United or divided in diversity? The heterogeneous effects of ethnic diversity on European and national identities.

In this article, we argue that the size and cultural proximity of immigrant populations in people's residential surroundings shape national and European identities. This means that the type of migrant population activates cultural threat perceptions and opportunities for contact to varying degrees. Geocoded survey data from the Netherlands suggests that large non-Western immigrant shares are associated with more exclusive national identities, while mixed contexts with Western and non-Western populations show more inclusive identities. These results suggest that highly diverse areas with mixed immigrant populations hold a potential for more tolerance. In contrast, exclusive national identities become strongly pronounced under the presence of sizeable culturally distant immigrant groups.

Introduction

Rising geographical inequalities in public opinion and voting have attracted a lot of attention (e.g. Carreras et al., 2019; Chalmers and Dellmuth, 2015; Huijsmans et al., 2021; Mason, 2016). While the geographical polarization of political behavior has been documented in various settings, we still lack an understanding of the contextual factors that can drive geographic divides in preferences. Cultural accounts emphasize the role of immigration and cultural threat perceptions to explain geographical polarization in political preferences (Carreras et al., 2019; Dinas and Van Spanje, 2011; Hopkins, 2010; Patana, 2020). The underlying socio-psychological explanations based on social identity theory diagnose a cultural geographical cleavage emerging along the lines of local immigration patterns. This paper speaks to this account and questions the uniform effect local ethnic diversity has on geographical political polarization.

We argue that not only the size, but also the types of immigrant populations in people's residential surroundings shape the extent of out-group threat perception. We conceptualize people’s in- and out-group perceptions via the development and strengthening of an exclusive national identity (ENI), which we define by whether people hold a European identity next to their national identity (Brewer and Pierce, 2005; Hooghe and Marks, 2009; Roccas and Brewer, 2002). European citizens’ layered identities are powerful antecedent causes of manifold political preferences, such as opinions on immigration or international cooperation (Clark and Rohrschneider, 2021; Curtis, 2014; Dalton, 2021). We argue that exclusive national identities are strongly activated under the local presence of large and culturally distant immigrant groups. Contrarily, highly diverse areas with culturally more proximate immigrant groups provide a potential for contact that weakens out-group thinking. These heterogeneous effects of local ethnic diversity on the degree of out-group thinking shape the activation of more inclusive or exclusive national identities.

We test this argument with geocoded data from a representative survey conducted with 1800 Dutch citizens in 2020. The data capture respondents’ national and European identities on a local level. A protected remote access option allows identifying the specific area of residence via the Dutch four-digit postcodes. We combine this geocoded survey data with geographical information about immigrant populations in Dutch postcode areas, provided in the Dutch neighbourhood and districts (Wijk- en Buurtkaart) database. The Wijk- en Buurtkaart database offers spatial data on the population structure of 4000 postcode areas, specifically with respect to immigrant populations. With the combination of these two geocoded datasets, we can investigate how local immigration structures and identities relate.

Findings from regression models of ENI show strong compositional effects of the local ethnic context. Highly diverse contexts with many immigrants from Western and non-Western countries are associated with more inclusive identities. In contrast, the presence of only non-Western immigrants is associated with strong exclusive national identities. This means that exclusive national identities are especially pronounced in locations where there is a high share of non-Western migrants and an absence of Western migrants. Highly diverse areas with similarly large shares of Western migrants, in contrast, are associated with more inclusive identities. These findings hold even if we control for area wealth and several other local conditions. Compositional effects of local ethnic diversity, therefore, appear to be pivotal for our understanding of current geographical polarization in political preferences.

This article contributes to the growing literature on determinants for identity-based conflicts and the role of geography in structuring this conflict. Especially, we highlight that structural factors pertaining to local, socio-demographic conditions are important for shaping inclusive identities. The type of local immigrant populations appears crucial to understanding the activation of more inclusive or exclusive political identities. Analysts ignoring the strong compositional effects of local ethnic diversity are likely to underestimate the role of the subnational immigration context for societal polarization. Our findings suggest that we need to account for the different local ethnic configurations to properly understand the geography in identity-based conflicts. Moreover, our results caution against approaches that depict group conflict in bipolar terms. Arguments on ‘immigrants against natives’ might overlook important configurations of local ethnic diversity that can foster inclusive identities. Our findings, thus, provide evidence on societal conditions that appear promising to prevent bipolar conflict.

European identities, immigration and the local context

Social identities and especially political identities have attracted scholarly attention for a long time, but have experienced increasing attention recently (Hobolt et al., 2021; Hooghe and Marks, 2009). Social and political identities are at the heart of political cleavages (Kriesi et al., 2006; Lipset and Rokkan, 1967) and can be based on more ‘fixed’ characteristics, such as age, skin colour or class, or based on fluid factors such as values and norms (Huddy, 2001; Inglehart, 1971). Political identities, however, are distinct from fixed identities: political identities, such as traditionalism and cosmopolitanism are – theoretically – not bound to an unchangeable attribute, meaning that these identities have permeable boundaries, granting the potential for mobility between different groups. Researchers traditionally assume that political identities are rather stable (Green et al., 2004). However, recent research in the US and the European Union (EU) context is questioning the claim of relatively stable identities (Egan, 2020; Kane et al., 2021; Negri et al., 2021).

