Religious Dimensions and Attitudes Towards Immigrants
in Cross-National Perspective
A comparative study of 29 European countries Master’s Thesis in Sociology Track: Migration and Ethnic Studies June 2016 by Danique van der Burg Student-id 11140941 E-mail address danique.vanderburg@student.uva.nl University of Amsterdam Graduate School of Social Sciences Department of Sociology Supervisor Dr. Bram Lancee Second reader Dr. Agnieszka KanasTable of contents
Abstract ... 2 1. Introduction ... 3 2. Theoretical framework ... 6 2.1 Previous research on dimensions of religiosity ... 6 2.2 Theories linking dimensions of religiosity to attitudes towards immigrants ... 7 2.2.1 Individual level theories ... 7 2.2.3 The influence of context – religious diversity ... 10 3. Data and measurement ... 13 3.2 Variables ... 14 3.2.1 Dependent variable ... 14 3.2.2 Independent variables ... 15 3.2.3 Control variables ... 17 3.3 Analytical strategy ... 18 4. Results ... 18 5. Conclusion & discussion ... 26 6. Acknowledgements ... 30 7. Literature ... 31 Appendix ... 36Abstract
This study examines the relationship between religiosity and attitudes towards immigrants in Europe. It contributes to existing research by distinguishing different religious dimensions as well as considering the influence of the religious diversity of a country. The three religious dimensions distinguished are religious belief, –practice, and –belonging. The theories used in this study relate to these three dimensions and predict different outcomes for each dimension. The differences in individuals’ attitudes towards immigrants between countries, it is argued, are partially explained by the varying religious diversity in European countries. In order to take both individual- and country-level measures into account, this study employs multi-religious diversity in European countries. In order to take both individual- and country-level regression models using data on 29 European countries from the sixth round of the European Social Survey. The main conclusion of this study is that religious dimensions relate to attitudes towards immigration in different ways. In line with expectations, religious belief is found to have a positive influence on attitudes towards immigrants. The relation between religious practice and attitudes towards immigrants appeared to be more complex than hypothesized and is dependent on the religious diversity in a country. In religiously homogenous countries, the influence of religious practice is negative and in religiously diverse countries it is positive. People that belong to a religious minority in their country are found to hold more positive attitudes towards immigrants than those who belong to the majority religion, regardless of how religiously diverse their country is.
1. Introduction
Nowadays, increasing immigration flows into Europe put pressure on countries to implement policies that support quick integration and adoption of newcomers into their societies. In these times, an understanding of how attitudes of the ethnic majority in a host country towards immigrants come about is more important than ever. Many scholars have been trying to disentangle country differences in attitudes of ethnic majority members towards immigrants and their changes over time. A vast amount of research exists on this topic, although based on different theoretical perspectives and with varying results. This stresses the need for more profound research that investigates why findings from previous studies are so divergent.
In this study, religion is taken as a starting point to study attitudes of ethnic majority members in European countries towards immigrants. Religion is very important to many people. It influences their actions and their view on how to treat others (McGuire, 2002). Although traditional religions have lost prevalence in many European countries over the past decades (Stark & Iannaccone, 1994), in the majority of countries the larger part of society is still religious to some extent. Furthermore, the majority of immigrants that European countries receive is also religious (Castles & Miller, 2009).
In multiple studies, religiosity of individuals has been found to influence attitudes and prejudices towards immigrants (Ben-Nun Bloom, Arikan, & Courtemanche, 2015; Boomgaarden & Freire, 2009; Doebler, 2014; Knoll, 2009; McDaniel, Nooruddin, & Shortle, 2011; Scheepers, Gijsberts, & Hello, 2002). The results are not in accordance with each other. While some conclude that religiosity positively affects attitudes towards immigrants (Boomgaarden & Freire, 2009; Knoll, 2009), others find a negative relation (McDaniel et al., 2011; Scheepers et al., 2002). This is probably caused by different ways in which religiosity is operationalized, such as measuring worship attendance (Knoll, 2009; McDaniel et al., 2011), combining different elements of religion in one measure (Boomgaarden & Freire, 2009), or studying different religious denominations (Doebler, 2014; Knoll, 2009). Another problem is that many studies do not take into account the influence of country characteristics on people’s attitudes towards immigrants. A comparative analysis of
different studies on the formation of attitudes towards immigrants and immigration has led to the conclusion that future research needs to diversify contextual level measures, take multiple levels into account and expand the number of countries in comparative studies (Ceobanu & Escandell, 2010).
This study aims to heed this demand and extend previous research in two ways. Firstly, by studying the interplay between individual characteristics, and structural conditions of countries that influence people’s attitudes towards immigrants. Studying the influence of individual religiosity and religious diversity in a country simultaneously, yields a better understanding of the underlying mechanism. In contrast to religious diversity, economic characteristics and minority group size have been studied extensively (Mayda, 2006; Ruist, 2016; Semyonov, Raijman, & Gorodzeisky, 2006). Both Semyonov et al. (2006) and Mayda (2006) stress the importance of taking non-economic factors into account in explaining attitudes towards immigrants and immigration. Studying religious diversity contributes to the existing knowledge of the influence of country characteristics and it can give new insights into how context influences people’s attitudes towards immigrants.
Secondly, this study distinguishes between different dimensions of religiosity. In most contemporary literature, religion is treated as a one-dimensional concept. However, the few studies that do distinguish between different dimensions all find that different elements of religion affect attitudes in different ways. For example, people that often attend religious meetings are different from those who do not, with regard to attitudes towards immigration. Following the literature (Ben-Nun Bloom et al., 2015; Doebler, 2014, 2015; Scheepers et al., 2002), it seems clear that measuring religiosity of individual without taking into account the different dimensions of religiosity, would be too short-sighted. Therefore, in this study, three dimensions will be studied separately.
The most widely used model for measuring different dimensions of religiosity is that of religious belief, belonging, and behaviour or practice, derived from the classical distinction by Stark and Glock (1968). Religious belief is understood as the extent to which people believe in a god or a spirit and are intrinsically motivated to outlive this religion (Allport & Ross, 1967). The belonging dimension of religion refers to the religious denomination that someone belongs to. In this study, Christianity, Islam, Judaism and Eastern religions, such as Buddhism and Hinduism, will be distinguished.
