Abstract
The authors argue that exposure to contextual diversity can prompt more complex, differentiated, and inclusive multiple in-group perceptions, that is, social identity complexity, with positive consequences for intergroup relations. Two unique, large-scale national surveys, involving respondents sampled from neighborhoods of varying degrees of diversity in Germany (Study 1; N = 1,381 drawn from 50 different neighborhoods) and England (Study 2; N = 580 drawn from 192 different neighborhoods), tested the prediction that people living in ethnically diverse neighborhoods would be higher in social identity complexity and, in turn, hold less negative intergroup attitudes. Results confirmed this hypothesis, showing that greater diversity was directly associated with higher social identity complexity (Studies 1 and 2) and indirectly associated with less in-group bias (Studies 1 and 2), and less social distance (Study 2), via social identity complexity. Findings are discussed with regard to their implications for the consequences of diversity for intergroup relations.
Many of us live, work, or socialize in ethnically, racially, or otherwise diverse environments, with various consequences for individuals, groups and societies at large (see e.g., Crisp & Turner, 2011; Van Knippenberg & Schippers, 2007 for reviews). For example, we now know that diversity in educational contexts can encourage complex thinking (e.g., Antonio et al., 2004), or that diversity in occupational settings can enhance work processes, including creativity (e.g., Webber & Donahue, 2001), and decision making (e.g., Homan et al., 2008). In this article, we extend this reasoning to argue that exposure to diversity may prompt individuals to expand and enhance their sense of self, and foster a more complex view of one’s multiple group memberships, with positive consequences for intergroup relations.
By focusing specifically on neighborhood diversity, we test the prediction that living in diverse neighborhoods is associated with: (1) greater social identity complexity, that is, the extent to which we perceive the multiple social groups we belong to in more complex and inclusive terms and (2) more positive intergroup orientations (less in-group bias and less social distance) by virtue of its positive covariation with social identity complexity.
Social identity complexity (Roccas & Brewer, 2002) rests upon the premise that we all belong to multiple social categories (see, e.g., Crisp & Hewstone, 2007, for a review), but that we may differ in the extent to which we perceive them as similar and overlapping, that is, in the extent to which we perceive them as interrelated. Social identity complexity thus defines individuals’ subjective cognitive representations of the interrelations between their multiple in-group categories, on a continuum from low to high complexity (Roccas & Brewer, 2002; Schmid & Hewstone, 2011). Variations in social identity complexity depend on the extent to which individuals perceive their multiple in-groups as similar in meaning to each other and/or as overlapping. Social identity complexity is thus conceptually divided into two related subcomponents: similarity complexity and overlap complexity (Roccas & Brewer, 2002). While overlap complexity refers to the perceived quantifiable boundaries between categories (i.e., the subjective perception of actual overlap in numbers or proportions between different categories), similarity complexity reflects the perceived similarities or differences in meaning associated with multiple in-groups (i.e., the perceived defining, prototypical or evaluative properties of categories). To illustrate, a New York Democrat might think that most New Yorkers are Democrats (high overlap complexity) and that being a New Yorker means more or less the same as being a Democrat (high similarity complexity), thus perceiving the categories “New Yorker” and “Democrat” as highly interrelated. This individual thus evidences relatively lower social identity complexity than another New York Democrat who perceives less convergence between these categories.
Individual variations in complex multiple in-group perceptions may depend on individual psychological variables such as mood or stress (Roccas & Brewer, 2002), cognitive style (e.g., Miller, Brewer, & Arbuckle, 2009), or situational factors such as perceived threat (e.g., Brewer & Pierce, 2005; Schmid, Hewstone, Tausch, Cairns, & Hughes, 2009). Previous research has also shown that having direct contact with out-group members is associated with greater social identity complexity (Schmid et al., 2009), suggesting that encountering out-group members may prompt differentiated in-group perceptions. This finding speaks in some ways to the as yet underexplored antecedent of social identity complexity considered in this article: diversity. We do not, however, wish to equate diversity of a spatial unit with subjective intergroup contact. Thus, although we know that individuals in diverse settings have a greater likelihood (i.e., opportunity) for intergroup contact (Wagner, Hewstone, & Machleit, 1989), this does not mean that every individual will take up these opportunities, or, importantly, that direct contact is the only way through which diversity can be experienced. In this article, we focus explicitly on neighborhood ethnic diversity, that is, the ethnic composition of individuals’ residential environments, and the extent to which exposure to this is, in and of itself, associated with social identity complexity.
