measurements) as well as changes since the start of the financial crisis (i.e. a comparison of 2010 with 2007). The migration context is captured by the inflow of foreign population (per 1,000 inhabitants). This variable is taken from the OECD statistics database, supplemented with Eurostat9 data (indicator migr_imm1ctz) for Greece, Cyprus and Lithuania. Labour market conditions are measured by harmonized overall unemployment and long-term unemployment (i.e. longer than 12 months) rates, taken from Eurostat data. To indicate the country’s economic situation (and the severity of the crisis), we include the annual GDP growth rate as reported by Eurostat.

In the Appendix, descriptive statistics for individual-level characteristics (Appendix 2A) as well as a matrix containing the contextual data are shown (Appendix 2B).

3.3 Analysis

In order to take into account the hierarchical structure of the ESS data (citizens are nested within countries) and to study the interplay between contextual and individual characteristics, we make use of two-level regression modelling. The multilevel analysis is conducted in several steps. After the estimation of an empty model (M0), we introduce individual socio-demographic characteristics, including work situation (M1) and perceptions of job deterioration and job security (M2). In subsequent steps, contextual predictors (M3), random slopes (M4), and cross-level interaction effects (M5) are tested for.

All analyses are weighted to correct for cross-national differences in sampling design (dweight) as well as to take into account the relative population sizes of the countries in the analysis. Multiple imputation (Rubin 1987) with five imputations is used to deal with item non-response.

To facilitate the interpretation of intercept and interaction terms, all predictors apart from dummy variables are grand mean centred. As a result, the intercept refers to an average respondent with regard to these characteristics.

4. Results

The results of the multilevel analysis are summarized in Table 2. Estimation of an empty model (not shown) confirms what is seen in Fig. 2: perceptions of ethnic threat vary significantly across countries (variance of the random intercept: 0.681; p < .01) as well as across individuals within countries (residual variance: 4.014; p < .001). The intra-class correlation—representing the proportion of variance that is located at the country level—equals 0.145, which is substantial and legitimizes the choice to use multilevel modelling.

Table 2.

Parameter estimates and level of significance of the stepwise multilevel analysis (ESS round five: for 23 EU countries). The dependent variable is ‘perceived ethnic threat’

Explanatory variables Model 1 Model 2 Model 3 Model 4 
FIXED EFFECTS         
Intercept 4.492 *** 4.490 *** 4.249 *** 4.255 *** 
Gender         
    Men −0.227 ** −0.213 ** −0.213 ** −0.212 ** 
    Women (no est.)         
Age         
    15–25 years 0.110  0.123  0.124  0.129  
    26–35 years 0.092  0.090  0.090  0.088  
    36–45 years (no est.)         
    46–55 years 0.084  0.093  0.093  0.093  
    56–65 years 0.038  0.080  0.080  0.074  
Education         
    None or primary 1.295 *** 1.260 *** 1.261 *** 1.297 *** 
    Lower secondary 1.111 *** 1.083 *** 1.083 *** 1.090 *** 
    Higher secondary 0.821 *** 0.804 *** 0.804 *** 0.807 *** 
    Tertiary (no est.)         
Urbanization −0.027 *** −0.030 *** −0.030 *** −0.031 *** 
Work situation         
    Unemployed 0.307 *** 0.258 ** 0.258 ** 0.307 *** 
    Self-employed −0.011  −0.015  −0.014  −0.022  
    Higher service class −0.379 *** −0.332 *** −0.332 *** −0.343 *** 
    Blue collar 0.263 ** 0.253 ** 0.253 ** 0.238 ** 
    White collar (no est.)         
Income deprivation   0.031 *** 0.031 *** 0.031 *** 
Job deterioration   −0.001  −0.001  −0.002  
Job security   −0.038 *** −0.038 *** −0.037 *** 
GDP growth (av. 2009–10)     −0.153 −0.147 
Unemployed x GDP growth       0.030 
RANDOM EFFECTS         
Variance RI 0.579 ** 0.524 ** 0.442 ** 0.441 ** 
Residual variance 3.600 *** 3.576 *** 3.576 *** 3.566 *** 
Variance RS(Unemployed)       0.014  
R2 country level 0.150 0.230 0.351 0.353 
R2 individual level 0.103 0.109 0.109 0.112 
Explanatory variables Model 1 Model 2 Model 3 Model 4 
FIXED EFFECTS         
Intercept 4.492 *** 4.490 *** 4.249 *** 4.255 *** 
Gender         
    Men −0.227 ** −0.213 ** −0.213 ** −0.212 ** 
    Women (no est.)         
Age         
    15–25 years 0.110  0.123  0.124  0.129  
    26–35 years 0.092  0.090  0.090  0.088  
    36–45 years (no est.)         
    46–55 years 0.084  0.093  0.093  0.093  
    56–65 years 0.038  0.080  0.080  0.074  
Education         
    None or primary 1.295 *** 1.260 *** 1.261 *** 1.297 *** 
    Lower secondary 1.111 *** 1.083 *** 1.083 *** 1.090 *** 
    Higher secondary 0.821 *** 0.804 *** 0.804 *** 0.807 *** 
    Tertiary (no est.)         
Urbanization −0.027 *** −0.030 *** −0.030 *** −0.031 *** 
Work situation         
    Unemployed 0.307 *** 0.258 ** 0.258 ** 0.307 *** 
    Self-employed −0.011  −0.015  −0.014  −0.022  
    Higher service class −0.379 *** −0.332 *** −0.332 *** −0.343 *** 
    Blue collar 0.263 ** 0.253 ** 0.253 ** 0.238 ** 
    White collar (no est.)         
Income deprivation   0.031 *** 0.031 *** 0.031 *** 
Job deterioration   −0.001  −0.001  −0.002  
Job security   −0.038 *** −0.038 *** −0.037 *** 
GDP growth (av. 2009–10)     −0.153 −0.147 
Unemployed x GDP growth       0.030 
RANDOM EFFECTS         
Variance RI 0.579 ** 0.524 ** 0.442 ** 0.441 ** 
Residual variance 3.600 *** 3.576 *** 3.576 *** 3.566 *** 
Variance RS(Unemployed)       0.014  
R2 country level 0.150 0.230 0.351 0.353 
R2 individual level 0.103 0.109 0.109 0.112 