Religious practice involves church attendance or other religious practices (Doebler, 2014; Stark & Glock, 1968)
By combining theories that address religiosity of individuals with theories that take structural conditions in countries as a starting point, differences in attitudes towards immigrants between European countries are better understood. A cross-national analysis is carried out to determine if religious diversity influences the effect of religion on attitudes towards immigration in European countries. With the use of data from the European Social Survey (ESS), 29 European countries are studied. The ESS data is appropriate for this study as it contains numerous questions about attitudes towards immigrants and includes different measures of religiosity. A multilevel regression analysis is performed to account for religious dimensions on the individual level and religious diversity on a country level, together with a number of control variables.
The mentioned considerations lead to the following research question: Is the relation between different dimensions of religiosity and attitudes towards immigrants affected by the religious diversity in a country?
Sub questions:
- To what extent does the relation between religiosity and attitudes towards immigrants differ between religious dimensions?
- Is the effect of religious dimensions (belief, belonging, practice) more pronounced in countries with a high religious diversity?
In the following sections, first the main theoretical perspectives, namely the religious compassion hypothesis, the social identity theory, and the religious marginalization perspective, are explained and linked to the various dimensions of religiosity. Subsequently, hypotheses about the interaction between religious diversity and the different dimensions of religiosity are derived from these theories. Finally, the research design is presented and the results of the multilevel regression models are discussed.
2. Theoretical framework
2.1 Previous research on dimensions of religiosity
A crucial finding from the contemporary literature is that it is important not only to distinguish between countries in terms of the presence of different religious groups, but also between different dimensions of religiosity. As mentioned earlier, most research conducted in Europe does not make this distinction and treats religion as a one-dimensional concept (Doebler, 2014). The first scholars who studied different elements of religion systematically and developed a system in which five core dimensions are distinguished were Stark and Glock (1968). In their study, they tried to find a way to measure religious commitment; an ambiguous concept that has a different meaning to different people. Linguistic analysis and survey research has led to the distinction of belief, practice, knowledge, experience and consequences (Stark & Glock, 1968). As mentioned before, the current study focuses on the first two of these dimensions and religious denomination. This decision is based on previous research on this topic (Ben-Nun Bloom et al., 2015; Doebler, 2014, 2015; Olson & Warber, 2008) and practical restrictions related to the ESS data that do not allow for distinguishing all five dimensions. The first dimension, religious belief, is a relevant dimension to study in relation to attitude formation because it refers to the acknowledgement of a supernatural realm (Stark & Glock, 1968). This does not necessarily need to be related to attending religious practices, it can also comprise people who contemplate their beliefs privately, at home (Doebler, 2014). Religious practice, on the other hand, does not refer to people’s inner experience of religion, but to religious participation or church membership. In this dimension, the influence of religious leaders and a religious community on people’s values and norms plays a role in the explanation of attitudes towards immigrants.
The last dimension of interest is religious belonging. A typical way in which religious belonging is used in research, is by comparing different religious denominations in terms of attitudes towards immigrants and immigration (Doebler, 2014; Knoll, 2009; Scheepers et al., 2002). In this study, religious belonging is used to determine whether people belong to a religious denomination that is a minority religion in their country.
2.2 Theories linking dimensions of religiosity to attitudes towards immigrants
The dimensions that are relevant for the current study have been discussed in the previous section and need to be linked to theories that can explain why there is a relation and what kind of relation there is between the religious dimensions and attitudes towards immigrants. These will be introduced in the following section.
2.2.1 Individual level theories
With regard to the belief dimension of religion, the religious compassion perspective is useful in explaining the influence of beliefs on attitudes towards immigrants (Ben-Nun Bloom et al., 2015). People with strong religious beliefs internalized the values of their religion. The inner experience of religion, as Allport & Ross (1967) address it. The majority of religions emphasize solidarity towards others, altruism and religious compassion (Ben-Nun Bloom et al., 2015; Knoll, 2009). Therefore, this theory argues that strong religious belief, in other words strong religious commitment, is related to more positive attitudes towards immigration (Ben-Nun Bloom et al., 2015; Boomgaarden & Freire, 2009).
Support for this theory has been found in previous research (Doebler, 2014; Knoll, 2009). Doebler (2014) reported a positive relation between the belief dimension and positive attitudes towards both ethnic and religious minorities. The results in the study of Ben-Nun Bloom et al. (2015) show a negative relation between beliefs and anti-immigration policies. Hence, the stronger a person believes, the less likely that person is to support anti-immigration policies. According to this literature, people with strong religious beliefs, who internalized religious values, are expected to have more positive attitudes towards immigrants than people with weaker religious believes, irrespectively of their religious denomination. Consequently, hypothesis one reads as follows: H1 – The stronger the religious beliefs of members of the ethnic majority in a country, the more positive their attitudes towards immigrants will be. The second theory, which relates to the behavioural dimension of religion, or so-called practice dimension, is the social identity theory which was originally developed by Tajfel
in the 1970’s (Hogg, 2006). It argues that if people belong to a group, they derive a sense of self-identity from that group. This relation gets stronger as a person gets a stronger feeling of belonging to the group, which is fostered by religious behaviour. If people attend religious meetings or other activities within a religious community, their sense of belonging to the group will become stronger (Ben-Nun Bloom et al., 2015). Belonging to a group in itself does not necessarily lead to expectations with regard to attitudes towards others. However, an important assumption of this theory is that people are motivated to derive a social identity from a group because of the need for self-enhancement and uncertainty reduction (Hogg, 2006). People tend to think in terms of in- and out-group and attach positive characteristics to one’s own group. People who frequently attend religious meetings are expected to have a stronger feeling of belonging to a group than those who do not. In this case, they will feel they belong to a religious community. Therefore, they are expected to have a stronger sense of out-group as well. Immigrants can be seen as threatening out-groups because of differences in culture or religion. Following this line of reasoning it can be expected people who frequently attend religious meetings are having more negative attitudes towards immigrants and immigration in general, seeing they will have a stronger sense of belonging to a group than people who do not frequently attend religious practices.
