ABSTRACT
Although Christian migrant groups make up a sizeable part of the immigrant population in Europe, little is known about their religiosity. This paper studies patterns of intergenerational change and proposes and tests hypotheses that specify when and why changes across generations are stronger. Using data from the European Social Survey (2002–2018) on 33 European countries, it is found that there is a strong pattern of intergenerational decline in the level of religiosity among Christian migrant groups in Europe. This process of religious decline is by no means universal. Results show that children from two foreign-born parents are much more religious than children from intermarried (foreign-born and native) couples. We also observe that intergenerational decline is much less pronounced in European countries that are more religious. Finally, when Christian migrant groups belong to a religious minority group, this is associated with higher levels of religiosity in both the first and second generation. It is argued that these insights can explain the ‘puzzling’ strong intergenerational religious transmission among Muslim migrant groups in Western European societies.
Background
Since the end of World War II, European societies have become increasingly secular (Bruce 2002). The process unfolds over generations, with each birth cohort being less religious than the previous one (Molteni and Biolcati 2018; Voas 2009; Voas and Chaves 2016). At the same time, Europe has witnessed a strong increase in the number of immigrants, many of them coming from highly religious countries (Van Tubergen and Sindradóttir 2011). Scholars of migration have noted the key role religion plays in the process of immigrant incorporation in European societies (Drouhot and Nee 2019; Foner and Alba 2008), and therefore an important empirical question is whether immigrants and their children follow the same secularization trajectory as ethnic majority populations (Kasselstrand and Mahmoudi 2020).
Previous work has largely focused on Muslim immigrant groups to answer this question (Connor 2010; Voas and Fleischmann 2012). Possible secularization trends have been studied in two ways: by directly comparing religious practices of the children with that of their parents, and by studying changes between ‘immigrant generations’, i.e. comparing those born abroad (1st generation) with those born in the host country (2nd generation). Studies generally find strong levels of intergenerational inheritance of religiosity in Muslim families (De Hoon and Van Tubergen 2014; Jacob and Kalter 2013; Molteni and Dimitriadis 2021). In France and Germany, studies report no or little change across generations (Diehl and Koenig 2009; Drouhot 2021; Soehl 2017b), whereas evidence suggests (some) signs of secularization across generations among Muslims in the Netherlands (Maliepaard et al.2010; Simsek et al.2018; Van De Pol and Van Tubergen 2014), Norway (Friberg and Sterri 2021) and the UK (Bisin et al.2008).
Much less is known about intergenerational changes among Christian immigrant groups in Europe. From a demographic perspective, this is surprising, because Christian immigrants are the largest religious group among ethnic minorities, and clearly outnumber Muslim immigrants. For example, figures for 2010 (Simsek et al.2018) suggest that among the children of immigrants the share of Christians is 38% in England (and 24% Muslim); 54% in Germany (31% Muslim), 34% in the Netherlands (27% Muslim) and 41% in Sweden (28% Muslim). To date, only a few studies have been done on Christian immigrant groups. The evidence so far suggests a pattern of secularization across immigrant generations in France (Drouhot 2021) and Norway (Friberg and Sterri 2021). A direct comparison of Christian parents with their children suggests intergenerational decline in religiosity in England, the Netherlands, Germany, and Sweden (Jacob and Kalter 2013; Simsek et al.2018).
We contribute to the literature on intergenerational changes among immigrant groups in Europe, in two ways. First, we make theoretical contributions. Previous work on Muslim and Christian immigrant groups has paid little, if any, attention to conditions that moderate the degree of intergenerational change. Most studies focus on the ‘average’ change across generations, without theorizing about and empirically testing when and why deviations from this overall pattern occur. We propose and test hypotheses that specify when, under which conditions, religious change is more likely to occur. Second, we expand the knowledge about Christian immigrant groups, about which little is known. By studying various Christian denominations (Protestants, Catholics, Eastern Orthodox, other), important insights can be obtained, which can also help understand the ‘puzzling’ strong intergenerational transmission of religiosity among Muslim groups – as we will argue in the conclusions of this paper. Compared to earlier work on Christian migrant groups, we go beyond the ‘single-country’ approach, combine nine rounds of the European Social Survey (2002–2018), and study patterns of intergenerational change among Christian groups in 33 European countries. To clarify, when we speak of ‘intergenerational change’, we refer to ‘immigrant generations’ and therefore focus on the relevance of the socialization context in terms of birthplace (origin or destination country) and not of birth cohort.
Theory and hypotheses
To understand patterns of change across immigrant generations, we test different hypotheses. These expectations are derived from different theoretical mechanisms. We start with assimilation theory, and subsequently introduce other theoretical arguments.
Most studies on intergenerational change rely on assimilation theory (Alba and Nee 1997), which argues that, over time, the cultural traits of immigrants and their children will become more like the mainstream culture of the host country. Because religious change largely happens across generations rather than within people's life-course (Voas and Chaves 2016), the assimilation theory predicts that each successive immigrant generation will more strongly adjust to the religious behavior and practices of the ethnic majority population. Findings indicate that on average immigrants in Europe are more religious than the ethnic majority population (Van Tubergen and Sindradóttir 2011). In the context of a secularizing Europe, this implies that on average one would expect to see a pattern of generational decline between the first and second generation. We therefore expect to see the following:
H1. The second generation is less religious than the first generation.
In this study, we go beyond the idea of a universal pattern of intergenerational change and theorize about conditions that modify the degree of change across changes. We consider conditions related to the family, the receiving country, and the majority-minority status of the immigrant groups.
