Acculturation, decoupling, or both? Migration’s impact on the linkage between religiosity and gender equality attitudes

ABSTRACT

Turkish migrants are one of the largest minority groups within Europe, the majority of which is Muslim. Radical right politicians stress the threat of Islam for what they consider European culture. Yet we know that migrant communities can adapt to the destination society culturally, albeit not always and in complex ways. This study aims advance our understand of this complex matter, as it analyses how such attitudinal developments take place and in what way religion plays a role. Through acculturation processes, where Muslim migrants become more secular and consequently more open to gender equal norms, and/or through decoupling, where migration leads to weaker connections between religious identity and patriarchal gender norms. With the use of the 2000Families data on Turkish Muslim European-migrants and Turkish-Muslim stayers, we find that the connection between religiosity and gender equality attitudes is gender-dependent. Across the board findings indicate limited support for both mechanisms. However, we do cautiously conclude that we find both acculturation and decoupling processes among migrant men with regard to individual religiosity, while we find decoupling between communal religiosity and gender-equality attitudes in migrant women. This suggests migrants undergo a range of acculturation processes simultaneously, but that the linkages for men and women differ between these dimensions.

KEYWORDS:

• Muslim migrants
• acculturation
• religion
• gender attitudes

1. Introduction

For over half a century, Europe has received migrants from across the world. Among them are about 4.5 million of Turkish descent (Ministry of Foreign Affairs, Turkey 2016). This makes Turks one of the largest ethnic minority groups in Western Europe (Guveli, Ganzeboom, Platt et al. 2016). Turks have been part of the migration debate since the 1960s (Van Klingeren, Zaslove, and Verbeek 2017); however, the debate as well as the general discourse on integration and immigration has changed massively since that time. Especially since the turn of the millennium, the debate in Western Europe no longer simply focusses on the integration of migrants as such, but often particularly emphasises Muslim migrants and cultural integration, thereby referring mostly to migrants of Turkish and Moroccan descent (Kogan and Weißmann 2019; Roggeband and Verloo 2007; Spierings et al. 2015).

More recently, the core symbol of the supposed cultural divide between Western Europe and its migrants has become gender role attitudes (Diehl and Koenig 2009; Shield 2017; Spierings et al. 2015). In these discussions, Muslim culture is linked to subordination, oppression, and abuse towards women. It is therefore considered incompatible with what is characterised as progressive West-European culture. Empirical studies have also shown a difference in average support for gender equality among European ‘natives’ and (certain groups of) Muslim migrants, finding piety to be negatively associated with gender equality attitudes as well (Davis and Greenstein 2009; Diehl and Koenig 2009; Diehl, Koenig, and Ruckdeschel 2009; Kretschmer 2018; Röder and Mühlau 2014; Spierings 2015, 2018).

Changes in both the degree of religiosity and more equalitarian gender attitudes have been linked to migration processes. On the one hand migration is argued to go hand in hand with maintenance of values stemming from the origin society at time of migration. These values serve as a stable marker for one’s identity (Diehl and Koenig 2009; Huschek, de Valk, and Liefbroer 2011), and are often intergenerationally transmitted (Diehl, Koenig, and Ruckdeschel 2009; Huschek, de Valk, and Liefbroer 2011; Kretschmer 2018). At the same time, immigrants tend to dissimilate from their origin society and acculturate into the culture of the destination society (Beek and Fleischmann 2019; Diehl, Koenig, and Ruckdeschel 2009; Guveli, Ganzeboom, Platt et al. 2016; Kogan and Weißmann 2019; Röder and Mühlau 2014; Spierings 2015). However, as Kogan, Fong, and Reitz (2019, 10) conclude in a recent overview, this relationship between religiosity and outcomes (like support for gender equality) among migrant communities is ‘complicated and multifaceted’.

Part of this complexity is the tension between acculturation and decoupling. The former has been discussed above already, but several studies also hint at another process: that is the changing role of religion in reshaping gender equality attitudes. A process that has hardly been studied empirically (cf. Banfi, Gianni, and Giugni 2016; Beek and Fleischmann 2019; Diehl and Koenig 2009; Diehl, Koenig, and Ruckdeschel 2009; Röder 2014). Of these and building on Diehl, Koening and colleagues’ observations, Röder (2014) was one of the first to systematically connect all three elements and to explicitly address the question what migration’s systematic impact is on the relationship between Islamic religiosity and gender equality attitudes.