Results from previous studies with regard to this dimension of religion are mixed. Some studies found that religious practice has a positive influence on attitudes towards immigration (Ben-Nun Bloom et al., 2015; Knoll, 2009). However, these studies were conducted in the United States, a country where the church, as an institution, plays a different role in society than in most European countries and where the religious background of immigrants is mostly Christian, whereas in Europe it is mostly Muslim (Foner & Alba, 2008). Furthermore, these studies investigated attitudes towards immigration policies, and not immigrants as such.
Another European study, focussing on tolerance towards immigrants specifically, found no statistically significant relation between religious practice and attitudes (Doebler, 2014). Others found that religious practice is negatively related to attitudes towards immigrants in another European study (Scheepers et al., 2002). However, there are some substantial differences between these studies in terms of measurement of the religious practice dimension and the countries that are included in the analysis. These differences might be responsible for the varying results. The absence of consistent
empirical findings might implicate that predicting the relation between religious practice and attitudes towards immigrants is practically impossible. However, the fact that up until now, a positive relation was only found in the U.S. provides ground for assuming that the social identity theory could be applicable in the case of Europe. Therefore, hypothesis two reads as follows: H2 - The more frequently members of the ethnic majority of a country attend religious meetings, the more negative their attitude towards immigrants will be. The last theory of importance applies to the third dimension, namely religious beloning. The religious marginalization perspective argues that minority groups tend to be more generous towards other minority groups because they share the experience of marginalization (Fetzer, 1998; Knoll, 2009). Studies investigating this perspective originate from the 1960’s and focused mainly on the marginalization of Jews and Catholics and its´ influence on voting behaviour (Ebersole, 1960). Later, this theory is extended by studying the influence of marginalization on attitudes towards ethnic minorities. The mechanism behind this theory is that people tend to sympathize with people that are similar to them. In this case, they share the experience of marginalization, since people that belong to a minority religion often experience religious discrimination (Knoll, 2009). Furthermore, it is argued that people also sympathize with potential victims of marginalization (Fetzer, 1998).
Previous studies found support for the hypothesis that religious marginalization leads to sympathy for other marginalized religious groups and marginalized ethnic groups. This was tested by examining people’s attitudes towards immigrant rights and immigration policies (Fetzer, 1998; Knoll, 2009). Following this research, it is expected that especially the people that belong to a religious minority group in their country have positive attitudes towards immigrants. However, it has to be mentioned that these studies were conducted in the U.S. and therefore results can be different when the theory is tested in European context.
H3 – Ethnic majority members that belong to a minority religion are expected to have more positive attitudes towards immigration than those that belong to the majority religion in a country or that are not religious.
2.2.3 The influence of context – religious diversity
As mentioned earlier, individual religiosity is expected to be an important predictor of attitudes towards immigrants. However, there is considerable evidence that contextual characteristics can explain between-country differences (Mayda, 2006; O’Rourke & Sinnott, 2006; Ruist, 2016). In the case of religiosity, the religious diversity in a country could be influencing this relationship in multiple ways. Up until now, there is no research available that investigates the influence of religious diversity on the relation between religiosity and attitude formation. However, theories originally used to explain the influence of the religious climate in a country on other aspects related to migration, like migration policies or attitude formation, are helpful in the deduction of hypotheses. One of the most common theories used in research on religious diversity is the religious market theory. The more diverse the religious context is, in terms of religious denominations that are present in a country, the more people can choose from. This theory states that increasing religious diversity brings about religious competition. This competition results in more investments of religious leaders, attempting to withhold people from switching religions. This means that they will proclaim some kind of religious particularism, which may lead to more negative attitudes towards other religions or other groups, on an individual level (Scheepers et al., 2002; Schneider, 2008). This theory implicitly stresses the importance of distinguishing between religious belief and -practice, for religious leaders can only influence those who attend religious meetings.
Another theory, that has not been related to religious diversity yet but could be applicable as well, is the classical contact theory, originated by Allport (1954). This theory argues that contact between different groups could reduce prejudice towards those groups. The higher the number of religious groups a country, the more likely it is for people to get in contact with other religious groups, especially religious groups that share other characteristics, such as being a minority or being discriminated against. Again, this shows that also this theory is applicable to a specific group of people. Attending religious meetings does not make a large difference with regard to the chance of having contact with other religious or ethnic groups.
As mentioned earlier, empirical evidence that supports these theories is scarce. However, studies that investigated religious diversity in relation attitudes in a different way can be useful for the deduction of hypotheses. Hello, Scheepers, and Gijsberts
(2002) are among the few scholars that investigated the religious diversity of European countries in relation to attitudes. However, this was not the main objective of the study, which was focused on the effect of educational level on ethnic prejudice. Religious diversity is included in their study to test whether it moderates this relationship or not. Their findings show that the religious diversity of a country at least partially explains differences between countries in the effect of educational level on prejudice.
In another study on ethnic prejudice in different European countries, religious diversity and different dimensions of religiosity are both included in one model, but only functions as a control variable together with a number of other country characteristics (Scheepers et al., 2002). A drawback of this study is that only eleven countries are included in the analysis and that the data were collected in 1998 and is therefore rather out-dated nowadays. Since then, the religious climate and presence of minority groups has changed in many European countries. Hence, a more rigorous test with the use of recent data is relevant.
Depending on the dimension, religious diversity is expected to have either no influence, or a positive influence on its’ relation with attitudes. The religious compassion hypothesis, that argues that people who strongly believe have more positive attitudes towards immigrants than people who do not strongly belief, is related to the belief dimension. The question is whether the level of religious diversity in a country affects this positive relation. The belief dimension of religion refers to intrinsic religiosity rather than extrinsic religiosity, that is referred to as the use of religion for people’s own ends (Allport & Ross, 1967). People that live their religion and internalized the values of their religion are expected to be less impressionable to change their attitudes towards other people by external factors. Religious diversity is an external factor and therefore not expected to have much influence on people that internalized religious values and experience it as something personal. Hence, hypothesis four: H4 - The relation between belief and attitudes towards immigrants is not affected by the religious diversity in a country As discussed before, religious practice is expected to have a negative effect on attitudes towards immigrants. The influence of religious diversity on this relation can be
hypothesized using both the theory of religious competition and the social identity theory. The first theory argues that religious diversity is associated with a competitive religious economy (Scheepers et al., 2002). If a country is religiously homogeneous, a church has a monopoly and the environment is non-competitive (Iannaccone, 1991). Just like with other markets, if there is a lot to choose from, competition increases. In countries with a high level of diversity, religious leaders may proclaim one particular religion because there is a threat from other religions. This can lead to religious particularism, which means that people consider their own religion as the only true religion (Scheepers et al., 2002). This line of reasoning is especially applicable to the practice dimension of religion, since people that attend religious practices more often are expected to be more likely to conform to what religious leaders proclaim.
