The European dimension of identity

Much of the scholarly work on identity is grounded in social identity theory, which posits that identity formation involves a stylised process of self-categorisation into a group, identification with that group, and the comparison with others. It refers to the process by which individuals recognise ‘the characteristic(s) shared by those with whom one psychologically belongs’, which may result in ‘multiple potential social identities’ (Citrin et al., Reference Citrin, Wong, Duff, Ashmore, Jussim and Wilder2001, 73). The identification with a specific group leads to the definition of an ‘us’ and consequently to a demarcation from others not belonging to this in-group, resulting in a distinction between the in- and out-group (see, eg Hainmueller and Hopkins, Reference Hainmueller and Hopkins2014; Kalin and Sambanis, Reference Kalin and Sambanis2018; Lindstam et al., Reference Lindstam, Mader and Schoen2021). This distinction is subject to an evaluative process, which relates positive characteristics to the in-group, while the perception of the out-group is not necessarily neutral because the out-group may be associated with stereotypes and negative perceptions (Tajfel, Reference Tajfel1981; Bergbauer, Reference Bergbauer2018; Royuela, Reference Royuela2019). As, for example, with regard to immigration, this negative perception can result in a feeling that immigrants are perceived as a threat to the economic interests or cultural identities of the host societies (Sides and Citrin, Reference Sides and Citrin2007; Wright, Reference Wright2011; Hellwig and Sinno, Reference Hellwig and Sinno2017). If so, individuals may adopt the view that the in-group should be protected from the out-group, which further contributes to negative attitudes towards immigration. The preferences resulting from this process are remarkably stable over time (Kustov et al., Reference Kustov, Laaker and Reller2021).

A much-studied distinction between in-groups and out-groups relates to nationalism (Eastwood and Greenfeld, Reference Eastwood, Greenfeld, Boix and Stokes2009; Erhardt et al., Reference Erhardt, Wamsler and Freitag2021). Studies going beyond national boundaries in the European context build on the contrast between European and national identities (McLaren, Reference McLaren2004; Del Ponte, Reference Del Ponte2021). The distinction between European and national identities assumes that there is only one form of identification with Europe and also ignores that local, national, and European identities are related to one another. Individuals do not exclusively identify with their nation-state or some form of Europe. Rather, national and European identities co-exist, interact, and are multifaceted (Romaniszyn, Reference Romaniszyn, Spohn and Triandafyllidou2003; Duchesne and Frognier, Reference Duchesne and Frognier2008; Guglielmi and Vezzoni, Reference Guglielmi, Vezzoni, Westle and Segatti2016; Dalton, Reference Dalton2021). This aligns with the described process of self-categorisation into several groups (Citrin et al., Reference Citrin, Wong, Duff, Ashmore, Jussim and Wilder2001), leading to individuals having multiple identities simultaneously (Roccas and Brewer, Reference Roccas and Brewer2002). However, these multiple identities are not equally salient at all times and are activated in relation to the environment in which they evolve (Lustick and Miodownik, Reference Lustick and Miodownik2002), as the sense of belonging to a specific group can be triggered by particular situational cues or contextual factors. When citizens act as voters, for example, they adapt their identities to align with their preferred political positions (Egan, Reference Egan2020), policies (Bergbauer, Reference Bergbauer2018), and political institutions such as electoral rules (Posner, Reference Posner2005). Crises can also affect identities: in Europe, for example, citizens expressed stronger attachment to the common European cause following Russia’s invasion of Ukraine (Steiner et al., Reference Steiner, Berlinschi, Farvaque, Fidrmuc, Harms, Mihailov, Neugart and Stanek2023; Nicoli et al., Reference Nicoli, van der Duin, Beetsma, Bremer, Burgoon, Kuhn, Meijers and de Ruijter2024).

