Praying on Brexit? Unpicking the Effect of Religion on Support for European Union Integration and Membership

This article examines how religious affiliation shapes support for European Union membership. While previous research has shown that Protestants are typically more eurosceptic than Catholics, little is known about the nature of this relationship: specifically, whether religion affects one's utilitarian assessments of the costs and benefits of membership, or one's affective attachment to the EU. Using the 2016 British Election Study Referendum Panel, this article shows that religious affiliation influences both sets of attitudes, suggesting that the values and shared history associated with one's religion shapes how a voter perceives the performance of the EU in delivering its policy objectives, and its operation as a legitimate institution. Moreover, some findings from previous research are challenged: Protestants are not as unified in their scepticism of the EU as is widely assumed, and the positive relationship between Catholicism and support for EU integration is not apparent in the UK.


Introduction
The United Kingdom's decision to leave the European Union presented a significant challenge to the progress of EU integration, and prompted expectations amongst eurosceptic politicians in Italy, Germany, Denmark and France that they could be next to leave. Explaining individual level euroscepticism, and how UK voters arrived at their decision, is important for not only understanding why such a challenge has occurred, but also to consider what could lead voters in other member states to demand similar referenda, or complete withdrawal from the EU, in the future. Most explanations for the 'Brexit vote' focus on national identity, economic deprivation, hostility to immigration and political alienation (see Clarke et al., 2017;Curtice, 2017;Fox and Pearce, 2018). Religious belief and affiliation play minor roles, despite the historic links between religion and many political parties central to the development of the EU (Bochel and Denver, 1970) and the effect of religion on many political attitudes, including euroscepticism (see Boomgaarden et al., 2011;Guth, 2003, 2015;Nelsen et al., 2011;Tilley, 2015). Previous studies identify significant variation in political attitudes between different denominations of Christianityby far the largest religious group in Europein addition to differences between 'religious' and 'non-religious' groups. While the magnitude of these differences is smaller than those associated with more common predictors of euroscepticism, any comprehensive explanation of public support for EU membership and integration cannot overlook religion.
There has been limited progress in explaining how religion influences attitudes towards European integration. It is unclear whether the fact that Protestants are typically more eurosceptic than Catholics stems from their assessment that the costs of EU integration and membership outweigh its benefits, or from their lack of affective attachment to the EU. Additionally, the robustness of causal frameworks is hindered by over-dependence on cross-sectional data. Finally, few studies have been able to explore any but the largest denominations of Christians, because of limited sample sizes from smaller groups (such as Methodists). This article addresses some of these shortcomings and provides a more nuanced analysis of how Christian religious affiliation affects support for EU membership.
First, it adopts a more nuanced conceptualization of euroscepticism, which recognizes the distinction between 'utilitarian' and 'affective' support for EU integration (Lindberg and Scheingold, 1970). Second, it uses the 2016 British Election Study (BES) Referendum Panel, which was collected before and after the referendum on the UK's EU membership, when the issue of European integration was most salient. The use of panel data allows a well-specified causal model to measure the impact of religious affiliation on utilitarian and affective support, and the impact of those on vote choice in the referendum. Finally, the large sample of the BES means that several Christian denominations can be studied, taking the analysis beyond a simplistic Protestant vs. Catholic dichotomy.
The analyses show significant effects on euroscepticism from association with different denominations, and that this reflects the effect of religion on both utilitarian and affective dimensions of support. As argued in previous literature, some Protestant groups tend to be particularly eurosceptic; however, there is considerable variation between them depending on their historic relationship with the national government and/or conceptions of national identity. Protestants who have traditionally had a strong connection to notions of UK national identity, such as Anglicans, are the most eurosceptic, while those that have been marginalized or persecuted by the state in the past are more supportive of the EU, possibly viewing it as a means by which their distinctive identity can be protected. In a further challenge to existing literature, Catholicism was not associated with greater support for EU membership in the UK once respondents' country of birth was accounted for.
The article does not only confirm, therefore, that religion is an important determinant of individual-level support for EU integration, but also illustrates the wider significance of religious affiliation in shaping political attitudes. Existing theories of support for European integration and the influence of religion in politics suggest that religious identity should primarily affect one's long-term values and attachment to the EU and one's national communityin other words, it should influence affective support for the EU. Our findings show, however, that religion affects both affective identification with the EU and voters' short-term assessments of its performance and competence. It not only shapes the way that voters relate to their national and regional communities and the EU, but also their political agendas and expectations, and consequently, their assessment of how successful the EU has been in promoting their interests. The influence of religion on attitudes towards the EU is more extensive and widespread than current theories suggest, and may well be more widespread than is commonly assumed in relation to other political traits as well.
