Effects of neighbourhood religious diversity and religious and national identity on neighbourhood trust

This paper examines the relationship between religious diversity, religious and national identity, and neighbourhood trust. Using data from 6,089 individuals in England matched to census-based statistical estimates for 300 local areas, we find that religious diversity is negatively associated with neighbourhood trust. Further analyses tested indirect relationships between religious diversity and neighbourhood trust to examine whether higher levels of religious diversity are associated with a stronger sense of religious identity, and whether a stronger sense of religious identity is associated with lower levels of neighbourhood trust. We simultaneously tested whether higher levels of religious diversity are associated with a weaker sense of national identity, and whether a weaker sense of national identity is associated with lower neighbourhood trust. Multigroup analyses were conducted to assess patterns across religious groups. Results indicate that, for Christians, religious diversity is associated with a stronger subordinate religious identification, which is in turn associated with lower neighbourhood trust. There were no associations between religious diversity, national identification, and neighbourhood trust. For the other religious groups, no significant associations were found between our variables of interest. We discuss the relevance and implication of increasing religious diversity in societies for the multiple groups involved.


Introduction
Religion is an important marker of group identity, with religious groups providing feelings of certainty, belonging, and inclusion to their members (Verkuyten & Yildiz, 2007). However, in a post-9/11 multicultural world, the salience of religion and religious social contexts has increased tremendously. Moreover, increased migration and settlement of individuals from non-Christian religious affiliations into traditionally Christian countries significantly changed the historically homogenous religious composition of local areas, and populist arguments claim that this change has increased intergroup tensions along religious lines (Caldwell, 2009). The political shift to populism has seen an increased narrative around the potential effects of these changes, as highlighted by events such as Brexit and the refugee crisis in Europe. Since religious identity can be considered a crucial marker and a divider between people, understanding the drivers of religious identification and the implications for social cohesion is crucial.
A large and expanding body of work across the social sciences has demonstrated that trust and social cohesion are lower in diverse communities (Abascal & Baldassarri, 2015;Beugelsdijk & Klasing, 2016;Portes, 2014;Putnam, 2000Putnam, , 2007. In meta-analytic reviews, both van der Meer and Tolsma (2014) and Dinesen et al. (2020) found that the specific form of social cohesion that was most consistently negatively affected by diversity was intraneighbourhood social cohesion in the form of trusting others living in the same neighbourhood, hence we focus on this form of trust in the present paper. Although it is expected that, with time, increasing social diversity will create opportunities for intergroup contact that mitigate initial negative effects on trust and social cohesion (Ramos et al., 2019; see also Li et al., 2021;Schmid et al., 2014), it has been shown that, in the short term, increases in religious diversity are associated with lower generalized trust in others (Ramos et al., 2019). The lack of trust and social cohesion in diverse communities and societies has been linked to other negative outcomes such as conflict (Esteban et al., 2012), poor economic growth (Easterly & Levine, 1997), and poor public goods provision (Baldwin & Huber, 2010).
Some commentators assert that diversity undermines the trust and solidarity necessary for cohesive societies (cf. Goodhart, 2013;Scheffer & Waters, 2011) because it reinforces separate ethnic (subordinate) identities rather than promoting a shared national (superordinate) identity.
In line with such arguments, social identity theory (Tajfel & Turner, 1979) suggests that diversity can lead individuals to identify more strongly with other ingroup members rather than with members of society more broadly, which could thereby restrict the development of a shared superordinate identity that promotes social cohesion.
We contribute to the social-psychological literature by investigating identity as a potential mediator of the relationship between diversity and trust. We also focus on the role of religious diversity and identity in a field that has been dominated by studies of ethnic diversity (for a review, see van der Meer & Tolsma, 2014). Previous research has demonstrated that the religious context plays an important role in identity salience. In the Netherlands,  found that mosque attendance was more frequent among Muslims living in areas with high versus low proportions of coethnics. Similarly, Maliepaard and Phalet (2012) found that Muslims with more contact with non-Muslim minority group members had increased religious practice and assertion, while Muslims with majority group contacts had decreased levels. Muslims who were more socially integrated in their communities with fellow Muslims also had higher levels of religious practice and assertion. In Belgium, Smits et al. (2010) found that Muslim religious participation was higher among immigrants who were socialized in a religious region in their country of origin, currently lived in areas with more mosques, and had coethnic social networks.