Diversity is a theoretically intriguing antecedent of social identity complexity since it speaks to two basic requirements that need to be met for individuals to reflect complex multiple in-group perceptions: awareness of multiple category existence and recognition of nonconvergence. Individuals thus need to be aware of the fact that their multiple in-groups constitute distinct social categories, and also recognize that these distinct social categories need not, in fact that they often cannot, completely converge. Exposure to and experience of diversity may thus provide the optimal basis for such increased awareness and recognition. In other words, being exposed to more complex social environments involving cross-cutting and diverse social groups, where individuals may share in-group membership on some categories but are out-group members with regard to others, should lead to increased cognitive differentiation processes and, consequently, more complex perceptions of one’s own multiple in-groups. Tentative support exists for this prediction. In a correlational study carried out in the United States, Miller et al. (2009) found that individuals who reported living in residential areas with a greater proportion of residents from different ethnic backgrounds were higher in social identity complexity. It needs to be kept in mind, however, that the focus of this study was on perceived diversity, which can say little about the implications of diversity as a more objective, contextual phenomenon surrounding a given spatial unit.
Diversity may be defined as the relative position of a subpopulation in a given spatial unit (e.g., neighborhood, school/classroom, workplace) in comparison to other subpopulations, where subpopulations are defined in terms of one or more key categorical attributes (e.g., ethnicity/race, religion; Harrison & Klein, 2007; Lieberson, 1969). Spatial units may thus be conceived of as ranging on a continuum from homogeneity to heterogeneity along key attributes (Lieberson, 1969; see also Budescu & Budescu, 2012). Thus, definitions of diversity, as an objectively measurable phenomenon, are intricately linked to its measurement (see Budescu & Budescu, 2012). A common approach to measuring diversity is in terms of a majority versus minority approach, such that diversity is operationalized in terms of the proportion of (minority) out-group members present in a social unit (e.g., Gurin, Dey, Hurtado, & Gurin, 2002; Hunt, Wise, Jipguep, Cozier, & Rosenberg, 2007). While this approach works reasonably well in situations defined by a clear two-group proportional representation (e.g., a U.S. high school that is 60% White and 40% Black), it fails to capture a more nuanced composition of a spatial unit defined by multiple out-groups (e.g., a neighborhood comprised by a multitude of different ethnic groups).
In this article, we operationalize diversity in terms of the ethnic fractionalization of a social unit (see e.g., Montalvo & Reynal-Querol, 2005), a diversity index that captures not only the representational proportion of the country’s ethnic minority groups in a given spatial unit but also considers the degree of heterogeneity among the different minority groups in this unit. It specifically captures the probability of two randomly selected individuals in a given unit belonging to a different ethnic group, and thus ranges from low (0) to high (1), with higher scores reflecting higher probability of encountering someone from any other background than one’s own in a given spatial unit.
Conceiving of diversity as an antecedent of social identity complexity holds additional theoretical implications for the consequences of diversity. One contentious view on the consequences of diversity holds that diversity may negatively affect intergroup relations, for example, that it leads to heightened prejudice between groups (e.g., Fosset & Kiecolt, 1989; Putnam, 2007; Quillian, 1995, 1996; but see Pettigrew, Wagner, & Christ et al., 2010; Schlüter & Wagner, 2008; Wagner, Christ, Pettigrew, Stellmacher, & Wolf, 2006). In this article, we argue that since diversity should lead individuals to think in more complex and inclusive terms about the multiple social groups they belong to, exposure to diversity should indirectly, and positively, influence intergroup attitudes. We thus argue that diversity holds indirect positive consequences for intergroup relations by virtue of its promotion of more differentiated multiple in-group perceptions, since social identity complexity typically covaries positively with intergroup attitudes. Social identity complexity is associated, for example, with more positive out-group attitudes (Miller, Brewer, & Arbuckle, 2009; Schmid et al., 2009), more tolerance (Brewer & Pierce, 2005), and greater support for multiculturalism and affirmative action policies (Brewer & Pierce, 2005). This stems, in part, from the fact that individuals high in complexity are able to recognize that others will share in-group membership on some of their categories, but will be out-group members on others, thereby allowing for greater inclusiveness of others in the individual’s representation of the self. Diversity, by virtue of its potential capacity to prompt greater social identity complexity, may thus bring about more distal, and positive, effects for intergroup relations (see Schmid & Hewstone, 2011).