The next section of this article reviews the literature on the relationship between religion (primarily Christianity) and euroscepticism, outlining the current theories and the shortcomings resulting from limitations in available survey data. The article then outlines the research design before results are presented. The conclusion summarizes the findings and considers their implications for the future study of public support for EU integration.

I. Literature Review
It is often claimed that as societies industrialize, religious habits gradually erode and people become indifferent to spiritual appeals, meaning religious attitudes have little impact on social and political values and decisions (Norris and Inglehart, 2004). This closely follows Lipset and Rokkan's argument that, over the last two centuries, politics in Europe has been shaped by the secular-clerical cleavage (Lipset and Rokkan, 1967), and that religion has gradually lost ground to secularismeven though the levels of secularization vary (Norris and Inglehart, 2004). In Britain, for example, religion (primarily the divide between Catholics and Protestants) was considered a major predictor of political behaviour until World War I (Wald, 1983). However, the declining number of people associating with the Christian church and of practicing Christians made it look increasingly irrelevant, especially alongside the rising membership of trade unions and the growing significance of social class (Butler and Stokes, 1969). This undermined the perceived importance of religion in Britain (see Rose and McAllister, 1986), and all but erased it from most contemporary analyses of electoral behaviour.
However, asserting the dominance of secular social cleavages and dropping religion altogether seems premature considering the historic links between key political and religious institutions in European democracies, and evidence of the importance of religious beliefs and background in shaping political attitudes and behaviours. In the UK, for example, the Conservative Party was historically very close to the Church of England, and the Liberal and Labour parties had roots in the traditions of dissenting Protestants and Catholics, relationships that continue to inform the ideology of those parties (Bochel and Denver, 1970). Similarly, in Germany, the Christian Democratic Uniona key political party in the foundation and development of the EU (Nelsen and Guth, 2003) has strong roots in both the Catholic and Protestant Christian traditions of civic and social values.
Furthermore, while formal religious practice such as service attendance is declining, new facets of religious identity are emerging that continue to hold sway over individuals' broader social and political beliefs (Davie, 1994). Identification with a certain religious group is no longer necessarily determined through engagement in religious practice, but can be based on shared socio-cultural values and experiences, moral principles and codes of behaviour (Beckford, 2003). This shared identity, and the social network, beliefs and values that underpin it, can have considerable influence over an individual's political attitudes and behaviour (Gutmann, 2003). Finally, religion still influences voter decisions. Tilley (2015) has demonstrated that even if a growing number of voters do not affiliate themselves with a particular religion, their parents and grandparents are more likely to do so, and they pass on the values, beliefs and shared history of their respective religions to their children during socialization. This means that religious groups can shape the political attitudes and values even of those who do not identify with them through family networks. These processes help to explain the continuing relationship between identification with religious groups and support for more conservative political ideologies and political parties, as well as vote choice in elections (Clements, 2015;Esmer and Pettersson, 2007;Tilley, 2015).

II. Religion and Euroscepticism
While research on the effect of identifying with a non-Christian religion on euroscepticism is rare (primarily because of the limited numbers of non-Christians in social surveys), numerous studies have shown significant differences between Christian denominations (Nelsen et al., 2011;Guth, 2003, 2017;Scherer, 2015). Catholics, for example, tend to be more supportive of EU integration than Protestants. Religious affiliation also shapes attitudes related to support for European integration, such as towards immigration or authoritarian values (Boomgaarden and Freire, 2009), and national identity (McLaren, 2002). Explanations for this relationship fall within two broad theories. The first focuses on the shared history of different Christian groups and their relationship with the Catholic Church and national governments, while the second considers the overlap between Christian values and beliefs and the political values that underpin the EU and its forebears.
Catholicism has historically been associated with membership of a community that transcends national borders and is overseen by a supra-national institution that constrains the sovereignty of national governments. While the EU is relatively young (compared with the Catholic Church), it nonetheless shares these characteristics with regard to policy areas in which integration has led it to take a leading role, and so reflects a political arrangement many Catholics are supportive of (Madeley, 2007;Nelsen and Guth, 2003). In addition, the international nature of Catholicism leads Catholics to recognize themselves as members of a borderless, international community (Nelsen et al., 2011).