*p < .05; **p < .01; ***p < .001 no est.: see note 7.

In a first step, socio-demographic variables, including work situation, are introduced into the model. Given the hypothesis regarding the impact of individual labour market positions on perceived threat (H1.1), our main interest is on work situation, whilst the other variables are considered as controls. Strongly significant effects are found for work situation. As predicted by group conflict theory, the highest levels of perceived ethnic threat are found among the unemployed. This group scores on average 0.307 higher than the grand mean (effect coding is used) on the THREAT scale. This is a difference of considerable size. In addition, blue-collar workers feel more threatened by immigration. Significantly lower levels of perceived threat are found among the higher service class and to a lesser extent also the white-collar workers.10 The THREAT score for the self-employed is situated near the grand mean of all work situation categories. The observation that the most vulnerable categories (blue-collar workers and the unemployed) feel more threatened than less vulnerable labour market positions (higher service class and white-collar workers) is in line with our expectations concerning the relationship between job market position and perceived threat. This means that Hypothesis H1.1 is not rejected.

Further, some of the control variables are found to be significantly related to perceived ethnic threat. Stronger threat perceptions are reported by women, the lower educated, and individuals living in less urbanized areas. No age effect is detected. Taken together, the predictors in this model are able to explain 10.3 per cent of the differences between individuals within countries as well as 15 per cent of the variation between countries (due to composition effects).

In Model 2, three additional indicators of individual-level economic conditions are introduced. Two of these variables turn out to have a significant effect on perceived threat. First, individuals who stated retrospectively that their income situation had become worse between 2007 and 2010 indicate significantly higher levels of perceived threat. Second, job security seems to provide a buffer against perceptions of ethnic threat.11 These findings are in line with our expectations, which means that Hypotheses H1.2 and H1.4 are not rejected. However, contrary to expectations, deterioration in job conditions does not significantly affect threat perceptions. This means that Hypothesis H1.3 is not supported. Several explanations are possible for this unexpected finding. It could suggest that the (reduced) income from work (captured by the variable income deprivation) or job insecurity, rather than the deteriorating conditions under which work is carried out, is relevant for the development of threat perceptions. However, the absence of a significant effect might also be due to the fact that ‘job deterioration’ shows relatively little variation (with most respondents reporting no or very little deterioration).

Even if two out of three hypotheses concerning job market position are not rejected, the explanatory power of the economic predictors introduced in Model 2 should not be overstated. These variables explain 0.6 percentage points of individual-level variation above that explained in Model 1, which is limited. At the country level, the increase in explained variance is eight percentage points, suggesting the presence of composition effects for the economic predictors.

The next step in the analysis involves testing the main effects of several contextual predictors. We focus on four indicators: inflow of foreign immigrants, GDP growth, unemployment, and long-term unemployment. For each of these indicators, we consider the prevailing situation (the average of the two years preceding the survey) as well as the change since the start of the crisis (2007 compared with 2010). Because we have a high number of potentially relevant contextual variables combined with a relatively small number of higher-level units (Meuleman and Billiet 2009), it is not possible to introduce all the contextual variables into the model at the same time. As an alternative therefore, we implement the following model selection procedure: First, each of the contextual variables is tested in a separate model (which includes all the individual-level predictors from Model 2). The obtained effect parameters of these models are reported in Table 3. Subsequently, the contextual predictors that are found to have a significant effect are introduced together into the model to test the robustness of the effects.

Table 3.

Effects of contextual variables on perceived threat

Contextual variables Par. Est. Sign. 
Inflow of foreign population (/1,000 inhabitants)   
    Change 2010 vs. 2007 −0.083  
    Average 2009–2010 0.021  
GDP growth rate   
    Change 2010 vs. 2007 −0.129 
    Average 2009–2010 −0.153 
Long-term unemployment rate   
    Change 2010 vs. 2007 0.069  
    Average 2009–2010 0.113  
Unemployment rate   
    Change 2010 vs. 2007 0.033  
    Average 2009–2010 0.028  
Contextual variables Par. Est. Sign. 
Inflow of foreign population (/1,000 inhabitants)   
    Change 2010 vs. 2007 −0.083  
    Average 2009–2010 0.021  
GDP growth rate   
    Change 2010 vs. 2007 −0.129 
    Average 2009–2010 −0.153 
Long-term unemployment rate   
    Change 2010 vs. 2007 0.069  
    Average 2009–2010 0.113  
Unemployment rate   
    Change 2010 vs. 2007 0.033  
    Average 2009–2010 0.028  

Note: This table displays the effects of various contextual variables on THREAT. The effects are estimated by means of multilevel models controlling for the individual variables mentioned in Table 2. Each contextual effect is obtained from a different model.