Political identities have important behavioural effects. Traditionalism and nationalism, for instance, are regarded as the root cause for the recent rise of populism (Norris and Inglehart, 2019). Another political identity is religion, which – when mobilized – is a strong predictor for partisanship in the US context (Layman, 1997). However, religion also demonstrates a crucial problem of conflicts between groups, namely that conflicts are usually not bipolar. There usually is a multipolar divide between different types of religions, but also between atheists and religious people.

It is not new that conflicts are often multipolar and that social identities can be overlapping (Esteban and Ray, 1994; Esteban and Schneider, 2008). Conflict research and social psychology research show that bipolarity is only the most extreme form of conflict: most disputes are multipolar and complex (Esteban and Schneider, 2008). People are members of different groups and sub-groups and in-and out-groups are defined in the respective context (Roccas and Brewer, 2002). For a Catholic, for instance, the out-group can be an Evangelist. Yet, Catholics and Evangelists belong to the category of religious persons, which can be the out-group for persons not believing in any kind of god. Multipolar identity divides have important behavioural consequences. Holding multiple complex identities increases perceived closeness towards out-groups (Brewer and Pierce, 2005). Therefore, the activation of more complex and inclusive identities is of key importance for the development of positive intergroup relations.

One prominent example of the development of an inclusive, nested identity is the case of a European identity, layered on top of a national identity (Hooghe and Marks, 2005, 2009; Risse, 2015; Risse-Kappen et al., 2004). European identity is an identity that allows everyone to be a member who subscribes to the norms and values of the ‘imagined community’ (Anderson, 2006). These European norms represent cosmopolitan values instead of the alternative notion of a ‘fortress Europe’ connotation (Hooghe and Marks, 2018; Schlenker, 2013). Similar to the notion of global citizens, the European identity focuses on similarities rather than differences between humans (Carmona et al., 2020).

In the European context, the superordinate (European) does not mitigate the subordinate (national) identity (Hooghe and Marks, 2009; Risse-Kappen et al., 2004), leading to a structure of nested identities. This means that people hold a national identity and additionally develop or hold a European identity. There is no need to reduce national identity to develop a European identity. Even in the context of the highly polarized Brexit referendum, most of the people do not experience a conflict between these nested identities (Sczepanski, 2021). Therefore, the European identity is an additional layer on a person's national identity (Hooghe and Marks, 2009).

The effects of these nested identities are important. For instance, holding a European identity in addition to a national identity was highly explanatory for voting to remain instead of voting to leave the EU in the case of Brexit (Carl et al., 2019; Hobolt, 2016). Next to the support for the political system, having an additional European identity has also other positive effects for social cohesion: Curtis (2014) has found that including European identity into the self-concept next to a national identity predicts positive attitudes towards immigrants. In contrast, having an exclusively national identity predicts preferences for closed borders (Karstens, 2020). Another key impact of inclusive national identity is the higher preferences for redistribution among Europeans (Nicoli et al., 2020). Thus, a European identity, next to a national identity, has a major impact on various political behaviours and attitudes.

Turning to the roots of the development and strengthening of such political identities, different causes are highlighted, especially socialization experiences and socio-structural conditions. Drawing on the research of partisanship, it has been suggested that people adapt their political identities early on and habitualize them. For example, Dinas (2014) shows that first-time voting for a candidate strengthens partisanship. In a similar vein, other research highlights how participation in EU-related elections strengthens the interest in the EU and identification with Europe, respectively (Schulte-Cloos, 2019; Sczepanski, 2021). Other research has focused on the role of socio-economic conditions for individuals’ identification with Europe. Correlational analysis has highlighted that European identity is a ‘class issue’ as people with higher levels of education and higher incomes also show high levels of an inclusive European identity (Fligstein, 2009; Schimmelfennig, 2010). Also, grand theoretical work on cognitive mobilization focused for example on the important role of education in shaping a European, cosmopolitan identity (Inglehart, 1970).

However, recent studies, using tools of causal inference, show that increases in education do not automatically translate into less Euroscepticism (Kunst et al., 2020). Moreover, results are mixed for the question of whether programs focusing on the development of European identity by a one-time cultural exchange, such as ERASMUS, produce higher levels of European identity (Hakhverdian et al., 2013; Kuhn, 2012b; Stoeckel, 2016). Thus, socio-structural conditions, as well as short-term socialization alone, have been proven to be insufficient to explain variation in the inclusivity of national identities.

In this article, we argue that the local context is crucial for a better understanding of the formation of nested identities. In the local context, socio-structural conditions and socialization come together. The local context enables direct contact with different social conditions, rather than only allowing a cognitive evaluation of reasons for or against a more inclusive identity. Because of this, behavioural studies of vote choice frequently rely on the local presence of the immigrant population to explain radical right party support (Evans and Ivaldi, 2021). However, findings on these local contextual effects are far from conclusive. Studies in the Netherlands have shown no effect of local non-Western immigrant shares on anti-immigrant attitudes and radical right voting (De Blok and Van der Meer, 2018; Janssen et al., 2019). However, studies looking at the ethnic context in the United Kingdom, Germany and Northern Ireland have found that increased identification with the neighbourhood is connected to positive outgroup attitudes in ethnic heterogeneous local contexts (Bai et al., 2020; Schmid et al., 2013; Stevenson et al., 2020). Thus, local contexts, inclusive identities and positive outgroup attitudes are deeply intertwined. In this article, we propose to take a step back and look at the geography in ethnic contextual composition and nested identities to address these mixed findings.