To begin, conditions within the family may play a role. The second generation is a heterogenous group, and a key dimension of variation to consider is the family formation of the parents. Research findings indicate that most immigrants are married endogamously (i.e. with in-group members), but also that exogamous marriages are quite common, particularly among Christian origin groups (Dribe and Lundh 2011; Hannemann et al.2018; Lucassen and Laarman 2009; Van Tubergen and Maas 2007). We argue that marriages with ethnic majority members accelerate the process of integration, as immigrants and their children are more strongly exposed to the mainstream culture via the partner (Alba and Nee 1997; Kasselstrand and Mahmoudi 2020). Children raised in foreign-born couples will inherit more strongly conservative religious norms, values, and behavior of their parents. We therefore hypothesize that:
H2. The intergenerational decline in religiosity is stronger for the second generation with intermarried ‘foreign-native’ parents than for the second generation with ‘foreign-foreign’ parents.
Intergenerational change can also depend on the receiving country. One key condition, we suspect, is the level of religiosity among the receiving population. Although European societies are on the path of secularization, their levels of religiosity still differ strongly. Some countries are already very secular, whereas in other countries the population is highly religious. To illustrate, in the Czech Republic, around 65% of the population were not affiliated with a religion in 2008, while the number for Poland is 5% (Molteni and Biolcati 2018). We argue that in religious countries like Poland, immigrants and their children do not experience the same degree of pressure towards secularization as in secular countries like the Czech Republic. Although the first generation can also be affected by the religiosity of the host country and adjust their religiosity (Van Tubergen 2006), we suspect that the second generation will be more strongly influenced by the mainstream culture than the first generation. The reason for this is that religiosity tends to be rather stable across the life course, and the second generation is more strongly socialized into the host-country culture than the first generation, as they attend school, learn the language, and have more interethnic contacts (Drouhot and Nee 2019). We therefore expect:
H3. The higher the religiosity in the country of destination, the lower the intergenerational decline in religiosity from the first to the second generation.1
Another condition that may modify intergenerational change is what we suggest calling the ‘minority-majority’ status of immigrant groups, and which indicates whether their religious affiliation is either a minority or majority in the host country. Some European countries are mainly Protestant, others are Catholic or Eastern Orthodox and some have a mixed Christian population. To illustrate, among those who were affiliated with a religion in 2008, 97% were Protestant in Denmark, 95% were Catholic in Spain, while in Germany 45% were Catholic and 52% Protestant (Molteni and Biolcati 2018). These religious traditions may be consequential for the religious practices of immigrants, and their children, who may either become a ‘religious minority group’ (i.e. their religion is the minority religion in the host country) or belong to the ‘religious majority group’ (i.e. their religion is like the religious mainstream). Immigrants from (Catholic) Argentina who migrate to Spain will become a majority religion, whereas if they migrate to Denmark, they will become a minority religion.
We argue that the religious majority-minority status impacts the religiosity of the first and second generation. Drawing on different theoretical lines of thought, we formulate two empirically opposite predictions.
First, one could argue that when immigrant groups belong to the religious majority, that they are then stronger embedded in a social context which respects their religion, and their corresponding norms, values, and practices. For these groups, such as the Catholic immigrants from Argentina who settled in Spain, their traditional religious practices, inherited from the country of origin, are the mainstream in the receiving context, and reinforced by the religious practices of the ethnic majority. In addition, majority religious groups will find a religious infrastructure, such as churches of their denomination, that is needed for religious meetings. In turn, these religious gatherings protect immigrant groups from the erosion of religious values and norms (Van Tubergen 2013).
Conversely, for immigrant groups who become a religious minority, such a strong religious infrastructure is not available, which makes it more difficult to maintain their religious practices. In addition, because of their cultural distinctiveness, they may be subject to stronger pressures from the ethnic majority group to give up their religious traditions and to assimilate to the cultural-religious mainstream. This need not imply religious switching (e.g. Orthodox immigrants converting to Catholicism), as such changes are rather extreme in the European context. Rather, such religious minority groups may face more difficulties to maintain the same level of religiosity, such as attending gatherings, praying.
Assuming, again, that religious change is mainly across generations, and that the second generation is more strongly influenced by the culture of the host country than the first generation, one would expect, for example, that the second generation from Argentina will be more religious in Spain than in Denmark because of their religious majority status. In summary, when drawing on this theoretical line of thought one would expect the following:
H4a. Religious majority groups experience smaller religious decline from the first to the second generation than religious minority groups.
An implicit assumption H4a makes is that both religious minority and majority groups have the same ‘susceptibility’ to be influenced by the ethnic majority. However, it could be argued that the religious minority group is less susceptible to pressures from by ethnic majority. This alternative theoretical idea is based on the mechanism of ‘bounded’ or ‘biased’ assimilation (Flache et al.2017). This proposition states that assimilation (also known as ‘positive social influence’) only occurs when people are sufficiently similar to each other. People adjust their cultural traits (opinions, values, and practices) to others when their traits are similar, but when their traits differ too much, nothing will happen.
Religious minority-majority status might be key in this respect. Religious majority groups, such as Argentinian immigrants in Spain, will be affected by the religious culture of the ethnic majority members. Because of their common religious identity, there will be more opportunities for cross-cultural contact and dialogue. In other words: immigrants who belong to a religious majority are more susceptible to social influence from the ethnic majority, i.e. to the mainstream ideas, views and practices. Consequently, this would mean, drawing on the idea of biased assimilation, that they will follow the path of secularization that is common in European countries.
When instead immigrant groups occupy the status of a religious minority in their host country, then something different may happen according to the biased assimilation model. Their distinctive religious identity may create a cleavage with the ethnic majority, which hampers intergroup contact, dialogue and exchange of cultural opinions, values, and belief systems. Even though the assimilation sentiments in the ethnic majority population might be particularly strong towards those immigrant groups that make up religious minorities-as they are more culturally distinct-, these ‘social pressures’ may be less consequential, because of lower susceptibility. Because religious minority members are too different from the ethnic majority group, they will be less influenced by them. The second generation will not adjust to the mainstream culture but instead strongly inherit the religious norms and practices of their own immigrant group. The biased assimilation model leads to the following testable hypothesis:
H4b. Religious majority groups experience larger religious decline from the first to the second generation than religious minority groups.