In this paper, we consider the extent to which individuals' exposure to different levels of religious diversity is associated with religious and national identification as well as trust. Focusing on residential areas, we test whether higher levels of religious diversity are associated with a stronger sense of religious identity, and whether a stronger sense of religious identity is associated with lower levels of neighbourhood trust. Further, we simultaneously test whether higher levels of religious diversity are associated with a weaker sense of national identity, and whether a weaker sense of national identity is associated with lower levels of neighbourhood trust. We also test these relationships across different religious groups, albeit in an exploratory manner. It is not known, for example, whether for the majority group in this study (i.e., Christians) these associations might be more pronounced given that increasing religious diversity may threaten their dominant status, and whether other minority religious groups might feel less threatened, and whether those living in contexts of increasing diversity will still evidence a similar pattern.

Religious Subgroup and National Superordinate Identification
According to the social identity perspective, which incorporates social identity theory (Tajfel & Turner, 1979) and self-categorization theory (Turner et al., 1987), individuals use and/or create distinct social categories to classify and organize their social worlds. Individuals compartmentalize themselves and others into these categories, thereby differentiating their ingroup ("us") from relevant outgroups ("them"; Tajfel & Turner, 1979). Moreover, individuals can identify with and attach value and meaning to the social categories they belong to, which can play a central role in shaping intergroup attitudes, behaviors, and social relations (e.g., Schmid et al., 2010).
Since individuals typically belong to many different groups, the identity groups they belong to can often be defined in terms of different levels of inclusiveness and may be hierarchically structured. Subgroup identities are less inclusive identities that are typically shared only with fellow members of the same group (e.g., Muslims, Christians), and that can be perceived as nested within a superordinate identity that is more inclusive and subsumes others belonging to various subgroups under a common, shared ingroup (e.g., British). Some have claimed that subordinate identities divide at the expense of a superordinate identity that may unite by promoting a common ingroup identity (Miller, 1995;Reeskens & Wright, 2013). However, the dual identity model (Gaertner & Dovidio, 2000) challenges the idea that there is necessarily a conflict between identities.
Various potential antecedents of social identification have been identified. The stability of status relations appears to influence group identification, such that individuals are more likely to identify with the ingroup when its status is unstable or under threat (Ellemers, 1993). There is also evidence to suggest that perceptions of increased ingroup superiority and status strengthen ingroup identification (Chow et al., 2008). Others have argued that identification with groups results from a desire to reduce uncertainty, or that individuals identify with social categories that allow them to optimally balance the need for belonging and the need for distinctiveness (for a review, see Hewstone, 2015). Selfcategorization theory specifically focuses on the cognitive underpinnings of social identity and argues that the salience of a social category within a given context or situation influences individuals' social identification (Turner et al., 1987). Specifically, identity may be flexible and influenced by salient features of a context, such as the composition and distribution of outgroup members. In particular, the situational salience (situational accessibility) of a particular social identity category within a social context is thought to be a powerful driver of individual identity categorization and strength (Oakes, 1987). When a social identity becomes salient in a context, intergroup differentiation and intragroup assimilation occur, which typically results in biases favoring the ingroup (cf. Brewer, 1979;Mullen et al., 1992).
Extrapolating from the social identity approach, we argue that contextual features of one's environment, such as varying degrees of religious diversity in one's neighbourhood, can heighten identity salience and thus influence the extent to which individuals identify with the groups they belong to, such as subordinate religious groups and superordinate national groups. However, a systematic investigation of the relationship between religious diversity and both religious and national identification has yet to be considered, despite the possibility that strong cultural or religious identities may pose barriers to integration or social cohesion because they may undermine a shared ingroup identity.

