Acceptable in the EU? Why some immigration restrictionists support European Union mobility

Abstract

Why do some Europeans support immigration from within the European Union, while rejecting immigration from elsewhere? Acceptance of intra-European Union mobility—even by those who wish to restrict immigration more generally—is important for popular support for the European Union itself. This paper identifies and attempts to explain the preferences of “EU-only inclusionists”: EU nationals who support relatively high levels of immigration, but only from within the European Union. We analyze an underexplored experimental module in the European Social Survey to explore European Union inclusionism in relation to other preference profiles. We find that identification with the European Union helps explain specific support for European Union mobility, while subnational (racial and religious) identities are associated with a preference for European migrants over non-Europeans, but not with specific support for intra-European Union movement.

Keywords

European Union immigration attitudes migration mobility supranational identity

Introduction

The European Union (EU) requires immigration policy that sharply differentiates between EU and non-EU nationals. Intra-EU mobility is a foundational right for EU nationals, one of the basic, non-negotiable “four freedoms”, alongside free movement of goods, capital, and services. By contrast, member states can and do develop policies to restrict immigration from non-EU countries. The EU even has a distinct nomenclature for each type of movement: “mobility” of EU nationals within EU member states is distinguished from “migration” by “Third Country Nationals (TCNs)” from outside the EU (Ruhs, 2017).

However, European publics may not make the same sharp distinction (Geddes and Hadj-Abdou, 2016). The disjuncture between public understandings of immigration and elite rhetoric and policy may pose a significant challenge to the ongoing public support for the EU itself. Anti-immigration public opinion was essential to the Brexit movement (Goodwin and Milazzo, 2017), and has created pressure for new restrictions on free movement in other EU states, which the EU has resisted (Ruhs, 2018). This environment makes it an urgent matter for policy-makers and scholars to understand whether EU citizens recognize the fundamental distinction between intra-EU mobility and non-EU migration, and what if anything generates support for this differentiation. However, little research examines EU citizens’ attitudes toward immigration across this fundamental policy dividing line.

This study addresses this gap in the literature. We focus on understanding what we call Europe-only and EU-only inclusionism: the pattern in which individuals support immigration from within Europe or the EU while opposing immigration from the rest of the world. Thus, we ask: why do some Europeans favor European immigration, while opposing immigration from other parts of the world?

We analyze attitudes of nationals/citizens across 20 countries of the EU and the European Free Trade Association (EFTA) with survey data from the European Social Survey (ESS). Our research design distinguishes between support for EU or European inflows, from a general support for immigration or a preference for non-EU/non-European inflows. We then use discrete choice models to analyze the determinants of distinct patterns of immigration preferences, particularly focusing on support for intra-European migration. Descriptively, we find that Europe-specific inclusionism is relatively rare, highlighting the challenge facing supporters of free movement. Our analysis further shows that support for immigration from Europe can come from supranational identification with the EU, but can also arise from exclusionary versions of more parochial subgroup identities, particularly along religious lines. Our study adds to a growing literature on the role of supranational or cosmopolitan identities (Hooghe and Marks, 2018; Teney et al., 2013) in support for the EU and its policies.

Explaining European or EU inclusionism

Much research addresses the determinants of attitudes toward immigration (Ceobanu and Escandell, 2010; Hainmueller and Hopkins, 2014). Europeans’ attitudes hinge on the extent to which they perceive immigration as a threat, to either concrete or symbolic resources (McLaren and Johnson, 2007). Threat perceptions, in turn, often depend on perceptions of who immigrants are. Public opinion distinguishes among immigrants according to their level of education or job skills (Hainmueller and Hiscox, 2010); reason for immigrating (Blinder, 2015); and racial, ethnic, national, or religious identities (Bansak et al., 2016; Ford, 2008; Gorodzeisky and Semyonov, 2016), all of which may relate to the degree and nature of threats that people perceive from immigration (Azrout and Wojcieszak, 2017).

Despite the political importance of the distinction between intra-EU and non-EU migration, the literature provides little guidance on whether migrants’ EU citizenship is one of the dimensions that shape attitudes. We lack up-to-date answers even for simple descriptive questions, such as how many EU citizens prefer intra-EU mobility to non-EU migration. Prior work suggests that “EU inclusionism” is uncommon: McLaren (2001) finds that most Europeans had the same attitudes toward immigrants from within and from outside the EU, even prior to EU enlargement in 2004; since then, increased economic and cultural heterogeneity would seem to decrease the likelihood that anti-immigration Europeans will make an exception for fellow EU nationals (Ruhs, 2017).

Even if selective support for European migration remains rare, there are important political and theoretical reasons to attempt to understand the determinants of the Europe-only and EU-only inclusionist patterns of migration preferences. Is EU inclusionism related to supranational forms of political identity, or to parochial forces such as nationalism and cultural chauvinism? Or perhaps EU inclusionism stems from broader economic and political drivers of immigration attitudes and EU support.

Identities

We begin with symbolic or identity-based explanations, investigating two contrasting identities that Europeans may hold. On one hand, identification with the EU itself may generate specific support for intra-EU mobility. On the other hand, support for European migration might reflect a more ethno-cultural vision of European identity, in which European migrants are preferred because they are seen as more similar on other identity dimensions such as race, ethnicity, or religion.

The EU is a locus of a supranational political identity for some citizens (Bruter, 2009; Kentmen-Cin and Erisen, 2017; Risse, 2010). Fligstein et al. (2012) show that European identity and a national identity can and do coexist within many EU citizens. However, many EU citizens develop strong EU identities; Kuhn (2015) finds that transnational experiences (such as living or studying abroad) increase individual identification with the EU, but only a small subset of EU nationals ever have these experiences. Likewise, supranational identification with the EU, as opposed to identification with one’s member state, is less common among EU nationals without a non-EU immigrant background (Erisen, 2017).

Prior work finds political meaning in this variation in EU identification, notably in predicting support for further EU integration (Hooghe and Marks, 2005). Extending this line of thought, we hypothesize that identification with the EU will also predict EU inclusionist immigration attitudes. Curtis (2014) found indirect evidence of this pattern, inferring from aggregate data on European and non-European migration flows.