Table 3 makes it clear that out of four contextual indicators, only economic growth seems to be related to perceived ethnic threat. A significant effect (p < .05) is found for average GDP growth as well as for change in GDP growth. Higher levels of perceived threat are found in those countries where the GDP per capita increase was lowest (or more so where the decrease in GDP was highest) and where economic growth had declined most strongly. This finding is in line with Hypotheses H2.1a and H2.1b. Because average GDP growth and increase or decrease in GDP growth largely overlap, we focus on the former variable (i.e. the one with the strongest effect) and introduce it into the stepwise analysis (cf. Model 3 in Table 2). Average GDP growth additionally explains 13 percentage points of country-level variation.

Fig. 3 shows in greater detail the relationship between country-level threat (here operationalized as the country-specific deviations from the overall intercept derived from Model 2) and average GDP growth. A negative trend (though far from perfectly linear) can be discerned. The three countries with the highest GDP growth (Poland, Sweden and Switzerland) are among those with the lowest levels of perceived threat. Conversely, the countries with the steepest economic decline (Lithuania, Estonia and Greece) are situated in the upper half of the country ranking of perceived threat.

Figure 3.

GDP growth by country-level residuals.

For an explanation of the country codes see Section 3.1.

Figure 3.

GDP growth by country-level residuals.

For an explanation of the country codes see Section 3.1.

Our finding that neither foreign immigration flows nor (long-term) unemployment rates (or their increase) are significantly related to threat perceptions contradicts Hypotheses H2.2a, H2.2b, H2.3a, and H2.3b, and is thus not in line with group conflict theory. However, these conclusions should be treated with caution. As Meuleman, Davidov and Billiet (2009) argued, it is more appropriate to study changes than absolute levels of threat, because the latter are the result of country-specific historical trajectories and therefore less sensitive to time-related contextual conditions such as immigration and economic conditions. In a similar vein, the significant effect of GDP growth cannot be interpreted in a strict causal fashion. Data from previous ESS rounds reveals that the Baltic States and Greece already scored comparatively high for perceived threat before the financial crisis that began in 2008 (Meuleman, Davidov and Billiet 2009).

As the final step of the multilevel analyses, we assess whether the effects of individual-level economic predictors are context dependent. We estimate three separate models (not shown here), including a random slope for each of the economic predictors. For income deprivation and job security, random slope variance is found to be insignificant. This indicates that, for these two variables, effects do not vary significantly over the 23 countries under study. However, significant random slope variance is found for the dummy indicating the unemployed. Concretely, this means that the gap in perceived threat between the unemployed and the grand mean of all work situations is different across countries. To find out why and how the effect of unemployment varies across contexts, we test for cross-level interactions between individual-level unemployment on the one hand, and economic growth and immigration flows on the other. The cross-level interaction for economic growth is found to be significant (see Model 4 in Table 2). The parameter estimate is positive, meaning that the threat-inducing effect of unemployment is even stronger in contexts where GDP growth is high. This finding obviously contradicts our hypothesis that the effect of individual economic situation would be stronger in economically unfavourable contexts (H3.1). Instead, it is precisely in economically less adverse conditions that the gap between the unemployed and the working is accentuated.

Fig. 4 represents the relationship between the size of the impact of unemployment on perceived threat (measured as country-specific deviations of the unemployment effect obtained from the random slope model) and economic growth. This illustrates that countries with a higher rate of GDP growth tend to show a stronger effect of unemployment (i.e. a positive deviation from the overall unemployment effect) on perceived ethnic treat. There are two clear exceptions: Spain and Germany. These countries respectively have a very weak and a very strong unemployment effect. Because of their average position with regard to economic growth, however, these outlying observations do not distort the regression line substantially.

Figure 4.

The effect of individual-level unemployment by GDP growth rate.

For an explanation of the country codes see Section 3.1.

Figure 4.

The effect of individual-level unemployment by GDP growth rate.

For an explanation of the country codes see Section 3.1.

5. Conclusions and discussion

The starting point of this study was that poor prevailing economic conditions could potentially affect intergroup relationships, and notably relationships between majority groups and immigrants. According to group conflict theory, anti-immigration sentiments and ethnic prejudice originate from socio-economic insecurity. In general terms, the theory predicts that socio-economically vulnerable individuals in particular are likely to develop negative attitudes toward immigration due to the perception of ethnic competition for scarce resources. The dynamic reformulation of ethnic competition implies that especially changing (i.e. worsening) economic conditions result in increasingly negative attitudes toward immigration. A considerable body of research provides empirical evidence for these theoretical arguments.

This study revisits the question of whether and how economic conditions at the individual as well as at the country level influence perceived ethnic threat among the majority population in the context of an economic crisis. In line with group conflict theory, three types of relationships are hypothesized: (1) relationships between perceived threat and economic predictors at the micro level, (2) effects of country-level variables on the perception of ethnic threat, and (3) moderating effects of national-level economic conditions on micro-level relationships between the economic situation and perceived ethnic threat (i.e. a cross-level interaction).