Local immigration exposure and nested identities

The local context is fundamental to the formation of group thinking and identity and thus immigration attitudes (Hopkins, 2010). The size and composition of local immigrant populations in neighbourhoods systematically shape natives’ opinions on immigration (Newman, 2013). Here, the accumulated evidence on the so-called ‘neighbourhood effect’ is largely supportive of the context-dependent cultural threat approach (Hainmueller and Hopkins, 2014). Translating these insights to the EU identity literature requires a closer look at the heterogeneous effects local immigration context can have on group conflict and individuals’ self-identification with groups.

Established theories of anti-immigration attitude and nested identities forward two countervailing accounts (Toshkov and Kortenska, 2015). First, the contact hypothesis suggests that local exposure to immigrants should increase natives’ tolerance due to a correction of stereotypes and prejudice (Allport, 1954; Pettigrew, 1998). In general, research finds overarching support for the effects of contact to reduce prejudice (Knappert et al., 2020; Pettigrew and Tropp, 2006; Wagner et al., 2006). However, research has left open pressing theoretical and empirical puzzles, for instance on the impacts of intergroup contact of adults on anti-immigration attitudes (Paluck et al., 2019).

Second, cultural threat theory suggests that local exposure to immigration generates diffuse fears among natives for the integrity of local languages, identities and norms (Hainmueller and Hopkins, 2014). This account suggests that the origin of the immigrant population is crucial, as a cultural threat should increase with a perceived widening cultural distance between native and immigrant populations (Brunner and Kuhn, 2018).

A central underlying mechanism of the cultural threat model is identity. Identities can be expressions of an ‘us versus them’ thinking that might be shaped by the local presence of immigrants. Variation in the local composition of immigrant populations can make cultural group conflicts more or less salient, which should have consequences for public opinion on EU integration. Indeed, the relative salience of types of immigration has been shown to shape EU support on the national level (Lutz, 2020).

We argue that local conditions, specifically the variation in local immigration exposure, shape the salience of the contact and cultural threat mechanisms. A perception of cultural threat from immigrants is often restricted to certain places that witness a politicization of the immigration debate (Hopkins, 2010). This politicization can result from local immigration rates, economic context and media framing (Hopkins, 2011). In this article, we argue that the local composition of the immigrant population is another factor that shapes the salience of cultural threat perceptions. We suggest that the local immigration context can either be conducive to less conflict and more tolerance due to contact with culturally less distant immigrants, or nurture group conflict and cultural threat due to the presence of sizeable minorities that are culturally more distant. We conceptualize cultural distance with a rather crude distinction between Western and non-Western immigrants. Areas can experience various degrees of culturally more distant or close immigration, and the local composition of the immigrant population could shape the opportunities for contact, as well as the projection of cultural threat.

However, the cultural threat hypothesis already assumes that not every immigrant population is perceived to impose the same level of threat. Researchers found large differences in how close natives feel towards various immigrant groups (Konitzer et al., 2019; Schalk-Soekar and Van de Vijver, 2008). For example, while Dutch natives place Turkish and Moroccan immigrants as more distant to themselves, Surinamers and Antilleans were placed closer to the Dutch natives (Schalk-Soekar and Van de Vijver, 2008). Although these differences in the perception of immigrants are important, most studies testing the impact of contextual effects of immigration have only focused on ethnic fragmentation (Bai et al., 2020; Schmid et al., 2013), or absolute numbers of immigrants in the local context (De Blok and Van der Meer, 2018; Maxwell, 2019), but not on the composition of the local immigrant population.

We argue that socially diverse contexts are associated with an inclusive national identity. Psychological research has shown that ethnic diversity produces less severe stereotypes of immigrants (Bai et al., 2020). Therefore, a diverse composition implies that people hardly perceive a bipolar competition between groups. As diversity in societies, and in this case localities, increases, people perceive overlap between different group identities. People observe similarities between the cultures of Western immigrants and their way of life, but also overlaps between the Western and non-Western immigrant populations. The various degrees of multiple overlaps minimize the tendency for ‘othering’ – perceiving the outgroup as well as the in-group as more homogenous (Schmid et al., 2013). Therefore, diversity not only produces opportunities for contact but also triggers people to perceive group boundaries as overlapping and inclusive. We therefore agree with the important note that ‘people can adjust to diversity’ (Bai et al., 2020: 12741), by translating surrounding complexity into social identity complexity (Roccas and Brewer, 2002; Schmid et al., 2013).