Our second set of identity-based hypotheses focuses on more exclusive, subgroup-based identities. Opposition to immigration is associated with negative attitudes toward demographic or cultural out-groups, especially along the lines of race/ethnicity, religion, and language (Creighton et al., 2018). These subgroup identity preferences might also be predictors of a preference for European over non-European immigration, since Europeans may perceive non-Europeans as less similar to themselves along these dimensions. Religion may be especially important; Muslims have been constructed as a threat to Europe and to citizens in many member states, leading to substantial negative sentiment applied specifically to Muslims (Azrout and Wojcieszak, 2017; Strabac and Listhaug, 2008).

We therefore hypothesize that Europeans who value shared ethnocultural characteristics will be more likely to hold Europe inclusionist preferences: accepting European immigration while opposing non-European inflows. Note that this differentiation focuses on Europe as a locus of ethnocultural identities rather than on the EU as a supranational political entity. The European-ness of potential migrants here acts as a proxy for other characteristics such as whiteness, Christianity, or linguistic or cultural similarity. While both supranational and subnational identities might encourage a preference for Europeans over other immigrants, these two sources of Europe-only inclusionism bear very different normative and political implications.

Resources

Aside from identity-based considerations, citizens may respond to immigration as a potential economic threat. In the labor market theory of immigration attitudes, each individual is more likely to welcome migrants who complement her own role in the labor market while opposing immigration of labor market substitutes (Mayda, 2006; O’Rourke and Sinnott, 2006). Low-skilled “native” workers should oppose the immigration of low-skilled workers but welcome high-skilled workers; high-skilled native workers should hold the opposite set of preferences.

In this theory, skills, not origins, matter; EU citizenship plays no direct role in explaining attitudes. However, EU citizenship may play an important indirect role as a proxy for labor market position. Intra-EU mobility, unlike non-EU migration, can be viewed as a likely route to low-skilled jobs. Because of free movement, EU member states cannot (directly) restrict the arrivals of low-skilled workers from within the EU. Fears of large flows of unskilled workers have long been a part of anti-migration, anti-EU discourse, crystallized in the widely circulated trope of the “Polish plumber”, expected by some to overwhelm domestic labor markets and drive down wages and employment among native workers (Favell, 2008; Spigelman, 2013). By contrast, EU member states limit immigration from outside the EU, including restrictions based on skill and income designed to attract only the highly qualified (Cerna, 2014).

These associations between EU status and immigrants’ skill levels, when combined with the labor market theory of immigration attitudes, generate additional hypotheses for EU inclusionism. In the labor market theory, low-skilled EU workers should be more likely to oppose intra-EU migration, whereas high-skilled workers should show greater support (while being less likely to support non-EU migration). This is because the hypothetical Polish plumber is a competitor for low-skilled native-born workers in France or Germany, but a complement (and service provider) for high-skilled native-born workers in those countries. The reverse relationships hold for the high-skilled non-EU immigrant worker coming to fill professional jobs as IT workers, doctors, or researchers, for example.

Alternate economic models of immigration preferences argue that restrictionist attitudes reflect sociotropic concerns about the impact of immigration on the national economy (Gerber et al., 2017; Valentino et al., 2017), or on public finances in particular (Facchini and Mayda, 2009). However, the existing literature does not provide adequate data to distinguish between EU and non-EU migrants’ real or perceived impacts on national economies or on national fiscal burdens. Thus, we do not generate hypotheses about European or EU inclusionist attitudes from these theories of economic impacts.

Political engagement

Finally, we develop hypotheses around “cognitive mobilization”, a prominent concept in prior literature on support for the EU and its political projects (Hobolt and de Vries, 2016). This theoretical perspective has roots in Zaller’s (1992) leading model of public opinion, in which citizens take cues from elites about what political attitudes they should hold. People who know more about politics are more likely to receive, understand, and accept cues from politicians and other elites (Zaller, 1992). Citizens who are most cognitively engaged in politics follow elite opinion most strongly, gravitating toward consensus on issues where elites agree, and polarizing when elites disagree.

In the EU case, political engagement has traditionally increased support for the EU and for EU integration, long seen as an elite project (Gabel, 1998), possibly because of exposure to political elites’ pro-EU messages (McLaren, 2001). To be sure, elites have become more divided on the EU, with the rise of Eurosceptic parties and shifts in ideology (Hooghe and Marks, 2009). Nonetheless, elites remain more supportive of the EU than the general public, while Euroscepticism springs disproportionately from ideological extremes with less attachment to mainstream parties (Elsas et al., 2016).

From this perspective on EU support, we would expect greater “cognitive mobilization” or engagement with politics to predict support for EU-only inclusionist preferences. As noted above, the EU and supporting political elites at the national level draw an extremely sharp distinction between intra-EU mobility and non-EU migration, in both policy and rhetoric. People who pay attention to politics will be more likely to absorb elites’ nuanced, elite-endorsed ranking of potential migration inflows. Less attentive citizens are thus expected to be more likely to miss these nuances, and to react to immigration in broader brush strokes.

Table 1 summarizes the key hypotheses. For each theoretical perspective, we list the factors associated with increased preference for immigration either from the EU specifically, or from Europe but not specifically from the EU. Again, these predictions refer to a relative preference for EU or European migrants, rather than general support for immigration.

Table 1.

Hypotheses.

	Expected preference
Category of hypotheses	Pro EU	Pro-European
Identity	EU identity (+)	Preference for ethnically and culturally similar immigrants (+)
Resources	High skilled (+)Low skilled (−)
Cognitive and elite “cues” mobilization	Political awareness and engagement (+)

Data

The empirical analysis relies on survey microdata from the seventh round of the ESS (2014) and considers nationals/citizens who are resident across 20 EU and EFTA member countries available.