Data and methods
Data
In this study, we use data from the European Social Survey (ESS). The ESS (2020, 2021) is a repeated cross-sectional survey, conducted for the first time in 2002, and then repeated once every two years, with the last round organized in 2018. We use data from all nine rounds (2002–2018). From this sample, we exclude (1) Israel and Turkey, (2) respondents whose parents were both born in the survey country, (3) those who do not self-identify as either Protestant, Catholic, Eastern Orthodox or belonging to other Christian groups on the question ‘Do you consider yourself as belonging to any particular religion or denomination? Which one?’. 2 This results in a sample of 25,224 respondents, from 192 origin groups in 33 receiving countries. 3
Dependent variables
Religiosity is a multidimensional concept, consisting of private and public aspects, believing and belonging. Therefore, especially in a multi denominational framework like that of this article, the optimal choice to perform is including different measures of religiosity in the analyses.
First, we study self-declared religiosity by simply referring to the question ‘How religious are you?’ to which respondents can answer by referring to a scale ranging from 0 (Not at all religious) to 10 (Very religious). Second, we study public practice by referring to the question asking interviewees ‘Apart from special occasions such as weddings and funerals, about how often do you attend religious services nowadays?’. For ease of interpretation, we transform this ordinal variable ranging from 1 (Every day) to 7 (Never) into a numerical one measuring the probability of attending church or another place of worship weekly. In doing so, we follow Hout and Greeley (1998) and assign a value of 0.99 to those who declare that they attend church or other places of worship weekly or more often (52 weeks over 52 weeks a year), a value of 0.23 to those who attend at least once a month (12 weeks over 52 weeks in a year), a value of 0.03 to those who attend only at special holy days, a value of 0.01 to those who attend less often and a value of 0.00 to those who never attend. Clearly, such a choice does not make the variable strictly cardinal but has the desired property of explicitly considering the distances between answer categories. Third, we study private prayer and, starting from the question ‘Apart from when you are at religious services, how often, if at all, do you pray?’, we follow a similar strategy as before and transform the ordinal variable into a numerical one measuring the probability of daily prayer (Every day = 0.99, More than once a week = 0.42, Once a week = 0.14, At least once a month = 0.03, Only on special Holy day = 0.008, Less often = 0.002, Never = 0). It should be emphasized that these values do not capture the full range, of course, and that more-precise measures of religiosity would be desirable.
Independent variables
To measure immigrant generation, we make a distinction between those born abroad from parents born abroad (i.e. first generation, and hereafter 1G) and those born in the survey country with at least one parent born abroad (i.e. second generation, and hereafter 2G). To capture intermarriage, we differentiate between (1) second generation with intermarried ‘foreign-native’ parents (i.e. those who have a foreign-born parent, and a parent born in the survey country, hereafter 2Gmixednative), and (2) those with intermarried ‘foreign-foreign’ parents (i.e. who have two foreign-born parents, hereafter 2Gforeign-foreign). Note that, due to low number of cases, it was impossible to empirically test for differences between those who have parents born in the same foreign country, and those whose parents were from different foreign countries.
We measure host-country religiosity with data aggregated from ESS. The aggregation is based on samples that exclude the first and second generation. For each country, we computed the average religiosity when pooling the nine waves. For the mixed-Christian countries (see below), these measures were constructed for the Protestant, Catholic, and Orthodox regions separately. In this way, three aggregated measures were created per country (or region): average self-declared religiosity (1–10), average weekly attendance (0–1), and average daily prayer (0–1). Depending on the analysis of the three different outcomes, we use the corresponding contextual measure.
The variable religious majority-minority status was constructed by taking individual-level information on self-identified religious affiliation (i.e. Protestant, Catholic, Eastern Orthodox) in ESS. We then combined micro-level data with a contextual measure of the dominant religious affiliation of the host country population, which was obtained from aggregating ESS data. Again, we excluded first and second generation, and averaged the religious affiliations across nine waves. As a threshold, we use 70% or more belonging to a certain denomination (among the religiously affiliated) to classify a country. In case no denomination reaches such a threshold (as in the case of Germany, the Netherlands, Switzerland, Great Britain, Estonia and Latvia), we split such countries into regions according to the religious majority within the single regions. For example, the region ‘NL-Prot’ comprises all the individuals living in Dutch regions where Protestantism is the most common denomination. Those declaring to belong to ‘other Christian denomination’ have been set as minority.
Table 1 presents an overview of the number of respondents (first and second generation) in each combination. The diagonal includes the religious majority cases (N = 18,015), the off-diagonal cases (plus the ‘Other Christian’) the number of respondents who belong to religious minority groups (N = 7209).
Religious affiliation respondent . | Religious majority receiving country . | Total . | ||
---|---|---|---|---|
Catholic . | Protestant . | Orthodox . | ||
Catholic | 10,395 | 2018 | 181 | 12,594 |
Protestant | 1513 | 2559 | 152 | 4224 |
Orthodox | 1440 | 608 | 5061 | 7109 |
Other Christian | 857 | 338 | 102 | 1297 |
Total | 14,205 | 5523 | 5496 | 25,224 |
Religious affiliation respondent . | Religious majority receiving country . | Total . | ||
---|---|---|---|---|
Catholic . | Protestant . | Orthodox . | ||
Catholic | 10,395 | 2018 | 181 | 12,594 |
Protestant | 1513 | 2559 | 152 | 4224 |
Orthodox | 1440 | 608 | 5061 | 7109 |
Other Christian | 857 | 338 | 102 | 1297 |
Total | 14,205 | 5523 | 5496 | 25,224 |
Note: Germany, Switzerland, the Netherlands, Great Britain, Estonia, and Latvia were divided into regions.