Related to identities is also the concept of citizenship, which typically has a territorial and jurisdictional component. More recent studies on citizenship emphasise the multilevel dimension, stating that citizenship is a ‘patchwork of different statuses operating simultaneously’ at the subnational, supranational, and transnational level (Maas, Reference Maas, Shachar, Bauböck, Bloemraad and Vink2017, 644). Like identity, citizenship is a constantly evolving concept with varying facets. In addition to local and subnational expressions, there is also a supranational citizenship with the EU as the most advanced example thereof (Maas, Reference Maas2005). This literature is a further testimony to the pertinence of the multilevel perspective in the European context. If we return to identities, a more recent strand of the literature explicitly theorises different forms of European identity (Pichler, Reference Pichler2008; Bergbauer, Reference Bergbauer2018; Voicu and Ramia, Reference Voicu and Ramia2021; König, Reference König2024). Our investigation builds on these studies, which submit that Europeans self-identify with a socially constructed collective of Europeans (Bruter, Reference Bruter2005; Risse, Reference Risse2010). A main insight of this literature is the absence of one single form of European identity that supersedes national and local identities (Pichler, Reference Pichler2008; Hooghe and Marks, Reference Hooghe and Marks2018; Clark and Rohrschneider, Reference Clark and Rohrschneider2021; Voicu and Ramia, Reference Voicu and Ramia2021). In a nutshell, two conceptions of Europeanness stand in contrast to each other: a civic identification with the EU and a cultural identification with Europe (Bruter, Reference Bruter2003; König, Reference König2024).

Our theoretical conceptualisation is also similar to the distinction between a civic and ethno-cultural identity applied in the context of national identity studies (Kohn, Reference Kohn2017; Erhardt et al., Reference Erhardt, Wamsler and Freitag2021; Lindstam et al., Reference Lindstam, Mader and Schoen2021; Piwoni & Mußotter, Reference Piwoni and Mußotter2023; König, Reference König2024). The literature studying the influence of national identity on immigration preferences submits that an ethno-cultural understanding of the nation tends to be more closely associated with restrictive immigration attitudes compared to the civic conception of national identity (Lindstam et al., Reference Lindstam, Mader and Schoen2021), although the directionality of the effect of the latter is contested (Schildkraut, Reference Schildkraut2013). In any case, aligning with the idea of multiple identities and multilevel citizenship, Lindstam et al. (Reference Lindstam, Mader and Schoen2021, 111) point out that ‘competing conceptions of the in-group are not confined to the nation’.

In sum, a growing and extensive scholarship studies the European dimension of identity and distinguishes between civic and cultural conceptions of Europeanness (Risse, Reference Risse2010; Glencross, Reference Glencross2020). This research is rich in conceptual and theoretical analyses, also because it can build on a vast literature on national identities, multilevel citizenship, and social identity theory, while, at the same time, few thorough empirical investigations of European identities exist. One reason for this lacuna is that established data provide only crude measures of civic and cultural European identities. A principal strength of our inquiry is that we distinctively conceptualise civic and cultural European identities and precisely measure them empirically with original data. Before we turn to this analysis, we investigate Eurobarometer data to get a first glimpse at the broader distinction between civics and culturals for all EU countries. The main motivation of the following analysis is to derive a country selection for our main empirical investigation, which we have to restrict to seven countries due to limited resources.

Observational data analysis and case selection

While standard Eurobarometer data do not allow for a nuanced measurement of European identities and immigration attitudes, they provide a broad coverage of all 27 EU countries. Also, they include two questions, which could be considered as (crude) proxies for civics and culturals, namely, whether respondents feel (a) attached to Europe and (b) to the EU. We code respondents who are ‘very attached’ to the EU but not to Europe as civics, and vice versa for culturals. With these data and coding, we can first provide a broad overview regarding the general immigration attitudes of culturals and civics and, second, derive an empirically informed selection of countries for the subsequent main analyses.

