Introduction

The citizen’s tolerance towards ethnic and religious out-groups is important for the social cohesion of Europe’s pluralistic societies. Research on this topic is of interest to social scientists and the European public. Even more so, as the number of immigrants has increased steadily since the 1970s, the largest group among them being Muslims (Kettani 2010).

Religion in Europe has often been associated with intolerance. Dissent over religious values and -identities has long been a part of European national histories. The ethno-religious conflicts throughout the 1990s and 2000s in South-Eastern European countries, the former Yugoslavia and Georgia are well-known examples. But religious intolerance is not just an Eastern European phenomenon that can be explained away by a legacy of conflict. The recent political disputes, the prohibition of the Muslim veil in France, Belgium and other countries, persistent Islamophobic campaigns of the extreme right across Western Europe that often result in anti-Muslim violence show that the relationship between religion and tolerance is difficult in the West as well. Strabac and Listhaug demonstrated using European Values Study data that these tensions are not simply explained by September 11 and its aftermath, as anti-Muslim prejudice was already found to be an issue in Western Europe in 1999 (Strabac and Listhaug 2008).

What is often forgotten in the heat of public debates over the legitimacy of the religious and cultural ‘other’ is that religion is not necessarily detrimental to tolerance. Europe’s religious denominations, through their moral teachings of neighborly love have the potential to foster tolerance. The Bible (Mark 12:31), Catechism of the Catholic Church (Vaticana 2011) and the famous open letter to Pope Benedict XVI, signed by 138 Islamic leaders in answer to his Regensburg lecture (Anonymus 2007) inform us that tolerance of others is an essential teaching of both Christianity and Islam.

This paper examines relationships between three dimensions of individual religiosity and intolerance towards Muslims and immigrants among the majority populations of Europe’s non-Muslim majority countries. The analysis is based on European Values Study data (wave four 2010). Since the majority of Muslims in non-Muslim majority countries also have a migration background (either by being the descendents of immigrants or being immigrants themselves), it is relevant whether the respondent’s levels of intolerance is different towards Muslims than towards immigrants.

Furthermore, this study is interested in whether religiosity is differently related to the two targets of intolerance. Anti-Muslim attitudes were found to be highly correlated with intolerance towards ethnic out-groups (Ford 2008; Strabac and Listhaug 2008). One might thus expect their distributions and covariates to be similar. Nonetheless, mere correlation does not imply that the two intolerance measures capture the same underlying concept. Kalkan, Uslaner and Layman (Kalkan et al. 2009) consequently distinguish between intolerance towards ethnic and cultural out-groups in America. It is plausible that intolerance towards Muslims is based on a rejection of cultural values that are perceived as incompatible with those of the Christian majority, while intolerance towards immigrants could be more related to general perceptions of ethnic threat and thus be unrelated to religion.

The analysis presented in this paper compares levels of intolerance towards Muslims and immigrants across countries and explores whether religion is differently related to the two. Careful attention will be paid to differential effects of three dimensions of religiosity. Theory suggests that the believing-, belonging, and practice dimensions of religiosity could be differently related to attitudes (Stark and Glock 1968; Huber 2007). Yet the vast majority of empirical studies on Europe so far have concentrated on one or two measures of religiosity, mostly church attendance and religious membership, thereby omitting the multi-dimensionality of religion. This paper contributes to the existing knowledge by analyzing differential effects of religious believing, belonging and practice in European comparison. The key questions of the analysis are: to what extent does the religiosity of Europeans living in non-Muslim majority countries influence their propensity to dislike Muslims and immigrants? Are the relationships similar across the two targets of intolerance? Do measures of religious believing, belonging and attendance differ in their relationship with intolerance towards immigrants and Muslims?

Three Dimensions of Religion: Theory and Hypotheses

The scientific study of ethnic tolerance and its relationship with religion goes back to American researchers of the mid-1960s (Allport 1966; Glock and Stark 1966; Allport and Ross 1967; Glock and Stark 1969; Herek 1987). Some of these early studies have already outlined multidimensional concepts of religion: Stark and Glock distinguished between religious belief, practice, experience and knowledge (Stark and Glock 1968). They found religious practice to be positively related to intolerance towards various out-groups. Allport and Ross found extrinsic, but not intrinsic forms of religiosity to be positively related to racial intolerance (Allport and Ross 1967).