Dependent variables

We construct two distinct dependent variables from 2014 ESS items, one that captures individual-level preferences for European over non-European immigration, and a second that provides leverage on the EU/non-EU distinction. The first dependent variable was created from two items asking how many people should be allowed to immigrate (many, some, few, or none), initially from poorer countries in Europe and then from poorer countries outside Europe. We use answers to these two items to place each respondent in one of four mutually exclusive categories, as follows: Individuals who support the arrival of many or some immigrants from both European and non-European countries are (a) general inclusionists. Those who want few or none to immigrate from both within and outside Europe are (b) general restrictionists. Meanwhile, others are selective inclusionists, welcoming substantial numbers of immigrants from one geographic area while preferring little or no immigration from the other. Europe inclusionists (c), our group of primary interest, prefer many or some immigrants from poorer European countries and few or no immigrants from poorer non-European countries. Non-Europe inclusionists (d) show the opposite pattern, preferring non-European to European immigrants. Table 2 illustrates the content of the categories and their frequency in the estimation sample. Descriptively, Europe inclusionists comprised only an estimated 9.3% of the population covered by ESS in 2014.

Table 2.

Support for Europe-only inclusionism (dependent A).

How many immigrants from?		Outside Europe
How many immigrants from?		Many	Some	A few	None
Europe	Many	General inclusionists		Europe inclusionists
	Some	50.2% n = 14,925		9.3% n = 3063
	A few	Non-Europe inclusionists		General restrictionists
	None	2.1% n = 656		38.4% n = 13,606

Note: Unweighted sample size and weighted percentages for countries included in sample of nationals. Percentages are weighted to correct for differences in population size across countries and for different probabilities of selection for different groups of the population as recommended (European Social Survey, 2014).¹

This measure might be argued to underestimate Europe inclusionism. The two items are asked consecutively, possibly inducing pressure on respondents to be consistently pro- or anti-immigration on both. Also, the items do not distinguish between EU and non-EU immigration from within Europe.

To address these limitations, we analyze data from an embedded survey experiment. Each respondent was asked how many of a particular type of immigrant she would like to “allow to come and live here” (many, some, a few, or none). Experimental versions of this question randomly varied the potential immigrant group along two dimensions: job skills (“unskilled laborers” or “professionals”) and place of origin (a European or non-European country). This design eliminates consistency pressure because each respondent was asked about only one of four types of immigration. We can then assess the impact of European immigrant origin on citizens’ preferences experimentally, by comparing aggregate responses to Europe and non-Europe conditions within each skill level.

The experiment revealed significant effects of both skill and origin, as Figure 1 shows. Around 37% of respondents supported allowing many or some unskilled workers from a poorer non-European country. This rose to 45%, a statistically significant increase (p < 0.001, two-tailed t-test), for those asked about unskilled workers from a poorer European country. For skilled workers, pro-inclusion stood at 70% for non-European migrants and 73% (p = 0.001) for European migrants.

In addition to individual level experimental variation, this design involved a country-level nonrandom source of variation that we use to distinguish between attitudes toward immigration from EU countries and immigration from European countries that are not part of the EU. The experimental survey question asked about immigrants from specific European or non-European countries, rather than from “Europe” in general. Each respondent was asked about immigrants coming from the poorer European or non-European country that provides the largest number of immigrants to that respondent’s country. For example, in Germany, respondents assigned to the “Europe” condition were asked about immigrants from Poland, the non-Europe condition asked about immigrants from Turkey. British and Swedish respondents in the Europe condition were also asked about immigrants from Poland, while the non-Europe condition asked about immigrants from India or Somalia, respectively. The Online appendix lists the immigrant-sending countries named in each country’s survey.

This feature provides some leverage on the role of EU status in generating support for immigrant flows. This is because the Europe treatments specify an EU member state for respondents in some countries and a non-EU European state for respondents in other countries. For example, respondents in Austria were asked about immigrants from Serbia, while those in Poland and Lithuania were asked about arrivals from Belarus.

Explanatory variables

We also test theoretical explanations for Europe- or EU-inclusionist preferences described above. Our first set of explanatory variables relate to the hypothesis that EU inclusionism comes from supranational identification with the EU. EU identification is represented by two ESS questions asking whether the respondent thinks EU unification has gone too far, and trusts the European Parliament (respondents place themselves on a scale from 0 to 10). These are proxies for EU identity rather than direct measures. Nonetheless, these indicators are conceptually and empirically linked to diffuse support for the EU as an institution, as opposed to specific support for policies (Beaudonnet and Di Mauro, 2012). Thus, these measures of identification with (diffuse) support for the EU can be used to predict more specific support for EU policy on issues including immigration.

In addition, we hypothesized that opposition to non-European immigration may also be more likely among respondents who prioritize various forms of national or subnational identities. Our indicators here assess precisely that tendency, in the context of immigration preferences. We include items on how important respondents think it is for immigrants to have fluency in the country’s official language, a Christian background, white racial identity, and a commitment to the country’s way of life, again measured by self-placement on scales from 0 (not important) to 10 (important). We add another binary item that taps into cultural chauvinism at a general level, asking to what extent respondents think some cultures are better than others or whether all cultures are equal. Finally, we take account of national identity with a question asking how close they feel to their country (not at all/not very/close/very close).

Collectively, these items allow us to assess whether preferences for one’s in-group translate into a preference for European over non-European migration. Crucially, we expect these variables to connect with Europe-only rather than EU-only inclusionism, reflecting preferences for Europeans on ethno-cultural lines rather than openness to others based on supranational political identification with the EU.

Moving from identities to economic factors, we measure the respondents’ occupation with International Standard Classification of Occupations (ISCO-08) categories, classifying workers as low-skilled (7–9), medium-skilled (4–6), high-skilled (1–3), or in armed forces occupations (0). For retired or unemployed people, we use the previous or most recent occupation. We also include a measure of economic activity: self-reported main activity in the last seven days, in four categories: (a) paid work; (b) education or training; (c) unemployed and actively looking for work; and (d) economically inactive. We control for subjective perceptions of socioeconomic well-being with items in which respondents assess their comfort with their current household income (very comfortable/comfortably coping/difficult to cope/very difficult) and health status (good/fair/bad/very bad). Finally, we include items asking whether it is important that immigrants (a) have good educational qualifications and (b) have needed skills (both 0–10 scales).