To test these hypotheses, we analyse comparative survey data from 23 country samples of the European Social Survey (ESS), round 5 (2010 to 2011). Change is not studied within individual cases, as ESS data is cross-sectional rather than longitudinal. As an alternative to ‘real’ longitudinal data, retrospective questions are used to measure the (subjective) perception of the change in respondents’ economic conditions during the period from 2008 to the end of 2010. Multilevel analysis is used. This choice is justified by the fact that variation in perceived ethnic threat at the country level is considerable, as indicated by the intra-class correlation.

Relatively strong effects are found for work situation. As predicted by group conflict theory, the highest levels of perceived ethnic threat are found among the unemployed. In addition, blue-collar workers, whose jobs are relatively vulnerable to competition from newcomers (and even more so in times of economic crisis), feel more threatened by immigration. Significantly lower levels of perceived threat are found among the higher service class and to a lesser extent also the white-collar workers. Two other variables that reflect the individuals’ economic position are found to exert a significant, although considerably smaller, effect on perceived threat. Individuals who stated retrospectively that at the end of 2010 and the beginning of 2011 their income situation had become ‘worse during the last three years’ report significantly higher levels of perceived threat. Job security on the other hand seems to provide a buffer against ethnic threat perceptions. However, the explanatory power of these economic predictors at the individual level is relatively weak. The increase in explained variance at the country level is stronger, suggesting the presence of composition effects for the economic predictors.

Of the four contextual indicators, only economic growth seems to be related to perceived ethnic threat. Higher levels of perceived threat are found in those countries where the decrease in GDP per capita was greatest and where economic growth had declined most strongly. Together with differences in population composition regarding the individual predictors, average GDP growth is able to explain the greatest part of country-level variation in perceived ethnic threat.

Our finding that neither immigration flows nor (long-term) unemployment rates (or their increase) are significantly related to threat perceptions contradicts four of our individual hypotheses. However, these conclusions should be treated with caution, as it is more appropriate to study changes than absolute levels of threat, as the latter are the result of country-specific historical trajectories and therefore less sensitive to time-related contextual conditions such as immigration and economic conditions. In a similar vein, the significant effect of GDP growth cannot be interpreted in a causal fashion.

As the final step of the multilevel analysis, we assess whether the effects of individual-level economic predictors are context dependent. For income deprivation and job security, random slope variance is insignificant, indicating that the effects of these two variables are similar across countries. Conversely, we find that the effect of being unemployed on perceived ethnic threat varies significantly across countries. In all countries, the unemployed report higher levels of threat compared with the grand mean, but the gap between the unemployed and the employed is considerably larger in some countries than in others. Contrary to our expectations, the biggest gap is found in countries with a higher rate of GDP growth. Unfavourable economic conditions therefore appear not to accentuate, but to cushion the perceptional gap between the economically advantaged and disadvantaged. This finding supports earlier results produced by Fachini and Mayda (2009), and Kunovich (2004). The latter suggested several explanations for the possibility that the relationships between individual characteristics and prejudice can become weaker when collective threats increase. Disadvantaged groups may respond with despair rather than with greater anti-immigrant prejudice when economic conditions deteriorate. Further, in poor economic conditions the anger among the disadvantaged groups might be directed away from immigrants toward the economic and political elites. Another explanation states that with worsening economic conditions, immigrants may fall further behind majority-group workers, which would decrease the threat posed by immigrants and thus reduce negative attitudes toward them. (Kunovich 2004: 39–40). Another similar explanation suggested by De Weerdt et al. (2007) points to the emergence of a generalized feeling of economic insecurity in countries that are hit severely by economic crisis. In these countries, strong feelings of economic insecurity—and the resulting levels of perceived ethnic threat—might also be present among the employed, thereby diminishing the gap between them and the unemployed. Conversely stated, favourable economic conditions may thus accentuate the gap between the employed and the unemployed. This unexpected finding can also be due to feelings of (relative) deprivation (De Weerdt et al. 2007). When the economic situation is improving for most people, the unemployed might feel (relatively) more deprived, as ‘everyone else seems to be doing better’, whereas they are still unemployed. This might increase their negative attitudes toward out-groups. These interpretations need closer examination in further research.

Although several indicators of individual economic conditions are significantly related to perceived threat, the explanatory power of these factors is somewhat limited. In earlier studies using ESS data, predictors such as religious involvement, participation in social activities, and social trust have been used instead of economic conditions (Billiet and Meuleman 2008; Meuleman and Billiet 2011). The increase in explained variance in ethnic threat was significantly higher in these studies than is found in the present one. The explanatory power also increases considerably when value orientations are introduced into the models (Davidov et al. 2008). This might suggest that the fear of conflict over values and culture, rather than economic and social competition, leads to increased ethnic prejudice (Schneider 2008).