However, the majority of the contemporary literature on Europe conceptualizes religion as one-dimensional. This may explain the inconsistent results: Some observe a positive relationship between church attendance and intolerance towards ethnic minorities (Scheepers et al. 2002; Guiso et al. 2003). Others find a positive relationship between religion and anti-Muslim attitudes only in Eastern-, but not in Western Europe (Strabac and Listhaug 2008, 280), and a number of contributions report negative relationships: Meulemann and Billiet (2011) observe in their study on 25 European countries that church attendance has a negative effect on ethnic threat perceptions in most countries. Likewise Billiet and de Witte (2008) find that non-religious are more likely than religious Belgians to express racist attitudes and to vote for the extreme right, and Coenders and Scheepers (2003, 332–333), and Billiet (1995) observe that regular churchgoers hold less exclusionary attitudes towards ethnic minorities than non-regular and non-churchgoers.

Only few studies make a distinction between believing and religious practice: In their work on the American context Froese et al. (Froese and Bader 2008; Froese et al. 2008; Mencken et al. 2009) report that belief in a wrathful as opposed to a loving God has negative effects on tolerance. Regarding the European context two studies are noteworthy: Eisinga et al. (1999) observe that neither religious believing, nor church attendance are positively related to ethnic intolerance. Scheepers et al. (2002) on the other hand find church attendance to be positively, but doctrinal believing to be negatively related to ethnic prejudice. Both studies are now a decade old. The latter is to the author’s knowledge the only cross-national study of ethnic intolerance in Europe that systematically compares relationships with religious believing, belonging and attendance.

We pick up on this work by applying a three-dimensional concept of religion, as suggested in some of the literature, using the latest European data.

Following Stark, Glock and others (Stark and Glock 1968; Davie 1990; Olson and Warber 2008; Scheepers et al. 2002), we distinguish between a believing- (religious beliefs), a belonging- (denominational affiliation, church membership), and a practice (church attendance, religious participation)—dimension of religion.

The most widely used measure of religious practice is church attendance, an indicator of participation in the activities of a moral community (Stark and Bainbridge 1996) and of social capital (Putnam and Campbell 2010). According to Stark and Bainbridge (1996), being integrated in the moral community of a church has beneficial effects on pro-social values, as church members acquire a shared set of morals through social interactions with religious peers. Active church members are thus more likely than others to internalize religious teachings of neighborliness and tolerance. This argument is shared by Billiet (1995) who finds that the moral teachings religious people acquire through regular church attendance indeed foster tolerance. Following the moral community argument, church attendance may well have a positive effect on tolerance even if it is unaccompanied by belief. According to Putnam and Campbell (2010), participation in church has beneficial effects on pro-social attitudes of non-religious people because the morals of the religious spill over to their non-believing peers. From this point of view, involvement in church, even non-religious volunteering rather than religious believing is key to increased tolerance. One may thus expect church attendance independent of believing to be negatively related to ethnic and religious intolerance. We thus hypothesize:

H1

People who attend church regularly are less likely to be intolerant of Muslims and immigrants than non-regular and non-churchgoers.

However, we find it unconvincing that simply going to church, out of mere habit or social obligation should be more influential than believing. Private contemplation of one’s beliefs, on the other hand, may well be influential even without going to church. Socialization theory states that religious beliefs and values are largely transmitted through early socialization in the family (Acock and Bengtson 1978; Kelley and de Graaf 1997). They are reinforced through social interactions later in life but were mostly obtained in childhood rather than in church. Moreover, in many European countries only a minority attends church regularly but the majority of the population still holds religious beliefs. Thus the dimension of religious beliefs is likely crucial for attitudes like tolerance.

In their classic theory Allport and Ross (1967) defined intrinsic religiosity as being religious for the sake of religion itself while the extrinsically religious utilize religion for worldly rewards like recognition and prestige: ‘the extrinsically motivated person uses his religion, whereas the intrinsically motivated lives his religion’ (Allport and Ross 1967: 434). Allport and Ross found that the extrinsically motivated tend to be more prejudiced and racist while the intrinsically motivated, who focus more on their inner beliefs tend to be more tolerant (Allport and Ross 1967). Since believing is more inwardly oriented, it can be seen as a measure of intrinsic religiosity. Allport’s and Ross I/E –Religiosity scale has been widely used in psychological studies (Donahue 1985; Genia 1993; Kirkpatrick 1993; Tiliopoulos et al. 2006) mostly on the American context. However, there is still a decided lack of cross-national comparisons incorporating Allport’s and Ross’ theory, or considering the believing dimension of religion at all.

The believing dimension can be operationalized via different beliefs in a Higher Being. The EVS-data allow for a distinction between a traditional belief in a personal God and a more modern, fuzzy belief in a Spirit/Life Force. The EVS also contains a measure of individualized religiosity (‘I have my own way of connecting with the Divine’). While belief in a personal God accords with the traditional doctrines of Christianity, the two latter beliefs deviate from them and are more individualized and modern. Since all major religions in Europe promote values of brotherly love, care and tolerance (Anonymus 2007; Vaticana 2011), we expect non-fundamentalist beliefs in God to be negatively related to both targets of intolerance.