To test the cognitive mobilization hypothesis, we include indicators of interest and engagement in politics. These include self-reported interest in politics (not at all/hardly/quite/very), self-reported turnout in the last national election (yes/no/not eligible), and feelings of closeness to a political party (yes/no). We also include a political participation scale, created by adding responses to seven different items asking whether or not the respondent took part in a given political action in the last 12 months, such as contacting a government official, working in a political party, wearing a campaign badge or sticker, or taking part in a demonstration or boycott. Scores ranged from zero, for taking part in none of these actions, to seven, for taking part in all of them.

In addition, we include education as a consistent correlate of political interest and attention to elite messages. We use a categorical variable based on the International Standard Classification of Education (ISCED): (a) low education if up to lower secondary (0–II); (b) medium education if up to upper secondary, postsecondary non-tertiary, or short tertiary (III–IV); and (c) high education if Bachelor’s Degree equivalent or higher (V–VI).

Demographic and other control variables

Our analysis takes into account a series of individual level demographic differences as known correlates of political attitudes rather than for any particular theoretical expectations. These include gender, age, birthplace (in country of interview or naturalized foreign-born), self-reported ethnic minority status, and rural or urban residence. We expect greater support for EU migration among women, younger people, urban residents, and ethnic minorities (Berg, 2010; Bogard and Sherrod, 2008; Burns and Gimpel, 2000; Pichler, 2010). But existing research provides no clear expectations that these variables will independently influence Europe inclusionism.

Analyzing Europe-only inclusionism

Since our first dependent variable includes four discrete, mutually exclusive, unranked categories, we employ maximum likelihood multinomial logit (MLML) regression. This allows us to estimate the effects of predictors on the likelihood that a respondent will be a general inclusionist, general restrictionist, Europe inclusionist, or non-Europe inclusionist. The MLML estimates a logistic regression model for each response category relative to a single reference category, and all within one estimation (Pasek et al., 2009). It yields estimated impacts on the likelihood of holding a particular attitude profile in relation to the chosen reference category (Kwak and Clayton-Matthews, 2002): “general inclusionists” in our case. This choice eliminates the known issues with using consecutive binary logit models to capture differences between multiple outcome categories: (a) comparing a specific outcome category to all other outcomes combined, which complicates the comparison of predictor effect sizes; or (b) comparing each outcome category to the same reference category by excluding two of the other categories in each estimation, which limits the ability to readily compare results across models.

The independent variables were those discussed in the Data section. We account for country differences¹ with the addition of country dummies in the specification, i.e. with fixed country effects.²

In MLML as in logit or probit models, coefficients represent changes in odds ratios. Further, they are linear with respect to odds ratios, but not with respect to actual outcomes (i.e. changes in the probability of the dependent variable taking on a particular value). Therefore, to demonstrate the impacts of our coefficients, we present results by plotting postestimation average marginal effects (AMEs), which approximate the amount of change in the probability of observing each dependent variable outcome that is associated with each change in the values of a predictor (Hanmer and Kalkan, 2013).

The AME plots thus show the average impact on the actual respondents in the sample. To produce the plots, we generate an individual-level prediction for each person in the sample by holding all independent variables constant at their actual values except for the independent variable whose effects we are plotting. For explanatory variables treated as continuous, the AME shows the predicted change associated with each unit increase in the values of that variable (e.g. each step toward the right end of a scale). When the predictor is nominal or binary, the AME shows the change relative to the base category.

Results from the first (multinomial logit) analysis support the identity-based hypotheses, while largely failing to support predictions from the resource-based and cognitive mobilization approaches.

First, we find that proxies for EU identities matter when it comes to demonstrating Europe inclusionist preferences. Figure 2 illustrates this relationship by plotting the change in the likelihood of opting for Europe inclusionism (bottom left panel) across the different levels of support for further EU unification. Each step towards thinking that EU unification has gone too far is associated with an average reduction in the probability of opting for Europe inclusionism of about 0.4 percentage points. The strongest opponents of further EU unification (respondents at 10 on the scale) are 2 percentage points less likely to be Europe inclusionists than the most positive toward further EU unification (0 on the scale).³ This effect seems small, but looms larger given the rarity of Europe inclusionists, who comprise only about 9% of the weighted sample. A similar but less pronounced effect (marginal effect of 0.2 rather than 0.4) is found for feelings of trust towards the European Parliament (illustrated in the Online appendix). The gradually increasing AME for support of general restrictions coupled with the almost null effect for non-Europe inclusionism suggests that people opposed to further EU unification move towards generalized restrictions; rather than switching to a preference for non-Europeans.

The second set of identity-based hypotheses also finds support. Here, European inclusionism is predicted by a preference for immigrants who share the receiving nation’s most prevalent cultural and ethnic identities. This is supported primarily for religious and national identity. Figure 3 shows the change in the probability of opting for each preference associated with the view that a Christian background is an important condition for immigrants to have. The probability of opting for Europe inclusionism slightly increases for those who think a Christian background is important, while the probability of general restrictionism is reduced. Notably, the same is not the case among those who value being white, speaking the language, or being committed to the country’s way of life as important. These factors instead differentiate between general inclusion and general restriction.

National identity also predicts Europe-only inclusionism: respondents who feel very close to their own country are about 3 percentage points more likely to opt for Europe inclusionism compared to those who reported feeling not at all close. Finally, people who think that some cultures are better than others are about 1.4 percentage points more likely to prefer European only inflows, and 4 percentage points more likely to be general restrictionists. People who view cultures as equal are more likely to be general inclusionists. Thus, the desire for identity-based exclusivity on religious and cultural dimensions predicts both a preference for European immigration and a preference for restricting immigration overall.