There is also overlap between the values of Catholicism and the nature of EU integration. It has been argued that European integration has a distinctly 'Catholic nature', which reflects the 'traditional idea of a unified moral leadership … safeguarding Christian values under the power of a central authority' Freire, 2009, p. 1243). Many of the ideals promoted by supporters of EU integration have clear overlaps with the values that lay at the heart of Catholic beliefs and ideal lifestyle, such as charity, peace and social justice (Madeley, 2007;Nelsen and Guth, 2003). Indeed, the original objectives of the European Coal and Steel Community included securing peace in Europe, and subsequent EU treaties and policies have emphasized its commitment to charity and social democratic visions of social justice (Coupland, 2004;Madeley, 2007). This overlap is also reflected in the fact that the European project has traditionally been led and strongly supported by political parties with strong links to Catholic values, such as Germany's Christian Democratic Union (CDU) (Madeley, 2007;Nelsen and Guth, 2015). Overall, therefore, Catholics are expected to be more supportive of the EU as a political institution, and more likely to view EU membership as beneficial to both the global Catholic community and the promotion of Catholic values, resulting in them being more supportive of EU membership and integration.
Protestants, in contrast, have historically been dependent upon strong and independent national governments to assure their protection from the dominance and persecution of the Catholic Church and its allies (Nelsen and Guth, 2015;Scherer, 2015). They have a historic interest in protecting the sovereignty and independence of the nation state, and suspicion of a supra-national institution constraining that sovereignty. The concessions to national sovereignty required by EU membership are, consequently, unappealing to many Protestants; as Coupland (2004) points out, many Protestants' euroscepticism originated in their concerns that European integration would lead to Catholic dominance.
In addition, while the values espoused by the EU and the notion of transnational citizenship are a source of support for integration amongst Catholics, the opposite is true for Protestants, many of whom are suspicious of the promotion of universalist values because of the potential impact on the distinct identity and beliefs of their church (Scherer, 2015). This is particularly apparent amongst members of churches that have historically played a prominent role in the articulation of national identity (such as the Church of England). For them, any challenge to national identity or the distinctiveness of the national government can be an indirect challenge to the distinctiveness and identity of their church (McLeod, 1999).
Not all Protestant churches, however, have the same close relationships with national governments or the same investment in national identity. In the UK, groups of Protestants broke away from the Church of England in the 17 th century because of disagreements over the interpretation of scripture and religious practice, forming new churches such as the Baptist and Presbyterian (Tombs, 2014). They were followed by Methodists in the 18 th century. These 'non-conformists' were persecuted for failing to adhere to the state religion of England represented by the Anglican Church. Hence, while they share many of the values of larger Protestant churches that differentiate them from Catholicism (such as a rejection of the Papacy, the Catholic Church and their universalist values), their relationships with national governments and the nation-state have been far more confrontational. Consequently, they are expected to be more hostile to EU membership than Catholics, although their weaker, historical attachment to the nation-state and national governments may mean that they are not quite as eurosceptic as members of the Church of England, for example. Their more localized nature and attachment to smaller community identities may make them relatively more supportive of the EU, if they perceive that their distinctive identity can be protected and promoted by EU policies.
In the context of the UK's referendum on EU membership, therefore, we expect Catholics to be the most pro-remain given their historic relationship with transnational institutions, and the overlap between their beliefs and those represented in EU integration and espoused by the EU. Protestants should be more eurosceptic and so more likely to support Brexit, though the smaller groups that do not have the same historic attachment to the nation-state or English/Scottish/Welsh national identity should be less so. This provides H1: H1. Anglicans are more likely to vote 'leave' than Baptists, Methodists and Presbyterians, who are in turn more likely to vote 'leave' than Catholics.

III. Religion and Dimensions of Support for the EU
To capture the complexity of support for EU membership, it is useful to distinguish between 'utilitarian' and 'affective' support (Lindberg and Scheingold, 1970). It reflects the distinction between specific (utilitarian) and diffuse (affective) support for a political institution (Easton, 1975), and between input-and output-oriented legitimacy (Down and Wilson, 2017;Scharpf, 1999). Although Wessels identified three (2007) and Boomgaarden et al. (2011) five such dimensions, a dichotomous utilitarian-affective distinction is sufficient to separate evaluations of the benefits of EU membership from more deeply held sentiments (Down and Wilson, 2017).
Utilitarian support is output-oriented and reflects an individual's cost-benefit analysis of EU membership and their evaluation of its performance in delivering policy goals. An individual who feels that the benefits of membership outweigh the costs, and that the EU delivers on its policy objectives, is more likely to have high utilitarian support for membership. It is reactive to indicators of the EU's performance, such as the strength of the economy, and so is relatively volatile (Boomgaarden et al., 2011;Down and Wilson, 2017). Affective support is input-oriented and based on more deep-seated attachments to the EU (Boomgaarden et al., 2011). Such attachments evolve more slowly and reflect one's fundamental beliefs and values about the EU as an institution. The failure of the EU to deliver a strong economy, for example, is less important than its adherence to the ideal of a worthy institution for this dimension of support (Down and Wilson, 2017).