Figure 1 shows the mechanism of the cultural threat hypothesis in different contexts. The typical setting of the cultural threat hypothesis is that people do not perceive any overlap between two groups. If cultures are perceived to be distant and thus, to be distinct, people are more likely to perceive group conflict in terms of ‘us versus them’. Therefore, identities associated with liberal values such as European identities are reduced. This perception of conflict is especially likely with cultures that are perceived as ‘non-Western’, as they imply larger perceived differences in shared norms and values.

Figure 1.

Cultural threat hypothesis under different conditions.

However, the setting changes in more diverse contexts, including different types of Western and non-Western immigrant populations. Western immigrants can be perceived as a common denominator between two distant cultures. Western immigrants share some common ground with cultures in the European context – shared history, media taste or more common cultural goods. However, Western immigrants also share common experiences with non-Western immigrants: coming into a country not being born in, adapting to new institutional rules and building new social contacts. Contact with Western immigrants can therefore lead to increased knowledge about immigration, inducing perspective-taking and empathy for out-groups (Pettigrew and Tropp, 2008). We argue that this perspective-taking is not limited to one immigrant group. As negative experiences of immigration are partly shared between Western and non-Western immigrants, perspective-taking and empathy can spill over from Western immigrants to non-Western immigrants.

Moreover, the complexity of highly diverse contexts can foster inclusive national identity building. As psychological research has shown, the complexity of a neighbourhood's social setting increases the number and strength of identities an individual holds (Brewer and Pierce, 2005; Miller et al., 2009; Roccas and Brewer, 2002; Schmid et al., 2013). As the localities become more complex in terms of groups, individuals are triggered to perceive more overlap between different social groups. Thus, the mere presence of Western immigrants might create a bridge, perceiving immigration more in continuous terms, rather than in categorical terms. The overlapping structure, we argue, hinders people to simply perceive an ‘us versus them’ pattern, but rather creates an overreaching inclusive national identity, which integrates a European next to the national identity.

To sum up the theoretical discussion, Table 1 presents how the local presence of Western and non-Western immigrant populations can interact. We expect that exclusive national identities are most strongly activated under a pronounced presence of a non-Western immigrant population (cell (2) in Table 1). The cultural distance increases threat perceptions and impedes social contact, for example, through language barriers, religious differences, etc. Therefore, exclusive national identities should be more pronounced under local contexts with sizeable and culturally distant immigrant groups. In contrast, a strong presence of Western immigrants produces, relatively speaking, less cultural threat and eases contact due to cultural proximity (cell (3) in Table 1). In fact, compared to areas with generally low immigrant shares (cell (1) in Table 1), tolerance towards immigrants could be more pronounced under higher shares of Western migration. The complete absence of immigrants and the resulting socio-demographic homogeneity tends to boost out-group discrimination (Fitzgerald, 2018). In contrast, we expect that high shares of Western immigration decrease out-group thinking, as more complex perceptions of immigration soften the boundaries of ethnic group identities. With respect to the diverse contexts with high Western and non-Western immigrant populations (cell (4) in Table 1), we expect more tolerance and less pronounced exclusive identities.

Table 1.

Interaction of culturally more close and distant local immigration exposure.

	Low non-Western	High non-Western
Low Western	(1) No cultural threat and no contact	(2) High cultural threat and impeded contact
High Western	(3) Low cultural threat and easier contact	(4) Low threat and easier contact

One potential alternative explanation to our argument might be residential sorting: people with more cosmopolitan attitudes sort themselves into specific residential contexts (Maxwell, 2019, 2020). We cannot account for sorting with our cross-section data. However, we reduce the effect of sorting by accounting for the theoretical explanations that are most predictive of cosmopolitan and anti-immigrant sorting. Namely, the socio-demographics and incomes of the respondent, as well as contextual factors of the locality, for example, the prosperity of a neighbourhood (Maxwell, 2019). Moreover, we think that sorting and the contextual effects we propose are both parts of a broader geography in political identity that is driven by local ethnic diversity.

Data and method

We test our argument with geocoded data from a representative survey conducted with 1800 Dutch citizens in 2020. The Netherlands is a typical case of a Northern European country with substantial numbers of Western and non-Western immigrant populations. Politics in the Netherlands has increasingly shifted to a cultural conflict over immigration and European integration (McDonnell and Werner, 2018), highlighting the representative character of the case. Official statistics report that 25% of the population in the Netherlands has a migrant background. EU citizens constitute the largest foreigner group, making up 28% of the migrant population. The largest migrant populations from non-EU countries come from Turkey, Morocco, Surinam, and Indonesia. The Online Appendix includes descriptive statistics on the Dutch immigrant population structure.

We survey identities using standard items from the literature. Respondents’ national and EU identities are measured on a scale from ‘0-feeling not attached at all’ to ‘10-feeling strongly attached’ to the Netherlands or the EU. Our measure of ENI is derived from subtracting the EU identity variable from the national identity. Positive values on the variable indicate that respondents identify more strongly with the Netherlands than the EU. This operationalization accounts for the multi-level structure of nested identities in the European context (Nicoli et al., 2020). In contrast to the categorical measures of exclusive national identities used in earlier work (e.g. Hooghe and Marks, 2005), our operationalization allows capturing the degree of exclusiveness. The majority of respondents tend to identify more intensely with the Netherlands, as the exclusive identity measure has a median of +2, with a minimum of −6 and a maximum of 10. The Online Appendix maps binned means in our ENI measure across the Netherlands.