We derive two hypotheses from the above:
Hypothesis1: Neighborhood diversity will be directly associated with greater social identity complexity.
Hypothesis2: Neighborhood diversity will be indirectly associated with more positive out-group orientations (less in-group bias and less social distance), via positive covariation with social identity complexity.
We conducted two studies to test these hypotheses; in each case, we used purposeful sampling from neighborhoods of varying degrees of diversity to capture diversity as an objective, macro-level measure. Study 1 uses data from Germany to examine the relationship between neighborhood diversity, social identity complexity (surrounding nationality and religion), and in-group bias (toward two minority out-groups: Turks and Russians). Study 2 examines the relationship between neighborhood diversity, social identity complexity (surrounding nationality, ethnicity, and religion), and both in-group bias and social distance (toward ethnic minorities).
Study 1 Method
Data
The data for this study were collected in Germany, using a cross-sectional survey. A professional survey organization collected the data by means of computer-assisted telephone-interviewing techniques, using trained social survey interviewers. We purposefully sampled respondents from neighborhoods varying in their proportional share of foreign residents, thereby setting up our data in a two-level hierarchical data structure with respondents nested in neighborhoods. We randomly sampled 50 neighborhoods (so-called Wohnviertel; minimum N = 2,800 residents, average N = 7,500 residents) from 16 different cities in Germany. From each neighborhood, we randomly selected first, 50 households, and second, one participant per household. The total achieved sample comprised 2,500 adults, of whom, in this article, we only selected respondents who self-categorized as both of German nationality and of Christian religion. The final sample size for this study was N within = 1,381 adults (M age = 56, SD = 16.83; age range: 18–93; 555 males and 826 females) nested within N between = 50 neighborhoods.
Neighborhood Diversity
Diversity was captured through the ethnic fractionalization index, which ranged from .06 (low) to .69 (high) across the 50 neighborhoods.
Individual-Level Measures
Social identity complexity
Our measure of social identity complexity reflected the similarity complexity subcomponent of social identity complexity, with items adapted from Roccas and Brewer (2002). We first made salient respondents’ relevant social identities by asking them to self-categorize in terms of their nationality and religion, and then followed this by 2 items surrounding these categories. The largest usable subsample for this study consisted of German Christians, who responded to the following 2 items: Being German means the same as being Christian, and The typical German person is similar to the typical Christian person (1 = strongly disagree, 5 = strongly agree; reverse coded; r = .47).
In-group bias was assessed using a series of feeling thermometers that asked respondents to indicate their feelings toward the in-group (How warm or cold do you feel about Germans?), and toward two out-groups (How warm or cold do you feel about people of a) Turkish background?’ and b) [. . . ] Russian background?. Feeling thermometer ratings were presented on a scale ranging from 0 = extremely unfavorable to 100 = extremely favorable (see Converse & Presser, 1986). The ratings for the two out-groups were significantly correlated (r = .55) and were subsequently treated as a combined index. In order to create an in-group bias score we subtracted the out-group from the in-group rating.
Results and Discussion
We estimated model parameters for all models reported in this article using Mplus version 6 (Muthén & Muthén, 1998–2010), with data entered as a covariance matrix, and full information maximum likelihood estimation to deal with missing data. Since our data were defined by a two-level structure, we conducted multilevel (hierarchical) modeling, making use of random-intercept modeling (see, e.g., Raudenbush & Bryk, 2002). Specifically, we entered social identity complexity and in-group bias as random effects into that part of the model that estimates the relationship between our variables of interest at the neighborhood, rather than the individual, level. This allowed us to test the direct relationships between diversity (the independent variable which has only between-neighborhood variation), and social identity complexity and in-group bias, respectively, as well as the indirect relationship between diversity and in-group bias via social identity complexity.
In order to control for any potential confounding influences on our hypothesized theoretical relationships, we included the following conventional demographic controls as independent predictors of social identity complexity and in-group bias in our analysis: age, gender, income, and educational background. 1
Figure 1 illustrates the estimated, theoretically relevant relationships between variables. Note that this figure depicts only that part of the model that holds between-neighborhood variation, since it is this set of relationships that is of key theoretical interest to this article.