Given the dynamic character of group conflict processes, a shortcoming of this study is the lack of a true panel design with repeated measurements for the same respondents. Unfortunately, such a design is not applied in either the ESS or other large-scale, cross-national survey projects for the period studied in this article and with as many general population samples. As an alternative, retrospective questions are used to gain information on changes in income deprivation or job deterioration. Such retrospective questions are not available for perceived threat. Obviously, this is an important hazard to the internal validity of our conclusions. Another limitation of our study relates to the use of yearly averages for economic performance as an explanatory factor. This time horizon might be too crude to pick up truly short-term fluctuations (and the associated media coverage) that affect public opinion.

Funding

This work was supported by EC 7th Framework Program Grant EPF 262208 “The European Social Survey-Data for a Changing Europe”, and by the Research Foundation Flanders (FW0) Grant G.0A25.12 “Solidarity in times of crisis? The impact of the economic crisis on attitudes toward redistribution and welfare state support”.

Acknowledgements

The authors wish to thank the Core Scientific Team and the local ESS teams for the collection and preparation of the data. They are also grateful to the Research Foundation of Flanders (FW0) for supporting the analysis of the data.

Conflict of interest statement. None declared.

Notes

1. Eurostat, Newsrelease – Euroindicators 24/2013, 14 February 2013 <http://epp.eurostat.ec.europa.eu/cache/ITY_PUBLIC/2-14022013-AP/EN/2-14022013-AP-EN.PDF>
2. For the EU27 countries, see Eurostat 24/2013. Eurostat, Unemployment Statistics − Statistics Explained <http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Unemployment_statistics>
3. Meuleman (2011) discussed possible reasons for the incompatibility of empirical results regarding the effects of national economic context on anti-immigration attitudes. First, the field suffers from vague theory formulation and a large theoretical gap between contextual indicators and individual attitudes (Goldthorpe 1997; Western 1996; Billiet 2013). Most studies have paid insufficient attention to unravelling the social mechanisms (in the sense of Hedström, 2005) that underlie the context−attitudes connection. Processes that might even have opposite effects, and could thus cancel each other out, can distort the findings considerably. Second, the inconsistencies might also be due to methodological issues. Often, multilevel models are used to analyse cross-cultural data, with countries as the higher-level units. However, statistical tests for contextual effects can lack reliability, given the relatively small numbers of available higher level units (Meuleman and Billiet 2009). Furthermore, previous research has often neglected the cross-cultural comparability of measurements.
4. Switzerland and Norway do not belong to the EU, but are part of the European Economic Area (members of the European Free Trade Association).
5. The country codes used in this study are between the brackets. These are the conventional codes that are applied in EUROSTAT and the ESS. The numbers refer to size of the effectively selected cases in each country sample.
6. In 2005, the ESS central team received the European Descartes collaborative research prize for its methodological rigour.
7. For an overview of the response rates per country, see Round 5 Summary and Deviations and the ESS 2010 Survey Documentation Report <http://www.europeansocialsurvey.org/data/deviations_5.html>.
8. The exact wordings of all questions in all languages can be found on the ESS website at <http://www.europeansocialsurvey.org/data/>.
9. Eurostat is a Directorate-General of the European Commission. Its main responsibilities are to provide statistical information to the institutions of the European Union (EU) and to promote the harmonization of statistical methods across its member states and candidates for accession as well as the European Free Trade Association: <http://en.wikipedia.org/wiki/European_Free_Trade_Association>.
10. No effect parameter is estimated for white-collar workers, because one category has to be left out of the analysis to avoid perfectly collinear predictors. However, because effect coding implies that the grand mean equals zero, it can be easily derived that the level of threat for this group is 0.181 lower than the grand mean.
11. The effect of job security is based only on working respondents, as this variable is interacted with a dummy indicating whether a person is employed or not.

References

Billiet
J
Harkness
J
Van de Vijver
F
Mohler
P
‘Cross-Cultural Equivalence with Structural Equation Modeling’
Cross-Cultural Survey Methods
 , 
2003
Hoboken, NJ
John Wiley and Sons
(pg. 
247
-
64
)
Billiet
J
Kenneth
P
‘Quantitative Methods with Survey Data in Comparative Research’
A Handbook of Comparative Social Policy
 , 
2013
2nd edn
Cheltenham
Edward Elgar
(pg. 
194
-
220
)
Billiet
J
Meuleman
B
‘Religious Diversity in Europe and its Relation to Social Attitudes and Value Orientations’
2008
 
Paper presented at the ‘Citizenship and Cultural Identities in the EU: Old Questions, New Answers’ International Conference in Istanbul (Turkey), 18–19 October 2007, pp. 83–104. Ankara (Turkey): Tübitak
Blalock
H M
Toward a Theory of Minority–Group Relations
 , 
1967
New York
John Wiley and Sons
Blumer
H
‘Race Prejudice as a Sense of Group Position’
The Pacific Sociological Review
 , 
1958
, vol. 
1
 (pg. 
3
-
7
)
Bobo
L
‘Whites’ Opposition to Busing: Symbolic Racism or Realistic Group Conflict?’
Journal of Personality and Social Psychology
 , 
1983
, vol. 
45
 
6
(pg. 
1196
-
210
)
Byrne
B M
Shavelson
R J
Muthén
B
‘Testing for the Equivalence of Factor Covariance and Mean Structure: The Issue of Partial Measurement Invariance’
Psychological Bulletin
 , 
1989
, vol. 
105
 (pg. 
456
-
66
)
Citrin
J
Green
D P
Muste
C
Wong
C
‘Public Opinion Toward Immigration Reform: The Role of Economic Motivations’
The Journal of Politics
 , 
1997
, vol. 
59
 
