Interfaith marriage attitudes in Muslim majority countries: A multilevel approach

ABSTRACT

This research uses multilevel structural equation modeling to examine Muslims’ attitudes toward interfaith marriage with Christians in 22 countries with a Muslim majority population (N = 21,373). Attitudes toward interfaith marriage, for sons and daughters separately, were measured with single items, and three binary items were used to measure participants’ religious beliefs. Overall attitudes were negative and more negative toward marriage of one’s daughter compared to one’s son. Stronger religious belief was associated with more negative attitudes, but less so for Muslims who perceived more similarities than differences between Islam and Christianity. Perceived religious similarity was associated with more positive attitudes. The proportion of Christians in a country was not associated with interfaith marriage attitudes. However, the association between belief and attitude was found to differ considerably across countries.

There are various media reports on the persecution of Christians in Muslim majority countries, and commentators have argued that there are tensions, hostilities, and conflicts between Muslims and Christians in these countries (Smith, 2016). Antipathy and hostility toward non-Muslims is evident in many countries where Muslims are in the majority and in many regions where they are a sizeable community. Some argue that these tensions and conflicts are due to the presence of radical or extremist Muslims and that the opinions and attitudes of ordinary Muslims differ (Esposito & Mogahed, 2007).

Social scientists typically view intermarriage as the “last taboo” in intergroup relations (Qian, 2005), and intermarriage rates and attitudes are considered the best indicators of the degree of societal integration of groups (Blau, Beeker, & Fitzpatrick, 1984; Gaines, Clark, & Afful, 2015; Glenn, 1982; Kalmijn, 1991, 1998; Lehrer, 1998; Sherkat, 2004). In general, cultural homophily is one of the strongest individual determinants of marital choice (Kalmijn, 1998; Lehrer, 1998). Marital relations envelop nearly every aspect of an individual’s life, and similarity in religious values and beliefs make a marriage more likely. In the context of the United States it has been found that the majority favors religious homogamy but that people of the major Protestant traditions (33%) and Jews (35%) are more often married or living with a partner of another religion than Hindus (9%), Mormons (18%), Muslims (21%), and Catholics (25%; Pew Research Centre, 2015). Differences in secularization and (sectarian) religious beliefs are likely to be important factors for these religious group differences (Sherkat, 2004; Wuthnow, 1993). However, existing research on interfaith marriage tends to focus on marriage across very broad Christian and Jewish groups (i.e., Catholic, Protestant, and Jewish) and is predominantly conducted in the context of the United States.

Very little is known about interfaith marriage rates in Muslim majority countries or countries with a substantial proportion of Muslims. Furthermore, to our knowledge there is little research on Muslim’s attitudes towards religious outgroups and toward interfaith marriage with Christians more specifically (Sadi & Basit, 2013). One example is a cross-national study in Western Europe that found quite similar interfaith marriage attitudes (marriage with a Muslim, with a non-Muslim) among native majority and Muslim minority respondents (Carol, 2013). We examined the interfaith marriage attitudes in 22 Muslim majority countries and regions with a substantial Muslim population. Compared with intermarriage rates, attitudinal measures have the advantage that they do not depend on the availability of opportunities. We investigated separately individuals’ attitude toward marriage of their son with a Christian women and their daughter with a Christian man. These attitudes were examined in relation to religious belief and perceived commonality between Islam and Christianity and the percentage of Christians living in the country.

Interfaith marriage attitudes and religiosity

Many Islamic countries are facing the challenges of globalization and modernity, and interfaith marriage is one of the issues that is “currently pitting staunch traditionalists against modern reformists within the Muslim community” (Leeman, 2009, p. 743). However, whereas most Islamic scholars agree that Muslim men are allowed to marry Christian women because they are “women of the book,” the marital choice of Muslim women is typically restricted to men who convert to Islam (Leeman, 2009). This gender distinction would stem from the idea that men hold authority over their wives. So when a Muslim man marries a Christian, this will be less problematic than when a Muslim woman marries a Christian man. In the latter case, a non-Muslim would hold authority over his Muslim wife and Shari’a law forbids this (Leeman, 2009). Considering that Shari’a law is a guideline for many Muslims around the world (Aslan, 2005), it can be expected that attitudes toward interfaith marriage not only will be quite negative but also reflect this gendered distinction. Thus, we expected the following:

H1:

Muslims will be less negative about their Muslim son marrying a Christian than their daughter marrying a Christian.

More negative attitudes are more likely for Muslims who are more religious (Carol, 2013). Religiosity is a multidimensional concept including, for example, religious belonging, religious belief, religious practices, and the involvement in a religious community (Bloom, Arikan, & Courtemanche, 2015; Dezutter, Soenens, & Hutsebaut, 2006; Kellstedt, Green, Guth, & Smidt, 1996). Here we focus on religious belief and how this relates to the attitude toward interfaith marriage of one’s son and daughter. The main reason for this focus is that people with strong religious beliefs will be more likely to link marital choices to religious qualities and a family’s capacity to maintain and transfer religious doctrines (Sherkat, 2004). Religious belief is taking these beliefs to be true, and this is what matters to the believer and what he or she wants to share and transfer to the next generation (Crane, 2017).