Random intercept model showing relationship between diversity, social identity complexity, and in-group bias for Study 1. Note. Standardized estimates and between-level part of the estimated model shown, with social identity complexity and in-group bias estimated as random intercepts; *p < .05. **p < .01. ***p < .001.
In support of Hypothesis 1, we obtained a significant direct relationship between diversity and social identity complexity (β = .52, p = .01), such that respondents living in more diverse neighborhoods were higher in social identity complexity. No direct relationship emerged between diversity and in-group bias (β = −.01, p > .05), but we obtained a significant direct relationship between social identity complexity and in-group bias (β = −.59, p = .04), showing that greater social identity complexity was associated with less in-group bias.
We probed for the hypothesized indirect relationship between diversity and in-group bias, via social identity complexity, using Sobel tests (see e.g., MacKinnon, Warsi, & Dwyer, 1995), which confirmed the expected indirect relationship between diversity and in-group bias (β = –.30, z = –1.99, p < .05). Living in more diverse neighborhoods was associated with less in-group bias, via positive variations in social identity complexity. The between-level part of the estimated model explained 27.30% of the variance in social identity complexity and 35.50% of the variance in in-group bias.
Study 1 thus supported our predictions, showing that living in more diverse neighborhoods was directly associated with more complex social identity structures and indirectly associated with less in-group bias, via social identity complexity.
Study 2
Study 2 sought to replicate, but also extend our first study in important ways. First, we sought to test the hypothesized relationships in a different context of intergroup relations, England. Second, since one may argue that Study 1 was limited in its use of only two categories for assessing a construct that seeks to capture complex multiple in-group perceptions, we considered an additional category, ethnicity, alongside nationality and religion, to measure social identity complexity in Study 2. Third, we also sought to establish whether the hypothesized set of relationships holds when considering a conceptually related, but distinct dependent variable. To this end, we added a measure of social distance alongside the simply evaluative attitude measure employed in Study 1.
Method
Data
As in Study 1, our data involved a two-level hierarchical structure, through purposeful sampling of respondents from neighborhoods (so-called middle layer super output areas, which are small geographical units derived by the Office for National Statistics in the United Kingdom; minimum population N = 5,000 residents, mean N = 7,200 residents). We used a random location quota sampling approach, and neighborhoods (as well as the households within each neighborhood) were selected based on a stratified design involving an Ethnic Density × Deprivation Stratification matrix to ensure that both density and degree of diversity were built into the design.
The total achieved sample comprised 1,686 adults, of whom we only selected respondents who were of British nationality, White ethnicity and Christian religion since these were the categories we focused on in our measurement of social identity complexity. The final sample comprised N within = 580 adults (M age = 50.48, SD = 19.22; age range: 16–97; 322 males, 258 females) nested within N between = 192 neighborhoods.
Data collection was subcontracted to a professional survey organization which employed computer-assisted personal interviewing by trained social survey interviewers, involving face-to-face interviews in respondents’ own homes.
Neighborhood Diversity
Diversity was based on the index of ethnic fractionalization, which ranged from .02 (low) to .83 (high) in the neighborhoods included in this study.
Individual-Level Measures
Social identity complexity
Similar to Study 1, we first made salient respondents’ relevant social identities by asking them to self-categorize in terms of their nationality, ethnicity, and religion, followed by social identity complexity items surrounding these categories. The largest usable subsample consisted of individuals who simultaneously self-categorized as British, White, and Christian. Items were preceded by the following instructions: “People belong to many different groups. Sometimes the groups people belong to are very similar to each other, and sometimes they are very different. We would now like to ask you about the groups you belong to and how similar you think the typical British person, the typical White person and the typical Christian person are.” Following this introduction, respondents answered the following three questions: “How similar are the typical British person and the typical White person to each other?” “How similar are the typical British person and the typical Christian person to each other?” and “How similar are the typical White person and the typical Christian person to each other?” (1 = not at all, 5 = very; reverse coded; α = .83).
In-group bias was assessed using two feeling thermometers (0 = extremely unfavorable; 100 = extremely favorable): “How warm or cold do you feel about White British people?” and “How warm or cold do you feel about ethnic minorities?” An in-group bias index was computed by subtracting the out-group from the in-group rating.