Culturals hold more restrictive immigration attitudes

The cultural conception is more rigid in its discrimination between the in-group and out-group, compared to the civic, because the othering (ie the definition of who does not belong to the European community) is based on a stronger demarcation line between the in-group and out-group (Blinder and Markaki, Reference Blinder and Markaki2019; Lindstam et al., Reference Lindstam, Mader and Schoen2021; König, Reference König2024). This rigid demarcation is not limited to European identities but also characterises national identity when conceived in cultural terms, where belonging is often tied to ancestry, heritage, or ethno-cultural markers (Smith, Reference Smith1998). In contrast, civic identity is based on values to which all individuals can subscribe (Wright et al., Reference Wright, Citrin and Wand2012). This suggests that culturals hold more restrictive attitudes towards immigration than civics (very generally speaking). To evaluate this hypothesis, we use as a dependent variable the Eurobarometer question asking respondents whether immigration of people from outside the EU evokes a positive or negative feeling in them. The answer categories are ‘very positive’, ‘fairly positive’, ‘fairly negative’, and ‘very negative’. The main explanatory variable is a broad distinction between culturals and civics, on the basis of the Eurobarometer, measured as attachments to Europe or the EU on a scale from 1 to 4.Footnote ¹ Besides the European identity variables, we include in the model measures for exclusive national identity (ie respondents who are ‘fairly attached’ or ‘very attached’ to their country but not to Europe and the EU), gender, place of residence (rural or urban), age, left-right self-placement, and education. We run a linear regression with country-fixed effects for the analysis of the difference between civics and culturals (while controlling for individual-level factors and country differences).

As expected, the estimates reported in Figure 1 show that civics are more supportive of immigration than culturals. Having a strong attachment to the EU correlates positively and strongly with a favourable attitude towards immigration from outside the EU (0.27) compared to individuals with a cultural European identity. This difference between civics and culturals is particularly notable as we also account for individuals’ national identity, which, not surprisingly, is also a very strong predictor of restrictive immigration attitudes.Footnote ²

Figure 1.

Linear regression estimates with country-fixed effects (positive effects indicate that a factor is associated with less restrictive immigration attitudes). Data: Eurobarometer 93.1 (European Commission, 2022a).

Country case selection

Due to limited resources, we had to restrict the data collection for our main empirical analysis with our own data to seven countries. For the selection of these seven countries, our attempt was to cover a broad range in terms of the distribution of culturals and civics among all 27 EU countries. With this strategy, we avoid biases stemming from the country case selection. To that end, we estimate the average attachments to Europe and the EU for all 27 EU countries using the Eurobarometer data. Figure 2 shows the distribution of EU member states along the two axes, displaying the mean attachment to Europe and the EU.

Figure 2.

Mean attachment to the EU and to Europe for EU member states in 2022. Data: Eurobarometer 96.3 (European Commission, 2022b).

The countries line up on the diagonal from weak to strong cultural and civic identities, suggesting that the two dimensions are not orthogonal. In the upper left corner are the populations of Sweden and Finland, which are, on average, more strongly attached to Europe than they are to the EU, reflecting Nordic Euroscepticism (Raunio, Reference Raunio2007). The lack of countries in the lower right corner shows that no population is, on average, more attached to the EU than to Europe. For the analysis with our own original data, we selected France, Germany, Greece, Hungary, Italy, Poland, and Sweden (coloured red in Figure 2).Footnote ³ Hungary and Poland represent countries with high attachments to the EU and Europe, while the Greeks have a low attachment to Europe and to the EU. The citizens of France, Germany, and Italy have, on average, moderate attachments to Europe and the EU. As discussed above, the Eurobarometer questions do not measure precisely the civic and cultural forms of European identities. Rather, they serve as approximations, and the data analyses suggest that we find differences in general immigration attitudes between culturals and civics, and they help us to derive a systematic selection of seven countries for our main analysis.

Theoretical considerations and expectations

As discussed above, we build on the different strands of literature on national identity, social identity, and multilevel citizenship with the ambition to advance the conception of European self-identification. The following theorises more clearly the civic and cultural European identity types, and how they relate to immigration attitudes.

Conjoint analysis

We fielded our surveys with Kantar from 15 March 2023 to 18 April 2023. The total sample includes 10,285 completed responses by individuals between the ages of 18 and 85 years who are citizens of the respective EU member states (slightly less than 1,500 respondents per country). More information on the survey can be found in the Appendix. The sampling strategy was based on quotas for age and gender, which were met, except for the oldest age bracket in Greece (75–85). Furthermore, we checked whether the shares of party sympathy per country corresponded to the party strength in the last election (to ensure that our sample is representative of the political landscape). This was the case for most parties and countries, with the exception of centre-right voters in Hungary and Poland, who are slightly under-represented. We present in the article the results from the unweighted sample as the main analysis and robust findings with stratified samples in Appendix D.