H2a

Belief in a Personal God and belief in a Spirit/Life Force are negatively related to both intolerance towards Muslims and immigrants.

However, modernization theory (Inglehart and Welzel 2005; Norris and Inglehart 2004) posits that modern, individualized and fuzzy religious beliefs which depart from traditional doctrines are associated with more tolerance towards various out-groups more than traditional belief. We thus hypothesize that:

H2b

Belief in a Spirit/Life Force is more strongly negatively related to intolerance of Muslims and immigrants than belief in a Personal God.

As argued above, religious beliefs are largely socialized in childhood in the family rather than through churchgoing later in life. Thus believing is likely to have stronger links to social attitudes like tolerance than church attendance:

H2c

Beliefs in God are more strongly negatively related to intolerance of Muslims and immigrants than church attendance.

When studying relationships between religion and intolerance it is crucial to distinguish non-fundamentalist from fundamentalist believing. Fundamentalism has been found to be an important predictor of ethnic intolerance (Glock and Stark 1966; Altemeyer and Hunsberger 1992; Wylie and Forest 1992; Eisinga et al. 1995; Laythe et al. 2002), political intolerance and homophobia (Laythe et al. 2002; Froese et al. 2008; Schwartz and Lindley 2009; Whitehead 2010; Eisenstein 2006). Fundamentalism is here defined as an exclusive truth-claim of one religion over others, expressed by the statement ‘there is only one true religion’.Footnote 1 This definition follows a convention based on prior literature (Kirkpatrick 1993; Leeming et al. 2010).Footnote 2 Based on the findings from the literature we hypothesize:

H3a

Religious fundamentalism is positively linked to both intolerance towards Muslims and immigrants.

Furthermore, since fundamentalism is an expression of closed-mindedness towards the truth claims of other religions (Altemeyer and Hunsberger 1992; Altemeyer 2003), we expect fundamentalist believers to be more intolerant towards religious, than ethnic out-groups.

H3b

Fundamentalist believing is more strongly positively related to intolerance towards Muslims than to intolerance towards immigrants.

Regarding the belonging dimension of religion, this research is interested in differences between members of Christian denominations and non-members. Denominational membership is an important marker of group-identity. It has been argued by identity theorists that the identification of an individual with a group takes place by delimiting the in-group from not accepted out-groups (Tajfel 1974; Kunovich and Hodson 1999; Seul 1999). This is likely to be true for people who identify with a denomination without being religious rather than the religiously devout members. We argued that across denominations the non-fundamentalist religious value Christian teachings of neighborly love and tolerance and are therefore less likely to be intolerant than the non-religious.

H4a

Members of different Christian denominations do not differ significantly in their levels of intolerance towards Muslims and immigrants.

It is, however, plausible that the so-called nominal Christians,—people who are not devout but identify with a denomination (Voas and Day 2010) may be more intolerant than their devout peers because they seek religious membership as a social identifier rather than for the sake of religion itself. If we find positive relationships between denominational belonging and ethnic/religious intolerance, it is therefore of interest whether devoutness is a moderator. If devout members are found to be more intolerant than non-devout and non-members we can assume that this is a true religiosity effect. If, however, belonging to a denomination is related to intolerance only for the non-devout, then this would point towards an association with religious belonging as an identity-marker,—not a true religiosity effect. This can be tested via interaction terms between denominational membership and church attendance.Footnote 3

H4b

If a positive relationship between religious membership and anti-Muslim/anti-immigrant intolerance occurs, the effect is driven by non-devout rather than the devout members.

The Influence of National Contexts

A number of contextual factors have been found to influence ethnic tolerance and therefore need to be controlled for: Modernization theory posits that a country’s level of wealth and political stability is positively related to its population’s inclination towards liberal values and tolerance (Inglehart and Welzel 2005; Inglehart and Welzel 2003; Norris and Inglehart 2004). Hence our models control for GDP and levels of political stability as measured by the World Bank (Kaufmann et al. 2009). Contact- and group-competition theories emphasize the import of the out-group’s size for the majority-population’s probability of tolerating them. Contact theorists (Pettigrew 1998, 2008; Schneider 2007; Wagner et al. 2006) argue that large numbers of immigrants enhance chances of inter-group contact and thus increase tolerance. Group competition theorists (Coenders and Scheepers 2003; Meuleman et al. 2009; Quillian 1995; Schlueter and Scheepers 2010) argue the opposite: the larger the number of immigrants, the more likely are members of the majority to perceive them as a threat. Since the most religious countries in Europe are also among the poorest and the least politically stable and are mostly sending-countries of migrants, it is necessary to control for these contextual factors in order to test our results for cross-national robustness.