Social distance was measured using 3 items: “How much would it bother you to . . . (1) marry someone from an ethnic minority background, or have someone in your family do so? (2) . . . have someone from an ethnic minority background as a doctor? and (3) . . . have people from ethnic minority backgrounds as neighbors on the same street? (1 = not at all, 5 = very; α = .73).
Results and Discussion
We again estimated a multilevel path model to test the hypothesized sets of relationships, estimating the direct paths between all variables as well as the indirect effects between diversity and in-group bias and social distance, respectively, via social identity complexity (see Figure 2). Age, gender, education, and income were again entered as demographic control variables. 2
Random intercept model showing relationship between diversity, social identity complexity, in-group bias, and social distance for Study 2. Note. Standardized estimates and between-level part of the estimated model shown, with social identity complexity, in-group bias, and social distance estimated as random intercepts; additional residual correlation between in-group bias and social distance r = .22, p < .01. *p < .05. **p < .01. ***p < .001.
In support of Hypothesis 1, and replicating Study 1, we found a significant relationship between diversity and social identity complexity (β = .51, p < .001), such that respondents living in more diverse neighborhoods were higher in social identity complexity. No statistically significant direct relationships of diversity with in-group bias (β = –.08, p > .05), nor with social distance (β = –.07, p > .05) emerged. However, social identity complexity was directly associated with less in-group bias (β = –.57, p < .01), and with less social distance (β = –.39, p < .05).
Using Sobel tests, we confirmed our hypothesized indirect relationships between diversity and in-group bias (β = –.29, z = –2.74, p = .01), and between diversity and social distance (β = –.20, z = –2.11, p = .04), via social identity complexity. Diversity was thus indirectly associated with more positive out-group attitudes as well as with less social distance via variations in greater social identity complexity. The between-level part of the estimated model accounted for 26.30% of the variance in social identity complexity, 32.60% of the variance in in-group bias, and 15.40% of the variance in social distance.
General Discussion
The two studies supported our predictions. First, individuals living in more diverse neighborhoods held more complex and inclusive multiple in-group perceptions, that is, higher social identity complexity (Studies 1 and 2). Second, greater neighborhood diversity was associated indirectly, and positively, with less in-group bias (Studies 1 and 2) and social distance (Study 2), accounted for by positive covariations with social identity complexity (Studies 1 and 2).
This research is the first empirical test of its kind to examine the prediction that exposure to diversity as a macro-level phenomenon reflecting the demographic makeup of a spatial unit is associated with more complex and differentiated multiple in-group perceptions (see Brewer & Pierce, 2005). Prior research has so far only confirmed a relationship between subjective rather than objective measures of diversity and social identity complexity (Miller et al., 2009), making it difficult to ascertain the contribution of diversity per se. By employing purposeful sampling, which allowed us to capture diversity as an objective, macro-level measure, and by generalizing our findings across two different contexts, our research constitutes a rigorous test of the link between macro-level diversity and social identity complexity. Moreover, our research confirmed a relationship between diversity and social identity complexity over and above any potential confounding effects of individual-level demographic background variables in both studies. This allows us to conclude with greater confidence that exposure to diverse social environments, in and of itself, may raise individuals’ awareness and recognition of the complex and nonoverlapping nature of social groups, and may thus affect how we think about the multiple social groups we belong to in more inclusive terms (see Roccas & Brewer, 2002; Schmid & Hewstone, 2011).
Confirming prior research, and adding to a small but growing body of research on the involvement of social identity complexity in shaping intergroup relations (Brewer & Pierce, 2005; Miller et al., 2009; Roccas & Brewer, 2002; Schmid et al., 2009), we also found social identity complexity to positively covary with out-group orientations. This highlights that how individuals perceive themselves in terms of their multiple in-groups has important consequences for how they perceive diverse others.
Our work also speaks more widely to the question of the potential consequences of diversity for intergroup relations. In both studies, diversity was indirectly associated with more positive out-group orientations (in-group bias in Studies 1 and 2, and social distance in Study 2), highlighting that it may not suffice to simply consider how individuals think about diverse others when seeking to disentangle the effects of diversity on intergroup relations, but also how they think about their multiple in-groups. This research thus adds to a body of research that has shown that diversity can indirectly affect intergroup attitudes, via key psychological processes. Previous research has found, for example, that living in diverse residential areas is associated with less prejudice via increased intergroup contact and reduced intergroup threat perceptions (e.g., Pettigrew et al., 2010; Schlüter & Scheepers, 2010; Wagner et al., 2006). The current findings highlight that social identity complexity is a key psychological process that also helps explain how exposure to diversity can influence intergroup attitudes. Indeed, social identity complexity may be thought of as a particular form of “deprovincialization,” a psychological term coined by Pettigrew (1997) to explain how individuals, when encountering diverse others, may come to reappraise their in-group in less provincial or ethnocentric terms, with positive consequences for intergroup perception (see also Brewer, 2008).