The term religious belief is used here to refer to the main body of Islamic following whereby the Quran is considered the revealed message from God, and the Prophet Muhammad the primary exemplar of what it means to be a Muslim. In religious belief there is an emphasis on a single, unchangeable interpretation of basic doctrines that are binding for all believers. For strong believers, one’s religion is considered an unchangeable entity laid down in sacred texts with the Qu’ran as the final and perfect revelation of God (Aslan, 2005; Esposito & Mogahed, 2007).

Strong belief implies a rather rigid group distinction between one’s “true” belief and those that have different religious or secular beliefs who might challenge what Muslims believe to be true. Research on the “worldview-conflict proposition” demonstrates that dissimilar values, beliefs, and moralities between groups contribute to intolerance (Brandt & Van Tongeren, 2015). People seek to affirm the validity of their own beliefs and worldview and therefore express intolerance toward groups whose beliefs and worldviews are dissimilar to their own. This has been found among individuals high and low on measures of religious belief (Brandt & Van Tongeren, 2015), but the intolerance is stronger among those with stronger beliefs. Research on Muslims in Europe has found that religious belief was the strongest driving force behind negative and hostile attitudes toward non-Muslims (Koopmans, 2015). Furthermore, approval of interfaith marriage is closely tied to religiosity among Muslims, and non-Muslims, in Western Europe (Carol, 2013). Therefore, Muslims who hold more strong beliefs are more likely to have negative outgroup attitudes toward those who do not hold the same beliefs, such as Christians. So we expected the following:

H2:

Muslims with stronger religious beliefs to have more negative attitudes toward interfaith marriage with Christians.

However, the strength of this association might depend on the perceived commonality between Christianity and Islam.

Muslims and Christians are both “people of the book,” and both religions are considered a continuation of the tradition of the Abrahamic faith. Islam and Christianity have much in common, but there are also important differences (Aslan, 2005; Esposito & Mogahed, 2007). An emphasis on the commonalities rather than the differences implies that religious group boundaries are less salient and that there is a shared identity making “worldview defence” less likely (Brandt & Van Tongeren, 2015). Religious groups that emphasize religious commonalties are less likely to reject outgroup members for intermarrying (Lehrer, 1998; Lehrer & Chiswick, 1993). Research on the common ingroup identity model indicates that when individuals see themselves and others as part of a larger category (people of the book) rather than as members of two very distinct groups (Muslims and Christians) their feelings and behaviors toward those others tend to be more positive (see Gaertner & Dividio, 2000). This means that the following can be expected:

H3:

The negative association between religious belief and interfaith marriage attitudes is weaker for Muslims who perceive many communalities, rather than many differences, between Islam and Christianity.

Proportion of Christians in society

There are two competing propositions for the possible association between proportion of Christians in society and attitudes of Muslims toward interfaith marriage. First, based on conflict theories (see Esses, Jackson, & Bennett-AbuAyyash, 2010), it can be argued that a larger proportion of Christians in a country leads among Muslims to more feelings of competition over scarce material and symbolic resources. Outgroup size is often used as a proxy for competition between ingroup and outgroup members (e.g., Quillian, 1995; Savelkoul, Scheepers, Tolsma, & Hagendoorn, 2010). When there is a relatively large proportion of a minority group, the majority might feel that there is more to compete with, and this tends to lead to more negative attitudes toward these minorities (Scheepers, Gijsberts, & Coenders, 2002). This could mean the following:

H4a:

When the proportion of Christians in a country is higher, the attitudes toward interfaith marriage for sons and daughters with a Christian will be more negative.

The alternative proposition is that the opportunity structure is an important determinant for intergroup marriage (Kalmijn, 1998). When there is a relatively large proportion of outgroup members in society (Christians), there are more opportunities for meeting potential outgroup marriage partners. Furthermore, a larger proportion implies more opportunities for intergroup contact, and positive contact has been found to lead to more favorable outgroup attitudes (see Pettigrew & Tropp, 2016), also in conflict situations (Wright, Tropp, & Mazziotta, 2017). This is not only found in numerous studies that examine actual contact but also in research that use the opportunity structure as a proxy for intergroup contact (e.g., Schmid, Al Ramiah, & Hewstone, 2014; Wagner, Christ, Pettigrew, Stellmacher, & Wolf, 2006). This reasoning would mean the following:

H4b:

When the proportion of Christians in a country is higher, the attitude towards interfaith marriage will be more positive.

Method

Data

The current study uses the Pew data set “The World’s Muslims: Religion, Politics and Society.” These data were gathered between 2011 and 2012 among Muslims in countries across the world, and the measures were designed by the Pew Research Forum to measure the political and social views of Muslims in these countries. These data were collected during different periods, but on average the collection took place during 3 weeks in each country. This was done through face-to-face interviews (approximately 1,000 per country) in the language or dialect of choice of the participant.