Data and Methods

The analysis is carried out using data from the fourth wave of the European Values Study (EVS 2010). The EVS comprises 47 European countries and is therefore the survey with the most comprehensive coverage of Europe to date. The data was obtained using stratified and multistage random sampling (GESIS 2012). Since this study is interested in intolerance towards Muslims and immigrants among the Christian and unchurched majority populations of Europe’s historically Christian countries, we excluded all foreign-bornFootnote 4 respondents and Muslims from the analysis. Thus 4,595 foreign-born respondents and 2,501 Muslims living in Christian majority countriesFootnote 5 were excluded. In addition, the 5,990 respondents living in the Muslim majority countries Azerbaijan, Northern Cyprus, Kosovo and Turkey were also excluded from the analysis. Because of Eastern Germany’s communist past and the resulting cultural differences, Eastern and Western Germany are treated as separate entities throughout the analysis. Thus, 54,700 respondents in 44 countries remain in the analysis.

Dependent Variables

Intolerance towards immigrants is measured by an affirmative answer to the statement ‘I would not like as neighbors: immigrants/foreign workers’. Intolerance towards Muslims is measured by the statement: ‘I would not like as neighbors: Muslims’.

Independent Variables

Table 1 displays the descriptive statistics of the dependent and independent variables of the models.

Table 1 Descriptive statistics of the dependent and independent variables
Full size table

Devout religious practice is measured by frequency of church attendance (‘how often do you attend religious services?’). In addition, as a measure of civic engagement outside church, a dummy variable for volunteering in an organization was included in the models.Footnote 6

Religious believing is operationalized via three types of belief in God: traditional belief in a personal God, belief in a Spirit/Life Force as a fuzzy, modern form of belief (Voas 2009) that is often associated with modernization and religious individualization.Footnote 7 In addition, the statement ‘I have my own way of connecting with the divine’ is included as a measure of individualized religiosity that according to individualization theorists is typical for a new generation of increasingly religiously unattached, yet still spiritual people (Pollack and Pickel 2007; Voas 2009). Lastly, fundamentalism, operationalized as the belief that ‘there is only one true religion’ is included in the models.

Religious Belonging is measured by denominational affiliation. The three denominations, Catholic, Protestant and Orthodox are included in the models with unchurched (having no affiliation) as the reference category. The sample has 53 Jewish, 5 Hindu, 35 Buddhist, and 1,016 respondents who ticked the category “other” with no further specification. They comprise the category “other denomination” in the models.

Controls

The following individual-level control variables were included: education (dummy variable: respondent has tertiary education), whether the respondent has experienced long-term unemployment of three months or more, the respondent’s age,Footnote 8 sex (male as the reference category), anomy as expressed by the feeling of having no or very little control over one’s life. Anomy is known to be related to intolerance (Billiet 1995).

A number of studies have found authority-mindedness and particularly right-wing authoritarianism not only to be strongly correlated with intolerance (Johnson 1977; Whitley and Bernard 1999; McFarland 2010; Asbrock et al. 2011) but also to be potential mediators of relationships between religion and intolerance (Laythe et al. 2002; Rowatt and Franklin 2004; Tsang and Rowatt 2007). It is therefore important to control for right-wing authoritarianism. The models include a measure for being right wing on a political left–right scale (1–10) and a measure of authority-mindedness, the statement ‘having a strong leader who does not have to bother with parliament and elections would be a good thing for future society’.

Country-Level Controls : Steps 4 of the multilevel models includes the country-level controls GDP per capita (IMF 2007), the World Bank’s political stability index for 2008 (Kaufmann et al. 2009), and the percentage of foreign-born (aggregated from the EVS-data).

Strategy and Analyses

The analysis sets out to compare effects of religious believing, belonging and practice on intolerance towards Muslims and immigrants across Europe. A multilevel analysis is the appropriate approach (Snijders and Bosker 1999), as large between-country differences in the overall levels of intolerance and in the effects of religion are to be expected. Because both dependent variables are binary, logistic multilevel regressions are carried out using the software package STATA.