Religious Diversity, Superordinate and Subordinate Identification, and Neighbourhood Trust
Figure 1 summarizes our hypotheses. Diversity is said to relate to experienced perceived or real threat. Conflict theory (Blalock, 1967;Blumer, 1958) suggests that the diversity in an area can stimulate perceptions of realistic/material or symbolic/cultural threat. Religious diversity may thus increase the salience of religious identity for people's sense of who they are and stimulate ingroup biases, translating into stronger subordinate religious identification and, consequently, weaker superordinate national identification. This relationship assumes that religious and national identities are mutually exclusive, in line with the prediction of classical social identity theory that a rise in salience of one identity comes at the expense of another identity. H1: Religious diversity is positively associated with strength of religious (subordinate) identity.
H2: Religious diversity is negatively associated with strength of national (superordinate) identity.
Populist arguments suggest that subordinate identities conflict with superordinate identities at the expense of trust and social cohesion. According to these claims, strong subordinate identity erodes trust and social cohesion (Goodhart, 2013;Scheffer & Waters, 2011), while strong superordinate national identity strengthens social cohesion (Miller, 1995). This assumes that strong subordinate identification results in greater solidarity, trust, and cooperation among one's ingroup that do not extend beyond these group boundaries to other outgroups. In contrast, a strong national identity is assumed to reflect a greater commitment to solidarity, trust, and cooperation more broadly. In other words, a strong religious subordinate identity is thought to divide a society, whereas a strong national identity is thought to unite it.
H3: Strength of religious (subordinate) identity is negatively associated with neighbourhood trust.
H4: Strength of national (superordinate) identity is positively associated with neighbourhood trust.

Majority Versus Minority Perspectives
Theoretically, there may also be differences between subgroups in terms of the role of religious diversity, depending on their status and prestige. The distribution of subordinate groups according to religion in England is such that Christianity is the numerical majority religion. As such, variations in the diversity of the social context typically indicate increases in the number of members of non-Christian subordinate groups through internal and international migration. For Christians, increases in religious diversity suggest that they are becoming a smaller group, and may trigger a potential threat unique for them as members of the historical majority group (cf. Craig & Richeson, 2014;Outten et al., 2012). These contextual changes may increase the sensitivity of Christians to these mechanisms as the historical majority group.
Previous work found that minority and majority group members perceive outcomes related to social identity and intergroup relations-such as intergroup contact (Saguy & Dovidio, 2013) and superordinate identities (Ufkes et al., 2012)-differently, and that group status can affect the relationship between social context and social identity. Majority/ high-status groups may view demographic change as an end to the relative advantage they have historically enjoyed, and consequently they feel more threatened (Scheepers & Ellemers, 2005), which may cause their identity to become more salient. One experimental study demonstrated that White majority members felt more anger and fear towards ethnic minorities if they viewed demographic projections of Whites no longer holding numerical majority status (Outten et al., 2012). This aligns with survey research demonstrating that greater actual outgroup size is associated with higher threat perceptions for the White majority (e.g., Quillian, 1995;Taylor, 1998), an effect also shown for perceived outgroup size (Semyonov et al., 2004). Likewise, it has been argued that the effect of change in the ethnic context on various outcomes such as trust and social cohesion is, in general, more influential for majority group members' attitudinal responses (Allport, 1979;Ziller, 2015). There is, however, research showing that demographic changes in societies require, at least in the short term, an adaptation from both majority and minority religious groups (Ramos et al., 2019). In this work, it was demonstrated that a short-term increase in countries' religious diversity was associated with lower generalized trust in others and lower well-being of all religious groups. These demographic shifts may unsettle the minority group's status as well, given that these groups may feel a need to compete for resources with other new groups. As such, in this study, we present our results with all religious groups included (controlling for group differences) and then provide a subgroup analysis to examine group patterns in more detail.

Methods
We used data from the 2008-2009 Citizenship Survey (CS2008) (NatCen & DCLG, 2010), which contains a large nationally representative sample of approximately 10,000 adults from England, alongside a minority boost sample of 4,962 ethnic minority respondents. We selected only those respondents for whom complete data for our outcome variables and independent variables were available, and who reported a British national identity (the superordinate identity), accounting for 6,089 respondents living in 300 local authorities in England. 1 Respondents who reported some other non-British national identity would be reporting national identification in reference to a different group, which may not necessarily be a superordinate group in the context of this particular study.