For this research, 22 of the total 26 countries are included: Albania, Algeria, Azerbaijan, Bangladesh, Bosnia-Herzegovina, Egypt, Indonesia, Iraq, Jordan, Kazakhstan, Kosovo, Kyrgyzstan, Lebanon, Malaysia, Morocco, Pakistan, Palestinian Territories, Russia, Tajikistan, Tunisia, Turkey, and Uzbekistan. Afghanistan, Iran, and Thailand were excluded because the dependent variable was not measured in these countries. Niger was excluded because it was the only sub-Saharan Africa country and our focus was on the population of Muslims in the Middle-East and Eurasia.

In Bosnia-Herzegovina and Russia, Muslims participants were specifically targeted by focusing on regions with higher concentrations of Muslims. This was done to represent the Muslims in Muslim majority areas despite the countries not having a Muslim majority. In all countries the sampling was done by using a probability design. Firstly, a proportional stratification was made based on region and city. Then a proportion based on the population size was selected as the primary sampling units. A random selection was made for the secondary and tertiary sampling units within these proportional sampling units. In addition, the data were weighted by the Pew Research Forum. Overall, this data set is reasonable representative for the Muslim population in these 22 countries. The sample (N = 27,620) consists of 41% female participants, and the age varies between 18 and 97 years (M = 37.8, SD = 14.8).

Individual-level variables

Similar to other studies (e.g., Carol, 2013; Huijnk & Liefbroer, 2012; Huijnk, Verkuyten, & Coenders, 2013; Johnson & Jacobson, 2005), attitudes toward interfaith marriage of one’s daughter and one’s son were measured with single, straightforward items that do not require much cognitive effort (see Campbell & Herman, 2015). The following questions were asked: “How comfortable would you be if a daughter of yours someday married a Christian?” and “How comfortable would you be if a son of yours someday married a Christian?” The response categories were provided on a 5-point scale and were recoded as follows: 1 (not at all comfortable, 2 (not too comfortable), 3 (depends on situation), 4 (somewhat comfortable), and 5 (very comfortable). A higher score indicates a more positive attitude toward interfaith marriage. Individuals who responded that they don’t know or refused to answer were recoded as missing, as was done for all the measures used (see Table 1).

Table 1. Descriptive statistics of the main and control variables.

Variables	N (Total)	Range	M/Proportion	SD	Missing
Individual-level variables
Attitude for daughter	27,620	1–5	1.75	1.23	4.0%
Attitude for son	27,620	1–5	2.13	1.37	4.3%
Religious belief	27,620	0–1	.73	0.31	2.6%
Commonality	27,620	0/1	.33	—	17.5%
Country-level variable
Proportion of Christians	22	0.1–73.3	9.97	—	—
Control variables individual level
Rural	27,620	0/1	.46	—	—
Female	27,620	0/1	.51	—	—
Age	27,620	18–99	37.97	14.80	—

Religious belief was measured by three binary items. Respondents were given two opposing statements and were asked to choose the one that best represents their view. Each item was recoded to reflect less strong beliefs (0) as opposed to more strong beliefs (1). This resulted in the following coding: “Many religions can lead to eternal life in heaven” versus “Islam is the one true faith”; “There is more than one way to interpret the teachings of my religion” versus “There is only one way to interpret the teachings of my religion”; “Shari’a is developed by men, based on the word of God” versus “Shari’a law is the revealed word of God.” Individuals who responded that both were equally accurate or neither was accurate were treated as missing for these items (see Appendix A).¹

Perceived commonality was measured using the following item: “From what you know, do you think that the Muslim religion and the Christian religion have a lot in common, or do you think that the Muslim religion and the Christian religion are very different?” The binary response categories were “They are very different” (0) and “They have a lot in common” (1). In total, 32% of the participants indicated that Islam and Christianity have a lot in common, and around two thirds of the Muslims perceived little overlap between both religions.

Country-level variable

Proportion of Christians was derived from the Religious Diversity Index created by the Pew Research Forum in 2010 (see also Hackett & Grim, 2011). The information on the proportion of Christians in each country was used from this index. On average, there were about 10% Christians living in the 22 countries, but the range of this variable is very large, with some countries (e.g., Morocco; 0.1%) having a very low percentage of Christians and other countries (e.g., Russia; 73.3%) having a large percentage of Christians (see Appendix B).