Unlike the identity-based hypotheses, the resource-based view gains very limited support. As Table 3 shows, people in medium- and high-skilled occupations are less likely to prefer general restrictions on immigration (relative to low skill occupations). However, the relationship between respondents’ skill level and European inclusionism does not follow theoretical expectations. Again, since European inflows are more likely to include low skilled workers, the labor market hypothesis suggests they would be welcomed by high skilled native workers and opposed by low skilled native workers. Attitudes toward non-European migrants, more likely to be highly skilled because of existing immigration restrictions, should show the opposite pattern. However, we find no difference by respondents’ occupational skill in the preference for European-only inflows.

Table 3.

Marginal effects of variables on inclusion preferences.

		General restrictionists		Europe inclusionists		Non-Europe inclusionists
Predictor	Values	dy/dx	P > \|z\|	dy/dx	P > \|z\|	dy/dx	P > \|z\|
Activity (vs. economically inactive)	In education	−0.058	0.000	0.028	0.025	0.005	0.318
	Unemployed	−0.044	0.003	0.019	0.074	−0.002	0.778
	In paid work	−0.001	0.904	0.000	0.998	0.001	0.711
Occupation (vs. low skill)	Medium skill	−0.018	0.021	−0.002	0.743	−0.004	0.088
Occupation (vs. low skill)	High skill	−0.037	0.000	0.002	0.596	−0.007	0.001
Feeling on current income (vs. very comfortable)	Very difficult	0.075	0.000	−0.016	0.070	0.001	0.805
Subjective health (vs. very good)	Very bad	0.037	0.156	0.011	0.588	0.012	0.269

Note: Estimation sample = 26,971 observations; pseudo R² = 0.18; marginal effect of factor on the probability of expressing each category of dependent A; mean effect with 95% CI; standard errors clustered by country and country fixed effects included alongside complete set of predictors.

Finally, the cognitive mobilization hypothesis does not find support. No measures of engagement or mobilization were significant predictors of the Europe inclusionism. Instead, greater interest and participation in politics were associated with general inclusionism.

Complete results and estimated AME for all predictors are available in the Online appendix. We also show the robustness of these substantive results to various specifications and estimation methods. Most important, we reestimated the main model using an alternative construction of the dependent variable that captures finer-grained differences in attitudes toward European vs. non-European immigration. The few differences that emerge tend to strengthen our general findings: high levels of job skills become inversely related with Europe inclusionism, against the expectations of the labor market hypothesis, although being in paid work becomes a positive predictor.

Analyzing support for EU and European migration

For our second dependent variable, we use multilevel modeling to estimate the probability of supporting inclusion of immigrants, with the generic experimental treatment conditions (professional vs. unskilled, European vs. non-European) included as independent variables. Multilevel models are ideal for this situation, with multiple, layered sources of variation in outcomes, and a goal of estimating effects on outcomes at the lowest level of variation (Steenbergen and Jones, 2002). We have individuals nested in countries, and we expect sources of variation at both of these levels. Country-level variation comes from differences in the specific immigrant-sending countries named in all four versions of the treatment, as well as any cross-national differences in attitudes toward either EU or non-EU immigrants.

We estimate these effects with a mixed-effects logistic regression model that includes all respondents (level 1) nested within countries (level 2). We use a binary form of the outcome variable, immigration preferences: 0 if the person opted for restriction (none/few), and 1 for inclusion (many/some). We then estimate the probability that an individual in country C will support inclusion of immigrants $(y = 1),$ as a function of (a) their individual characteristics; and (b) the treatment they received; we also include (c) random effects at level 2 for differences between countries and the differential impacts of the treatment at the country level.

Including a random coefficient for each treatment at the country level ensures that the estimated effects of our variables of interest are not confounded by country-level variation in which specific immigrant-sending countries where mentioned in the treatment wording. This variation was nonrandom and therefore needs to be controlled for statistically. We specify the covariance in the model as independent, since respondents cannot be part of more than one treatment or more than one country sample, simultaneously. This isolates the likelihood of preferring restriction or inclusion of immigrants depending on the treatment, while controlling for other country and individual-level differences. The specifications are otherwise identical across all estimations as discussed in the Data section.

We again use AMEs to estimate the impact of independent variables on the predicted probability of inclusionist immigration preferences. Since the treatment design does not vary EU membership experimentally, we use model estimates to derive the impact of receiving a treatment in which the sending countries is within the EU. Recall that half of respondents were asked to evaluate immigrants from a European sending country, but that designated sending country was an EU member only for a subset of receiving countries. Therefore, we estimate average marginal probabilities of supporting inclusion for respondents who were asked to evaluate immigrants from an EU member separately from respondents who received a non-EU sending country. The model already controls for other potential explanatory variables, allowing us to estimate the impact of EU inclusion on immigration preferences, above and beyond other explanatory factors that might increase support for migration from within Europe. We present the most relevant results for our argument below. The Online appendix includes complete results and robustness checks.

As Table 4 shows, we find a statistically significant advantage associated with EU status, compared with both non-EU European and non-European inflows. Support for allowing many or some immigrants is predicted at 74% when asked about skilled workers from EU sending countries in Europe and 47% for unskilled workers from the same sending country. The equivalent for those asked about European non-EU countries is estimated at 64% (skilled) and 35% (unskilled).

Table 4.

Predicted probability of preferring inclusion of inflows by treatment and EU/non-EU country mentioned.

Treatment fixed effect	Predictive margin	[95% CI]
Skilled from Europe (non-EU)	0.64	0.60	0.68
Skilled from Europe (EU)	0.74	0.70	0.77
Skilled from outside Europe	0.65	0.61	0.70
Unskilled from Europe (non-EU)	0.35	0.30	0.41
Unskilled from Europe (EU)	0.47	0.41	0.52
Unskilled from outside Europe	0.34	0.29	0.39

Note: Estimation sample = 28,599 observations; full model correctly predicts the outcome in 76% of cases, Pearson residuals larger than ±2 in 3.8% of estimation sample (residuals mean = 0.007); values shown represent predicted margins of probability of expressing support for allowing many or some immigrants; fixed portion of mixed effects logit. For contrast tests in differences between probabilities see the Online appendix.