Once missing data were accounted for, the final eligible sample for analysis was 7,751 (5,948 missing cases). As mentioned, only respondents who reported a British national identity were eligible for the analysis, resulting in 7,211 dropped cases. Of the other ineligible cases, a large proportion were not eligible due to data limitations, as opposed to questions that respondents refused to answer; 2,155 respondents were aged 70 and over and were not asked about their education. Thus, our analysis is only based on respondents aged 16-69 years. A further 595 respondents were removed as they lived in Wales, where it was not possible to gather information on their neighbourhoods. Lastly, 135 respondents were dropped as the local authority boundaries in which they lived changed between 2001 and 2011, making 2008 contextual-level estimations impossible for religious heterogeneity, urban location, and area deprivation. These data restrictions resulted in a final sample of 6,089 respondents. Table 1 presents the descriptive statistics for the full citizenship survey and the eligible study sample, and suggests no substantive differences.
We Local authorities (LAs) are the Level 2 unit of analysis, as this is the lowest level of analysis available for matching. The average population size of a local authority is around 330,000 people. There are 300 LAs included in this analysis out of 326 in the 2011 Census.

Key Individual-Level Variables
Religious and national identity. We investigate the effect of religious diversity on religious and national identification. Group identification is central to a person's self-concept (Turner et al., 1987), and the importance of group identification is one dimension thereof (cf. Ashmore et al., 2004;Leach et al., 2008). Respondents were first asked to identify their religion with the following question: "What is your religion even if you are not currently practicing?" and were then asked, "How important is religion to your sense of who you are?" Respondents were also asked, "How important is your national identity to your sense of who you are?" (0 = not at all important, 1 = not very important, 2 = quite important, 3 = very important; "do not know," nonresponses, and "nonapplicable" were dropped: 255 cases for religious identity and 186 cases for national identity).
Trust. Trust was measured with the question "Would you say that . . ." (1 = none of the people in your neighbourhood can be trusted, 2 = a few can be trusted, 3 = some can be trusted, 4 = many of the people in your neighbourhood can be trusted).

Contextual-Level Variables
Basic descriptive statistics are displayed in Table 1 and a correlation matrix is displayed in Table 2. Religious diversity of local authorities was measured using derived data from the 2001 SAMS and the 2011 Census, defined as the Herfindahl index. The religious diversity measures were created using the following census religious categories: Christian, Buddhist, Hindu, Jewish, Muslim, Sikh, other religion, and no religion. The Herfindahl index is calculated as follows: Where S i is the proportion of people who profess religion i in local authority j. This index ranges from 0 to 1, indicating the probability that two randomly selected individuals in a country belong to different religious groups. The index increases with both the number of religious groups and the evenness of the distribution of individuals across groups. Our analyses control for poverty at the local authority level to account for the possibility that in deprived areas there might be more competition over resources, which may amplify the associations between religious diversity, religious identity, and neighbourhood trust. We used the Index of Multiple Deprivation (IMD) measure to indicate contextual-level poverty, which is a composite poverty measure created using the rankings of the local authority according to the following characteristics: income; employment; health deprivation and disability; education, skills, and training; barriers to housing and services; crime; the living environment.

Individual-Level Control Variables
Respondents were asked to self-categorize in terms of their religious group: Christian,  Buddhist, Hindu, Jewish, Muslim, Sikh, any other religion, no religion at all. Due to small sample sizes for Hindus, Sikhs, and Buddhists, we were forced to combine these categories into the "Other" category. This resulted in four categories: Christian (N = 3,138), Muslim (N = 1,300), Other (N = 951), No religion (N = 700).