Control measures

Control variables were included only for the individual level. Three demographic variables were included: gender (0 = male, 1 = female), age (in years), and whether individuals live in urban (0) or rural (1) settings.²

Method of analysis

To be able to compare the different models during multilevel structural equation modeling, participants with missing values were excluded using the procedure of listwise deletion. This resulted in a smaller sample (n = 21,373; 22.6% smaller than the initial sample). The missing values were excluded in MPlus, which did not allow for inspection of patterns of missing data. However, most of the missing values were due to participants not answering the question of perceived commonality between Islam and Christianity. The reason for this is unclear, but it might be because they did not know or refused to answer this question.³ The control variables and country-level variable did not contain any missing values. Further, treating the binary items for religious belief as a latent factor proved problematic. The model fit statistics indicated that the model was not identified due to a larger number of parameters than clusters, and this lead to an error. Thus, for the analyses, a mean score was created for the three items of religious belief, and this score was treated as an observed variable. Third, multilevel structural equation modeling was conducted using MPlus 7.3 and using the ML estimator. This method allows for simultaneously testing of direct effects and moderation effects at both the individual and country level (Hox, 2002). The analyses were conducted simultaneously for the two dependent variables, namely, attitudes toward interfaith marriage for sons and attitudes toward interfaith marriage for daughters. Across all models the residual variance of the dependent variables was allowed to correlate.

Results

Descriptive findings

As a first step, we examined whether the attitudes toward interfaith marriage of daughters and sons had sufficient variation. Although a majority of participants indicated to be “not at all in favor of interfaith marriage with a Christian” (63.1% for daughters, 47.3% for sons), others did not share this view. For daughters, 13.6% responded with “very” or “somewhat” comfortable, 6.2% with “depends on the situation,” and 13.2% with “not too comfortable.” For sons, these percentages were 21.3%, 8.5%, and 18.6%, respectively. This indicates that not all Muslims felt (very) negative toward interfaith marriage and that it is possible to examine the associations as proposed. The correlation between the attitudes toward marriage of daughters and sons was r = .71, which indicates that 49.6% of the variance in attitudes is not shared. This means that it is useful to use attitudes toward interfaith marriage for sons and attitudes toward interfaith marriage for daughters as two distinct dependent variables. Furthermore, as expected, there was a difference between the means of both attitudes with participants having more negative attitudes toward interfaith marriage of a daughter compared to a son, Wald χ²(1) = 1074.963, p < .001.

The attitudes toward interfaith marriage for daughters and sons were negatively related to religious belief (r = −.23, and r = −.18) and positively to perceived religious commonality (r = .14 and r = .17). Further, the negative correlation between religious belief and perceived commonality (r = −.15) indicates that those with stronger beliefs perceived less similarity between Islam and Christianity. Lastly, the correlations for the control variables were all < .052.

Explaining attitudes toward interfaith marriage for daughters and sons

Multilevel structural equation models were tested with the individual-level control variables included.⁴ The first model contained the individual-level variables and included both the direct effect of perceived commonality on attitudes toward interfaith marriage for daughters and sons and the interaction effect of religious belief and perceived commonality (Model 1, Table 2). This model was just-identified, and therefore the absolute fit of the model could not be assessed. However, the results of such a model can be interpreted and discussed (Muthen, 2012). The main effect of religious belief on attitudes toward interfaith marriage for both daughters and sons was significant and negative. This indicates that having stronger religious beliefs was associated with less positive attitudes toward interfaith marriage. It was also tested whether religious beliefs differed significantly in its association with the attitude toward daughters and sons, and this proved not to be the case, Wald χ²(1) = 0.213, p = .645. Further, the interaction effect of religious belief and perceived commonality was negative and significant for both attitudes. This indicates that individuals who had more strong beliefs were more negative toward interfaith marriage for both daughters and sons, but less so when they perceived Islam and Christianity to have a lot in common (b_daughters = −0.454, b_sons = −0.425), compared to individuals who considered Christianity and Islam to be very different (b_daughters = −0.695, b_sons = −0.610). Also, older compared to younger individuals had more negative attitudes toward interfaith marriage, and individuals living in rural areas had more negative attitudes toward interfaith marriage than those living in urban areas. Gender was found to have no effect on attitudes toward interfaith marriage.

Table 2. Multilevel models predicting attitudes towards interfaith marriage for daughters and sons.

	Model 1 Fixed effects individual level including moderation^a		Model 2 Fixed effects individual and country level^a
	Attitude for daughters	Attitude for sons	Attitude for daughters	Attitude for sons
	B (SE)	B (SE)	B (SE)	B (SE)
Fixed effects
Individual level
Intercept	2.199 (.183)***	2.504 (.148)***	2.059 (.262)***	2.392 (.232)***
Main predictor
Religious belief	−0.454 (.087)***	−0.425 (.065)***	−0.453 (.031)***	−0.424 (.037)***
Control variables
Rural	−0.113 (.030)***	−0.095 (.027)***	−0.113 (.016)***	−0.096 (.018)***
Female	0.007 (.021)	−0.021 (.029)	0.007 (.015)	−0.021 (.017)
Age	−0.002 (.001)*	−0002 (.001)*	−0.002 (.000)***	−0.002 (.001)***
Moderation
Perceived commonality	0.381 (.077)***	0.463 (.075)***	0.380 (.038)***	0.463 (.045)***
Belief × Commonality	−0.239 (.080)**	−0.182 (.088)*	−0.239 (.049)***	−0.181 (.057)**
Country level
Proportion of Christians			0.011 (.012)	0.009 (.010)
Variance components
σ^2 individual level	1.143 (.141)***	1.577 (.098)***	1.142 (.011)***	1.576 (.015)***
σ^2 country level	0.337 (.193)	0.253 (.148)	1.043 (.097)***	0.808 (.075)***
Covariance individual level between daughter and son	0.896 (.115)***		0.895 (.011)***
Covariance country level between daughter and son	0.271 (.170)		0.895 (.011)***
Fit statistics
Likelihood (df)	−60,758.188 (0)		−60,762.875 (1)
Deviance	121,516.376		121,525.750
AIC	11,556.375		121,557.197
CFI	1.000		0.999
RMSEA	0.000		0.023