The relationship between religious affiliation and support for EU integration is rooted primarily in the historic relationship between the Catholic and Protestant churches and national governments, as well as the values and beliefs regarding politics, religion and community held by various Christian denominations. This means that religion should have a stronger impact on affective support for integration. An individual's religious affiliation is expected to shape the level of their affective attachment towards the EU. In other words, the historic relationships between Christian churches and the EU and national governments, the perceived overlap between their values, and the EU's role in promoting or challenging religious and national identity are expected to shape the affective assessments individuals form regarding the legitimacy and worthiness of the EU as a political institution, as shown in H2: H2. Catholics have stronger affective attachment to the EU than Baptists, Methodists and Presbyterians, who in turn have stronger affective attachments than Anglicans.
On the other hand, religion is not expected to have a significant effect on utilitarian support. There is little reason to believe that members of different denominations are likely to assess the policy performance of the EU in areas such as promoting economic prosperity or reducing conflict differently because of their religious beliefs once the impact of their affective support for integration has been controlled for.
This provides H3: H3. There is no effect from religious affiliation on utilitarian support for EU membership.
Finally, representing the expectation that the effect of religious affiliation on affective support for integration is the cause of the relationship between religious affiliation and euroscepticism: H4. There are no significant differences between religious groups' support for 'leave' once affective assessments have been accounted for.

IV. Research Design
This study uses the BES Referendum Panel waves 7-9 (collected online between April and July 2016) to study the effect of affiliation with a Christian denomination on respondents' utilitarian and affective attitudes towards the EU, and subsequently on their support for Brexit. Support for Brexit is taken as a behavioural expression of euroscepticism, which is in turn expected to be largely (though not exclusively) caused by individual attitudes towards the EU and EU membership. The UK's referendum is currently the only instance in which a large sample from a range of Christian denominations were surveyed regarding their attitudes towards EU membership and integration. The BES Referendum Panel is the only source of panel data that allows causal relationships between religion and eurosceptic attitudes (and for the potentially delayed impact of causal factors to be accounted for) to be explored in such a context. It is not, of course, a typical circumstance in which to study attitudes towards EU integration, as the UK is a particularly eurosceptic member state (Down and Wilson, 2017;Nelsen and Guth, 2003), and the referendum was the first of its kind since that in Greenland in 1979. Furthermore, the campaign was particularly heated and dominated public debate, thus making the issue of EU membership unusually salient for voters. That said, these data can provide a reliable indication of the 'normal' relationship between religious affiliation and euroscepticism in the EU: i) There is no reason to believe that the campaign had a distinctive effect on religious citizensindeed, religion was all but absent from the public debate; ii) Studies of the correlates of support for Brexit and euroscepticism support the assumption that voting 'leave' in the referendum is a behavioural manifestation of euroscepticism (Clarke et al., 2017;Curtice, 2017); iii) The correlates of euroscepticismincluding religious affiliationare similar across member states (Boomgaarden et al., 2011;Nelsen and Guth, 2017), which makes the conclusions of this study to some extent applicable beyond the UK.The dataset includes a sample of 62,763 individual responses, including data about personal characteristics, political attitudes and behaviour, and support for EU membership. The main dependent variable is a dichotomous measure of whether the respondent supported 'leave' in the Referendum (wave 9 of the BES). The analyses also use two exogenous dependent variables measuring utilitarian and affective support for EU membership.
Utilitarian support is captured through a composite measure of responses to six questions regarding views of the political and policy successes of the EU (waves 7 and 8). They include respondents' agreement with the view that the EU has prevented war in Europe, made Britain more prosperous, brought people from different EU countries closer together, undermined the powers of the UK Parliament, created more red tape for business, and allowed too many members to join. These variables were recoded to be positively correlated, and latent structure analysis confirmed that they indicate the same latent construct. 1 The scores were combined into a composite indicator, in which higher scores imply a more positive utilitarian assessment of EU membership.
There are no direct measures in the BES of respondents' assessments of the legitimacy of the EU or of trust in it, which are typically used as indicators of affective attachment (Down and Wilson, 2017). Following Boomgaarden et al. (2011), we use a scale of self-identification as 'European' (with data collected in waves 7, 8 and 9). Like political trust, identification with political institutions is a stable trait (Down and Wilson, 2017). Self-identification as 'European' is indicative of a stable, positive attachment to the EU described by the affective dimensions of support. As expected of such a measure, the correlation between European identification (on a scale from 0 to 7, with 7 indicating 'very strong European') and support for Brexit is À0.69. Owing to the well-established argument that European and national identities tend to be positively correlated (Bruter, 2003), our measure captures the intensity of feeling 'European' independently from other facets of national or ethnic identity, thus, allowing them to vary independently and accounting for the possibility of multiple political identities.