We included a measure that gauges the degree to which the respondent agrees that religion is an important reason for living in their neighbourhood with the question, "To what extent do you agree or disagree that your religion affects where you live?" (0 = strongly disagree, 4 = strongly agree). We assigned atheists a "0" on this scale as they were not asked this question and we assumed that religion is the least important to where atheists live (this also seems to be the assumption made by the survey organizers by not asking atheists). 2 This variable may control for self-selection into an area according to religious preferences.
In order to control for any effects of mixing/contact with outgroup members on identity, and as a proxy for levels of segregation, we used the question, "[I]n the last year . . . how often, if at all, have you mixed socially with people from different ethnic and religious groups to yourself . . . at your home or their home?" (1 = never, 2 = less often than once a year, 3 = at least once a year, 4 = monthly, 5 = weekly, 6 = daily). We also controlled for individual-level demographic variables capturing sex, age, marital status, education, and place of birth (born in the UK or outside UK).
As can be seen in Table 1, the mean of neighbourhood trust is 3.12, religious identity 2.86, and national identity 3.27. The mean level of religious diversity is 57% across neighbourhoods, with a minimum of 27% and a maximum of 76%. Table 2 presents the correlations among all variables. Religious diversity was negatively associated with neighbourhood trust (r = −.26, p < .001). Religious and national identity were both negatively related to neighbourhood trust (r = −.15, p < .001; r = −.04, p < .001, respectively). Religious diversity had a substantively greater association with religious identity (r = .33, p < .001) than national identity (r = .07, p < .001).

Analytical Strategy
We estimated two-level hierarchical path models to account for the clustering of respondents in local authorities (StataCorp, 2017). This was done using STATA Version 15 and the multilevel GSEM (generalized structural equation modeling) suite of commands. GSEM uses maximum likelihood and enables us to specify a probit coefficient estimator for our ordinal outcome variables (trust, religious identity, and national identity). In addition to the pathways outlined in Figure 1, our models also estimate the direct path from religious diversity to trust (see Ramos et al., 2019), and allow religious identity and national identity to be correlated with each other. Finally, we included all of our individual-level and contextual-level control variables in the regressions for all of the endogenous variables in our models (i.e., religious identity, national identity, and trust).

Results
We present the main findings for the pathways between religious diversity, religious identity, national identity, and neighbourhood trust in Figures 2-7, but also present the full models in Tables 3 and 4. Figure 2 (and Table 3) presents the results of a pooled analysis where all religious groups are combined. Figures 2-6 present the multigroup comparison for each religious group separately (Christian, Muslim, other, nonreligious). As a sensitivity check, we also estimated a model for all non-Christian respondents combined, to increase the sample size and statistical power (Figure 7 and Table 4; results do not change between sample specifications).
Figures 2-7 (see Tables 3 and 4 for full results) present the main results of our path models estimating the relationship between religious diversity, religious and national identity, and neighbourhood trust. Before discussing the main substantive results, we draw attention to the finding that there is a positive correlation between religious and national identity (b = 0.22, p < .001), suggesting that these two identities are not mutually exclusive. The multiple-group analysis (Figures 3-7   Note. Path coefficients are unstandardized, the corresponding standardized coefficients are shown in parentheses. Controls in each pathway are not shown (see Table 3   Note. Path coefficients are unstandardized, the corresponding standardized coefficients are shown in parentheses. The estimated indirect effect of religious diversity on trust via religious identity is −0.02 (p = .070). Controls in each pathway are not shown (see Table 4 for full results): sex, age, education, marital status, born in the UK, religion is important for place to live, mix with outgroups at home, area deprivation.

Pooled Analysis With All Religious Groups
Religious diversity and identity. Figure 2 (Table 3) presents the results for the relationship between religious diversity and religious and national identity for the pooled sample including respondents of all religious groups in the same model. Hypothesis 1 stated that the religious diversity of an area would be positively associated with subordinate religious identity, while Hypothesis 2 stated that it would be negatively associated with the superordinate national identity. There was support for Hypothesis 1. Results showed that respondents living in areas with higher religious diversity were more likely to report stronger subordinate religious  Note. Path coefficients are unstandardized, the corresponding standardized coefficients are shown in parentheses. Controls in each pathway are not shown (see Table 4 for full results): sex, age, education, marital status, born in the UK, religion is important for place to live, mix with outgroups at home, area deprivation.  Note. Path coefficients are unstandardized, the corresponding standardized coefficients are shown in parentheses. Controls in each pathway are not shown (see Table 4  identification (b = 0.64, p < .001). In contrast, there was no support for Hypothesis 2, as no significant relationship between religious diversity and national identity emerged (b = −0.06, p = .71). Taken together, these findings show an association between religious diversity and a subordinate religious, but not superordinate national, identification.