The full random intercept model has the following equation:

$$ \begin{aligned} Logit \left[ {\frac{pij}{1 - pij}} \right] & = \beta 0j cons + \beta 1Catholic + \beta 2Protestantij + \beta 3 Orthodoxij + \beta 4 other ij \\ & \quad + \beta 5Church\,Attendanceij + \beta 6 Volunteeringij + \beta 7Belief{:}Peronal\,Godij \\ & \quad + \beta 8 Belief{:}SpiritLife\,Fij + \beta 9 Fundamentalism + \beta 10 Individualized\,Rel.ij \\ & \quad + \beta 11 Educationij + \beta 12Unemploymentij + \beta 13 Ageij + \beta 14 Age2ij \\ & \quad + \beta 15 Femaleij + \beta 16Anomyij + \beta 17Right\,Wingij + \beta 17stromg\,Leaderij \\ & \quad + \beta 18GDPij \, + \beta 19\,\% \, foreign \hbox{-} born + \beta 20Political \, Stabilityij \, \\ \end{aligned} $$

All models were tested for outliers through careful normality—and residual checks and multicollinearity diagnostics were carried out for all variables that were included in each model. When ‘would not like immigrants’ is the outcome, Northern Ireland and Iceland showed to be potential outliers. Tests of the effect of each on the model coefficients indicated that they are not influential cases. However, when ‘would not like as neighbors: Muslims’ is the outcome, Iceland is an influential case. In order to control for this outlier without losing statistical power, a country-dummy was included in the models and the intercept set to zero for Iceland following the suggestions of Van der Meer, Grotenhuis and Pelzer (Van der Meer et al. 2010).

Treatment of Missing Values

The analysis has to deal with missing values: of the 54,700 respondents in the dataset, 7,807 have missing values on one or more variables when intolerance towards Muslims is the outcome. When intolerance towards immigrants is the outcome, 8,551 cases have missing values. Thus, a missing-data analysis testing for MNAR (Enders 2010: 13) was carried out. All models were then run in two sets, first as a complete case analysis and secondly after applying multiple imputation using chained equations in STATA mi. The imputation model includes income, subjective health and life-satisfaction as auxiliary variables. 40 imputed datasets were created and stepwise random intercepts models run across the imputed data. However, the differences in the estimates and their standard errors between the imputed models and the complete case analysis are very small, yielding no significant changes in magnitude or significance of the results of the analysis. We are therefore confident that the results of the complete case analysis are not biased by the missing data. Since multiple imputation made no difference to the results of the models and imputed multilevel models are disproportionally computationally intensive to run, we decided to present the complete case analysis.

Results

Intolerance towards Muslims and towards immigrants are correlated, but clearly do not measure the same underlying concept. Pearson’s phi correlation coefficient is 0.479 and Cronbach’s alpha is 0.650, thus the two indicators would not constitute a reliable scale. Across Europe, only a small majority (54 %) of those who say they dislike Muslims also say the same about immigrants.

A first comparison of the country percentages of respondents saying they would not like Muslims as neighbors with respondents who would not like immigrants indicates two things: first, intolerance towards Muslims is more prevalent across most of Europe than intolerance towards immigrants. This confirms Strabac and Listhaug’s earlier findings with older EVS data from 1999 (Strabac and Listhaug 2008). The statistical significance of the difference in the percentages between the two items was tested across countries using paired t tests. Table 2 (third row) shows significant differences between the two items in most countries. In most of Europe, except Albania, Hungary and Russia, immigrants are more accepted than Muslims. Secondly, the Scandinavian and Western European countries are the least intolerant of both Muslims and immigrants, while the South-Eastern and Eastern European countries exhibit the highest overall levels of intolerance in comparison.

Table 2 Percentages of the two dependent variables and the difference between them per country
Full size table

The bivariate relationships between the religion measures are as expected: all are moderately but not highly correlated. Spearman’s rho is 0.448 between church attendance and belief in a personal God, and −0.061 for belief in a Spirit/Life Force. Fundamentalism is correlated 0.297 with church attendance, 0.366 with belief in a personal God and −0.165 with belief in a Spirit/Life Force. Of the respondents who attend church at least once a month, 66 % believe in a personal God, 25 % in a Spirit/Life Force and 36 % say there is ‘only one true religion’. Of those, who believe in a personal God, only 49 % attend church regularly. Likewise only half of the respondents (50 %), who make the fundamentalist statement, attend church regularly. A single dimension of religiosity can therefore not be assumed, the different religiosity measures may well show different relationships with ethnic and religious intolerance (See Tables 3, 4).

Table 3 Intolerance towards Muslims, Random Intercept Models
Full size table
Table 4 Intolerance towards Immigrants, Random Intercept Models
Full size table

The Multilevel Models

We proceed with the results of the multilevel analysis. The models for ‘would not like Muslims’ and ‘would not like immigrants’ are interpreted next to each other. For both outcomes the same models were run stepwise. In a first step each religion variable was included on its own in a separate random intercept model in order to ensure that no religion effect is controlled away by other variables. These bivariate coefficients are provided in Table 5a and b in the appendix. The bivariate analysis already suggests that religious believing matters more for intolerance towards Muslims and immigrants than belonging or practice. Only the coefficients of belief in God and of the fundamentalist statement ‘there is only one true religion’ are statistically significant.