Note. The independent variable (religious belief) is based on an observed mean score. Unstandardized coefficients are reported. The comparative fit index (CFI) and root mean square error of approximation (RMSEA) model fit statistics are included in the table but not discussed for Model 1 because it is just-identified and the fit statistics cannot be used to assess absolute model fit (Kenny, 2015). AIC = Akaike information criterion.

^a N = 21,373.

*p < .05; **p < .01; ***p < .001.

Overall, this model accounted for 2.7% of the explained variance for attitudes toward interfaith marriage for daughters on the individual level and 2.8% of the explained variance for attitudes toward interfaith marriage for sons. On the country level, the explained variance was relatively high (19.8% for attitudes toward interfaith marriage for daughters and 21.7% for attitudes toward interfaith marriage for sons). This suggests that there are substantial country-level differences, and this could not be accounted for by differences in religious belief, perceived commonality, and the control variables included in this analysis.

In the analysis the country-level predictor of proportion of Christians was added in Model 2. This second model fit the data slightly worse (Akaike information criterion = 121,557.197) in comparison to the previous model. However, we continued to analyze the role of the proportion of Christians, as the absolute fit of this model was good (comparative fit index = .999, = .023). For both attitudes the effects of the individual-level predictors remained the same. On the country level, no significant relationship was found between the proportion of Christians and attitudes toward interfaith marriage for both a daughter and son.

Additional analyses

We examined whether the associations between religious belief and attitudes toward interfaith marriage were constant across countries. Therefore, additional models were run including random slopes for religious belief (see Appendix C). This had to be done separately for attitudes toward daughters and attitudes towards sons.⁵

The significance of both the random slopes for religious belief indicated that the association between religious belief and attitudes toward interfaith marriage for daughters and sons varied across countries, as was suggested by the high level of explained variance at the country level discussed earlier. The average effect of religious belief on attitudes toward interfaith marriage for daughters (b = −0.433) was predicted to fall between −1.284 and 0.418 in 95% of the countries. The effect of religious belief on attitudes toward interfaith marriage for sons (b = −0.411) was predicted to fall between −1.504 and 0.682 in 95% of the countries. These ranges indicate that, although the average effect of religious belief on attitudes toward interfaith marriage for daughters and sons was negative, the effect of religious belief was more negative in some countries and even positive in others. To try to explain the variation between countries, a model was run including a cross-level interaction with the proportion of Christians (Appendix C). The cross-level interactions for both attitudes were significant, and part of the variance between countries of the association between religious belief and attitudes toward interfaith marriage could be explained by the proportion of Christians in the country.

Discussion

The rate of interfaith marriage and individual interfaith marriage attitudes are important indicators of hostile or more peaceful relations between religious groups (Glenn, 1982; Kalmijn, 1991; Lehrer, 1998; Sherkat, 2004). Research on interfaith marriage is mainly conducted in Western countries and among Christians. The current study examined Muslims’ attitudes toward marriage with Christians across 22 Middle Eastern and Eurasian countries. Further, the respondents were asked about their feelings toward their daughter marrying a Christian husband and their son marrying a Christian wife, rather than about their general intermarriage attitudes that are more sensitive to social desirable responding (Campbell & Herman, 2015).

Overall, Muslims were quite negative about these sort of marriages and more so for daughters compared to sons. This indicates that there are relatively strong religious group boundaries between Muslims and Christians in these countries. However, a substantial number of Muslims did not have negative attitudes toward these interfaith marriages, which means that there were important individuals differences in attitudes. To explain these differences, we examined religious beliefs and perceived commonality between Islam and Christianity. Individuals with more strong beliefs were found to be more negative about both daughters and sons marrying a Christian. This finding is in line with research on intergroup relations among Muslim migrants in Europe that demonstrates that stronger religious belief is associated with more negative intermarriage attitudes (Carol, 2013), and outgroup attitudes in general (Koopmans, 2015). The association between religious belief and interfaith marriage attitude did not differ for a daughter or a son. This indicates that a more strong belief has a more general impact on attitudes toward interfaith marriage.