Identity and trust. Hypotheses 3 and 4 concern the association between religious and national identity and neighbourhood trust, whereby stronger religious identification is associated with lower levels of trust (Hypothesis 3), while stronger national identification is supposed to have the opposite relationship (Hypothesis 4). These hypotheses presume that subordinate religious identity divides society, whereas superordinate national identity unites it. Figure  2 ( Table 3) shows that there is marginal support for Hypothesis 3; religious identity is negatively associated with trust (b = −0.02, p = .057). There was no support for Hypothesis 4, as the pathway between national identity and trust was statistically nonsignificant (b = 0.00, p = .740), although the coefficient was positive, in line with the hypothesis. Note. Path coefficients are unstandardized, the corresponding standardized coefficients are shown in parentheses. Controls in each pathway are not shown (see Table 4    Note. Path coefficients are unstandardized, the corresponding standardized coefficients are shown in parentheses. Controls in each pathway are not shown (see Table 4

Subgroup Analysis
With a subgroup analysis, we examined patterns across different religious groups. The results of this analysis are shown in Figures 3-7 (Table 4).
In Figure 3, results demonstrate that Christians reported stronger religious identification in areas of higher religious diversity (b = 1.12, p < .001). However, Christians' national identity was not associated with these contextual characteristics (b = 0.17, p = .315). Results for Christians mirror those of the pooled analysis. For all other religious minority groups (and the pooled sample of religious minority groups), these relationships were not observed (bs < .43 and ps > .131; see Figures 4-7 and Table 4). In terms of the relationship between religious and national identity and trust, we found that, for Christians, a stronger religious identity was associated with lower neighbourhood trust (b = −0.03, p = .040). We also observed a negative indirect effect (b = −0.02, p = .070) of religious diversity on trust via religious identity, which approaches significance at the 10% level. Among religious minority respondents, we did not find any statistically significant pathways between identity and trust (bs < .04 and ps > .492). These results suggest that religious identity may be more relevant in driving trust for the Christian majority subpopulation. A coefficient difference test supported this claim by showing that religious diversity pathways vary significantly between groups. Specifically, the pathway between religious diversity and trust is statistically different between groups, χ 2 (3) = 16.40, p < .001, as is the pathway between religious diversity and religious identification, χ 2 (3) = 15.79, p = .001. These results are consistent with the claim that subordinate religious identities may be detrimental for trust and social cohesion, albeit only for the majority sample.

Discussion
This research examined the relationships between the religious diversity of local areas, religious and national identity, and neighbourhood trust. We derived hypotheses based on the social identity theory literature (e.g., Tajfel & Turner, 1979) to suggest that religious diversity would relate differently to subordinate and superordinate identification. We also tested the idea that subordinate and superordinate identification would have negative and positive implications for neighbourhood trust, respectively. Specifically, we argued that religious diversity would be positively associated with religious subgroup identification (Hypothesis 1), but negatively related to national identification (Hypothesis 2). We argued further that religious subgroup identification would be negatively related to neighbourhood trust (Hypothesis 3), whereas national identification would be positively related to this form of trust (Hypothesis 4).
Our hypothesis that the religious diversity of an area would be positively associated with subordinate religious identity (Hypothesis 1) was      confirmed in our pooled analysis and subgroup analysis for Christians. We found that a greater religious neighbourhood diversity was related with stronger subordinate religious identity. Although support for this hypothesis was found in the pooled analysis, we did not find a significant relationship for the Muslim, other, or nonreligious respondents. We further predicted that religious diversity would be associated with lower national identification for all groups (Hypothesis 2), but we found no support for this hypothesis across all of our analyses.