The key independent variable is respondents' self-reported religious affiliation (measured in waves 7, 8 and 9). There were sufficient numbers identifying as Anglican (Church of England, Anglican or Episcopal), Roman Catholic, Presbyterian, Methodist and Baptist to allow reliable analyses of them as separate groups. The sample sizes for other groups (such as members of the United Reformed Church) were too small and, thus, were coded as 'Other Christians'. Respondents from non-Christian groups were also too few for reliable analysis and were coded as 'Non-Christians'. Those with no religious affiliation served as the reference category. 2 The research would have been aided further with a measure of 'religiosity', which would have allowed for a distinction between practicing and non-practicing Christians. Although it is not available in this BES dataset, this does not undermine the value of the analyses. The expectation is that by identifying as a 'member' of a particular religious denomination, respondents are likely to differ in their values, identity and history they exhibit, which will shape their attitudes towards the EU.
The analyses were conducted on (long) panel data, pooling observations from Waves 7-9 of the BES and stacking them by wave (Wooldridge, 2013). The first analysis used random-effect logistic (Gaussian) regression models with cluster robust inference on the dependent variable to estimate the effect of religious affiliation on the likelihood of voting 'leave' in the EU Referendum. The model controlled for characteristics associated with individual-level euroscepticism, including age, ethnicity, attention to politics, political ideology and party identification (Boomgaarden et al., 2011;Fox and Pearce, 2018;Franklin et al., 1995;Gabel, 1998;Hooghe and Marks, 2005;McLaren, 2002). 3 The second set of analyses explored the effect of religious affiliation on utilitarian and affective support for EU membership, using generalized least squares (GLM) random effects models with cluster robust inference, with data taken from all three waves and including the same controls. The final analysis explored the effect of religious affiliation on 'leave' voting while controlling for utilitarian and affective support and controls. The structure of the data and the nature of the independent variables, many of which are time-invariant, meant that the analyses were best conducted using random effects, as long as one can 'plausibly assume the unobserved effect is uncorrelated with all explanatory variables' (Wooldridge, 2013, pp. 474-478). The specifications of our data closely fit the assumptions for using random effects models (Becketti, 2013). Given the large sample size, the random effects estimators also have smaller standard errors than the corresponding pooled OLS estimators (Wooldridge, 2013).
Clustering applied to the quasi-time-demeaned equation is used to get robust standard errors and test statistics. Cluster robust inference was applied to account for serial correlation, which is common practice when the number of observations is much larger than the number of time periods (Becketti, 2013). Finally, the data used in the models were unweighted, firstly because the focus of the study is on assuring the internal validity of relationships between religious affiliation and euroscepticism rather than on generalizability, for which probability weights are less important. Secondly, such weights can unnecessarily over-complicate models Fieldhouse et al., 2016;Gelman, 2007).

V. Analysis and Results
Table 1 presents descriptive statistics showing the vote preferences of respondents within each religious group (and those of no religion) in the EU referendum.
The data show clear differences, with the most pro-EU by some margin those of no religious affiliation (as well as those of non-Christian religion), with 56 per cent voting 'remain'. Consistent with H1, the most pro-EU denomination were Catholics, 52 per cent of whom voted 'remain', and by far the strongest supporters of Brexit were Anglicans, 60 per cent of whom voted 'leave'. Presbyterians, Baptists and Methodists sat in between, more or less evenly split, though the differences between them and Catholics is small. Table 2 reports the analysis predicting support for Brexit for each religious group while controlling for factors associated with euroscepticism (Model I). 4 The control variables had the anticipated effects: older, working class, white non-graduates and those identifying with the eurosceptic United Kingdom Independence Party (UKIP) were more supportive of Brexit. Those living in Scotland or born in another member state were more likely to vote 'remain', though the effect is very small for the latter. Against the expectations of the literature, higher levels of political engagement (represented by attention to politics and interest in the Referendum) did not affect vote choice.