In addition, we found that when Muslims perceive Islam and Christianity to have many commonalities (compared to differences), they were less negative about interfaith marriage for daughters and sons. Furthermore, the perception of commonalities made the association between religious belief and interfaith marriage attitudes less strong. These findings suggest that the perception of commonalities and belief similarities make group boundaries less salient and indicate a shared category (people of the book) that can improve outgroup attitudes (Gaertner & Dovidio, 2000).

We found no association between the proportion of Christians in the country and interfaith marriage attitudes. Thus, we found no support for either the intergroup competition proposition (Scheepers et al., 2002) or the intergroup contact proposition (Pettigrew & Tropp, 2006). One possible reason is that both processes of competition and of contact are relevant at the same time and therefore statistically cancel each other out. Furthermore, research has found that it is not so much the actual percentage of an outgroup that leads to negative outgroup attitudes but rather the perceived percentage (Wagner et al., 2006). Another possible reason is that the proportion of Christians is a too general proxy for either competition or contact. It does not capture whether there is, for example, actual segregation between Christians and Muslims. Focusing on smaller geographical areas—for example, specific areas instead of the country as a whole—might yield different results (Wagner et al., 2006).

However, further analyses showed that there were significant differences between countries with regard to the association between religious belief and attitudes toward interfaith marriage. A possible reason is that countries differ in terms of their interpretation and implementation of Shari’a law (Otto, 2008). This could influence the degree to which interfaith marriage is seen as acceptable, which in turn could influence the attitudes that Muslims hold with regard to interfaith marriage. Unfortunately, we were not able to take these country differences systematically into account because there is no reliable source on how different countries have incorporated Shari’a law in their legal system and everyday lives of its citizens. Also, the historical, political, and economic relations between Christians and Muslims in the different countries could not be taken into account because there were no reliable indicators available for all countries involved.

There are several other limitations to the current analysis. First, the attitudes toward interfaith marriage and perceived commonality were measured with single items, which raises questions about the reliability and validity of these measures. However, the use of these single items is common in the research literature (e.g., Carol, 2013; Huijnk & Liefbroer, 2012; Johnson & Jacobson, 2005) and the items focus on personalized (own son or daughter) rather than general acceptance of interfaith marriage. Furthermore, a study comparing different ways of measuring intermarriage concluded that “we must ask questions that are straightforward and do not require excessive cognitive effort” (Campbell & Herman, 2015, p. 727). This is especially important in cross-national research and in non-Western contexts in which the level of education of respondents can be quite low.

Second, we were not able to control for social class or education in the analyses (see note 2). This could mean that part of the findings is related to educational differences. In the context of the United States, educational attainment has been found to be associated with lower religious belief and increased likelihood of interfaith marriage (Kalmijn, 1998; Sherkat, 2004). However, educational factors do not seem to have an impact on intermarriage in, for example, Ireland (O’Leary, 2001), which suggests that the religious, historical, and educational context is important. Religious factors channel educational attainment both in terms of quality and quantity. In most Muslim countries there is a school system that serves as both an educational institution for Islamic education and a market of homogenous marriages. This makes it less likely that educational attainment has a strong influence on people’s attitudes toward interfaith marriage. Moreover, family, relatives, and friends influence marital choices through approval and disapproval. Research in the Western world has demonstrated that the opinions and actions of parents, siblings, and other kin can have a strong impact on individual attitudes toward interfaith marriage, independent of educational attainment and income (Huijnk, Verkuyten, & Coenders, 2010; Huijnk et al., 2013; Kalmijn, 1998). Such an impact is also very likely in non-Western Muslim majority countries in which people strongly value religious community life, religious socialization, and family integrity (Pew Research Centre, 2013).

Third, for measuring religious belief we could use only three binary statements, which is not ideal. This might be a reason for why the correlation between religious belief and the attitudes toward interfaith marriage were rather low. A more extensive set of Likert-type questions could result in stronger but, perhaps, also weaker associations. Using a more extensive measure would provide us with a better measurement of religious belief and, with that, a more reliable and better understanding of how this is associated with interfaith marriage attitudes.

Future studies should examine the role of religious belief further by using a more extensive measure. Future research could also consider the role of other dimensions of religiosity, such as religious belonging, religious behavior, and the involvement in one’s religious community (Bloom et al., 2015; Dezutter et al., 2006; Kellstedt et al., 1996). It might be the case, for example, that Muslims’ attitudes toward interfaith marriage are associated with their involvement in the Muslim community or with the degree to which they practice their religion. For example, research shows that individuals who regularly partake in religious practices tend to be less accepting of people of another religion than those who are not involved in such practices (e.g., Beatty & Walter, 1984; Sullivan, Piereson, & Marcus, 1982). The more emphasis a religion places on strict religious rules and practices, the less accepting the followers will be toward groups that do not follow these rules and practices. Islam is a religion with a strong emphasis on orthopraxis, presenting precise rules, practices, and guidelines for living in accordance with the will of Allah. These rules and practices are the most visible markers of Islamic faith, and higher involvement in these practices might form a strong barrier against interfaith marriage.⁶