We also derived hypotheses about the potentially conflicting relationships that religious subordinate identification and national superordinate identification would have with neighbourhood trust. Specifically, we hypothesized that a strong religious identification would be divisive for society and be negatively associated with trust (Hypothesis 3), while a strong national identification would be a uniting factor for society and be positively related to trust (Hypothesis 4). Our results showed that religious identity was negatively related to neighbourhood trust in our pooled analysis and subgroup analysis for Christians only (this relationship did not exist for Muslims, other religions, and the nonreligious). No relationship emerged between national identification and trust across the pooled and subgroup analyses.
The subgroup analysis suggests that the effects for Christians may be stronger than for the other religious groups. This is consistent with the claim that subordinate religious identities may be detrimental for trust and social cohesion, albeit only for the Christian majority sample. For the minority religious groups, our key variables appeared to be unrelated to religious diversity. Consistent with predictions of conflict theory (Blalock, 1967;Blumer, 1958), integrated threat theory (Stephan & Stephan, 2000), and reactive identity theory (Rumbaut, 1994), greater diversity at a contextual level may be associated with increased threat among majority group members, which could be involved in identification. For minority groups, however, greater heterogeneity appears to be associated with distancing from religious subgroup identities. Future research should seek to extend this work to understand the wider implications of these findings for outcomes related to intergroup relations, such as intergroup contact and threat. What remains unclear is how identification affects key outcomes for both groups. Since identification can affect both positive (e.g., positive effects on health and well-being) and negative (e.g., intergroup bias) outcomes, policies aimed at promoting positive outcomes for individuals and communities will need to find ways of enhancing the benefits of identification in diverse communities without generating threat perceptions.
Our results have broad implications, which can be summarized into three general themes. First, religious diversity is a salient and important characteristic of the environment in which individuals live. Much of the existing literature has focused on the ethnic context, but this paper emphasizes the necessity of considering the religious context. Our finding that the religious diversity of an area affects the majority group in particular is similar to findings of previous research (Allport, 1979;Ziller, 2015). A first key implication of our findings is that they may fuel concerns that social cohesion may be undermined through a potential "reactive" Christian identity among the religious majority in areas of high religious outgroup density (Rumbaut, 1994). This could lead majority group members to behave in ways that protect their status. As such, it is possible that the "reactive" Christianity relationships identified at levels of higher religious diversity may well disappear at lower levels of religious diversity, perhaps because of threshold effects of religious diversity that cannot be identified here. Unfortunately, however, the actual size and distribution of the Christian population across local authorities in England prevent the estimation of effects of religious diversity for Christians living in very religiously diverse areas.
Second, the findings also suggest that a strong subordinate religious identity does not weaken the strength of the superordinate national identity. The lack of a weakened superordinate ingroup identity amongst those with a strong subordinate ingroup identity is in line with the dual identity model (Gaertner & Dovidio, 2000) and the integrative model of subgroup relations (Hornsey & Hogg, 2000), both of which suggest that individuals can simultaneously have salient superordinate and subordinate ingroup identities, as is also reflected in the significant correlation between religious and national identification in our sample. Religious and national identities can be considered complementary identities in multicultural England, and thus strong religious identities do not appear to weaken the common national identity that is believed to be necessary for social cohesion (Gaertner & Dovidio, 2000).
Third, our findings fail to support populist arguments, focused on minority groups, suggesting that subordinate identities erode trust and social cohesion (Goodhart, 2013;Scheffer & Waters, 2011). In fact, we found that it was only for majority group Christians that religious diversity was associated with a stronger subordinate religious identification, which, in turn, was associated with lower neighbourhood trust. For the other religious groups, we found no significant associations between our variables of interest. A policy implication of this finding may be that more attention should be focused on responding to and reducing majority group members' perceived threats, rather than insisting that minority members give up, or loosen, their religious identities in order to "fit in" and promote cohesion.