The social identity theory, as discussed before, predicts a similar pattern. First of all, purely on an individual level, people who often attend religious meetings are expected to have negative attitudes because immigrants are seen as threatening out-groups because they often have a different religion. It might seem self-evident, but in a religiously diverse country there are more threatening out-groups than in religiously homogeneous countries. Therefore, the negative effect of religious practice on attitudes towards immigrants is expected to be stronger in countries with a high religious diversity.
H5 - The relation between religious practice and attitudes towards immigrants becomes stronger when religious diversity increases.
With regard to the belonging dimension of religion, following the religious marginalization perspective, it is expected that people who belong to a religious minority group have more positive attitudes than people who belong to a religious majority group. Religious diversity is likely to have a positive effect on this relation. The line of reasoning that leads to this expectation is as follows. In a religiously diverse country, people are used to being surrounded by people with different religions and cultures and, than in religiously homogeneous countries. This can bring about a more welcoming climate for immigrants, since people are already used to minority groups in their society. Seeing that minority groups are expected to have more sympathy for other minorities, this welcoming climate is expected to strengthen this already existing
propensity. Thus, religious diversity positively influences the effect of belonging to a minority religion and attitudes towards immigrants. Subsequently, hypothesis 6 reads as follows:
H6 - The relation between belonging to a religious minority and attitudes towards immigrants is expected to become stronger when religious diversity increases.
3. Data and measurement
3.1 Data and sampleTo test the hypotheses that have been discussed in the previous paragraphs, analysis is carried out using data from the sixth round of the European Social Survey (ESS). This biannual survey is conducted in a large number of European countries, of which the majority participated every round since the start in 2002. In this study, all 29 countries that participated in this ESS round will be included in the sample. Every ESS-questionnaire consists of a set of standard questions and a so-called rotating module that differs every round and is included on request. Questions about a respondent’s attitude towards immigration, immigrants, and religion are included in the core questionnaire. In this study, the most recent wave of which the data for all countries is available is used for the analysis. Data for the sixth round were gathered in 2012. The selection of respondents for the sample in each country is based on random probability. Respondents have to be at least 15 years old to participate. Since sampling methods are random, but not the same for all countries, the data contains weights that are applied to the analyses to correct for the differences in sampling methods and the probability for respondents of being selected for the sample. Countries with a population size of less than 2 million are allowed to interview 800 respondents per round. For all other countries, the minimum sample size is 1500. These conditions are not always met, but for the majority of the countries that are included, the dataset contains more than 1500 cases.
The initial sample consists of 54673 respondents. However, the theories that are used in this study explain ethnic majority members’ attitudes towards immigrants. Therefore, only citizens that do not belong to an ethnic minority group in their country are included in the analysis. After selecting only respondents that stated to be citizens of
the country they live in and do not belong to an ethnic minority group, 48974 respondents are left in the sample. By selecting respondents based on citizenship and ethnic minority status, the work of Meuleman, Davidov and Billiet (2009) is followed.
With regard to missing values on the variables that are used in the analysis, the decision has been made to exclude all the cases that have a missing value on one of the items. The dependent variable ‘attitude towards immigrants’ is the only exception to this rule since this measure is a compound of 6 variables. Therefore, on this measure, one missing is allowed. After removal of all the cases that do not meet the criteria for the dependent variable, 45821 respondents are left in the sample. This means that 6.44 per cent of the cases are excluded because of missing values on more than one item in the attitude measure. After excluding all cases with a missing on one of the other variables that are used in the analysis, 44636 respondents are left in the sample. In table A1 in the appendix an overview is given of the number of respondents in the sample per country after list wise deletion of all cases with missing values.
3.2 Variables
3.2.1 Dependent variable
The six questions that are used to create a measure for attitudes towards immigrants are listed in table 1. The first three items that are used are in line with previous research (Meuleman et al., 2009). The latter three items are added to create a more reliable scale. Answer categories are not the same for all items in the attitude measure. Three questions are coded 1-10 and the other three are coded 1-4. Therefore, the scales are standardized to 0-1 scales in which zero means ‘negative’ and one ‘positive’. To control for the reliability of the scale, a principal component analysis is performed. This test shows that the six questions that are used to create the scale have one underlying component and an eigenvector of 0.3881 or higher. The alpha or the scale is 0.8819 for the pooled sample. For single countries, the Chronbach’s Alpha varies between 0.775 and 0.902. Table A1 in the appendix shows the Chronbach’s Alpha for single countries.
Table 1 -Questionnaire items used in the analysis
Question Variable
Now, using this card, to what extent do you think [country] should allow people of the same race or ethnic group as most [country]’s people to come and live here?
Attitude
How about people of a different race or ethnic group from most people? Attitude
How about people from poorer countries outside Europe? Attitude
Would you say it is generally bad or good for [country]’s economy that people come to live here from other countries?
Attitude And, using this card, would you say that [country]’s cultural life is generally
undermined or enriched by people coming to live here from other countries? Attitude Is [country] made a worse or a better place to live by the people coming to live here from other countries? Attitude Do you consider yourself as belonging to any particular religion or denomination? Religion (belonging) (If yes) Which one? Religion (belonging) Regardless whether you belong to a particular religion, how religious would you say you are? Religion (belief) Apart from special occasions such as weddings and funerals, about how often do you attend religious services nowadays? Religion (practice) Are you a citizen of [country]? Citizenship Do you belong to a minority ethnic group in [country]? Ethnic majority In what year were you born? Control
About how many years of education have you completed, whether full-time or part-time? Control Sex (male/female) Control Were you born in [country]? Control, foreign born Source: ESS data archive 2012 3.2.2 Independent variables
With the use of various questions from the survey, a variable is constructed for each dimension of religion. The question ‘Regardless whether you belong to a particular religion, how religious would you say you are?’ is used to measure religious belief. Since the ESS questionnaire does not contain many in-depth questions about the religiosity of a respondent, this is the only suitable question that can be used to construct the variable for the belief dimension of religion.