Notwithstanding the contributions of our work, we wish to acknowledge some methodological and conceptual limitations. First, our data were drawn from cross-sectional surveys, preventing us from making inferences about causality. Although we estimated a series of alternative models, which showed that by and large the coefficients of reverse paths were of lower magnitude than in our original model, we stress that such tests of alternative models remain insufficient grounds for ruling out alternative causal orders to our proposed optimal model. Any model estimation based on cross-sectional data needs to rely heavily on both the theoretical and the substantive plausibility of the tested set of relationships, based on available previous research, which is what we have done here. We highly recommend that future work tests the hypothesized relationships identified in this article using longitudinal investigations that will allow for greater, albeit not unequivocal, approximation of the causal order between variables. Future research might also test whether findings replicate across different settings and types of diversity, for example, considering educational or workplace diversity and, if possible, direct manipulations of diversity.
Second, it should also be kept in mind that our studies were conducted in contexts characterized, by and large, by nonconflictual intergroup relations between different ethnic groups. It remains to be seen whether the relationships tested in this article hold in contexts of extreme and violent intergroup conflict, or in situations where norms surrounding diversity may be very different from other contexts. For example, it may be that in situations of extreme conflict, involving disputes about territory and land, a greater presence of out-group members may be perceived as threatening which may prompt a reduction in social identity complexity if a single in-group category becomes, or is consistently, salient/dominant (see also Roccas & Brewer, 2002). An interesting line of research that should also be further pursued concerns diversity beliefs (i.e., the extent to which individuals value and endorse diversity; see e.g., van Knippenberg, Haslam, & Platow, 2007; van Dick, van Knippenberg, Hägele, Guillaume, & Brodbeck, 2008), since it is plausible that the positive effects of macro-level diversity on social identity complexity depend on subjective endorsement of positive diversity beliefs. Moreover, since our studies relied on samples involving the respective country’s majority population only, future research should seek to explore whether these relationships hold for (ethnic) minority groups also, as well as explore potential differences between groups of different status or power.
Finally, although our research focused on social identity complexity, we were only able to focus on one of the two subcomponents underlying this construct, similarity complexity. Although similarity and overlap complexity have been found to be positively correlated (Schmid et al., 2009), future research should aim to confirm the tested set of relationships using an overlap complexity measure also, or ideally, including both types of complexity in a single investigation. In addition, future research should consider using a more comprehensive list of categories, and indeed measures to assess social identity complexity. Future research should also consider including additional measures of intergroup orientations toward a range of additional out-groups, also toward categories not included in the measures of social identity complexity. Since we collected our current data within the realm of two larger surveys on diversity and intergroup relations among general population samples, we were unable to include a comprehensive set of measures to assess social identity complexity, nor indeed intergroup attitudes. This shortcoming is, however, offset by our purposely designed surveys employing large samples of the general population, which are extremely rare in the social–psychological literature, but are of high value in terms of their external validity, when compared with laboratory studies. These samples were, moreover, drawn specifically from areas selected to represent a wide range of diversity, based on objective demographic statistics.
In conclusion, our work has important implications for how we might think about diversity. Our findings highlight that being exposed to diverse settings may influence how we think about ourselves, and the multiple social groups we belong to, in more complex, differentiated and inclusive terms, with positive consequences for intergroup relations.
Footnotes
Acknowledgments
The authors acknowledge and thank their collaborators: Neli Demireva, Anthony Heath, Serena Hussein, Ceri Peach, Soeren Petermann, Thomas Schmitt, Karen Schoenwaelder, Sarah Spencer, Dietlind Stolle and, especially, Steven Vertovec.
Declaration of Conflict of Interest
The author(s) declared no conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research reported in this article was funded by a grant on ‘Ethno-religious diversity and trust in educational and residential settings’ from the Leverhulme Trust, as well as the Max-Planck-Institute for the study of religious and ethnic diversity, Göttingen, Germany.