Control variables
We include the following controls: age (continuous), gender (female = 1), education (ISCED6, ‘Less than lower secondary education (ISCED 01)’ as reference), marital status (4 categories: ‘Married’ (reference), ‘Separated/divorced’, ‘Widowed’, ‘Never married’), employment (6 categories: ‘Employed’ (reference), ‘Student’, ‘Unemployed’, ‘Retired’, ‘Homemaker’, ‘Other’), and survey year (9 categories from 2002 to 2018). In addition, in models testing hypotheses 1 and 2, we include fixed effects for country of origin and destination.
Table 2 provides an overview of the descriptive statistics of the dependent and independent variables (without the control variables). The supplemental materials present data on the number of first and second generation per country of destination and religious affiliation (Table S1).
. | N (unique) . | Mean/ proportion . | SD . | Min . | Max . |
---|---|---|---|---|---|
Dependent | |||||
Self-declared religiosity | 25,084 | 6.22 | 2.42 | 0.00 | 10.00 |
Weekly attendance (probability) | 25,139 | 0.26 | 0.38 | 0.00 | 0.99 |
Daily prayer (probability) | 24,844 | 0.39 | 0.44 | 0.00 | 0.99 |
Generation | |||||
2G | 12,527 | 0.50 | |||
1G | 12,697 | 0.50 | |||
Generation 2 | |||||
2Gmixednative | 8868 | 0.35 | |||
2Gforeign-foreign | 3604 | 0.14 | |||
1G | 12,697 | 0.50 | |||
Denomination | |||||
Catholic | 12,594 | 0.50 | |||
Protestant | 4224 | 0.17 | |||
Orthodox | 7109 | 0.28 | |||
Other Christian | 1297 | 0.05 | |||
Religious majority/minority status | |||||
Religious majority | 18,015 | 0.71 | |||
Religious minority | 7209 | 0.29 | |||
Host country - average religiosity | |||||
Average self-declared religiosity | 39 | 4.72 | 1.07 | 2.39 | 6.99 |
Weekly attendance (probability) | 39 | 0.18 | 0.13 | 0.04 | 0.56 |
Daily prayer (probability) | 39 | 0.25 | 0.13 | 0.06 | 0.52 |
. | N (unique) . | Mean/ proportion . | SD . | Min . | Max . |
---|---|---|---|---|---|
Dependent | |||||
Self-declared religiosity | 25,084 | 6.22 | 2.42 | 0.00 | 10.00 |
Weekly attendance (probability) | 25,139 | 0.26 | 0.38 | 0.00 | 0.99 |
Daily prayer (probability) | 24,844 | 0.39 | 0.44 | 0.00 | 0.99 |
Generation | |||||
2G | 12,527 | 0.50 | |||
1G | 12,697 | 0.50 | |||
Generation 2 | |||||
2Gmixednative | 8868 | 0.35 | |||
2Gforeign-foreign | 3604 | 0.14 | |||
1G | 12,697 | 0.50 | |||
Denomination | |||||
Catholic | 12,594 | 0.50 | |||
Protestant | 4224 | 0.17 | |||
Orthodox | 7109 | 0.28 | |||
Other Christian | 1297 | 0.05 | |||
Religious majority/minority status | |||||
Religious majority | 18,015 | 0.71 | |||
Religious minority | 7209 | 0.29 | |||
Host country - average religiosity | |||||
Average self-declared religiosity | 39 | 4.72 | 1.07 | 2.39 | 6.99 |
Weekly attendance (probability) | 39 | 0.18 | 0.13 | 0.04 | 0.56 |
Daily prayer (probability) | 39 | 0.25 | 0.13 | 0.06 | 0.52 |
Notes: 2G = second generation; 1G = first generation; 2Gmixednative = those who have a foreign-born parent, and a parent born in the survey country; 2Gforeign-foreign = those with intermarried foreign-born parents.
Results
Unconditional mean differences in religiosity between immigrant generations and ethnic majority.
Unconditional mean differences in religiosity between immigrant generations and ethnic majority.
When analyzed per country, one can observe that most countries follow this general pattern, but also that few exceptions are present. Among them, the most striking one is Greece, which is the only country in which the second generation displays statistically significantly higher levels of religiosity on all the three dimensions than the first generation. A potential pitfall of comparing unconditional means of religiosity across immigrant generations is that unobserved heterogeneity can lead to artificial differences across generations or suppresses real change. It could be, for example, that in Greece, there has been a recent inflow of immigrants from rather secular countries -leading to ‘puzzling’ higher religiosity levels among the second generation. And, vice versa, it can also be the case that the first generation in a country is largely made up of a recently arrived, highly religious origin group, whereas the second generation is from other, less religious countries. Those from the same origin group might then have become less religious than their parents were, but because of the changing immigrant population, such change is artificially suppressed.
To account for heterogeneity, regression models were estimated that include origin and destination country fixed effects and individual-level controls (Table 3). Model 1 presents the bivariate results, Model 2 adds individual controls, Model 3 also includes fixed effects for the receiving contexts, and Model 4 adds fixed effects for origin country. While differences get smaller while moving from M1 to M4 (thus suggesting that both origin and destination contexts, their interplay and the corresponding migratory patterns play a role in shaping immigrants’ religiosity) the reading that Table 3 provides always goes in the direction of second-generation to be less religious than first generations, no matter the dimensions we are focusing on.