To precisely measure civics and culturals, our survey includes four questions about the different dimensions of the civic and cultural conceptions based on the theoretical elaboration discussed above (two for each). More precisely, we ask respondents how important on a scale from 1 (not important at all) to 5 (absolutely essential) the following two dimensions are for their identity as Europeans:

Civic: (a) a common belief in democratic institutions and the rule of law and (b) values such as respect for human rights, freedom, and equality

Cultural: (a) a common European cultural and historical heritage and (b) a common Christian tradition

To derive a relative estimate ranging from -8 (distinctively cultural) to 8 (distinctively civic), we estimate the sum of these two dimensions and the differences between the two scores. Accordingly, we classify respondents with negative values on this relative score as culturals and those with positive values as civics. This measure also lends itself to a straightforward classification of a reference group with a score of 0, which we call the undefined group, as respondents of this group subscribe to both sets of values to an equal extent. Our sample includes 1,139 culturals, 6,245 civics, and 2,284 with an undefined European identity. Civics thus clearly outnumber culturals, and a substantial share of respondents does not prioritise either civic or cultural values. Figure 3 shows the distribution of our respondents on the identity spectrum with different colouring shades for culturals, undefined, and civics. This distribution also aligns with the conception of an assimilationist and ethno-cultural understanding of identity presented by Schildkraut (Reference Schildkraut2013) in the American context.

Figure 3.

Distribution of culturals, undefined, and civics.

Of course, identities are not unrelated to other factors, such as left-right political orientation, gender, or education. Figure 4 shows the distributions of self-placement on the left-right scale for civics, culturals, and the undefined. The differences are as expected (ie culturals are the most right-wing), but even for this political variable, for which we find the strongest correlation with the identity types, the differences are small. Appendix A2 reports how civics, culturals, and the undefined differ along political and socio-demographic characteristics, providing a clearer picture of who is part of these groups. Overall, we do find differences, as one would expect them, but these differences are small, suggesting that our identity-categorisation is not mimicking one specific variable but captures a genuine, independent factor. It is also important to emphasise that the common characteristic of the undefined is not that they do not share European values. Rather, they subscribe to the core values of culturals and civics to the same extent, meaning that they do not discriminate between the two sets of values, which makes this an excellent reference group for the investigation of the differences between culturals and civics (see Appendix A2).

Figure 4.

Distribution of left-right self-placement by identity groups.

For the analysis of how culturals and civics relate to immigration attitudes, we rely on conjoint experiments in which we ask respondents whom they would rather give priority to being admitted to their country. Conjoint experiments are a powerful method to investigate multidimensional preferences (Hainmueller et al., Reference Hainmueller, Hopkins and Yamamoto2014, Reference Hainmueller, Hangartner and Yamamoto2015), and they have been applied in research on immigration preferences (Hainmueller and Hopkins, Reference Hainmueller and Hopkins2015; Bansak et al., Reference Bansak, Hainmueller and Hangartner2016; Spilker et al., Reference Spilker, Nguyen, Koubi and Böhmelt2020; Jeannet et al., Reference Jeannet, Heidland and Ruhs2021; Letki et al., Reference Letki, Walentek, Dinesen and Liebe2025). The profiles of immigrants in our surveys are composed of the attributes theorised above, which are adherence to cultural or civic values, country of origin, religion, education, gender and family attributes, and education. Following the conjoint approach, we show pairwise comparisons and randomly vary the expressions of these attributes and the order in which they are presented. Each respondent evaluates seven pairs of hypothetical immigrant profiles, deciding which one they would rather see admitted.Footnote ⁴ Table 1 provides an overview of the immigrant attributes and possible expressions thereof, including the reference categories (in bold). Appendix G presents diagnostic checks following Hainmueller et al. (Reference Hainmueller, Hopkins and Yamamoto2014) and Leeper et al. (Reference Leeper, Hobolt and Tilley2020) for subgroup analyses and a power analysis for our main results.Footnote ⁵

Table 1.