Table 5 The coefficients of the religion measures when included separately in the multilevel models
Full size table

In the next steps individual- and country-level controls are included in the models.

Table 3 presents the random intercept model for ‘would not like Muslims’. Model M1, in the first row shows the coefficients of the religion variables without controls. M 2 includes the individual-level controls and M 3 includes the individual-level and country level controls wealth (GDP), percentage of foreign-born per country, and political stability (Kaufmann et al. 2009).

Table 3 demonstrates that of the three dimensions of religiosity, believing clearly matters most for the European’s likelihood of being intolerant towards Muslims. The coefficients of non-fundamentalist believing are significantly negative, indicating that people who believe in a Higher Being (a personal God or a ‘Spirit/Life Force’) are less inclined than non-believers to say they would not want to live next door to a Muslim: taking the anti-log of the model coefficients reveals that when holding the other variables constant, believers in a personal God are 12 % and believers in a Spirit/Life Force 15 % less likely than non-believers to be intolerant of Muslims.

The findings thus support H2a. The coefficient of ‘Spirit/Life Force’ is noticeably larger than the coefficient of belief in a personal God, indicating that modern fuzzy believing is more conductive of tolerance than traditional belief, as modernization theory would predict. The finding supports H2b. Individualized religiosity (‘I have my own way of connecting with the divine’), too, is significantly negatively related to intolerance towards Muslims.

Fundamentalist believing on the other hand is strongly positively associated with anti-Muslim intolerance, as predicted by H3a: when controlling for the other variables, fundamentalists are 37 % more likely than non-fundamentalists to be intolerant towards Muslims. All relationships are robust when controlling for country-level wealth, political stability, and the percentage of foreign-born (M3).

However, church attendance is not statistically related to intolerance towards Muslims. The effect size is negligible even when tested on its own without including controls (“Appendix”). H1 is therefore not supported by the data with respect to Muslims, but H2c is: believing matters more for religious intolerance than church attendance. The moral community argument thus finds little support from the European data. Secular volunteering on the other hand is indeed negatively related to anti-Muslim intolerance, as predicted by social capital theory.

As to the dimension of religious belonging, Catholics, Protestants and Orthodox do not differ significantly in their levels of intolerance towards Muslims. H4a is therefore partly supported: religious belonging is not a predictor of intolerance towards Muslims in Europe. Only ‘other denomination’ shows a statistically significant coefficient and the relationship is strongly negative.

The result is not surprising. Since the respondents in this category are all members of religious minorities, who may well themselves feel discriminated against it makes sense for them to be compassionate of other religious minorities.

The controls show the expected relationships: women and highly educated people are less likely to be intolerant of Muslims than men and the lower educated. Individuals suffering from anomy, right-wing supporters and people who prefer a strong leader over a democracy are more likely to be intolerant. On the country-level, poverty and political instability are strong predictors of intolerance towards Muslims, while the proportion of foreign-born among the population is not statistically significant. The finding confirms prior findings by Strabac and Listhaug (Strabac and Listhaug 2008) and contradicts both contact theory and group-competition/group-size theory.Footnote 9

Most importantly, including the country-level controls does not change the coefficients of the religion measures. The effects of individual-level religiosity are robust across countries. As a further test for cross-country robustness, random slopes were fitted for each statistically significant religion coefficient, thus allowing the effect to vary across countries. The random slope coefficients for both outcomes are supplied in Table 6 in the appendix, together with visualizations of the country-slopes of religious believing (Figs. 1, 2). The random slope models demonstrate some small cross-country variation, but only in the size, not the direction of the effects. The relationships are therefore robust across the vast majority of countries.

Table 6 Random slope variances of each religion variable
Full size table
Fig. 1
figure 1

The random coefficient of belief in a personal God. The country abbreviations used in the EVS-2008-data follow the standard ISO 3166-1 alpha-2 codes

Full size image
Fig. 2
figure 2

The random coefficient of fundamentalism

Full size image

The analysis moves on to intolerance towards immigrants. Table 4 contains the results of the random intercept models for ‘would not like as neighbors: immigrants’. The models show very similar results to the models with intolerance towards Muslims as the outcome: the believing-dimension again shows the strongest relationships. The log-odds reveal that believers in a personal God are 9 % and believers in a Spirit/Life Force 16 % less likely than non-believers to dislike immigrants. Fundamentalists, unsurprisingly, are 28 % more likely than other people to be intolerant. Church attendance is not statistically significantly related to the outcome but secular volunteering is. Social capital indeed seems to benefit people’s ethnic and religious tolerance, but religious attendance does not provide a religious ‘booster effect’ as Putnam and Campbell (2010, 445) had theorized.