3
(pg. 
858
-
81
)
Coenders
M
Nationalistic Attitudes and Ethnic Exclusionism in a Comparative Perspective. An Empirical Study of Attitudes Toward the Country and Ethnic Immigrants in 22 Countries
 , 
2001
Nijmegen
Interuniversity Center for Social Science Theory and Methodology
Cummings
S
‘White Ethnics, Racial Prejudice, and Labor Market Segmentation’
The American Journal of Sociology
 , 
1980
, vol. 
85
 
4
(pg. 
938
-
50
)
Davidov
E
Meuleman
B
Billiet
J
Schmidt
P
‘Values and Support for Immigration: A Cross-Country Comparison’
European Sociological Review
 , 
2008
, vol. 
24
 
5
(pg. 
583
-
99
)
Davidov
E
‘Explaining Attitudes Towards Immigration Policies in European Countries: The Role of Human Values’
Journal of Ethnic and Migration Studies
 , 
2012
, vol. 
38
 
5
(pg. 
757
-
75
)
De Weerdt
Y
Catellani
P
De Witte
H
Milesi
P
Flecker
J
‘Perceived Socio-Economic Change and Right-Wing Extremism: Results of the Siren-Survey among European Workers’
Changing Working Life and the Appeal of the Extreme Right
 , 
2007
Aldershot
Ashgate
(pg. 
63
-
83
)
De Witte
H
Hagendoorn
L
Nekuee
S
‘Everyday Racism in Belgium: An Overview of the Research and an Interpretation of its Link with Education’
Education and Racism: A Cross-National Inventory of Positive Effects of Education on Ethnic Tolerance
 , 
1999
Aldershot
Ashgate
(pg. 
47
-
74
)
De Witte
H
Meuleman
B
Loosveldt
G
Swyngedouw
M
Cambré
B
‘Job Insecurity and Voting for an Extreme Right-Wing Party’
Measuring Meaningful Data in Social Research
 , 
2007
Leuven
Acco
(pg. 
93
-
112
)
Dustmann
C
Preston
I
Racial and Economic Factors in Attitudes to Immigration
 , 
2004
London
Centre for Research and Analysis of Migration
Fachini
G
Mayda
A
‘Does the Welfare State Affect Individual Attitudes towards Immigration? Evidence Across Countries’
Review of Economics and Statistics
 , 
2009
, vol. 
91
 (pg. 
295
-
314
)
Fetzer
J S
‘Economic Self-Interest or Cultural Marginality? Anti-Immigration Sentiment and Nativist Political Movements in France, Germany and the USA’
Journal of Ethnic and Migration Studies
 , 
2000
, vol. 
26
 
1
(pg. 
5
-
23
)
Fetzer
J S
Rosenblum
M R
Tichenor
D J
‘Public Opinion and Populism’
The Oxford Handbook of The Politics of International Migration
 , 
2012
Oxford
Oxford University Press
(pg. 
229
-
323
)
Ganzeboom
H
Treiman
D J
‘Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations’
Social Science Research
 , 
1996
, vol. 
25
 (pg. 
201
-
39
)
Goldthorpe
J H
‘Current Issues in Comparative Macrosociology: A Debate on Methodological Issues’
Comparative Social Research
 , 
1997
, vol. 
16
 (pg. 
1
-
26
)
Hagendoorn
L
Nekuee
S
Education and Racism: A Cross-National Inventory of Positive Effects of Education on Ethnic Tolerance
 , 
1999
Aldershot
Ashgate
Hainmueller
J
Hiscox
M
‘Educated Preferences: Explaining Attitudes Towards Immigration in Europe’
International Organisation
 , 
2007
, vol. 
61
 (pg. 
399
-
442
)
Harkness
J
Van de Vijver
F
Mohler
P
Cross-Cultural Survey Methods
 , 
2003
Hoboken, NJ
John Wiley and Sons
Hedström
P
Dissecting the Social: On the Principles of Analytical Sociology
 , 
2005
Cambridge
Cambridge University Press
Hello
E
Gijsberts
M
Scheepers
P
‘Education and Ethnic Prejudice in Europe: Explanations for Cross-National Variances in the Educational Effect on Ethnic Prejudice’
Scandinavian Journal of Educational Research
 , 
2002
, vol. 
46
 
1
(pg. 
5
-
24
)
Hemerijck
A
Knapen
B
van Doorne
E
Aftershocks: Economic Crisis and Institutional Choice
 , 
2010
Amsterdam
Amsterdam University Press
Jackson
J W
‘Realistic Group Conflict Theory: A Review and Evaluation of the Theoretical and Empirical Literature’
The Psychological Record
 , 
1993
, vol. 
43
 
3
(pg. 
395
-
413
)
Jones
F
Lambert
W
‘Occupational Rank and Attitudes Toward Immigrants’
Public Opinion Quarterly
 , 
1965
, vol. 
29
 (pg. 
137
-
44
)
Kunovich
R
‘Social Structural Sources of Anti-immigrant Prejudice in Europe’
International Journal of Sociology
 , 
2002
, vol. 
32
 (pg. 
39
-
57
)
Kunovich
R
‘Social Structural Position and Prejudice: an Exploration of Cross-national Differences in Regression Slopes’
Social Science Research
 , 
2004
, vol. 
33
 (pg. 
20
-
44
)
Lancee
B
Pardos-Prado
S
‘Group Conflict Theory in a Longitudinal Perspective: Analysing the Dynamic Side of Ethnic Competition’
International Migration Review
 , 
2013
, vol. 
47
 