Profiles of immigrants in the conjoint survey

Findings

We report the average marginal component effects (AMCEs), as they allow for an evaluation of the causal effects of the different characteristics of immigrants on the admission decision. Comparing the AMCEs between culturals, undefined, and civics shows which characteristics of immigrants influence the choice of respondents to give priority in admission decisions and whether there are differences in these effects between the three groups of European (for the descriptive levels of preferences with respect to specific attributes, see the marginal means reported in Appendix D2). When comparing subgroups, AMCEs are sensitive to the chosen reference category, but they allow us to investigate the causal effects of immigrant attributes while still contextualising the general subgroup preference levels (Leeper et al., Reference Leeper, Hobolt and Tilley2020).

The main results of the conjoint analysis are reported in Figure 5 (see Appendix D1 for the estimation tables). The estimates in the left plot indicate whether (and by how much) the specific attribute increases or decreases the probability of a profile being chosen by culturals, civics, and the reference category. The right plot indicates whether the effects are statistically different for civics and culturals compared to the reference group (ie the undefined). The comparison to the reference group in the right plot is particularly relevant for our analysis, as it allows for a precise evaluation of whether we find statistically significant differences between culturals and civics.

Figure 5.

Average marginal component effects (AMCEs) of immigrant profile attributes on support for admission (left plot). Estimates with 95% confidence intervals show changes in the probability of giving priority to the admission of immigrants with a specific attribute, compared to the reference attribute in that group (see Table 1 above). The right plot reports whether the effects for civics and culturals are statistically different compared to the reference category of the undefined.

Let us start with the novel hypotheses of this inquiry on value congruence. E1 predicts that Europeans prefer immigrants who share the in-group values associated with both conceptions of European identities. The empirical findings corroborate that adherence to the cultural or civic set of values strongly increases the probability of a positive admission decision in all subgroups. The marginal means further emphasise the importance and strength of these effects (see Figure A7 in the Appendix), as they show that all respondents prioritise immigrants whose values are known (and correspond to European values) over immigrants whose values are unknown. It is noteworthy that the preference for an immigrant with no indication of values is as low as for immigrants with no formal education, illustrating the substantive size of this effect. Europeans thus strongly prioritise immigrants who share cultural or civic European values, as theorised by E1.

If we further investigate, as stipulated by E2 and E3, whether civics more strongly prioritise immigrants with civic values and culturals those with cultural values, the results are more nuanced. The left plot of Figure 5 shows that for both subgroups (and the undefined), adherence to one or the other set of values is important, while civics make a difference between cultural and civic values. The effect of adherence to civic values is much stronger than the effect of cultural values, supporting hypothesis E2. Culturals (and the undefined), in contrast, do not differentiate between the two value sets. We thus do not find empirical evidence for E3. The right plot further supports this finding by showing that for civics (compared to the undefined), sharing civic values has a clear positive effect on their prioritisation of immigrants with civic values. We thus find evidence for the more nuanced formulation of the value congruence as far as the civics are concerned (E2) but not for culturals (E3). Taken together, adherence of immigrants to civic and cultural values is very relevant in admission decisions across the board, while civics make a more systematic distinction by giving stronger priority to immigrants with civic values.

Finally, E4 predicts that Europeans derive perceptions about immigrants from personal characteristics, such as the country of origin, and E5 stipulates that these effects are stronger for culturals compared to civics. The left plot of Figure 5 shows that being from Afghanistan and Albania has a negative effect (compared to India) on the admission decision, while being from Ukraine and the United Kingdom (UK) has a positive effect for respondents of all three subgroups. We find, as expected by E4, a clear selection pattern based on cultural closeness – with Afghanistan as the most distant country of origin, India and Albania in the middle, and the UK and Ukraine as closest (see also the marginal means in Figure A7 of the Appendix). That EU citizens evaluate Albanians like Indians is noteworthy, given that Albania has been granted EU candidate status.