Tests between mean differences in probability of support suggest that being asked about skilled immigrants from EU countries is associated with a 7% increase in inclusion compared to European but non-EU countries and 8% compared to skilled immigrants from outside Europe. Among those responding to unskilled inflows, EU origin is associated with an 8% increase in support for inclusion compared to non-EU European countries and 12% over unskilled inflows from outside Europe. Although modest in size, the advantage associated with EU origins is statistically significant in both the skilled and unskilled conditions.

Since it became apparent that intra-EU immigration drew more support than non-EU immigration (even from within Europe), our further analysis of the model focuses on accounting for support for this form of migration, both in general and in relation to support for non-EU European inflows. We return to the variables representing our initial hypotheses that support for EU immigration in particular might be explained by (a) identification with the EU; (b) economic responses to a source of low-skilled immigration; or (c) acceptance of elite cues about the acceptability of intra-EU mobility.

Figures 4 to 6 visualize the average probability of support for inclusion of inflows with its associated error range (95% confidence interval (CI)) for respondents in the European sending country treatment groups, across the values of a selected predictor. We show probabilities separately for those treated with EU countries and those treated with non-EU European countries. Non-Europe treatments are included in the model but not illustrated in the figures. The first key results are the changes (or lack thereof) in probability of support for EU inflows across the values of a given predictor, which show how that factor affects support for EU mobility in general. We also focus on the disparities in probability between the EU and European non-EU versions of the treatments, to understand whether some factors are associated with people differentiating more or less between EU and non-EU European immigration. A lack of overlap between the intervals of two lines (representing treatments) indicates that the probability of inclusion of those inflows statistically differs for these two treatments at the 95% confidence level.

We find again that proxies for EU identification explain both support for immigration and the tendency to differentiate between immigration flows by geographical origin. As Figure 4 shows, greater opposition to further EU unification is associated with reduced support for inclusion across skill levels and for both EU and non-EU European sending countries. More important, it is the people who support EU unification (lower end of scale) who are driving the aggregate gap, favoring EU immigration more than non-EU European immigration. This gap disappears among those opposed to further unification, as seen in the narrowing gap between the intervals across the values of the scale.

Average support for inclusion is estimated at 66% among respondents who were maximally positive to the EU and who were asked about unskilled inflows from EU countries, but 27% among people maximally opposed to EU unification asked about the same inflows—a drop of 39 percentage points in probability of support. The equivalent gap was roughly 29 percentage points when considering unskilled non-EU inflows from Europe.

This suggests that EU identity has a dual effect on immigration attitudes: it does some work in support of a particular conception of intra-EU mobility as free movement, while also representing a more positive viewpoint toward European immigration more broadly. This pattern of results is similar for the independent variable measuring trust in the EU parliament as an additional and perhaps more distant proxy for EU identity.

In contrast, the subgroup identity-based predictors of Europe-only inclusionism from the earlier analysis do not predict additional support for EU over non-EU European immigrants, as illustrated in Figure 5. This is consistent with expectations shown in Table 1. Unsurprisingly, respondents who think that whiteness or Christianity are important qualifications for immigrants are less likely to support immigration generally. For people who say that any of these cultural criteria are important, the estimated probability of supporting inclusion is almost identical for EU and non-EU European inflows (small or no gaps between lines/treatments). This supports our expectations that racial or religious biases do not encourage distinctions between EU and non-EU European migrants. This is largely true of other subgroup identity variables as well; with a possible exception at the low end of the scales for those considering language proficiency and commitment to way of life as unimportant characteristics for migrants.

Resource-based views again receive limited support. Consistent with the labor market hypothesis, support for unskilled immigrant workers is particularly low among respondents who themselves work low-skill occupations; however, the opposite prediction does not hold for skilled respondents’ preferences. Respondents in highly skilled occupations exhibit the most positive attitudes toward other skilled inflows, especially from EU countries, rather than being particularly likely to exhibit feelings of competition stemming from labor market displacement or lowered wages.

Meanwhile, cognitive and elite cue mobilization contributes to higher support for inflows across all wording treatment groups, but not to differentiating EU inflows from others. Having voted in the last national election is not related to immigration preferences across all skills-origin treatments.

Conclusion

Using social survey data across 20 European countries, we have examined the incidence and determinants of Europe and EU inclusionism in Europeans’ preferences for immigration flows. We identify occasions where respondents express preference for more immigrants to be allowed from within Europe and/or the EU than from outside, although these represent only about 1-in-10 respondents. Further, we examined the determinants of patterns of immigration preferences, testing explanations derived from several prominent theories of immigration attitudes and EU support.

Our descriptive findings alone have important political implications. The political project of the EU depends on broad acceptance of intra-EU mobility, as Brexit illustrates. Yet anti-immigration sentiment is commonplace in immigrant-receiving societies (Duffy and Frere-Smith, 2014), and this does not seem likely to change. Therefore, broadening support for the EU project requires that some portion of the public who generally feel negatively about immigration will make an exception for fellow EU nationals

Our descriptive results show the weakness of support for such an exception. The vast majority of support for EU mobility comes from people who support immigration from everywhere; less than 10% of ESS respondents support immigration from Europe but not from outside Europe. Similarly, when skills and origins are manipulated experimentally and respondents are asked about either EU or non-EU European immigrants, we found a relatively modest preference for immigration from EU countries over non-EU European countries, all else equal. These patterns contrast with the normative position entrenched in EU institutions and rhetoric in which EU mobility is decidedly favored over non-EU immigration.

Theoretically, our results highlight the explanatory power of both supranational and subgroup identities, and downplay the role of labor market position and cognitive mobilization, staples of research on immigration attitudes and EU integration, respectively. Identification with the EU has real if limited impact on specific support for intra-EU mobility. On the other hand, a preference for immigrants with shared subgroup identities such as a Christian background, is linked to European inclusionism but not to a specific preference for EU immigrants over other Europeans. Thus, the estimated effects of supranational EU identities and sub-group identification with Europeans’ most common religion have partially overlapping and partially diverging effects.