Religious practice is measured with the use of the question: ‘Apart from special occasions such as weddings and funerals, about how often do you attend religious services nowadays?’ This question has answer categories ranging from every day (1) to never (7) and is treated as a continuous variable in the analysis.
The third dimension of religion, religious belonging, is used in the analysis to determine whether belonging to a minority religion in a country is related to attitudes towards immigration. Dummy variables are constructed for belonging to the majority religion, minority religion and not belonging to any religious denomination. The
questions ‘Do you consider yourself as belonging to any particular religion or denomination?’ and ‘Which one’ are used to calculate the proportions of each religious denomination per country. This allows for determination of the majority religion in the different countries and, subsequently, assignment of people to the majority- or minority religion category. In this measure, people that do not belong to any religious denomination are treated as a separate group. The consideration behind this decision is that including the category ‘no religion’ as a religion would mean that in most countries all religious denominations would become a minority since in many European countries the majority of the people do not consider themselves as belonging to any particular religious denomination.
In all models, belonging to the majority religion is used as a reference group. There are two reasons for this decision. Firstly, the main point of interest is the difference between people belonging to the minority and the majority religion in a country. Secondly, the fact that the category ‘majority religion’ contains the highest number of cases makes it a more reliable variable to treat as a reference group than the categories ‘minority-’ or ‘no religion’.
The last independent variable is the religious diversity in a country. Measures of the religious diversity in different European countries are not readily available. Therefore, individual religious denominations in the data set are aggregated to create a scale for religious diversity. In order to obtain an index, the proportions of each denomination and the number of denominations have to be taken into account. The formula that is used to create this index is the ‘Herfindahl index of concentration’. This is a common index for the measurement of religious diversity (Lancee & Dronkers, 2011; Paldam, 2001; Ruiter & van Tubergen, 2009; Zaleski & Zech, 1995). Two questions from the ESS survey are used to create this scale. The first one is ‘Do you consider yourself as belonging to any particular religion or denomination?’ and the second one is ‘Which one?’ Respondents who reported not to belong to any particular religion are labelled as ‘non-religious’ and treated as a religious group in the diversity measure.
Unfortunately, no distinction is made between different religious movements within the Islam in the survey, whereas there is for different Christian movements. To prevent countries that are mainly Christian from being treated as more diverse than, for example, mainly Islamic countries because of the different Christians movements, all
Christian movements are taken together in the analysis. This results in a variable that contains six categories: non-religious, Christian, Islamic, Jewish, Eastern religion and other non-Christian religions. The diversity index is calculated using post-stratification and design weights. The sum of the squared proportions -1 is the index that will be used in the analysis. By subtracting one from the total sum, a high value reflects a high religious diversity in a country. The lowest value of the index is 0.0753 in Cyprus. This country is mainly Christian, with only 8 respondents who stated to be Muslim and 31 that reported to be non-religious. The country with the highest religious diversity in the sample is Albania, with an index of 0.578. For this country the majority religion is Islam, but many respondents also reported to be non-religious or Christian. An overview of the religious diversity per country is given in table A1 in the Appendix. 3.2.3 Control variables
Six variables will be added to the models to control for both individual and country characteristics. Control variables on the individual level are gender, number of years of education and age. On a country level, GDP, the percentage of foreign born people, and unemployment rate are added to the models. A high GDP has been found to be related to pro-immigration attitudes (Mayda, 2006). Therefore, it is important to control for differences in GDP between countries. Unemployment rate is added to the models following the research of Ruist (2016) , who found this to be a significant predictor of various elements of attitudes towards immigration and immigration policies. These variables are not included in the original data set and therefore added from another source (for resulting indicators see table A1 in the appendix). The GDP for every country in 2012, the year that the ESS data was collected, is obtained from the World Bank’s World Development Indicators (World Bank, 2012). Unemployment rates are, with a few exceptions, obtained from the OECD database (OECD, 2016). For three countries, the data are not available in this database and therefore other sources are used. Data for Albania and Ukraine are obtained from www.statista.com and the unemployment rate in Kosovo is obtained from the Worldbank database.
Another control variable that is added to the models is the percentage of foreign-born people in a country. This variable is used as a control variable in many studies (Ruist, 2016; Schlueter, Meuleman, & Davidov, 2013; Schneider, 2008; Semyonov et al., 2006) and mostly found to be significantly related to attitudes towards immigrants.
This control variable is added to the models to control for the possibility that the ethnic composition of a country is a more important predictor of attitudes than religious diversity. The variable is constructed by calculating the percentage of people that are not born their country of residence using the question ‘were you born in [country]?’
3.3 Analytical strategy
To test the hypotheses that are derived from the theories in the previous section, analyses are conducted using both individual level and country level data. An ordinary least squares regression model is not sufficient for the analysis of data that contains information about individuals and countries. Therefore, a multilevel ordinary least squares regression is the best suitable method to test the hypotheses. With the use of this method, it is possible to account for individual differences within countries in the analysis (Rabe-Hesketh & Skrondal, 2006). After constructing all the variables and the removal of cases with missing values, 44636 cases remain on the first- and 29 cases on the second level. The consequence of using multilevel modelling techniques to test the hypotheses is that the interpretation of the intercepts in the models changes compared to ordinary least squares regression (Paccagnella, 2006). It is necessary to center the continuous variables in order to be able to interpret the obtained coefficients in the regression analysis. This research does not aim to explain differences within countries in attitudes towards migration but differences between countries. Therefore, variables are centered around the overall mean and not on the country mean, thereby following the work of Paccanella (2006).