The Dutch survey allows exceptional precision in capturing the local context due to a protected remote access option for identifying respondents’ specific area of residence via the Dutch four-digit postcodes. The four-digit postcodes provide a very fine-grained picture of local conditions. The Netherlands has around 4000 postcode areas on the four-digit level. The median population of a postcode area in 2019 was just 2835, with a minimum of 750 residence in the first quartile and 7100 in the third quartile. Therefore, the postcode level allows capturing individual villages or neighbourhoods. This granularity captures people's individual surroundings and living context much more precisely than regions or municipalities. This spatial granularity has been shown to be crucial for the identification of contextual effects of ethnic diversity (Dinesen et al., 2020; Dinesen and Sønderskov, 2015).

We combine our geocoded survey data with detailed geographical information about Dutch districts and neighbourhoods, provided in the Wijk- en Buurtkaart database. The Wijk- en Buurtkaart database offers spatial data on the population structure at the four-digit postcode level. Based on these data, we derive postcode-level measures of the share of Western and non-Western immigrants. The definitions of Western and non-Western immigrants are developed by Statistics Netherlands and are mostly based on the continent of origin, with some exceptions, such as Turkey or the former Dutch oversee territories (e.g. Indonesia). Western immigrants are defined as migrants from Europe (excluding Turkey), North America, Oceania, Indonesia, or Japan. Non-Western immigrants are defined as migrants from Africa, Latin America, Asia (excluding Indonesia and Japan), and Turkey. The local immigration shares are strongly left-skewed. We therefore take the natural logarithm to arrive at a more normal distribution. To account for local changes in immigration, we use the Wijk- en Buurtkaart database 2010 to calculate the change from 2010 to 2019 for both, Western and non-Western immigrant shares.

Omitted variable bias is an obvious issue for the cross-sectional relationship between local immigration shares and identities. Our opportunities to account for this are limited, as data availability on the fine-grained postcode level is rather limited. A major concern is that local immigration rates strongly co-vary with local wealth differences. Poorer and richer areas attract different kinds of immigrants, which, in turn, is associated with differences in the local prevalence of exclusive identities.

Here we were able to merge data on area wealth, captured as the average household incomes across the Dutch four-digit postcode areas. This data is generated from the Dutch regional income survey. It allows us to measure the average wealth in each postcode area to block spurious relationships that might emerge from local wealth differences. We also control for an area’s distance to the national border, as it has been suggested that border distance is related to cosmopolitan attitudes (Kuhn, 2012a). Finally, we account for spatial effects by accounting for the average immigration rates in neighbouring areas, as this might shape preferences as well (Evans and Ivaldi, 2021). In all models, we control for individual-level socio-demographic factors, such as gender, age, education, migratory background and income. The Online Appendix presents descriptive statistics for all variables included in our analyses.

Figure 2 maps the bivariate distribution of Western and non-Western immigrant populations across postcode areas. It shows that immigration shares strongly vary on this highly disaggregated level. More aggregated measures of local immigration context run the risk of missing this variation. Furthermore, the bivariate map highlights various configurations of local immigration context. Across the Netherlands, we find residential surroundings for all combinations of the two local immigration variables. This underlines the importance of considering the types of local ethnic diversity to improve our understanding of contextual effects on preferences.

Figure 2.

Bivariate map of postcode level Western and non-Western migrant shares.

One should note that our survey data is sparse in its geographical coverage. The representative sample provides survey responses from a variety of postcode areas scattered across the Netherlands. We observe 1 to 6 respondents clustered within 1248 postcode areas. At the mean, we have sampled 1.5 respondents per postcode area. Many of the 1800 respondents, therefore, are uniquely defined by a postcode area. This has consequences for the statistical model specification. On the one hand, the data is clustered hierarchically as respondents are nested within postcode areas, suggesting the use of multi-level models. On the other hand, the average of 1.5 hinders an adequate estimation of a random effect variance. We therefore use simple ordinary least squares (OLS) regression to estimate individuals’ level of nested identity.

Results

Table 2 presents bivariate correlations for the postcode-level variables used in our analysis. It shows that an area's wealth is negatively correlated with the presence of immigrants. Yet, non-Western immigrant shares are more strongly correlated with area wealth than Western immigrant shares. The direction of the correlation shows that non-Western immigrants more frequently reside in poorer areas. This suggests that area wealth is an important covariate, as it is substantially related to the geography of immigration. Border distance is moderately low correlated with the presence of Western migrants. There might be a tendency of Western migrants to live close to the border of their home country, such as Germany, Belgium, or France. Residing in border areas, in turn, has been argued to affect cosmopolitan beliefs (Kuhn, 2012a). Therefore, we control for border distance of a postcode area to account for this potential backdoor path. Moreover, Table 2 shows that Western and non-Western immigration shares have a substantial positive correlation. However, the correlation is only moderately high and Figure 2 demonstrates that areas can have distinct levels of Western and non-Western immigrant shares. Finally, Table 2 shows that the measures of local changes (Δ) in immigrant shares are mostly correlated with the cross-sectional measures of today's local migrant shares.