With these controls accounted for, there was a strong, statistically significant effect from identifying as Anglican on the probability of voting 'leave', making them the most eurosceptic group (supporting H1). While the effects from other denominations were not significant, the direction of the relationships was not entirely as expected. Contrary to the expectation that Protestants would be more likely to support 'leave' than Catholics and those with no religion, Baptists and Methodists were marginally more likely to support 'remain'. In addition, against the expectation that Catholics should be more pro-EU, they were more likely to support 'leave' than other groups except Anglicans. Table 3 demonstrates the effect of religious affiliation on affective (Model II) and utilitarian (Model III) support. Model II shows significant differences between the groups' affective attachments to the EU. Consistent with H2, Anglicans and Presbyterians had the weakest attachment, whereby their religious affiliation decreased their affective 4 Numerous formulations of the models were tested and assessed on the basis of fit statistics and theory to determine the most appropriate for testing the hypotheses. Different ways of modelling region were tested. For the most part, region had no significant effect once the other factors were controlled for, except for living in Scotland.  Source: BES, Wave 7-9. * effect statistically significant at 95% confidence level; † significant at 99% confidence level; ‡ significant at 99.9% confidence level. sentiment by coefficients of À0.21 and À 0.19, respectively (both significant). This supports the theory that the Protestant groups with the strongest attachments to a national identity (English and Scottish) are more eurosceptic. The effects for Methodists and Baptists were non-significant, but matched expectations in showing these groups as sitting between Anglicans and Catholics. While the coefficient for Catholics is non-significant, once the control for being born in another EU member state is removed from the model, the coefficient increases substantially and becomes significant. This suggests that it is the fact that most Catholics were born in other member states that is more important for explaining their opposition to Brexit than their religious belief, challenging the theory of H2 that Catholics would be the most affectively supportive group. This suggests that affiliation with the Catholic Church in Britain drives more negative assessments of the costs and benefits of EU membership and undermines affective attachment to the institution somewhat. Model III shows that, against the expectations of H3, affiliations with the Church of England, the Presbyterian Church of Scotland and the Catholic Church substantively decreased one's approval of the EU's policy performancewith coefficients of À0.41, À0.59 and À 0.32, respectively (all statistically significant). The coefficient for Methodists of À0.15 suggested they had a lower utilitarian support than the non-religious and Baptists, but was non-significant. The model suggests that the groups with the most utilitarian support for EU membership are Baptists, Methodists, 'other Christians', non-Christians and the non-religious. While Catholics were more positive in their assessments than Presbyterians and Anglicans, they were more sceptical than other Protestants, and considerably more so than those with no religious affiliation. A further surprising finding is that Presbyterians were more critical than Anglicans. The vast majority of Presbyterians live in Scotland, which was overwhelmingly opposed to Brexit during the Referendum, but since the model controls for living in Scotland, this is not a regional effect. Table 3 also contains important results relating to the effects of the control variables. Those with more economic resources and human capital (in the form of working in middle class occupations and having a degree) had higher utilitarian and affective support, likely reflecting their greater capacity to take advantage of the economic opportunities of EU membership and not to find themselves in competition with migrants for low skilled jobs (Gabel, 1998). There was no significant effect from gender, while those from minority ethnic backgrounds were significantly more supportive of EU membership on both dimensions. Identifying either as ideologically right-wing or with the anti-EU UKIP depressed both utilitarian and affective support substantially, and the opposite was true for identifying as ideologically left-wing or with the more pro-EU Labour and Liberal Democrat parties.
There are also some controls that had different effects on utilitarian and affective support. Age, while having a very weak effect, was significantly and negatively associated with utilitarian support but positively with affective support. The utilitarian coefficient is likely a reflection of the fact that older people are less likely to be in a position to take advantage of many of the opportunities of EU membership because of their commitments to work and family (Down and Wilson, 2017). It is unclear, however, why older people would exhibit stronger affective support for EU membership given that they are typically more eurosceptic (Fox and Pearce, 2018). A further important difference is apparent in the effect of political engagement; those with more interest in politics had higher utilitarian support but lower affective support, once their engagement with the referendum in particular was controlled for. Gabel (1998) and Hooghe and Marks (2005) suggest that people with greater political engagement should be more supportive of EU membership because of their greater familiarity with the role of the EU in domestic politics. A positive effect from political interest on utilitarian support makes sense, therefore, but this cannot explain the negative effect on affective support. Unpicking precisely what is being captured by this coefficient would require further research that is beyond the scope of this study, though it possibly reflects the persistently higher levels of euroscepticism exhibited by UK citizens and the failure of the engaged British electorate to develop an affective attachment to the European project in the way that citizens in other member states have.
Finally, Table 4 adds the utilitarian and affective measures of support to the referendum vote choice model and tests the expectation that the effect of religion would disappear once its impact on affective sentiments towards the EU were controlled for (H4).