Of course, given the potential for endogeneity in the relationship between EU identity and immigration attitudes, our findings must be acknowledged to demonstrate associations rather than causation. In particular, it has been suggested that, while EU identity might shape immigration attitudes, at the same time immigration attitudes might play a role in determining attitudes toward the EU (Stockemer et al., 2018). We suggest that this reverse causal pathway seems less likely given our descriptive findings showing that relatively few Europeans distinguish between EU and non-EU immigration. However, for those that do make this distinction, we acknowledge that approval of the EU might be triggered by support for the EU’s preferred pattern of immigration restrictions, just as EU identity might boost support for intra-EU migration.

Meanwhile, our non-findings bear emphasis as well. The weakness of the cognitive mobilization theory contrasts with earlier findings, although this may not be surprising given increased elite disagreement on EU issues. Further research might aim to distinguish whether this result reflects a lack of elite influence on disaffected citizens, or, instead, ongoing influence by elites who are more polarized on issues of immigration and Europe. Our results also reaffirm the weakness of objective labor market position as an explanation for attitudes toward immigration, consistent with recent research (Jeannet, 2018) and reviews (Hainmueller and Hopkins, 2014). Low-skilled workers are not particularly opposed to intra-EU mobility, the largest source of low skilled migrant inflows that they face. We thus provide further reason for economic explanations of immigration attitudes to shift away from the labor market hypothesis and towards theories emphasizing fiscal burdens or sociotropic perceptions of impacts on national economies.

Our tests of these theories indicate limited but real avenues for understanding EU exceptionalism. The only consistent predictors of specific support for intra-EU mobility involve support for or identification with the EU itself. Education and occupational variables are associated with pro-migration attitudes generally, but—against prior expectations generated by existing theories—these variables are not associated with EU-only inclusionism. Thus, increasing specific support for intra-EU mobility depends on generating increased EU identity among those who do not generally support immigration, a difficult task to be sure.

Supplemental Material

Supplemental Material1 - Supplemental material for Acceptable in the EU? Why some immigration restrictionists support European Union mobility

Supplemental material, Supplemental Material1 for Acceptable in the EU? Why some immigration restrictionists support European Union mobility by Scott Blinder and Yvonni Markaki: the EuroHYP-1 investigators in European Union Politics

Supplemental Material

Supplemental Material2 - Supplemental material for Acceptable in the EU? Why some immigration restrictionists support European Union mobility

Supplemental material, Supplemental Material2 for Acceptable in the EU? Why some immigration restrictionists support European Union mobility by Scott Blinder and Yvonni Markaki: the EuroHYP-1 investigators in European Union Politics

Footnotes

Acknowledgements

The authors thank the researchers and staff of the REMINDER (Role of European Mobility and its Impacts in Narratives, Debates and EU Reforms) project consortium and participants in the Centre of Migration, Policy and Society (COMPAS) Works-in-Progress Seminar. Additional special thanks to Michael Donnelly, Christine Melzer, Martin Ruhs, Meredith Rolfe and Chiara van Praag for comments and assistance. Earlier versions of this paper were presented at the Annual Scientific Meeting of the International Society of Political Psychology, San Antonio, TX, 4–7 July 2018 and at the American Political Science Association Annual Meetings, Boston, MA, 30 August—2 September 2018.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 727072.

Supplemental material

Supplemental material for this article is available online.

Notes

References

Azrout

Wojcieszak

(2017) What’s Islam got to do with it? Attitudes toward specific religious and national out-groups, and support for EU policies. European Union Politics 18(1): 51–72.

Bansak

Hainmueller

Hangartner

(2016) How economic, humanitarian, and religious concerns shape European attitudes toward asylum seekers. Science 354(6309): 217–222.

Beaudonnet

Di Mauro

(2012) Support for Europe: Assessing the Complexity of Individual Attitudes. ID 2191447, SSRN Scholarly Paper. Rochester, NY: Social Science Research Network. Available at: https://papers.ssrn.com/abstract=2191447 (accessed 1 March 2019).

Berg

(2010) Race, class, gender, and social space: Using an intersectional approach to study immigration attitudes. The Sociological Quarterly 51(2): 278–302.

Blinder

(2015) Imagined immigration: The impact of different meanings of ‘Immigrants’ in public opinion and policy debates in Britain. Political Studies 63(1): 80–100.

Bogard

Sherrod

(2008) Citizenship attitudes and allegiances in diverse youth. Cultural Diversity and Ethnic Minority Psychology 14(4): 286–296.

Bruter

(2009) Time bomb? The dynamic effect of news and symbols on the political identity of European citizens. Comparative Political Studies 42(12): 1498–1536.

Burns

Gimpel

(2000) Economic insecurity, prejudicial stereotypes, and public opinion on immigration policy. Political Science Quarterly 115(2): 201–225.

Ceobanu

Escandell

(2010) Comparative analyses of public attitudes toward immigrants and immigration using multinational survey data: A review of theories and research. Annual Review of Sociology 36: 309–328.

10.

Cerna

(2014) Attracting high-skilled immigrants: Policies in comparative perspective. International Migration 52(3): 69–84.

11.

Creighton

Schmidt

Zavala‐Rojas

(2018) Race, wealth and the masking of opposition to immigrants in the Netherlands. International Migration EarlyView online.

12.

Curtis

(2014) Inclusive versus exclusive: A cross-national comparison of the effects of subnational, national, and supranational identity. European Union Politics 15(4): 521–546.

13.

Duffy

Frere-Smith

(2014) Perceptions and Reality: Public attitudes to immigration. London: Ipsos-MORI Research Institute.

14.

Elsas EJ van, Hakhverdian A and Brug W van der (2016) United against a common foe? The nature and origins of Euroscepticism among left-wing and right-wing citizens. West European Politics 39(6): 1181–1204.

15.

Erisen

(2017) Seeking refuge in a superordinate group: Non-EU immigration heritage and European identification. European Union Politics 18(1): 26–50.

16.

European Social Survey (2014). ESS-7 2014 Documentation Report. Edition 3.2. Bergen, European Social Survey Data Archive, NSD - Norwegian Centre for Research Data for ESS ERIC. doi:10.21338/NSD-ESS7-2014.

17.