Table 2.

Correlations between postcode area variables.

	Western	Non-Western	Area wealth	ΔWestern	ΔNon-Western	Border distance
Western	1	0.604	−0.247	0.468	0.182	−0.263
Non-Western	0.604	1	−0.454	0.323	0.329	0.084
Area wealth	−0.247	−0.454	1	−0.205	−0.048	0.076
ΔWestern	0.468	0.323	−0.205	1	−0.018	−0.032
ΔNon-Western	0.182	0.329	−0.048	−0.018	1	−0.022
Border distance	−0.263	0.084	0.076	−0.032	−0.022	1

Table 3 presents OLS regressions of the ENI variable. All models include a set of individual-level controls. Age has a strong non-linear impact on the probability of having a high ENI. The coefficient signs indicate an inverted U-shaped relationship, with the highest probability for exclusive identities among middle-aged respondents. Moreover, tertiary education is negatively associated with holding a strong ENI. Finally, the individual-level covariates suggest a significantly lower probability of holding ENIs among respondents with a migrant background. The estimated coefficients on these individual-level controls remain highly stable across all models. The coefficients on the effects of the main postcode area variables are statistically significant. ENIs appear more pronounced in wealthier areas and areas with sizeable non-Western migrant shares. In contrast, Western migrant shares are negatively associated with the probability of holding an ENI.

Table 3.

Ordinary least squares (OLS) regressions of exclusive national identity.

	(1)	(2)	(3)	(4)	(5)
Individual-level variables
Female	0.007	0.015	0.009	0.009	0.009
	(0.127)	(0.127)	(0.127)	(0.128)	(0.128)
Age	1.453***	1.440***	1.452***	1.453***	1.454***
	(0.389)	(0.389)	(0.389)	(0.390)	(0.390)
Age2	−1.352***	−1.345***	−1.355***	−1.350***	−1.351***
	(0.382)	(0.382)	(0.382)	(0.383)	(0.383)
Tertiary education	−0.802***	−0.790***	−0.796***	−0.812***	−0.812***
	(0.131)	(0.131)	(0.131)	(0.132)	(0.132)
Income	−0.023	−0.028	−0.029	−0.029	−0.029
	(0.062)	(0.062)	(0.062)	(0.062)	(0.062)
Migrant background	−0.557***	−0.533***	−0.537***	−0.536***	−0.536***
	(0.174)	(0.174)	(0.174)	(0.174)	(0.174)
Postcode-level variables
Log non-Western	0.250**	1.136***	1.063***	1.356***	1.357***
	(0.108)	(0.401)	(0.405)	(0.434)	(0.434)
Log Western	−0.730***	−0.027	0.060	0.099	0.098
	(0.173)	(0.352)	(0.358)	(0.360)	(0.360)
Area wealth	0.153**	0.131*	0.118	0.134*	0.134*
	(0.072)	(0.072)	(0.073)	(0.074)	(0.074)
Border distance			0.006	0.004	0.004
			(0.004)	(0.004)	(0.004)
ΔNon-Western				−0.047	−0.049
				(0.031)	(0.042)
ΔWestern				0.059	0.057
				(0.043)	(0.048)
Log non-Western×log Western		−0.380**	−0.371**	−0.465***	−0.465***
		(0.166)	(0.166)	(0.177)	(0.177)
ΔNon-Western×ΔWestern					0.001
					(0.010)
Constant	4.454***	2.888***	2.686***	2.519***	2.520***
	(0.345)	(0.765)	(0.779)	(0.788)	(0.788)
Observations	1612	1612	1612	1605	1605
R 2	0.058	0.061	0.062	0.065	0.065
Adjusted R2	0.053	0.055	0.056	0.057	0.056

Note: *p < 0.1; **p < 0.05; ***p < 0.01; Δ = temporal change 2010–2019; standard errors in parenthesis.

To account for the interaction of Western and non-Western immigration theorized in Table 1, Model 2 of Table 3 adds an interaction for the local immigration measures. The interaction term is negative and statistically significant. Figure 3 plots predicted values for a substantive interpretation of the interaction and shows that increasing shares of non-Western immigration in a postcode area strongly boost ENIs, but only if Western immigration shares remain low. This effect is sizable. Moving from zero immigration to the maximal share of exclusive non-Western immigration increases the predicted level of ENI by 5 points.

Figure 3.

Interaction effect of Western and non-Western immigration on exclusive national identity.

This finding aligns with the cultural threat theory. However, it adds to previous research by demonstrating that subnational cultural threat perceptions rest on the presence of a clearly identifiable, culturally more distant out-group. High immigration shares per se do not activate exclusive identities. In fact, predicted levels of exclusive identity slightly decrease for areas with large Western and non-Western immigration shares. Moving from a non-immigration context to a highly diverse setting of large Western and non-Western immigrant shares decreases the exclusive identity score by about 2 points. This suggests that the absence of a clear out-group and more diverse opportunities for contact in very diverse areas dampen out-group thinking. In contrast, very homogenous areas with few immigrants show intermediate levels of exclusive identity. This intermediate level of out-group discrimination can be explained by less contact with immigrants in the immediate surrounding compared to more diverse areas, but also less cultural threat compared to areas with many non-Western immigrants.