Self-declared religiosity | M1 | M2 | M3 | M4 |
2G (ref = 1G) | −0.432*** | −0.321*** | −0.397*** | −0.259*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 6.438*** | −31.340*** | −12.304 | −4.790 |
N | 25,084 | 24,363 | 24,363 | 24,363 |
Weekly attendance (probability) | M1 | M2 | M3 | M4 |
2G (ref = 1G) | −0.046*** | −0.034*** | −0.046*** | −0.025*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.278*** | 0.760 | 1.899 | 3.293** |
N | 25,139 | 24,411 | 24,411 | 24,411 |
Daily prayer (probability) | M1 | M2 | M3 | M4 |
2G (ref = 1G) | −0.089*** | −0.071*** | −0.068*** | −0.044*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.433*** | −0.164 | −0.556 | 0.348 |
N | 24,844 | 24,128 | 24,128 | 24,128 |
Self-declared religiosity | M1 | M2 | M3 | M4 |
2G (ref = 1G) | −0.432*** | −0.321*** | −0.397*** | −0.259*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 6.438*** | −31.340*** | −12.304 | −4.790 |
N | 25,084 | 24,363 | 24,363 | 24,363 |
Weekly attendance (probability) | M1 | M2 | M3 | M4 |
2G (ref = 1G) | −0.046*** | −0.034*** | −0.046*** | −0.025*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.278*** | 0.760 | 1.899 | 3.293** |
N | 25,139 | 24,411 | 24,411 | 24,411 |
Daily prayer (probability) | M1 | M2 | M3 | M4 |
2G (ref = 1G) | −0.089*** | −0.071*** | −0.068*** | −0.044*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.433*** | −0.164 | −0.556 | 0.348 |
N | 24,844 | 24,128 | 24,128 | 24,128 |
Note: * p < .05; ** p < .01; *** p < .001 (two-sided tests). 2G = second generation; 1G = first generation.
Controls: age, gender, education, marital status, employment, religious denomination, year of interview.
The analyses of religiosity do not take into account patterns of disaffiliation. Because we focused on the religiosity of Christian migrant groups, we excluded everyone without a Christian religion in both the first and second generation. This could create bias when comparing generations. For example, it could be that two foreign-born parents are both Protestants, but that their children, being raised in a secular host country, no longer identify themselves as Protestants. In Western European countries, a sizable share of the population no longer identify themselves with a religion, and hence immigrant assimilation might imply that the second generation follow this pattern of disaffiliation. If this happens on a larger scale, it would mean that the second generation with Christian parents is actually less religious than we find in our paper.
To examine this issue in detail, information is needed on the religious upbringing of the second generation. With these data, it would be possible to quantify the rate of disaffiliation among the second generation, for those who were raised in Christian families. However, the ESS only asked respondents whether they ever considered themselves belonging to a religion or denomination in case respondents indicated to have no current religion or denomination. Therefore, it is impossible to precisely assess how many second-generation members with Christian roots have become disaffiliated. However, we can assume that -with just a few exceptions- those who are currently Christian, have been raised as Christians. With this assumption, we calculated how many in the first and second generation have been raised as Christian, and how prevalent disaffiliation is in both generations. For example, among 2G respondent in our dataset there are 6489 currently affiliated Catholics and 1606 disaffiliated who indicated that they considered themselves Catholic in the past. This leads to a rate of disaffiliation among Catholics of (1606 / (6489 + 1606)) 19.84%. Table 4 presents for each Christian denomination the disaffiliation rates for the ethnic majority group, 1G and 2G. Results show that the overall disaffiliation rate among 2G (17.75%) is higher than among 1G (14.27%). When analyzed per denomination, we find this pattern for Catholics, Protestants and ‘other Christian’. The only exception is the Orthodox Christian group, among rates of disaffiliation are low among both 2G (6.20%) and 1G (7.26%).
. | Ethnic majority . | 2G . | 1G . |
---|---|---|---|
Past and present denomination | |||
Catholic | 110,023 | 6489 | 6105 |
Protestant | 46,046 | 2323 | 1901 |
Orthodox | 28,906 | 3254 | 3855 |
Other Christian | 3721 | 461 | 836 |
TOTAL | 188,696 | 12,527 | 12,697 |
Past denomination | |||
Catholic (now disaffiliated) | 16,731 | 1606 | 1182 |
Protestant (now disaffiliated) | 11,103 | 772 | 463 |
Orthodox (now disaffiliated) | 623 | 215 | 302 |
Other Christian (now disaffiliated) | 662 | 111 | 166 |
TOTAL | 29,119 | 2704 | 2113 |
Disaffiliation rate (%) | |||
Catholic | 13.20 | 19.84 | 16.22 |
Protestant | 19.43 | 24.94 | 19.59 |
Orthodox | 2.11 | 6.20 | 7.26 |
Other Christian | 15.10 | 19.41 | 16.57 |
TOTAL | 13.37 | 17.75 | 14.27 |
. | Ethnic majority . | 2G . | 1G . |
---|---|---|---|
Past and present denomination | |||
Catholic | 110,023 | 6489 | 6105 |
Protestant | 46,046 | 2323 | 1901 |
Orthodox | 28,906 | 3254 | 3855 |
Other Christian | 3721 | 461 | 836 |
TOTAL | 188,696 | 12,527 | 12,697 |
Past denomination | |||
Catholic (now disaffiliated) | 16,731 | 1606 | 1182 |
Protestant (now disaffiliated) | 11,103 | 772 | 463 |
Orthodox (now disaffiliated) | 623 | 215 | 302 |
Other Christian (now disaffiliated) | 662 | 111 | 166 |
TOTAL | 29,119 | 2704 | 2113 |
Disaffiliation rate (%) | |||
Catholic | 13.20 | 19.84 | 16.22 |
Protestant | 19.43 | 24.94 | 19.59 |
Orthodox | 2.11 | 6.20 | 7.26 |
Other Christian | 15.10 | 19.41 | 16.57 |
TOTAL | 13.37 | 17.75 | 14.27 |
2G = second generation; 1G = first generation.
Taken together, our findings show that 2G is less religious than 1G, and that disaffiliation is more common in 2G as well. We accept hypothesis 1.