4. Results
In table 2, a descriptive overview of the variables that are used in the analysis is given. Weights are not applied to the statistics in this table in order to give an impression of what the data look like. In all other figures and tables, design weights are applied.Table 2 - Descriptive statistics
Variable Min Max Mean SD N
Attitude towards immigrants 0 1 0.518 0.224 45821 Religious belief 0 10 4.67 3.08 48500 Religious practice 1 7 2.56 1.52 48593 Majority religion 0 1 0.59 0.492 48512 Minority religion 0 1 0.0201 0.14 48512 No religion 0 1 0.39 0.488 48512 Religious diversity 0.075 0.578 0.436 0.123 48974 Female 0 1 0.545 0.498 48959 Years of education 0 51 12.6 4.01 48595 Age 15 103 48.7 18.7 48862 GDP 3601 101564 33516 21698 48974 Unemployment rate 3.1 30 9.83 5.06 48974 Percentage foreign born 0.413 23.4 7.94 5.51 48974 Source: ESS Round 6 Source Questionnaire Figure 1 shows the mean for each country on the attitude measure compared with the score for that country on the religious diversity index. A line is plotted that shows that most countries are centered around this line, with some outliers at both ends of the religious diversity axis. A remarkable outlier is Cyprus, which is located in the bottom left corner of the graph. The multi-level regression model assumes linearity. In response to Figure 1, quadratic and cubic functions are added to the models to test if a linear model indeed fits the attitudes measure best. There are no indications that the attitude measure has a quadratic or cubic form. An explanation of the country codes used in figure 1 can be found in table A1 in the appendix.
Figure 4 is sorted on the magnitude of the majority religion in a country and shows approximately the same order as figures 2 and 3. The Pearson correlation between religious belief and religious practice is 0.602. This is a rather strong correlation, but not so high that a multi-collinearity problem will occur when both variables are added to the models.
Figures 2, 3, and 4 show the mean for every country on each dimension of religion sorted by highest to lowest score. Among the countries located in the upper half of the figures, the order is quite stable over the different figures. In the lower half of the figures, there is some variation visible. For example, Sweden scores 3.110 on the belief dimension and therefore is the second lowest scoring country, whereas on the practice dimension this country is close to the middle of the figure with a mean of 2.138. For Belgium, it is the other way around.
The first step in building the multi-level model is to investigate if the expected variance between countries is also found in the data. Comparing the empty fixed intercepts model with a model that allows intercepts for different countries to vary gives an estimated deviance of 6777,49. This is a significant improvement in model-fit and signifies that the difference in attitudes towards immigrants between countries is statistically significant. 14.97% of the variance is explained by country differences (intra class correlation). With all independent variables added to the model the intra class correlation is reduced to 13.08%. Additional models are estimated in which the independent variables are included separately (not shown) to ensure that effects are not controlled away by adding these variables in the same model.
After investigation of the intra class correlation and a test of the deviance between the empty fixed intercepts model and the 2-level random intercepts model (shown in table 3), more elaborate models can be estimated. In the next step, individual level predictors are added to the model. These variables are the measures for the different dimensions of religiosity. Adding religious diversity to the model as a contextual level predictor does not change the coefficients of the other variables in the model and the coefficient itself is not statistically significant either. Table 3 - Model-fit overview Models Log Likelihood 2*Log Likelihood Δm d.f d.f.= Δm P-value 1 Intercept 3488,583 0 2 1 + random variation at country level 6877,330 6777,494 3 3 0,000*** 3 2 + dimensions of religiosity 6913,480 72,300 6 3 0,000*** 4 3 + religious diversity 6915,750 4,540 8 2 0,045* 5a 4 + individual level controls 8267,547 2703,594 11 3 0,000*** 5 5a + country level controls 8269,869 4,644 14 3 0,019* 6 5 + random slope for religous belief 8363,776 187,814 15 1 0,003** 7 5 + random slope for relgious practice 8336,748 133,757 15 1 0,005** 8 5 + random slope for minority rel. dummies 8400,262 260,786 16 2 0,000*** 9 5 + random slope for all dimensions 8477,410 415,082 18 4 0,000*** 10 9 + interaction belief*diversity 8477,440 0,060 19 1 0,962 11 9 + interaction practice* diversity 8484,270 13,719 19 1 0,046* 12 9 + interactions min. rel. dummies*diversity 8479,783 4,745 20 2 0,042* 13 9 + all interactions 8486,520 18,221 22 4 0,000***
In models 5a and 5, the control variables are added to model 4. First, individual level control variables are added. Whether a respondent is male of female does not seem to be an important predictor of attitudes and is not statistically significant across all estimated models. Both age and education are statistically significant and significantly improve the model-fit. The model fit statistics for all estimated models are displayed in table 3.
Adding the contextual level control variables does not have a big impact on the coefficients and p-values of the different measures of religiosity and the religious diversity variable (not shown). GDP is statistically significant in all the estimated models whereas unemployment rate on the other hand is not statistically significant at all. This result is not very surprising. In the research of Schlueter et al. (2013) unemployment rate appeared not to be an important predictor of attitudes and in the study that Ruist (2016) conducted, the effects differed depending on the dependent variable that was used.
Models 6 to 9 (table 3) show that allowing random slopes for the variables that measure the different dimensions of religiosity significantly improves the model fit. Furthermore, all the random slope coefficients are significant. Therefore, interaction variables that test hypotheses 4, 5 and 6 will be added to model 9. Random intercepts coefficients are not reported in table four, for they are all significant and do not add much information to the models.
Model 9, shown in table 4, is used to test hypotheses 1, 2, and 3. The first hypothesis relates the belief dimension of religion to attitudes towards immigrants and predicts a positive relation between belief and attitudes towards immigrants. Model 9 shows a statistically significant positive effect of religious belief on attitudes towards immigrants. This outcome confirms hypothesis one.
Hypothesis two predicts that the more often people attend religious meetings, the less positive their attitudes towards immigrants will be. Model 9 shows no statistically significant effect of attending religious meetings on attitudes towards immigrants. Therefore hypothesis two needs to be rejected.