Overall, Model 2 of Table 3 suggests that the share and cultural distance of immigrant populations in residential areas shape the extent of out-group thinking. ENIs are stronger in areas with clearly identifiable out-groups of non-Western migrants. In contrast, highly diverse settings with Western and non-Western immigrants weaken exclusive identities. A strong presence of Western migrants in the Dutch postcode areas is strongly associated with more tolerance and less exclusive identities, even if we account for area wealth and education levels. This shows that local immigration exposure can have heterogeneous effects on identity.

Models 3 to 5 in Table 3 provide a number of robustness checks to our main finding. Model 3 adds the border distance measure. Border distance is not systematically associated with ENI and the findings for our local immigration measures remain the same. Models 4 and 5 subsequently add the local changes in immigrant shares and their interaction. The change terms and their interaction are not systematically associated with the strength of ENIs. We provide additional robustness checks in the Online Appendix, where we show that the findings are robust to the inclusion of a measure of satisfaction with the economy, a measure of the postcode area population size, and fixed effects for the 12 Dutch provinces.

Table 4 presents results for national and European identities separately. It shows that the effect of local ethnic diversity is restricted to EU identity. National identities are not systematically associated with any of our postcode-level variables. This aligns with existing accounts that depict national identities as rather stable and resilient, also because they are more available and present for individuals (Anderson and McGregor, 2016: 92). Indeed, national identity does not vary that much across people. It is rather the inclusiveness or exclusiveness of national identity that becomes polarized across society.

Table 4.

Ordinary least squares (OLS) regressions of national and EU identity.

	(1)	(2)
	National identity	EU identity
Individual-level variables
Female	0.212**	0.204
	(0.100)	(0.125)
Age	0.718**	−0.735*
	(0.305)	(0.383)
Age2	−0.507*	0.843**
	(0.300)	(0.376)
Tertiary education	0.125	0.937***
	(0.103)	(0.129)
Income	0.061	0.091
	(0.048)	(0.061)
Migrant background	−0.337**	0.199
	(0.137)	(0.171)
Postcode-level variables
Log non-Western	−0.097	−1.453***
	(0.340)	(0.426)
Log Western	−0.349	−0.447
	(0.282)	(0.353)
Area wealth	0.078	−0.056
	(0.058)	(0.072)
Border distance	−0.002	−0.007
	(0.003)	(0.004)
ΔNon-Western	−0.014	0.033
	(0.024)	(0.030)
ΔWestern	−0.002	−0.060
	(0.034)	(0.042)
Log non-Western×log Western	0.078	0.543***
	(0.139)	(0.174)
Constant	8.232***	5.713***
	(0.617)	(0.774)
Observations	1605	1605
R 2	0.024	0.056
Adjusted R2	0.016	0.049

Note: *p < 0.1; **p < 0.05; ***p < 0.01; Δ = temporal change 2010–2019; standard errors in parenthesis.

Figure 4 presents the interaction effect of the local immigrant variables on EU identity. While EU identity is moderately high in ethnically homogenous areas, it becomes strongly polarized across the different types of ethnic diversity. Self-identification with the EU is more pronounced in highly diverse areas characterized by large Western and non-Western migrant populations. Contrarily, EU identity strongly decreases in areas with a sizeable and exclusively non-Western migrant population. Therefore, the findings of Table 4 resemble our conclusions from the regressions of ENI in Table 3. Table 4, however, shows that local ethnic diversity shapes ENIs through the willingness of people to incorporate a European identity on top of their national identity. We do not see that local ethnic diversity shapes the basic self-identification with the nation-state.

Figure 4.

Interaction effect of Western and non-Western immigration on EU identity.

Overall, our empirical findings suggest that local ethnic diversity has heterogeneous effects on political identities. On the one hand, national identities become more inclusive to broader territorial identities in contexts of very diverse ethnic configurations, in which large Western and non-Western immigrant populations mix. On the other hand, the national identity becomes strongly exclusive under the single presence of a sizable non-Western population. Ethnic diversity, therefore, can be a cause of both – more inclusive and exclusive national identities. The type of local ethnic diversity is crucial to disentangle this heterogeneity.

Moreover, our empirical findings demonstrate that ethnic diversity does not shape individuals’ basic identification with their nation-state, but makes them more willing to further integrate a more cosmopolitan European identity into their self-concept. Local ethnic diversity strongly structures how willing people are to accept a supranational identity in addition to their national one. This adds another component to existing models, which predict that higher ethnic heterogeneity might lead to larger shares of people with ENI, especially among the poor (Shayo, 2009).

Our findings are highly consequential for our understanding of current societal conflicts in Western democracies. In line with previous research (Brewer and Pierce, 2005; Dalton, 2021; Hooghe and Marks, 2009; Roccas and Brewer, 2002; Ziblatt et al., 2021), we believe that nested identities are a powerful antecedent cause of a range of political preferences that have become polarized over the recent decades. Table 5, for instance, presents the effect of ENIs on respondents’ support for EU membership and globalization. All continuous variables in Table 5 are standardized, which allows a comparison of effect sizes. In both models, ENI ranges among the strongest coefficients and is highly significant. ENI is strongly associated with lower EU membership support and opposition towards globalization. This underlines the relevance of our main findings and suggests that broader divisions across the communitarian-cosmopolitan divide might be – at least partially – driven by the geography in ethnic diversity.