To test hypothesis 2, we use the same modelling strategy as before, but this time by referring to a more refined version of the variable about immigrant generations (Table 5). When looking at M3, which includes individual-level controls and fixed effects for country of destination, we clearly see that for each dimension of religiosity, there is a statistically significant and substantively meaningful difference between those having both foreign-born parents, and children of one foreign-born and one native parent. Interestingly, when adding country of origin fixed effects, the differences become much smaller. The reason for this is that this variable captures (part of) the effect of mixed marriages with natives, as these native parents are (by definition) born in the destination country and less-religious (in general) than parents born abroad. These findings strongly support H2.
Self-declared religiosity | M1 | M2 | M3 | M4 |
2G mixed native (Ref: 2Gforeign-foreign) | −0.299*** | −0.289*** | −0.213*** | −0.097* |
1G (Ref: 2Gforeign-foreign) | 0.221*** | 0.119** | 0.251*** | 0.195*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 6.217*** | −31.694*** | −12.544 | −5.081 |
N | 25,030 | 24,316 | 24,316 | 24,316 |
Weekly attendance (probability) | M1 | M2 | M3 | M4 |
2G mixed native (Ref: 2Gforeign-foreign) | −0.017* | −0.032*** | −0.043*** | −0.030** |
1G (Ref: 2Gforeign-foreign) | 0.034*** | 0.011 | 0.016* | 0.004 |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.244*** | 0.696 | 1.921 | 3.295** |
N | 25,085 | 24,364 | 24,364 | 24,364 |
Daily prayer (probability) | M1 | M2 | M3 | M4 |
2G mixed native (Ref: 2Gforeign-foreign) | −0.030*** | −0.043*** | −0.043*** | −0.024** |
1G (Ref: 2Gforeign-foreign) | 0.068*** | 0.041*** | 0.038*** | 0.028*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.365*** | −0.23 | −0.518 | 0.353 |
N | 24,791 | 24,081 | 24,081 | 24,081 |
Self-declared religiosity | M1 | M2 | M3 | M4 |
2G mixed native (Ref: 2Gforeign-foreign) | −0.299*** | −0.289*** | −0.213*** | −0.097* |
1G (Ref: 2Gforeign-foreign) | 0.221*** | 0.119** | 0.251*** | 0.195*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 6.217*** | −31.694*** | −12.544 | −5.081 |
N | 25,030 | 24,316 | 24,316 | 24,316 |
Weekly attendance (probability) | M1 | M2 | M3 | M4 |
2G mixed native (Ref: 2Gforeign-foreign) | −0.017* | −0.032*** | −0.043*** | −0.030** |
1G (Ref: 2Gforeign-foreign) | 0.034*** | 0.011 | 0.016* | 0.004 |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.244*** | 0.696 | 1.921 | 3.295** |
N | 25,085 | 24,364 | 24,364 | 24,364 |
Daily prayer (probability) | M1 | M2 | M3 | M4 |
2G mixed native (Ref: 2Gforeign-foreign) | −0.030*** | −0.043*** | −0.043*** | −0.024** |
1G (Ref: 2Gforeign-foreign) | 0.068*** | 0.041*** | 0.038*** | 0.028*** |
Control variables | Yes | Yes | Yes | |
Destination (Fixed Effects) | Yes | Yes | ||
Origin (Fixed Effects) | Yes | |||
Constant | 0.365*** | −0.23 | −0.518 | 0.353 |
N | 24,791 | 24,081 | 24,081 | 24,081 |
Notes: * p < .05; ** p < .01; *** p < .001 (two-sided tests). 2G = second generation; 1G = first generation; 2Gmixednative = those who have a foreign-born parent, and a parent born in the survey country; 2Gforeign-foreign = those with intermarried foreign-born parents.
Controls: age, gender, education, marital status, employment, religious denomination, year of interview.
To test hypotheses H3 to H4b, we use multilevel random intercept models in which individuals (level 1) are nested into countries or regions, in case of mixed religious nations (level 2). Table 6 presents the results for all models.4
. | H3 . | H4a vs. H4b . | ||||
---|---|---|---|---|---|---|
M1 . | M2 . | M3 . | M4 . | M5 . | M6 . | |
Self-declared religiosity . | Weekly attendance (probability) . | Daily prayer (probability) . | Self-declared religiosity . | Weekly attendance (probability) . | Daily prayer (probability) . | |
Fixed part | ||||||
2G (ref: 1G) | −0.823*** | −0.046*** | −0.125*** | −0.386*** | −0.037*** | −0.061*** |
Destination country self dec. rel. | 0.157* | |||||
Destination country weekly att. | 0.631*** | |||||
Destination country daily prayer | 0.442*** | |||||
2G#Destination country self dec. rel. | 0.089** | |||||
2G#Destination country weekly att. | 0.024 | |||||
2G#Destination country daily prayer | 0.234*** | |||||
Religious minority (ref: majority) | 0.249*** | 0.038*** | 0.061*** | |||
2G#Religious minority | 0.053 | 0.004 | 0.01 | |||
Controls | Yes | Yes | Yes | Yes | Yes | Yes |
Constant | 5.708*** | 0.137*** | 0.147*** | 6.337*** | 0.233*** | 0.225*** |
RP2 (country split) | ||||||
var(cons) | 0.191 | 0.003 | 0.003 | 0.249 | 0.009 | 0.009 |
RP1 (individuals) | ||||||
var(cons) | 5.392 | 0.135 | 0.169 | 5.381 | 0.135 | 0.168 |
N (individuals) | 24,363 | 24,411 | 24,128 | 24,363 | 24,411 | 24,128 |
N (countries/regions) | 39 | 39 | 39 | 39 | 39 | 39 |
. | H3 . | H4a vs. H4b . | ||||
---|---|---|---|---|---|---|
M1 . | M2 . | M3 . | M4 . | M5 . | M6 . | |
Self-declared religiosity . | Weekly attendance (probability) . | Daily prayer (probability) . | Self-declared religiosity . | Weekly attendance (probability) . | Daily prayer (probability) . | |
Fixed part | ||||||
2G (ref: 1G) | −0.823*** | −0.046*** | −0.125*** | −0.386*** | −0.037*** | −0.061*** |
Destination country self dec. rel. | 0.157* | |||||
Destination country weekly att. | 0.631*** | |||||
Destination country daily prayer | 0.442*** | |||||
2G#Destination country self dec. rel. | 0.089** | |||||
2G#Destination country weekly att. | 0.024 | |||||
2G#Destination country daily prayer | 0.234*** | |||||
Religious minority (ref: majority) | 0.249*** | 0.038*** | 0.061*** | |||
2G#Religious minority | 0.053 | 0.004 | 0.01 | |||
Controls | Yes | Yes | Yes | Yes | Yes | Yes |
Constant | 5.708*** | 0.137*** | 0.147*** | 6.337*** | 0.233*** | 0.225*** |
RP2 (country split) | ||||||
var(cons) | 0.191 | 0.003 | 0.003 | 0.249 | 0.009 | 0.009 |
RP1 (individuals) | ||||||
var(cons) | 5.392 | 0.135 | 0.169 | 5.381 | 0.135 | 0.168 |
N (individuals) | 24,363 | 24,411 | 24,128 | 24,363 | 24,411 | 24,128 |
N (countries/regions) | 39 | 39 | 39 | 39 | 39 | 39 |
Notes: * p < .05; ** p < .01; *** p < .001 (two-sided tests) 2G = second generation; 1G = first generation.