H4 is modified as the result of the rejection of H3 (that is, the finding that religious affiliation affected utilitarian assessments of EU membership). It now stipulates that controlling for affective and utilitarian support for EU membership should account for the direct effect of religious affiliation on vote choice in Model IV.
Unsurprisingly, the results show that the more negative one's utilitarian and affective assessments of EU membership, the greater the probability of voting 'leave', with both measures having negative and highly significant effects. The effect of being Anglican was greatly reduced and became non-significant compared with that of Model I; the effect of being Presbyterian was reversed (with a non-significant coefficient of À0.3); and the effect of being Catholic remained non-significant but was reduced notably in magnitude.
For the most part, the data support the modified H4: once the relationships between religious affiliation and utilitarian and affective sentiments towards the EU are accounted Source: BES, Wave 7-9. * effect statistically significant at 95% confidence level; † significant at 99% confidence level; ‡ significant at 99.9% confidence level.
for, there is no effect from religious affiliation on vote choice in the Referendum. The exceptions are Baptists and Methodists: there was virtually no effect on their relative probability of supporting 'leave' once utilitarian and affective sentiments were controlled for, suggesting that their assessments of EU membership are driven by something that is not captured by these two dimensions of support.

Conclusion
This research contributes to the literature highlighting the importance of religious affiliation and values in shaping one's political attitudes and behaviourin this case with regard to EU membership and integration. It confirms the importance of religion for explaining support for Brexit in the UK's referendum through a causal model and using panel data. It also enhances our understanding of how religion affects support for the EU by assessing support for utilitarian and affective dimensions of euroscepticism. Finally, it challenges findings regarding the euroscepticism of Protestants by studying smaller Protestant groups in the UK in their own right rather than as part of a larger 'Protestant' bloc.
The study confirms that as elsewhere in Europe, the UK's Protestant groups with historically close relationships to national government or national identity, such as the Church of England and the Presbyterian Church of Scotland, are more eurosceptic than other Christians, non-Christians and the non-religious. This supports the theoretical expectation that the distinctive history and values of these Protestant groups has left them with a lasting propensity to be more hostile towards EU integration (that is, to have lower affective support for the EU) because of its impact on the sovereignty of national governments. Developing the existing literature further, however, this research has also shown that there are important differences between Protestant groups, with smaller groups that were persecuted by the national government and larger Protestant churches in the past being more supportive of EU membership. This challenges the assumption that Protestant churches throughout Europe have a shared euroscepticism that sets them apart from Catholics and the non-religious. Instead, the history and values of a specific Protestant church are vital in shaping their relationship with the EU and in differentiating between Protestant groups.
We also found that the particularly eurosceptic Protestants (Anglicans and Presbyterians) are likely to have low utilitarian support for EU integration as well. This goes against our expectations of the way religion should affect political support based on existing literature. There are three likely explanations for this. First, the depth of their affective hostility to the EU is so extensive that it influences their assessments of the EU's policy performance as well, in a comparable manner to the way having a strong partisan attachment to a party that is not in government can colour the way people interpret the policy performance of the government. Second, the distinction between affective and utilitarian support is more subtle than assumed, meaning that there is more of a link, or potentially a reciprocal relationship, between one's affective attachments to the EU and assessments of the costs and benefits of membership. Finally, another possibility is that the differences between Anglicans' and Presbyterians' policy priorities and those of other Protestants, Catholics and the non-religious are more substantial than expected. Consequently, it would be possible for the EU to meet the policy priorities of one group (such as Catholics) while failing to meet the priorities of the Anglicans or Presbyterians, resulting in lower utilitarian support from the latter.
Investigating this unexpected finding further would be a fruitful avenue of further research. In any case, the analysis confirms that most Protestants' hostility to EU integration and membership is born out of their weak affective attachment to the institution, and their negative assessments of its policy performance alongside the perceived costs of membership. These results also challenge the widely held view that British Catholics typically have stronger affective attachment to the EU than Protestants. The argument that the history and values of Catholicism make Catholics more likely to support the EU is plausible, however British Catholics have more negative assessments of the costs and benefits of EU membership and weaker affective attachments once the fact that many were born outside of the UK has been accounted for. This suggests either that there is something distinctive about British Catholicism that makes them less supportive of the EU (although it is difficult to consider what that might be), or that the positive relationship between Catholicism and support for EU integration in other European countries may well be a result of stronger public support for the EU amongst particular communities that happen to have large numbers of Catholics (such as supporters of the CPU in Germany), but that has little to do with their religious identification. Further research in other national contexts with large numbers of Catholics is required to solve this conundrum.