We acknowledge some limitations of this research regarding causality and selection, to the extent that the difference in associations found for religious diversity among religious groups could be due to selection effects on residential mobility. Members of minority religions who choose to live in areas with high proportions of nonminorities may be more assimilated into mainstream society and so have weaker religious identities prior to selecting into these high outgroup areas. Similarly, the theoretical framework that we have applied reflects a dynamic process, and our cross-sectional data do not allow us to explore individual-level and contextual-level changes over time (see Li et al., (2021); Laurence & Bentley, 2016;Ramos et al., 2019).
We are also unable to determine the precise mechanisms behind the associations reported in this research. The estimates for the direct paths between religious diversity and neighbourhood trust display strong negative associations across all of our models. These results are reasonably consistent with existing British literature on the direct relationship between ethnic heterogeneity and social cohesion; and neighbourhood trust, specifically, is the form of social cohesion most consistently negatively affected by diversity (Dinesen et al., 2020;van der Meer & Tolsma, 2014).
The persistence of the strong direct relationship between religious diversity and neighbourhood trust-having accounted for the role of identity (both religious and national)-suggests that there are other mechanisms that might be at play here. We suggested that one additional mechanism may be that of increased perceived threat when greater proportions of subordinate outgroup members live in the same area (Stephan et al., 1999). However, future work is needed to test this explanation. The use of complementary research strategies, such as experimental methods, may also aid our understanding of the causes underlying different aspects of identity.
Moreover, we used the available data on religious diversity at the local authority level. These geographical areas are wider than what individuals may typically perceive as a neighbourhood, and had we had access to contextual data at smaller geographical units, this would have allowed us to more precisely match individual experiences to the demographic data. Such a more fine-grained analysis would have allowed us to disentangle other relevant variables such as segregation that, due to the effect it has on reducing intergroup contact, may amplify the effects reported in our study. Nonetheless, we do not consider this a major concern for the present results, given that previous work has found an association between country-level religious diversity and perceived trust (Ramos et al., 2019). It has, however, been argued that effect sizes of diversity tend to be larger in analyses with smaller geographical units that are closer to the individual (Dinesen & Sønderskov, 2012;Laurence & Bentley, 2016).
Our findings also raise other questions such as how proximate forms of cultural diversity may be less affected by heterogeneity compared to more culturally distant ones; for example, if Pakistanis share a colonial legacy, they might see themselves as more British compared to Eastern European immigrants to the UK. Likewise, this research should be extended to other countries where Christians are the majority, but also to countries where other religious groups are the majority, to see whether the findings uncovered here generalize to other contexts.
A final limitation of the study concerns the reliability of the single-item measures of religious identity, national identity, and neighbourhood trust. Identification is a complex construct with multiple underlying dimensions that single-item measures may fail to capture. It has, however, been suggested that social identification can be operationalized using a single-item measure (Postmes et al., 2013). Most of the research on different types of trust has only used single-item measures of each (e.g., Schmid et al., 2014), and future research should test the reliability of the single-item measures used here with multiple-item measures of the same constructs.
To conclude, our study has provided important insights into the relationships between religious diversity, religious subordinate and national superordinate identification, and neighbourhood trust. Our research further suggests that Christian majority group members in particular may appear to react more strongly to changes in religious diversity. These reactions are manifested in decreased trust in others who share their neighbourhood, which could lead to animosity against particular groups in society.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs
Matthew R. Bennett https://orcid.org/0000-0002-9523-6053 Katharina Schmid https://orcid.org/0000-0001-6018-9245 Notes 1. This restriction is based on the following survey question: "What do you consider your national identity to be? Please choose as many or as few as apply." The answer options are English, Scottish, Welsh, Irish, British, and Other. Among the 14,322 respondents living in England, 51.1% declared a British national identity, 84.6% declared a British or English national identity, and 9.5% both. Including those respondents who do not consider their national identity to be British or English in our analyses does not substantially alter our results. Likewise, the results do not change when including respondents who consider their national identity to be British or English (available upon request). 2. To be sure our coding did not interfere with the study's results, we ran our models without this variable and all findings remained largely comparable.