In conclusion, the aim of the present research was to make a contribution to our understanding of the relations between Muslims and Christians in Muslim majority countries. To our knowledge, our study is the first to systematically examine attitudes toward interfaith marriage among Muslims across the world. Relations between Muslims and Christians have become increasingly polarized in many countries, and there is a clear need for serious interfaith engagement. We found that Muslims were negative toward marriages with Christians and more so for their daughter than their son. However, there also was a substantial number of Muslims who had a more positive attitude toward interfaith marriage. Stronger religious belief was associated with more negative attitudes but not in all countries examined. In some countries, stronger belief was not associated with interfaith marriage attitudes, and in other countries even had a positive association. This suggests that it depends on the national context whether and how religious belief is associated with interfaith marriage attitudes. Future research could examine country characteristics that are responsible for these differences. Furthermore, the association between religious belief and attitudes toward interfaith marriage was less strong when people perceived communalities rather than differences between Islam and Christianity. The perception of communalities was also associated with less negative interfaith marriage attitudes. This suggests that it is useful to emphasize similarities between religions for improving interreligious attitudes. This might be particularly important in our current times in which many Islamic countries and communities are facing the challenges of modernity and social change.

Notes

¹ An additional sensitivity analysis was conducted to test if the coding of the religious belief items affected the findings. The category of neither/both, equally/depends was recoded (0.5) between the extremes of less strong beliefs (0) and strong beliefs (1). This recoding did not led to a different pattern of findings.

² We also tried to include level of education as a control variable. However, in the Pew data set, each country had a different coding for education and many of the categories that were included were not in the International Standard Classification of Education scale, making a comprehensive and comparable education variable problematic.

³ Additional analysis without this variable with a larger sample yielded similar findings for the other predictors.

⁴ In the main analyses, the number of children that one has was not controlled for. This is because this variable contained many missing values and the sample was significantly reduced when including this (N = 15,609). However, in an additional analysis this variable was included as a control (continuous variable ranging between 1 and 14) and was found to have a significant effect on attitudes toward interfaith marriage for daughters and a marginally significant effect on attitudes toward interfaith marriage for sons. Individuals with more children were somewhat less negative toward interfaith marriage for daughters and sons. Important to note, in this additional analysis the findings for the other variables were similar as presented.

⁵ In the model including random slopes, testing both dependent variables simultaneously resulted in a larger amount of parameters than clusters. This led to a nonidentification of the model. Therefore, it was decided to run separate models for attitudes toward interfaith marriage for daughters and attitudes toward interfaith marriage for sons.

⁶ Although the Pew survey contained many questions, it did not contain extensive questions regarding religious belonging, religious behavior, and religious belief. For example, only one item was used to measure religious belonging, and the three questions on religious behavior were related only to prayer. In a preliminary analysis we nevertheless examined whether it was statistically possible to make a distinction between religious belief, belonging and behavior as three separate aspects of religiosity. A three-factor confirmatory factor analysis was run using the robust weighted least squares and variance adjusted estimator, as many of the items were categorical. The model fit of this three-factor model was very poor, χ²(df) = 359,136.667 (df = 13), comparative fit index = 0.000, Tucker–Lewis index = −34.869, root mean square error of approximation = 1.000. The items did not clearly load on different factors, and only the three items for religious belief loaded on a single factor, indicating that it formed the only clear distinct and reliable construct to use in our cross-national analysis. Moreover, multilevel analyses with other measures of religiosity (e.g., religious belonging) did not show different results. Furthermore, a confirmatory factor analysis was run for the three items of religious belief, and the reliability of the single factor for religious belief was ρ = 0.60. This is rather modest, but with this data set these items were the best measures available.

APPENDIX A

Table A1. Percentage per category for the three religious belief items.

Items for religious beliefs	Less strong beliefs (0)	Strong beliefs (1)	Neither/both equally/depends (999)	Don’t know (999)	Refused (999)
Belief item 1	13.8%	77.9%	3.7%	4.0%	0.6%
Belief item 2	27.5%	60.1%	3.2%	7.9%	1.3%
Belief item 3	25.8%	58.8%	6%	8.4%	1.0%

APPENDIX B

Table B1. Descriptive statistics for the countries separately.