Facchini

Mayda

(2009) Does the welfare state affect individual attitudes toward immigrants? Evidence across countries. Review of Economics and Statistics 91(2): 295–314.

18.

Favell

(2008) The new face of east–west migration in Europe. Journal of Ethnic and Migration Studies 34(5): 701–716.

19.

Fligstein

Polyakova

Sandholtz

(2012) European integration, nationalism and European identity. JCMS: Journal of Common Market Studies 50(s1): 106–122.

20.

Ford

(2008) Is racial prejudice declining in Britain? The British Journal of Sociology 59(4): 609–636.

21.

Gabel

(1998) Public support for European integration: An empirical test of five theories. The Journal of Politics 60(2): 333–354.

22.

Geddes

Hadj-Abdou

(2016) An unstable equilibrium: Freedom of movement and the welfare state in the European Union. In: Gary P. Freeman and NIkola Mirilovic (eds) Handbook on Migration and Social Policy. Cheltenham, UK: Elgar, pp. 222–238.

23.

Gerber

Huber

Biggers

, et al. (2017) Self-interest, beliefs, and policy opinions: Understanding how economic beliefs affect immigration policy preferences. Political Research Quarterly 70(1): 155–171.

24.

Goodwin

Milazzo

(2017) Taking back control? Investigating the role of immigration in the 2016 vote for Brexit. The British Journal of Politics and International Relations 19(3): 450–464.

25.

Gorodzeisky

Semyonov

(2016) Not only competitive threat but also racial prejudice: Sources of anti-immigrant attitudes in European societies. International Journal of Public Opinion Research 28(3): 331–354.

26.

Hainmueller

Hiscox

(2010) Attitudes toward highly skilled and low-skilled immigration: Evidence from a survey experiment. American Political Science Review 104(1): 61–84.

27.

Hainmueller

Hopkins

(2014) Public attitudes toward immigration. Annual Review of Political Science 17(1): 225–249.

28.

Hanmer

Kalkan

(2013) Behind the curve: Clarifying the best approach to calculating predicted probabilities and marginal effects from limited dependent variable models. American Journal of Political Science 57(1): 263–277.

29.

Hobolt

de Vries

(2016) Public support for European integration. Annual Review of Political Science 19: 413–432.

30.

Hooghe

Marks

(2005) Calculation, community and cues: Public opinion on European integration. European Union Politics 6(4): 419–443.

31.

Hooghe

Marks

(2009) A postfunctionalist theory of European integration: From permissive consensus to constraining dissensus. British Journal of Political Science 39(1): 1–23.

32.

Hooghe

Marks

(2018) Cleavage theory meets Europe’s crises: Lipset, Rokkan, and the transnational cleavage. Journal of European Public Policy 25(1): 109–135.

33.

Jeannet

A-M

(2018) Revisiting the labor market competition hypothesis in a comparative perspective: Does retirement affect opinion about immigration? Research & Politics 5(3): 2053168018784503.

34.

Kentmen-Cin

Erisen

(2017) Anti-immigration attitudes and the opposition to European integration: A critical assessment. European Union Politics 18(1): 3–25.

35.

Kuhn

(2015) Experiencing European Integration: Transnational Lives and European Identity. Oxford: Oxford University Press.

36.

Kwak

Clayton-Matthews

(2002) Multinomial logistic regression. Nursing Research 51(6): 404.

37.

Mayda

(2006) Who is against immigration? A cross-country investigation of individual attitudes toward immigrants. Review of Economics and Statistics 88(3): 510–530.

38.

McLaren

Johnson

(2007) Resources, group conflict and symbols: Explaining anti-immigration hostility in Britain. Political Studies 55(4): 709–732.

39.

McLaren

(2001) Immigration and the new politics of inclusion and exclusion in the European Union: The effect of elites and the EU on individual–level opinions regarding European and non–European immigrants. European Journal of Political Research 39(1): 81–108.

40.

O’Rourke

Sinnott

(2006) The determinants of individual attitudes towards immigration. European Journal of Political Economy 22(4): 838–861.

41.

Pasek

Tahk

Lelkes

, et al. (2009) Determinants of turnout and candidate choice in the 2008 U.S. Presidential election. Public Opinion Quarterly 73(5): 943–994.

42.

Pichler

(2010) Foundations of anti-immigrant sentiment: The variable nature of perceived group threat across changing European societies, 2002-2006. International Journal of Comparative Sociology 51(6): 445–469.

43.

Risse

(2010) A Community of Europeans? Transnational Identities and Public Spheres. Ithaca, NY: Cornell University Press.

44.

Ruhs

(2017) Free movement in the European Union: National institutions vs common policies? International Migration 55: 22–38.

45.

Ruhs

(2018) Labor immigration policies in high-income countries: Variations across political regimes and varieties of capitalism. The Journal of Legal Studies 47(S1): S89–S127.

46.

Spigelman

(2013) The depiction of Polish migrants in the United Kingdom by the British press after Poland’s accession to the European Union. International Journal of Sociology and Social Policy 33(1/2): 98–113.

47.

Steenbergen

Jones

(2002) Modeling multilevel data structures. American Journal of Political Science 46(1): 218–237.

48.

Stockemer

Niemann

Rabenschlag

, et al. (2018) Immigration, anti-immigrant attitudes and Eurosceptism: A meta-analysis. French Politics 16(3): 328–340.

49.

Strabac

Listhaug

(2008) Anti-Muslim prejudice in Europe: A multilevel analysis of survey data from 30 countries. Social Science Research 37(1): 268–286.

50.

Teney

Lacewell

De Wilde

(2013) Winners and losers of globalization in Europe: Attitudes and ideologies. European Political Science Review 6(4): 575–595.

51.

Valentino

Soroka

Iyengar

, et al. (2017) Economic and cultural drivers of immigrant support worldwide. British Journal of Political Science. Epub ahead of print 1 November 2017. DOI: 10.1017/S000712341700031X

52.

Zaller

(1992) The Nature and Origins of Mass Opinion. New York, NY: Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

14.32 MB

0.00 MB

0.37 MB