Table 5.

Logit regressions of cosmopolitan political preferences.

	EU membership	Globalization
	(1)	(2)
Female	−0.195	−0.141
	(0.119)	(0.113)
Age	−0.536	−1.009***
	(0.364)	(0.338)
Age2	0.478	0.845**
	(0.358)	(0.334)
Tertiary education	1.104***	0.537***
	(0.124)	(0.115)
Income	0.062	0.012
	(0.053)	(0.061)
Migrant background	−0.384**	−0.082
	(0.158)	(0.150)
Satisfaction economy	0.619***	0.455***
	(0.066)	(0.066)
Exclusive national identity	−0.887***	−0.463***
	(0.068)	(0.062)
Constant	−0.006	−0.835***
	(0.105)	(0.104)
Observations	1605	1605

Note: *p < 0.1; **p < 0.05; ***p < 0.01; standard errors in parenthesis.

Conclusion

Can we diagnose a cultural geographical cleavage along the lines of local immigration patterns? This paper argues that local immigration exposure can have heterogeneous effects on nested identities, as not only the size, but also the cultural proximity of immigrant populations shapes cultural threat perceptions and opportunities for contact. Culturally more similar immigrant populations provide a potential for contact that weakens out-group thinking. Contrarily, the presence of large and clearly identifiable out-groups in people's residential context drives cultural threat perceptions, which boost the development of an ENI. These diverse effects of immigration on out-group thinking shape variation in inclusive national identities.

Regression models of ENI among 1800 respondents show that the local presence of Western migrants weakens exclusive national identities. In contrast, the local presence of non-Western migrants is associated with a stronger ENI. Exclusive national identities are especially dominant in locations with segregated, distant cultures. This suggests that clearly identifiably out-groups in people's immediate surroundings activate out-group thinking. A separate analysis of national and European identity shows that the local ethnic diversity primarily shapes European identity, but not national identities. The immigration context, therefore, does not activate the strength of national identity, but structures the inclusiveness of people's nested identities.

This study contributes theoretically to an ongoing debate about the origins of inclusive identities. Our findings suggest that Western immigrants can build a perceptional bridge between native and non-Western populations, as they share characteristics with both groups. We therefore emphasize the important role of multi-polar contexts for the activation of an inclusive national identity – a European identity layering on top of a national identity.

However, our findings also highlight how a pronounced geography in ethnic diversity can boost exclusive group perceptions. The positive effect of diversity on inclusive identities ceases if culturally distant immigrants strongly cluster spatially. Thus, people should carefully observe the regional clustering of immigrant populations to understand the polarization of social identities.

We want to be explicit about the strength and weaknesses of our study. Our article provides a novel argument about the heterogeneous effect of local ethnic diversity on identity conflict and tests this with fine-grained geo-coded data. The highly disaggregated data is crucial to capture the societal effects of ethnic diversity (Dinesen and Sønderskov, 2015), but it comes at the cost of limited abilities to account for confounding factors. Additionally, employing cross-sectional data, we cannot completely account for the self-selection of persons with cosmopolitan attitudes into more diverse areas. Moreover, we cannot investigate the implied theoretical mechanisms. For instance, we are not able to show how prejudice, residential sorting, or identity change play out empirically. We therefore encourage future research to address the important effects of contextual surroundings on identities in longitudinal and experimental settings as well.

Click here for additional data file.

Supplemental material, sj-docx-1-eup-10.1177_14651165211063770 for United or divided in diversity? The heterogeneous effects of ethnic diversity on European and national identities by Dominik Schraff and Ronja Sczepanski in European Union Politics

Click here for additional data file.

Supplemental material, sj-txt-2-eup-10.1177_14651165211063770 for United or divided in diversity? The heterogeneous effects of ethnic diversity on European and national identities by Dominik Schraff and Ronja Sczepanski in European Union Politics

Click here for additional data file.

Supplemental material, sj-R-3-eup-10.1177_14651165211063770 for United or divided in diversity? The heterogeneous effects of ethnic diversity on European and national identities by Dominik Schraff and Ronja Sczepanski in European Union Politics

Click here for additional data file.

Supplemental material, sj-csv-4-eup-10.1177_14651165211063770 for United or divided in diversity? The heterogeneous effects of ethnic diversity on European and national identities by Dominik Schraff and Ronja Sczepanski in European Union Politics

Click here for additional data file.

Supplemental material, sj-shp-5-eup-10.1177_14651165211063770 for United or divided in diversity? The heterogeneous effects of ethnic diversity on European and national identities by Dominik Schraff and Ronja Sczepanski in European Union Politics

Click here for additional data file.

Supplemental material, sj-shp-6-eup-10.1177_14651165211063770 for United or divided in diversity? The heterogeneous effects of ethnic diversity on European and national identities by Dominik Schraff and Ronja Sczepanski in European Union Politics