1
(pg. 
106
-
31
)
Mayda
A M
‘Who is Against Immigration? A Cross-Country Investigation of Individual Attitudes Toward Immigrants’
The Review of Economics and Statistics
 , 
2006
, vol. 
88
 
3
(pg. 
510
-
30
)
Meuleman
B
Davidov
E
Schmidt
P
Billiet
J
‘Perceived Economic Threat and Anti-Immigration Attitudes: Effects of Immigrant Group Size and Economic Conditions Revisited’
Cross-Cultural Analysis: Methods and Applications
 , 
2011
New York
Routledge
(pg. 
218
-
310
)
Meuleman
B
Billiet
J
‘A Monte Carlo Sample Size Study: How Many Countries are Needed for Accurate Multilevel SEM?’
Survey Research Methods
 , 
2009
, vol. 
3
 
1
(pg. 
45
-
58
)
Meuleman
B
Billiet
J
Davidov
E
Schmidt
P
Billiet
J
‘Religious Involvement: Its Relation to Values and Social Attitudes’
Cross-Cultural Analysis. Methods and Applications
 , 
2011
New York
Routledge
(pg. 
175
-
208
)
Meuleman
B
Davidov
E
Billiet
J
‘Changing Attitudes Toward Immigration in Europe, 2002-2007: A Dynamic Group Conflict Theory Approach’
Social Science Research
 , 
2009
, vol. 
38
 
2
(pg. 
352
-
65
)
Meuleman
B
Davidov
E
Billiet
J
Changing Attitudes Toward Immigration In Times Of Crisis: A Cross-National and Over-Time Comparison
 , 
2013
 
Paper presented at the 5th Conference of the European Survey Research Association, 15–19 July 2013, Ljubljana
OECD
Classifying Educational Programmes: Manual for ISCED-97 Implementation in OECD Countries
 , 
1999
Paris
OECD
Olzak
S
Dynamics of Ethnic Competition and Conflict
 , 
1992
Stanford, CA
Stanford University Libraries
O’Rourke
K H
Sinnott
R
‘The Determinants of Individual Attitudes Towards Immigration’
European Journal of Political Economy
 , 
2006
, vol. 
22
 (pg. 
838
-
61
)
Quillian
L
‘Prejudice as a Response to Perceived Group Threat: Population Composition and Anti-Immigrant and Racial Prejudice in Europe’
American Sociological Review
 , 
1995
, vol. 
60
 
4
(pg. 
586
-
611
)
Rubin
D B
Multiple Imputation for Nonresponse in Survey
 , 
1987
New York
John Wiley and Sons
Scheepers
P
Gijsberts
M
Coenders
M
‘Ethnic Exclusionism in European Countries. Public Opposition to Grant Civil Rights to Legal Migrants as a Response to Perceived Ethnic Threat’
European Sociological Review
 , 
2002
, vol. 
18
 
1
(pg. 
1
-
18
)
Scheve
K F
Slaughter
M J
‘Labor Market Competition and Individual Preferences of Immigration Policy’
The Review of Economic and Statistics
 , 
2001
, vol. 
83
 
1
(pg. 
133
-
45
)
Schneider
S L
‘Anti-Immigrant Attitudes in Europe: Outgroup Size and Perceived Ethnic Threat’
European Sociological Review
 , 
2008
, vol. 
24
 
1
(pg. 
53
-
67
)
Semyonov
M
Raijman
R
Yom Tov
A
Schmidt
P
‘Population Size, Perceived Threat, and Exclusion: A Multiple-Indicators Analysis of Attitudes Toward Foreigners in Germany’
Social Science Research
 , 
2004
, vol. 
33
 (pg. 
681
-
701
)
Semyonov
M
Gorodzeisky
A
‘The Rise of Anti-Foreigner Sentiment in European Societies, 1988–2000’
American Sociological Review
 , 
2006
, vol. 
71
 (pg. 
426
-
49
)
Steenkamp
J E
Baumgartner
H
‘Assessing Measurement Invariance in Cross-National Consumer Research’
Journal of Consumer Research
 , 
1998
, vol. 
25
 (pg. 
78
-
90
)
Stephan
W G
Ybarra
O
Martínez
C
Schwarzwald
J
Tur-Kaspa
M
‘Prejudice Toward Immigrants to Spain and Israel: An Integrated Threat Theory Analysis’
Journal of Cross-Cultural Psychology
 , 
1998
, vol. 
29
 
4
(pg. 
559
-
76
)
Stephan
W G
Bachman
G
‘Prejudice Toward Immigrants’
Journal of Applied Psychology
 , 
1999
, vol. 
29
 
11
(pg. 
2221
-
37
)
Strabac
Z
Listhaug
O
‘Anti-Muslim Prejudice in Europe: A Multi-Level Analysis of Survey Data from 30 Countries’
Social Science Research
 , 
2008
, vol. 
37
 (pg. 
268
-
86
)
Zarate
M A
Garcia
B
Garza
A A
Hitlan
R T
‘Cultural Threat and Perceived Realistic Group Conflict as Dual Predictors of Prejudice’
Journal of Experimental Social Psychology
 , 
2004
, vol. 
40
 
1
(pg. 
99
-
105
)

Appendix 1

Measurement model for the latent variables ‘perceived ethnic THREAT’, ‘JOB deterioration’, and ‘INCOME deterioration’

Figure A.1.