Table 4 - multilevel models, dependent variable ‘attitude towards immigrants’
Model 9 Model 10 Model 11 Model 12 Model 13
Religious belief 0.003** 0.003 0.003** 0.003** 0.006 (0.001) (0.004) (0.001) (0.001) (0.004) Religious practice -0.000 -0.000 -0.016*** 0.000 -0.016*** (0.001) (0.001) (0.004) (0.001) (0.004) No religion 0.019*** 0.019*** 0.019*** 0.049** 0.046* (0.004) (0.004) (0.004) (0.019) (0.018) Minority religion 0.052** 0.052** 0.052** -0.043 -0.039 (0.019) (0.019) (0.019) (0.108) (0.107) Majority religion (ref.) Religious diversity 0.134 0.134 0.130 0.162 0.000 (0.184) (0.184) (0.183) (0.186) (.) Female 0.016 0.015 0.024 0.015 0.021 (0.341) (0.341) (0.340) (0.341) (0.340) Age -0.100*** -0.100*** -0.099*** -0.100*** -0.099*** (0.018) (0.018) (0.018) (0.018) (0.018) Years of education 0.114*** 0.114*** 0.114*** 0.114*** 0.114*** (0.008) (0.008) (0.008) (0.008) (0.008) GDP 0.019** 0.019** 0.019** 0.019** 0.019** (0.006) (0.006) (0.006) (0.006) (0.006) % Foreign born -0.359 -0.360 -0.360 -0.349 -0.352 (0.253) (0.253) (0.253) (0.252) (0.252) Unemployment rate 0.151 0.151 0.147 0.163 0.156 (0.239) (0.240) (0.241) (0.236) (0.239) Cross-level interactions Belief*diversity -0.002 -0.007 (0.007) (0.007) Practice*diversity 0.038*** 0.038*** (0.008) (0.009) No religion*diversity -0.065+ -0.058 (0.039) (0.038) Minority religion*diversity 0.198 0.190 (0.205) (0.205) Constant 0.456*** 0.456*** 0.458*** 0.444*** 0.447*** (0.086) (0.086) (0.086) (0.087) (0.087) N 44636 44636 44636 44636 44636 se in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 The third hypothesis predicts that people that belong to a minority religion hold more positive attitudes towards immigrants than people who belong to the majority religion or consider themselves non-religious. Model 9 shows that, compared to belonging to a majority religion, belonging to a minority religion is positively related to positive attitudes towards immigrants. For both people that belong to a minority religion and those who do not belong to any religious denomination, the effects are statistically
significant and point in the same direction. However, the positive effect of belonging to a minority religion is bigger than for the non-religious group. Separate models are estimated with non-religious people as the reference group (not shown). These models show that people that belong to a minority religion are not statistically different from people that are not religious. However, as expected, people that belong to the majority religion in their country hold significantly more negative attitudes towards immigrants than religious minority members. Model 9 appeared to have the best model fit (see table 3) and is therefore used as the base-model to which the interaction terms are added. Models 10 to 13 extend model 9 by including cross-level interaction variables for each dimension of religion separately. These are used to test hypotheses 4, 5, and 6 (see table 4). In model 14, all the interaction terms are included in one model. Even though this could be problematical in term of degrees of freedom since level two only counts 29 cases, the estimates do not change much in this final model. Moreover, the effects of the interaction terms are stable as well. This indicates that these effects are not depending on each other.
In model 12, the interaction term for religious belief is included. Adding this variable to the model makes, contrary to what is found in the other estimated models, the direct effect of belief on attitudes statistically insignificant. The interaction term it self is not significant either. This implicates that hypothesis four, predicting no significant effect of diversity on the effect of belief on attitudes, is confirmed by this data. As theorized, the relation between belief and diversity is not influenced by religious diversity. However, seeing that once the interaction is added to the model, the statistically significant relationship between religious belief and attitudes towards immigrants disappears, raises questions about the influence of this variable on the model.
As discussed before, model 9 shows that religious practice is negatively related to attitudes towards immigrants. In model 11, an interaction term for religious practice and religious diversity is added to this model to test hypothesis 5. This hypothesis predicts a positive effect of religious diversity on the relation between religious practice and attitudes towards immigrants. This means that the interaction term is expected to be negative, so the negative effect of religious practice becomes stronger. However, model 11 shows an effect that that is the opposite of what is predicted by hypothesis 5.
The main effect of religious practice on attitudes towards immigrants is still negative, like expected, and becomes statistically significant in this model, but the coefficient of the interaction term is positive and significant. This means that, in a religiously homogeneous country, the effect of religious practice on attitudes is negative, but in a religiously diverse country attending religious meetings has a positive effect on attitudes. The results are visualized in figure 5, based on model 11, which shows that the average marginal effect of religious practice on attitudes towards immigrants becomes positive when the religious diversity in a country is higher than 0.45 on the Herfindahl index. Model 12 includes the interaction term for belonging to the minority-, majority- or no religious denomination. The sixth hypothesis predicts a positive effect of religious diversity on the relation between belonging to a religious minority and attitudes towards immigration. Adding an interaction for all three dummy variables that are included for the religious minority measure does not influence the model substantially. Both interaction terms are not statistically significant at a 0.05 confidence level. However, the model fit improves a little bit when interactions for this dimension are added to model 9. All in all, the results do not support hypothesis 6, which therefore has to be rejected.
In order to ascertain that the results in the estimated models are not depending on one particular country, models 9 to 12 are checked for robustness by leaving each
country out of the analysis one by one. The results in these additional tests are to a large extent the same as the models presented in table 4, except for the models that are estimated with Poland excluded from the sample. In these models, the direct effect of religious diversity attitudes towards immigration becomes statistically significant. However, the effects of the interaction terms in the different models with Poland excluded do not change. Furthermore, the coefficients and statistical significance of the religious minority measure changes when some countries are left out. This result is not very surprising considering the fact that this measure contains very little cases in some countries. Analysis of the residuals is performed by plotting the standardized residuals of the fitted values against their normal scores (Hox, 2010). These tests show that the residuals are approximately normally distributed for all countries, indicating that the normality assumption of the multi-level regression model is not violated.
5. Conclusion & discussion
This study examined to what extent different dimensions of religiosity are related to attitudes towards immigrants and how this relation is affected by the religious diversity in a country. The religious dimensions distinguished are religious belief, religious practice and religious denomination. The latter is used to determine whether people belong to the majority or minority religion in their country. In order to study the relation between these dimensions and attitudes towards immigrants, individuals in 29 European countries are compared with the use of data from the European Social Survey. The results indicate different effects on attitudes towards immigrants for the different dimensions of religiosity that are distinguished. There is evidence to suggest that religious belief is an important predictor of attitudes, but country characteristics do not seem to influence this relation. The influence of religious practice on attitudes towards immigrants is found to be dependent on the religious diversity in a country. For people that belong to a minority religion, the results are somewhat mixed. Overall, the results indicate that distinguishing between different dimensions of religiosity is of importance in the study of attitudes towards immigrants. These results are in accordance with other studies that followed this distinction (Doebler, 2014, 2015; Olson & Warber, 2008; Scheepers et al., 2002; Stark & Glock, 1968).