As to the belonging dimension, only the coefficient of Protestant denomination (Table 4) is statistically significant: Protestants are 12 % more likely than unchurched people to be intolerant towards immigrants. Since 85 % of the Protestants in the sample live in the wealthy immigration countries of Western Europe, it is important to control for national contexts. Model 3 shows that the relationship is robust when controlling for GDP, political stability and the percentage of foreign-born. The robustness of the effect is confirmed further when fitting a random coefficient, allowing the effect-size of Protestant denomination to vary across countries (appendix, Table 5, last row): no statistically significant between-country variation was found.

This finding is counterintuitive. The literature gives no reason to expect Protestants to be less tolerant than others. Moreover, the coefficient of Protestant denomination was not statistically significant in the uncontrolled model (Table 5b, appendix), but becomes strongly significant when holding the effect of religiosity (church attendance, believing) constant. Hence the question arises whether this is a true religiosity effect. To test this, an interaction term between Protestant membership and church attendance was included in the controlled model (Table 4, M4). Model M4 (Table 4, last column) shows that the interaction term is negative and statistically significant. As hypothesized in H4b, it is the non-devout Protestants who are more likely to be intolerant. The more a Protestant goes to church, the less likely is s/he to be intolerant towards immigrants. Congruent with identity theory it can therefore be said that the positive coefficient of Protestant denomination is due to non-religious Protestants, who use their membership as a social identifier. H4b is therefore confirmed by the data.

In summary, the results of the analyses lead to a rejection of hypothesis H1and support H2a, H2b, H2c, H3a, H3b and H4b. Believing matters more for both targets of intolerance than practice and belonging (H2c). Beliefs in God are negatively (H2a, H2b) and fundamentalist truth-claims positively related to both targets of intolerance (H3a). As hypothesized in H3b, fundamentalism is more strongly related to intolerance towards Muslims than immigrants (H3b). H4a is only partly confirmed by the data: most large denominations do not differ in their member’s propensity to dislike immigrants and Muslims, but members of religious minorities (‘other denomination’) are less intolerant towards both Muslims and immigrants than other people. With regard to immigrants, Protestants are the exception: the model coefficients in Table 3 show that when fixing the other variables, Protestants who do not practice their religion are 12 % more likely than non-religious people and 10 % more likely than Protestants who practice their religion to be intolerant towards immigrants. Thus H4b is supported by the data.

Discussion

The analysis has shown that more Europeans express intolerance towards Muslims than immigrants. The result is not surprising given the history of post- 9/11 and the persistent Islamophobia that has been reinforced through mass media for a decade. Nonetheless, as others have emphasized (Ford 2008; Strabac and Listhaug 2008), ethnic and religious intolerance are highly correlated and although the two indicators are not suited to capture one common scale of ethno-religious intolerance, our models demonstrated that both share the same predictors. Both intolerance towards Muslims and towards immigrants show largely the same patterns of relationships with religion, social-structural and contextual variables.

Our main finding is that of the three dimensions of religion believing matters most for the European citizen’s inclination to tolerate members of ethnic and religious out-groups. Both traditional and modern fuzzy beliefs in God were shown to be strongly negatively related to intolerance towards Muslims and immigrants. The models demonstrated that this is the case in the vast majority of countries independent of wealth, political stability and the percentage of immigrants. Non-fundamentalist religious believers indeed seem to internalize Christian moral teachings of neighborly love and this in turn seems to foster tolerance towards ethnic and religious out-groups. The finding confirms older findings by Scheepers, Gijsberts and Hello (Scheepers et al. 2002), who also found that doctrinal believing is negatively related to ethnic prejudice. Furthermore, the coefficient of belief in a Spirit/Life Force is larger than that of belief in a Personal God for both outcomes. This indicates congruent with modernization theory (Inglehart and Welzel 2005; Norris and Inglehart 2004) that modern, fuzzy believers tend to be more tolerant. However, in both sets of multilevel models the coefficient of ‘I have my own way of connecting with the divine’ is significantly smaller than the coefficients of the two God-beliefs. This might be due to the less clear-cut phrasing of the statement, which may have impacted measurement. Still, here too, the relationships are clearly negative for both outcomes.

As expected based on prior literature (Glock and Stark 1966; Altemeyer and Hunsberger 1992; Wylie and Forest 1992; Eisinga et al. 1995; Laythe et al. 2002), fundamentalism is strongly positively linked to intolerance towards Muslims and towards immigrants. Fundamentalism has the strongest coefficient of all believing measures. Its effect is even as large as the effects of education and right-wing authoritarianism, both traditionally strong predictors of intolerance. People who are closed-minded towards the truth-claims of other religions are indeed more likely to also be intolerant of their members and more likely to be intolerant of ethnic out-groups in general. The models also show that fundamentalism is more strongly related to intolerance towards Muslims than intolerance towards immigrants. The findings are robust across Europe.