The study also reveals an important finding regarding the non-religious, namely that they tend to be more supportive of EU membership and integration than religious respondents (though the differences are not large). Given that age is controlled for in the models, this is not a function of the tendency of younger people to be both more supportive of EU membership and not to affiliate with religion. Given that this group includes a diverse range of respondents, it is difficult to pin down precisely what is causing this more pro-EU outlook. It may simply reflect the lack of the historical and value-based reasons to be sceptical of EU integration exhibited by some Protestants, or a trait associated with being non-religious that affects other attitudes correlated with support for EU membership that is not accounted for in the analyses, such as a distinct moral or ethical philosophy not rooted in religious belief. It is impossible to explain this tendency from these data, but this does highlight a further avenue of inquiry for explaining the relationship between religion and euroscepticism, as understanding how the consequences of lacking a religious identity are related to euroscepticism is surely as important as understanding the consequences of having such an identity.
The findings of this study have considerable implications for the future study of euroscepticism. First, it is clear that religious affiliation is an important driver of support for EU membership. Moreover, this research suggests that the effect is widespread, affecting both utilitarian assessments and affective sentiment. A clear limitation of this study is that the data are limited to the atypical case of the UK and during the referendum on EU membership. While this does not undermine the validity of these findings regarding the broader relationship between religion and euroscepticism, there is a need to corroborate them in other contexts, should the data become available. Such research could explore in more detail why religious affiliation affects both utilitarian and affective dimensions of support, and consider whether this reflects the history and values of Protestantism and Catholicism, or there is another explanation entirely that has yet to be considered in the literature. Second, the importance of accounting for the multi-dimensional nature of euroscepticism is made clear. It is apparent that affiliation with different religious groups, as well as age, political engagement and ideology, have different effects on utilitarian and affective support for European integration. If these different dimensions are not accounted for, these relationships will be obscured, producing potentially misleading conclusions about the religious, social, economic and political predictors of support for EU integration.
Finally, there are also implications from these findings for the broader understanding of how religion affects political attitudes. Specifically, it re-opens the discussion regarding the significance and empirical application of social and political cleavages. Classic cleavage theories emphasize the diminishing importance of religion in and growing secularization of politics, especially in Europe, resulting from the separation of church and state and increasing liberalization of social attitudes (Lipset and Rokkan, 1967;Norris and Inglehart, 2004). Our empirical findings, however, suggest that announcing that 'religion is dead' is somewhat premature, as there are significant effects from religious affiliation on contemporary attitudes, such as euroscepticism. Moreover, that effect is widespread, influencing both the way people identify with and attach to the EU, and the way that they assess the costs and benefits of membership and its policy performance. In other words, religion has a significant effect on the full range of factors voters take into account when reaching a decision about their approval (or not) of EU membership. While religious practice, belief and affiliation may be in decline, this does not mean that the influence of denominational, self-identified religious membership on socio-political attitudes and behaviour is declining or negligible, as the case of the UK's 2016 Referendum demonstrates.
This direction of research is important for both empirical and normative reasons. Empirically, there is very little clarity in the literature regarding the explanations as to why affiliation with a certain religious denomination influences political attitudes and behaviours. Most of the explanations for these effectsin this and previous research (see Clements, 2015;Nelsen and Guth, 2003;Tilley, 2015) draw on the analysis of the historical backgrounds, organizational structures, social and moral values of religious groups (see Coupland, 2004;Madeley, 2007;McLeod, 1999;). However, there is a lot more data on major Protestant groups and Catholics rather than minor groups such as Methodists and Baptists, even though they might be dominant in specific localities and capable of swinging a result of an electoral poll in some constituencies. A finer categorization of religious groups (and the inclusion of those of no religion) and a more nuanced approach to the analysis of the effect of religion building on the ever-increasing opportunities for data linkage and big data analysis will eventually enable us to develop a more detailed and precise explanatory models regarding the role of religion in politics. Normatively, that would enable us to improve theories of religious engagement, their applicability to new cases and forecasting potential by drawing on reliable and replicable empirical research.

Funding details
This study is based on research supported by the Wales Institute of Social & Economic Research, Data & Methods (WISERD). WISERD is a collaborative venture between the Universities of Aberystwyth, Bangor, Cardiff, South Wales and Swansea. The research that this publication relates to was undertaken through the WISERD/Civil Society Centre

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Appendix A1: Measuring utilitarian sentiment towards the EU Table A1: Correlation coefficients between assessments of EU performance Table A2: Mokken Scale Analysis results Appendix B: Religious composition of the sample and comparison with the 2011 Census Table B1: Religious affiliation in BES sample compared with 2011 census Appendix C: Justification for control variables