The religious compassion hypothesis is used to link religious belief to attitudes towards immigrants. This theory predicts a positive effect of religious belief on attitudes towards immigrants, based on the fact that most traditional religions emphasize solidarity towards others. Since religious belief refers to individual religiosity, regardless of someone’s religious denomination or attendance of religious meetings, external factors are not expected to have much influence on this dimension of religion. Therefore, religious diversity is expected to have no effect on the relation between religious belief and attitudes towards immigrants either. The results indicate that this is indeed the case and support hypotheses one and four. This result is in line with the work that has been done by other scholars, in both European and American context (Ben-Nun Bloom et al., 2015; Doebler, 2014; Knoll, 2009) and contributes to existing reserach by extending the knowledge about the influence of country characteristics on this relationship.
Contrary to expectations, there is no evidence to suggest a direct negative influence of religious practice on attitudes towards immigrants. However, an effect is found for the influence of religious diversity on this dimension. Although the relation is different from what is expected, the results show an interesting pattern. In religiously homogeneous countries, religious practice has a negative effect on attitudes towards immigrants, whereas in religiously diverse countries this relation is positive. Meaning that attending religious meetings has a positive effect when people live in a religiously diverse country. This finding might explain the mixed results that are found by other scholars studying this topic (Ben-Nun Bloom et al., 2015; Doebler, 2014; Knoll, 2009; Scheepers et al., 2002). Previous research has been conducted in different countries, with different levels of religious diversity, and most studies did not take this characteristic into account. The one study that did consider religious diversity found similar effects (Scheepers et al., 2002).
The social identity theory and religious competition theory were both used to predict the effects of religious practice on attitudes towards immigrants and seem not to be of use in this study. A possible explanation for the contradictory results may be found in the decline in church attendance that many European countries face today. Although scholars do not agree on whether Europe is secularizing or not, the decline in church attendance is a phenomenon that they do agree on and it is noticeable in many European countries (Halman & Draulans, 2006; Stark & Iannaccone, 1994). This could
change the notion of what people consider to be ‘their own’ group, which is the explaining mechanism in the social identity theory. Future research should therefore focus on how people define their in- and out-group. If there is a shift towards ‘religious’ versus ‘non-religious’ as group identity instead of particular religious denominations that define each other as out-groups, this could explain why church attendance does not seem to be negatively related to attitudes towards immigrants. After all, as discussed before, the majority of immigrants that move to European countries are religious (Castles & Miller, 2009; Foner & Alba, 2008).
Another explanation for the finding that in religiously diverse countries, religious practices and attitudes towards immigrants are positively related, may lie in the fact that most theories that are used to predict the relationship are developed in the United States. Applying these theories to European context could be problematical because of the different role that religion plays in society in most European countries (Foner & Alba, 2008). Especially the assumption from the religious competition perspective that religious leaders increasingly proclaim one particular religion when religious competition increases is arguable and may not be true for most European countries.
The religious marginalization perspective is used to relate the belonging dimension of religion to attitudes towards immigrants. This theory predicts that people that belong to a minority religion hold more positive attitudes towards immigrants than people that belong to the majority religion or not to any religion because they share the experience of marginalization (Ebersole, 1960; Fetzer, 1998; Knoll, 2009). The results indicate that a distinction needs to be made between members of the majority religion and non-religious people. Most models show that religious minority members hold more positive attitudes towards immigrants than religious majority members. However, this effect disappears when interacted with religious diversity and showed to be unstable in the robustness checks. It therefore has to be interpreted with some caution. Another finding is that non-religious people also hold more positive attitudes towards immigrants, compared to religious majority members. In the view of this theory, this result is interesting considering the religious marginalization perspective can explain the effect that is found for religious minority members, but not for non-religious people. Moreover, no effect is found for religious diversity.
A remark has to be made about the distribution of the variable that is used to test the two hypotheses associated with this theory. This study is interested in attitudes of
ethnic majority members towards immigrants and therefore ethnic minorities are excluded from the analyses. However, the removal of many religious minorities from the sample is an inevitable consequence. In the religious minority category, only ethnic majorities that belong to a minority religion are included. For most countries, this group is rather small. For five countries this category contains less than 10 cases with Lithuania and Slovenia only counting just one case. This could account, at least partially, for the relatively small effects found for this group. Testing this theory with data that contains more cases is recommended to be able to draw conclusions based on results that are less vulnerable for type 1 errors.
A final suggestion for future research is to analyse the influence of religious diversity with data that includes more information about immigration flows to particular countries. Data about the religiosity and religious denomination of immigrants can be used to control for the difference between the religious denomination of the ethnic majority in a country and that of the immigrants a particular country receives. Seeing that migration flows are not static, information about the religious denomination of immigrants or ethnic minorities that participated in the ESS survey is not sufficient, for there is no information about the time of migration included in the ESS data set. Therefore, these controls could not be added to the analyses in the current study.
Nonetheless, this study has contributed to the scientific knowledge of the relation between attitudes towards immigrants and religiosity. In line with other studies, it demonstrates that distinguishing between different dimensions of religiosity is important (Ben-Nun Bloom et al., 2015; Doebler, 2014; Knoll, 2009; Stark & Glock, 1968). Moreover, even though religious diversity is not found to be influential on all dimensions of religiosity, there are strong indications that this contextual-level factor has an effect on individuals’ attitudes towards immigrants and should be taken into account in studies that attempt to explain differences in attitudes towards immigrants between countries.
6. Acknowledgements
Firstly, I would like to thank Dr. Bram Lancee for his supervision during the process of writing this master’s thesis. Without his valuable comments, and encouraging attitude I would not have been able to produce this work. Special acknowledgement to Marijn Keijzer for explaining me the basics of multi level modelling in Stata. I would also like to thank everyone that took the time to read my thesis or provided me with critical comments while discussing the topic during coffee breaks. The process has been very instructive thanks to all these people.