When we zoom in on the differences among subgroups, we find two significant results in the right plot. First, the negative effect on admission decisions for Afghans is less pronounced for civics and strongest for culturals. This supports E5, which submits that culturals draw a stronger demarcation line based on the non-European origin of Afghan immigrants compared to civics. Second, if we look at the UK, we find an additional interesting difference: being from the UK leads to a more critical evaluation by civics and a strongly positive one by culturals. We did not anticipate this finding, which points to distinct territorial evaluations by Europeans that, nonetheless, align with our theoretical differentiation between culturals and civics. Culturals build more on Europe as a boundary and draw a stronger demarcation to immigrants from non-European countries, such as Afghanistan. Civics, on the contrary, integrate the EU as a relevant border in their supranational self-identification, as the more recent literature on multilevel citizenship suggests (Maas, Reference Maas, Shachar, Bauböck, Bloemraad and Vink2017). This may explain why civics, who are strongly tied to the EU institutions and its territory, have a more unfavourable view towards citizens from the UK, after Brexit. They take the withdrawal from EU membership negatively into account (also on the level of individuals). For culturals, however, the relevant territorial border is (non-)European, not (non-)EU, and therefore they have a positive stance towards immigrants from the UK.

Religion also matters as hypothesised by E4 and E5. All three subgroups are much more critical of Muslims and give priority to Christians, as compared to immigrants with no religion. This is in line with the established findings of an ‘anti-Muslim bias’ (Kalkan et al., Reference Kalkan, Layman and Uslaner2009; Bansak et al., Reference Bansak, Hainmueller and Hangartner2016; König, Reference König2024). What is interesting is that the effect of Christianity, as compared to no religion, is also strong for civics, who subscribe to values that have no religious connotation. In line with E5, this pro-Christianity effect is weaker among civics compared to the reference group (ie the undefined, see the right plot of Figure 5). We also stipulate that women with children resemble the Christian family ideal and might, therefore, be given priority (E4), particularly by culturals (E5). The findings show that a woman (with children) indeed has a higher probability of being given priority in admission than a man for all three subgroups. The effects of the gender and family attributes are, however, not significantly different among the subgroups, with one exception: women with children. Here, the effect is even stronger for civics (compared to the undefined and culturals), which is contrary to E5. Thus, when it comes to gender and family characteristics, our theoretical expectations with respect to the difference between civics and culturals do not hold. An explanation for why civics prioritise women with children so strongly could be that they attest a higher vulnerability to women with children and a stronger need for protection and help. Lastly, the literature has established that education and associated labour market skills are relevant for admission decisions (Hainmueller and Hopkins, Reference Hainmueller and Hopkins2015; König, Reference König2024). Our results confirm this finding. All subgroups clearly prefer well-educated professionals, while this effect is strongest for civics.

To evaluate whether the reported differences between civics and culturals can be explained by other factors, we also separate our sample according to political orientation (left-right) and education levels and run the same conjoint analyses (see Appendix E). Three findings are noteworthy. First, the correlation between identity types and left-right political orientation is rather low (at 0.3). Second, while left voters have similar preferences as civics, the differences between left and right voters are much less pronounced compared to the differences between civics and culturals, particularly as far as the value congruence hypotheses are concerned. Third, we see even smaller effects in the comparison between highly and less educated citizens, suggesting that socio-economic determinants are less relevant for our analysis. These additional results, reported in Appendix E, suggest that other factors do not drive our results (but the results also show that identity is not unrelated to political and socio-economic factors). Finally, we run our main analyses for each country separately to test whether our findings are only observable in a subset of countries (see Appendix F). Overall, the results show again that the main effects are not driven by one or two specific countries in the sample.

Taking all of these findings together, we find strong evidence that Europeans prefer immigrants with similar values (E1). Civics put more emphasis on value congruence, as they are particularly benevolent with immigrants who share their civic values (E2), while culturals do not prioritise their cultural set of values (over civic values), which is evidence against E3. Europeans from both identity types evaluate immigrants based on their closeness to the in-group (E4). Ukrainians and UK citizens are most likely to be admitted, while the rejection is highest for Afghans and Muslims. As far as the distinction between culturals and civics is concerned, we find that culturals delineate more strongly in terms of the country of origin and religion (E5). However, we also find that civics are particularly benevolent towards women with children and well-educated immigrants, which was not anticipated by our theory.