Church attendance on its own is not statistically significantly related to intolerance towards Muslims and immigrants. Contrary to expectations based on social capital theory, being actively involved in church does not statistically significantly impact on religious and ethnic tolerance while volunteering in secular organizations does. Considering that church attendance rates have been steadily declining in most of Europe for many decades, this finding is perhaps not surprising. It adds support to prior observations by secularization-theorists (Pollack and Pickel 2007; Voas 2009): the traditional church has lost its significance as a social force in Europe. However, private intrinsic believing is still an important influence on civic attitudes.

As to the belonging dimension, Catholics, Orthodox and unchurched do not differ significantly in their propensity to tolerate Muslims and immigrants as neighbors. Only Protestants, who do not attend church regularly are significantly more likely than others to be intolerant towards the two out-groups. The finding can be understood in the light of identity theories (Kunovich and Hodson 1999; Tajfel 1974) and Allport’s and Ross’ (1967) theory of extrinsic religiosity: it is driven by non-devout Protestants who do not practice their religion, but utilize their church membership as an (extrinsic) identity marker against ethnic out-groups.

Why this secular Protestant identity effect is only observed with regard to immigrants, not Muslims remains a puzzle. If the story behind this relationship was a simple ‘us’ versus ‘them’ demarcation, we would expect to find the same relationship with regard to Muslims,—the main religious out-group in Europe. The data provide no easy explanation. A possible interpretation could be that members of religious denominations, even if they are not devout, tend to be more compassionate towards other religions than towards immigrants, because emphasizing a religious identity is a characteristic they have in common with Muslims. Immigrants, on the other hand, are an out-group they do not share an obvious common trait with. However, this still does not explain why Protestants in particular are more intolerant towards immigrants than Catholics or Orthodox. One plausible explanation may lie in the growing secularism within European Protestantism: European Protestants are on average less devout than members of other Christian denominations (only 18 % attend church regularly compared with 46 % of Catholics and 32 % of Orthodox). Congruent with our other findings,—the religious are less likely to be intolerant than the non-religious, it is plausible that denominations that are on average more religious, also tend to be more tolerant towards ethnic and religious out-groups. Another plausible explanation may be the historical tendency of Protestantism towards particularism. The history of Protestantism in Europe is largely a history of sectarianism while the Catholic and Orthodox Church have traditionally emphasized inclusiveness and a universality claim that aims at crossing borders. Thus intolerance towards others might generally be more strongly discouraged in Catholic and Orthodox pews than in Protestant sects. However a further testing of these possible explanations is beyond the scope of this paper. The Protestant finding merits further explorations in future case studies of Protestant, Catholic and Orthodox communities using in-depth questions on how the believer’s attitudes towards ethnic and religious out-groups are affected by religious convictions and contexts.

All in all, our findings show that when examining relationships between religion and attitudes like tolerance, it is important to distinguish between different dimensions of religiosity. Although our measurement of multi-dimensional religiosity is somewhat limited by the available data, our analyses come to a clear result: For both anti-Muslim- and anti-Immigrant intolerance, is the believing-dimension of religion that matters most: non-fundamentalist believing is clearly negatively and fundamentalist believing clearly positively related to both outcomes in the vast majority of countries.

Conclusion

This paper tried to examine to what extent individual religiosity is related to intolerance towards immigrants and Muslims, the main ethnic and religious out-groups in Europe. To this end, a distinction was made between a believing-, a belonging-, and a practice dimension of religiosity. The results show largely the same patterns of relationships for both outcomes: Religious believing matters greatly for European’s inclination to tolerate immigrants and Muslims. In the vast majority of countries and independent of country-level wealth, political stability and the number of immigrants, traditional monotheistic and fuzzy modern beliefs in God are strongly negatively related to both targets of intolerance, while the relationship between fundamentalism and intolerance is strongly positive. Non-fundamentalist religious believers, no matter whether they are traditional or modern, seem to contemplate the moral teachings of their religion and are therefore less likely to be intolerant towards cultural and ethnic out-groups than non-believers.

Across denominations and social strata in Europe, religious people are considerably less likely to be intolerant of ethnic and religious pout-groups than non-religious people. The results show that although the traditional churches have lost much of their social significance in Europe, private religious contemplation is still a strong social force influencing pro-social attitudes like tolerance towards ethnic and religious out-groups.