Controls: age, gender, education, marital status, employment, year of interview.
Models M1-M3 include variables for destination-country religiosity, a dummy variable for 2G (versus 1G) and interactions between generation and destination-country religiosity (and controls). We find that destination-country religiosity is positively and statistically significantly correlated with the religiosity of the respondents and that such an effect is as big as to explain most of the individual variation in the probability of weekly attendance and a relevant portion of the individual probability of daily prayer. This main effect can be read as the pattern for the first generation, and it could be interpreted as an immigration selection effect (e.g. more-religious immigrants being more likely to migrate to more-religious countries instead of more-secular nations), a remigration selection effect (e.g. immigrants whose religiosity is similar to the host country are less-likely to remigrate), and an incorporation-assimilation effect (e.g. immigrants adjusting their religiosity to the religiosity of the native mainstream). Regardless of the underlying mechanisms, we find this pattern for all three dimensions of religiosity.
The interaction variables between destination-country religiosity and 2G can be interpreted as how much, if any, the second generation differs in the relationship between destination-country religiosity and their own religiosity. We find for both self-defined religiosity and praying a positive statistically significant and quite strong relationship, with the effect size of the interaction variable being particularly pronounced for the latter. In fact, the probability of daily prayer is 0.125 lower for 2G compared to 1G when the level of host country religiosity is close to zero while only 0.009 lower when it is around 0.5 (the highest value in our sample of countries). A similar figure for the self-defined religiosity measure ranges from 0.64–0.20 (on a 0–10 scale). This means that, for these two dimensions, the religiosity of 2G positively correlates with the religiosity of the native mainstream -and it does so more strongly than for 1G. Consequently, this implies that the higher (lower) the religiosity in the country of destination, the lower (higher) the intergenerational decline in religiosity from the first to the second generation. With respect to self-identified religiosity and praying, our results therefore confirm H3.
In models M4-M6, we look at the role of being a religious minority. We formulated two opposing hypotheses on the potential effect of being a majority or minority religion for intergenerational changes: H4a and H4b. While H4a states that religious majority groups experience lower religious decline from 1G to 2G, H4b expects the opposite (based on the mechanism of ‘biased assimilation’). The empirical tests for these two competing hypotheses are represented by the interaction coefficients between 2G and the minority status provided in the models from M4 and M6. We do not find any statistically significant interaction effect.
Interestingly, however, we find a consistent main effect for being a religious minority. This means that 1G Christians who belong to the religious minority are already more religious than those who belong to the religious majority: they have higher religious self-identification, attend religious meetings more frequently, and also pray more often than Christian immigrants who have the same religion as the mainstream religion in the host country. The absence of any interaction effects means that for both 1G and 2G alike, being a minority religion is associated with higher levels of religiosity.
Conclusions and discussion
The motivation for this study was twofold. First, to increase our knowledge of intergenerational changes among Christian immigrant groups, which have received little attention in scientific research despite being far bigger in population size than Muslim groups. Second, to go beyond the study of the average pattern of intergenerational change, and to examine conditions that modify the degree of changes across generations. Three main conclusions can be drawn from this study.
First, we find a pattern of intergenerational decline in the level of religiosity among Christian migrant groups in Europe. We observe this secularization process regarding all three dimensions of religiosity. Relative to the first generation, those from the second generation score substantially lower on self-declared religiosity, they attend religious meetings less often, and they pray less frequently. We also find that disaffiliation is more common in the second generation. This ‘average’ change is in line with the incorporation/assimilation hypothesis of immigrant integration (Alba and Nee 1997; Drouhot and Nee 2019), as this process of religious decline among minority groups makes them more similar to the secularizing majority.
Second, this process of religious decline is by no means universal among Christian migrant groups. There are deviations to this average pattern, which are related to conditions in the family and the host country. Results show that children from two foreign-born parents are much more religious than children from intermarried (foreign-born and native) couples. We also observe that intergenerational decline is much less pronounced in European countries that are more religious.