		Attitude for daughters	Attitude for sons	Religious belief
	N	M (SD)	M (SD)	M (SD)	Perceived commonality: Proportion	% of Christians
Albania	525	4.00 (1.25)	4.08 (1.18)	.55 (0.44)	.36	18.0
Algeria	1,011	2.063(1.36)	2.266(1.47)	.72 (0.32)	.34	0.2
Azerbaijan	718	1.31(0.77)	1.70(1.08)	.69 (0.31)	.19	3.0
Bangladesh	1,555	1.88 (1.20)	2.02 (1.27)	.79 (0.26)	.19	0.2
Bosnia and Herzegovina	837	1.91 (1.15)	2.03 (1.17)	.67 (0.32)	.61	52.3
Egypt	1,538	1.10 (0.39)	2.15 (1.17)	.85 (0.24)	.34	5.1
Indonesia	1,814	1.27 (0.63)	1.47 (0.95)	.76 (0.26)	.12	9.9
Iraq	591	1.44 (0.85)	1.92 (1.15)	.74 (0.27)	.65	0.8
Jordan	739	1.08 (0.36)	1.74 (1.02)	.82 (0.25)	.23	2.2
Kazakhstan	790	2.50 (1.39)	2.65 (1.41)	.47 (0.38)	.62	24.8
Kosovo	922	1.90 (1.35)	2.01 (0.35)	.67 (0.35)	.32	11.4
Kyrgyzstan	1,069	2.05 (1.32)	2.24 (1.36)	.76 (0.32)	.41	11.4
Lebanon	497	1.81 (1.30)	2.20 (1.45)	.64 (0.34)	.38	38.3
Malaysia	1,101	1.82 (1.15)	2.04 (1.24)	.79 (0.24)	.12	9.4
Morocco	979	1.75 (1.34)	2.38 (1.62)	.72 (0.23)	.43	0.1
Pakistan	1,244	1.18 (0.64)	1.41 (1.02)	.89 (0.19)	.10	1.6
Palestinian Territories	792	1.24 (0.68)	1.70 (1.13)	.72 (0.27)	.52	2.4
Russia	773	2.59 (1.55)	3.09 (1.45)	.60 (0.40)	.56	73.3
Tajikistan	1,084	1.62 (0.97)	2.24 (1.31)	.79 (0.29)	.26	1.6
Tunisia	1,055	1.739 (1.16)	2.34 (1.36)	.58 (0.33)	.41	0.2
Turkey	1,140	2.02 (1.42)	2.23 (1.51)	.72 (0.35)	.33	0.4
Uzbekistan	599	1.69 (1.20)	2.03 (1.35)	.76 (0.34)	.29	2.3

Notes. The range indicates the range of the variables that is present in the sample. Percentage of Christians per country retrieved from Pew Research Forum (2010).

APPENDIX C

Table C1. Models including random slopes for religious belief and cross-level interaction with proportion of Christians.

	Model 1a: Model including random slopes for religious belief^a	Model 1b: Model including random slopes for religious belief^b	Model 2a: Model including cross-level interaction with the proportion of Christians^a	Model 2b: Model including cross level interaction with the proportion of Christians^b
	Attitude for daughters	Attitude for sons	Attitude for daughters	Attitude for sons
	B (SE)	B (SE)	B (SE)	B (SE)
Fixed effects
Individual level
Intercept	2.134 (.190)***	2.433 (.152)***	1.910 (.177)***	2.306 (.147)***
Main predictor
Religious belief	−0.433 (.098)***	−0.411 (.076)***	−0.260 (.093)*	−0.299 (.076)***
Control variables
Rural	−0.113 (.016)***	−0.094 (.018)***	−0.112 (.016)***	−0.094 (.018)***
Female	0.004 (.015)	−0.025 (.017)	0.004 (.015)	−0.025 (.017)
Age	−0.002 (.000)***	−0003 (.001)***	−0.002 (.000)***	−0.003 (.001)***
Moderation
Perceived Commonality	0.302 (.039)***	0.424 (.046)***	0.298 (.039)***	0.463 (.045)***
Belief × Commonality	−0.135 (.050)**	−0.124 (.059)*	−0.128 (.050)*	−0.113 (.059)
Country level
Proportion of Christians	0.003 (.007)	0.005 (.006)	0.021 (.008)**	0.016 (.006)*
Belief × Proportion			−0.014 (.004)***	−0.010 (.003)**
Variance components
σ^2 individual level	1.126 (.011)***	1.577 (.015)***	1.126 (.011)***	1.577 (.015)***
σ^2 country level	0.588 (.204)**	0.347 (.118)**	0.465 (.143)**	0.304 (.095)**
σ^2 slope belief	0.188 (.062)**	0.097 (.037)**	0.133 (.039)**	0.062 (.026)*
Fit statistics
Likelihood (free parameters)	−32,001.302 (12)	−35,529.398 (12)	−31,996.260 (13)	−35,525.637 (13)
Deviance	64,002.604	71,058.796	63,992.520	71,051.274
AIC	64,026.605	71,082.797	64,018.520	71,077.274

Note. The independent variable (religious belief) is based on an observed mean score. Unstandardized coefficients are reported. The degrees of freedom cannot be obtained from MPlus, and therefore the free parameters are reported. With these analyses, only the Akaike information criterion (AIC) fit statistic is available for model comparison in MPlus.

^a N = 21,587.

^b N = 21,532.

*p < .05; **p < .01; ***p < .001.