Conceptual presentation of the measurement model. For all countries, the common (invariant) and deviant (intercept and slope) parameters are shown in Table A.2.

Figure A.1.

Conceptual presentation of the measurement model. For all countries, the common (invariant) and deviant (intercept and slope) parameters are shown in Table A.2.

Table A1.

Full scalar (and metric) invariant model (M0) compared with the retained quasi-invariant model (M13)

Model Chi-square Df RMSEA p-value of close fit CFI 
Model 0 4,582.011 816 .080 .000 .946 
Model 13 2,121.165 802 .048 .948 .981 
Model Chi-square Df RMSEA p-value of close fit CFI 
Model 0 4,582.011 816 .080 .000 .946 
Model 13 2,121.165 802 .048 .948 .981 
Table A2.

Parameters of the (quasi) scalar invariant measurement model

Countries INCOME DEPRIVATION
 
JOB DETERIORATION
 
PERCEIVED THREAT
 
Intercepts G8 G9 G10 G58 G59 G60 G38 G39 G40 
CZ     0.274     
DK    0.426      
EE     0.333     
FI   −0.920 0.117     −0.395 
GR     0.673     
IE    −0.871      
LT     0.578     
NO    0.920      
SE    0.681      
All other −0.545 −0.578 −0.504 −0.199 −0.341 −0.137 −0.199 −0.084 −0.090 
Slopes G8 G9 G10 G58 G59 G60 G38 G39 G40 
BU   0.497       
CZ         0.733 
GR   0.645       
All other 0.867 0.858 0.854 0.525 0.844 0.646 0.724 0.831 0.808 
Countries INCOME DEPRIVATION
 
JOB DETERIORATION
 
PERCEIVED THREAT
 
Intercepts G8 G9 G10 G58 G59 G60 G38 G39 G40 
CZ     0.274     
DK    0.426      
EE     0.333     
FI   −0.920 0.117     −0.395 
GR     0.673     
IE    −0.871      
LT     0.578     
NO    0.920      
SE    0.681      
All other −0.545 −0.578 −0.504 −0.199 −0.341 −0.137 −0.199 −0.084 −0.090 
Slopes G8 G9 G10 G58 G59 G60 G38 G39 G40 
BU   0.497       
CZ         0.733 
GR   0.645       
All other 0.867 0.858 0.854 0.525 0.844 0.646 0.724 0.831 0.808 
Appendix 2A.

Descriptive statistics for the individual-level variables

 Mean Std. Min. Max. 
Perceived threat 19,012 4.992619 2.086446 10 
Urbanization 20,197 5.253627 3.091491 10 
Job security 14,559 5.861666 3.487186 10 
Job deterioration 17,421 2.241739 2.812862 10 
Income deprivation 19,670 3.918658 3.181857 10 
 Freq Percent    
Gender      
    Female 9,807 48.5    
    Male 10,413 51.5    
    Total 20,220 100    
Age      
    15–25 years 1,891 9.37    
    26–35 years 4,451 22.06    
    36–45 years 5,359 26.56    
    46–55 years 5,419 26.86    
    56–65 years 3,058 15.16    
    Total 20,178 100    
Education      
    None or primary 1,116 5.54    
    Lower secondary 2,131 10.57    
    Higher secondary 8,824 43.79    
    Tertiary 8,081 40.1    
    Total 20,152 100    
Work situation      
    Unemployed 2,497 12.43    
    Self-employed 2,316 11.53    
    Higher service class 2,497 12.43    
    White collar 7,447 37.06    
    Blue collar 5,337 26.56    
    Total 20,094 100    
 Mean Std. Min. Max. 
Perceived threat 19,012 4.992619 2.086446 10 
Urbanization 20,197 5.253627 3.091491 10 
Job security 14,559 5.861666 3.487186 10 
Job deterioration 17,421 2.241739 2.812862 10 
Income deprivation 19,670 3.918658 3.181857 10 
 Freq Percent    
Gender      
    Female 9,807 48.5    
    Male 10,413 51.5    
    Total 20,220 100    
Age      
    15–25 years 1,891 9.37    
    26–35 years 4,451 22.06    
    36–45 years 5,359 26.56    
    46–55 years 5,419 26.86    
    56–65 years 3,058 15.16    
    Total 20,178 100    
Education      
    None or primary 1,116 5.54    
    Lower secondary 2,131 10.57    
    Higher secondary 8,824 43.79    
    Tertiary 8,081 40.1    
    Total 20,152 100    
Work situation      
    Unemployed 2,497 12.43    
    Self-employed 2,316 11.53    
    Higher service class 2,497 12.43    
    White collar 7,447 37.06    
    Blue collar 5,337 26.56    
    Total 20,094 100    
Appendix 2B.

Context data matrix

graphic 
graphic