Third, findings from our study on different Christian denominations (Protestant, Catholic, Eastern Orthodox, other), suggest that when Christian migrant groups belong to a religious minority group (e.g. Protestant migrants in Catholic Italy) rather than a religious majority group (e.g. Protestant migrants in Protestant Denmark), this is associated with higher levels of religiosity in both the first and second generation. This finding could be due to selective (re)migration or result from ‘reactive religiosity’ processes. Another possible explanation is provided with the ‘biased assimilation’ model (Flache et al.2017). This model argues that social influence only occurs when people are ‘sufficiently’ similar to each other, assuming a threshold beyond which assimilative forces start to work. Belonging to a religious minority might be a strong barrier for social influences from the cultural mainstream.
This study has certain limitations. First, the European Social Survey is not specifically designed to study immigrant populations. This might particularly affect non-response among the foreign-born, because of the lack of destination-language skills. Possibly, such patterns of non-response correlate with religiosity. Second, we focused on intergenerational change in the sense of immigrant generations (place of birth) but not in the demographic sense, i.e. changes in religiosity between parent and child cohorts. Third, the cross-sectional nature of the data does not allow us to say more about selectivity at arrival in the host country, and about selective return migration. It is conceivable, for example, that immigrants who are not feeling comfortable with the secular values and practices in their host country, are more likely to remigrate. We speculate that these issues lead to an underestimation of the religiosity of the first generation, and hence would imply that the intergenerational decline in religiosity among Christian groups is even larger than we find in this study.
The findings of this study on the effect of religious minority/majority status can possibly help explaining the ‘puzzling’ strong intergenerational religious transmission among Muslim migrant groups in Western European societies -as compared with the overall pattern of religious decline among Christian migrant groups. The ‘deviant’ pattern among Muslims is puzzling considering the strong pressures from the ethnic majority population to assimilate. Because of their cultural distinctiveness, Muslim migrants may be subject to stronger pressures from the ethnic majority group than Christian migrants to give up their religious traditions and to assimilate to the cultural-religious mainstream. But, contrary to what is expected, the religiosity among Muslim groups, also in the second generation, is higher than among Christian groups.
A possible explanation for this ‘puzzle’, based on our findings, is that the distinctive Muslim identity may create a cleavage with the ethnic majority, which hampers intergroup contact, dialogue and exchange of cultural opinions, values, and belief systems. Even though the assimilation sentiments in the ethnic majority population might be particularly strong towards Muslim migrants, these ‘social pressures’ may be less consequential, because of lower susceptibility. Because Muslims, and other religious minority members, are more different from the ethnic majority group than religious majority groups, they will be less influenced by the ethnic majority.
Fewer cross-ethnic contacts and larger cultural differences with the mainstream contribute to the ‘lower susceptibility’ among Muslim migrants -as compared with Christian migrant groups. For example, research findings indicate that Muslim migrant groups have higher endogamy rates (Carol 2016; Huschek et al.2012; Kalmijn and Van Tubergen 2006), and more-conservative values (Soehl 2017a) than Christian groups -thereby providing a ‘barrier’ to secularization forces. Muslim migrant groups are always the religious minority group in Western European countries -unlike Christian groups, which may or may not have a religion similar to the mainstream majority. The cultural-religious similarity to the mainstream may be below the threshold for assimilation to work for the first-generation Muslims, and possibly also for the second generation. The Muslim migrant groups are at the pole of the continuum of religious-cultural similarity, even further away from the majority culture than the Christian minority groups. And this cultural dissimilarity provides a strong counterforce against secularization.
The ‘strong’ version of this biased assimilation model argues that social influence only occurs when people are ‘sufficiently’ similar to each other, but a more ‘flexible’ version can do without assuming such a threshold (which is difficult to define a priori). Based on such a more-flexible model, one could argue that when migrants are more similar in their cultural-religious practices to the mainstream cultural-religious practices in their host country, they are more strongly affected by the religious changes that take place over there. To test this idea, we encourage future work on the role of not only religious affiliation as an element of cultural-religious similarity, but also by looking at similarity in language and cultural values, and patterns of cross-ethnic contact.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Footnotes
There are also migrants who were born and raised in less-religious societies, and who migrated to European countries that have higher levels of religiosity. When strictly following the adaptation-assimilation mechanism, one would then expect to see that for these groups the second generation would become more religious than the first generation. However, in the ESS data we use, there are few such ‘deviant’ cases. They occur in Greece and Poland -two religious nations- but hardly elsewhere. Given the small number of cases, it is impossible to study if patterns of intergenerational change look different. Furthermore, theoretical and empirical work in the sociology of religion suggest that religious capital acquired from the parents (i.e., being born and raised in a religious family) is needed to become a religious person (Iannaccone 1990; Molteni 2020), and therefore we do not expect such intergenerational increase in religiosity. Rather, we expect to see no change in these cases, or only limited religious decline across generations.
In addition to these major criteria, we also removed those observations in which either the respondent's country of birth or his father's or mother's one is set to 77 (Refuse to answer), 88 (Don't know), 99 (No answer, including those excluded for anonymity reasons) plus those younger than 15 or older than 85.
All the replication materials can be found here: https://osf.io/gx3uf/
For ease of presentation the coefficients for the control variables have not been reported here but in the supplementary material (Table S3)
References
Frank van Tubergen is a professor of sociology at Utrecht University, Netherlands. His research interests include immigration, demography, religion, and big data. He is a fellow of the European Academy of Sociology and the author of ‘Introduction to Sociology’ (Routledge, 2020).
Francesco Molteni is an Assistant Professor at the University of Milan. His research interests include religion and religious change, immigration, and human values. He is member of the Italian EVS (European Values Study) group and of the SPS Trend group. He is the author of ‘A Need for Religion: Insecurity and Religiosity in the Contemporary World’ (Brill, 2020).
Author notes
EDITED BY Patrick Präg
Supplemental data for this article can be accessed at https://doi.org/10.1080/14616696.2022.2044067