Röder asks whether migration might lead to changes in religiosity, and therefore to changes in gender attitudes. And/or whether migration leads to weaker connections between religious identity and patriarchal gender norms. Thereby presenting the beginnings of a new perspective on the relationship between migration and gender equality attitudes: one of decoupling. Since the two mechanisms are not mutually exclusive, her findings do indicate support for the decoupling claim in addition to acculturation. She finds religiosity is no longer, or less strongly negatively, connected to gender equality attitudes among women who have resided in Western countries longer.

In the current study, we delve deeper into decoupling mechanisms by theorising them more systematically and analysing how the linkage between both communal- and individual religiosity on the one hand, and gender equality attitudes on the other hand might have changed due to migration. We also pay particular attention to how this process itself might be gendered, as recent studies have demonstrated that the link between religiosity and gender equality attitudes differs between men and women (Glas, Spierings, and Scheepers 2018; Idema and Phalet 2007; Röder and Mühlau 2014).

Empirically, our study is based on the 2000Families data (Ganzeboom et al. 2015; Guveli, Ganzeboom, Baykara-Krumme et al. 2016; Guveli, Ganzeboom, Platt et al. 2016; Guveli et al. 2017) on Turkish Muslim migrants and non-migrants. These data are collected starting from five high-sending regions in Turkey, and finally interviewing 5980 respondents that migrated across Europe, particularly to Germany, the Netherlands, Belgium, and Austria. ¹ Following the respondents’ family lineage, information on both migrant and non-migrant family members is collected, allowing us to compare those who stayed to those who migrated. A crucial feature, as the decoupling thesis assumes citizens with a migrant background develop other patterns than those who stayed in the country of origin.

The focus on citizens with a Turkish background also means we zoom in on the possibility of decoupling among a large group of European migrants. The rather strong Turkish ethno-religious infrastructure in Europe and their connection to the Turkish authorities is crucial in terms of the wider relevance of this selection. This makes it a rather conservative test. If we are to find evidence for decoupling among this group, it is very plausible that it is found among other migrant groups from conservative Muslim-majority societies.

In sum, we thus add to the literature by further theorising and testing the decoupling thesis, by zooming in on specific large-migration contexts, by investigating gender-specific effects, and by comparing migrants to those who stayed in the origin country.

2. Theoretical background

2.1. Religiosity and gender equality attitudes

Both people’s religiosity and norm-based attitudes are considered part of their identity. People are socialised into this by their direct surroundings, either in their formative childhood years (Glenn 2003; Inglehart 1997) or after (Lubbers, Jaspers, and Ultee 2009; Spierings 2015). Neither of the two, however, is constant, and throughout one’s life both one’s religiosity and (gender-equality) attitudes can change.

The two concepts are generally treated as part and parcel of a larger cultural cluster, as the claim that piety coincides with holding less liberal attitudes has been supported by empirical studies in the past. Because most dominant interpretations of (monotheistic) religions contain strong patriarchal norms, people for whom religion plays an important role in their life are likely to be socialised in the norms and values set by this religious institution (Burdette, Ellison, and Hill 2005).

Zooming in on Islam, Alexander and Welzel (2011), for example, found that ‘Muslim support for patriarchal values is robust against various controls’, identifying ‘mosque attendance as a mechanism to sustain Muslim support for patriarchy in non-Muslim societies’ (249). Similarly, in a more refined analysis of this linkage, Glas, Spierings, and Scheepers (2018) show that several dimensions of Islamic religiosity (e.g. attending religious service and devotion) fuel opposition towards gender equality in the Middle East. Hence, socialisation in dominant Islamic interpretations appears to weaken support for gender equality.

At the same time, there have been deviating findings, and it has become clear that different dimensions of religiosity play different roles (Glas, Spierings, and Scheepers 2018; Spierings 2018). To understand how religiosity, gender equality, and migration interrelate, first we have to conceptually distinguish between two core dimensions of religiosity (cf. Glas, Spierings, and Scheepers 2018; Guveli 2015): individual religiosity and religious communal integration.

Individual religiosity encompasses different manifestations of the extent to which one defines oneself as religious and acts accordingly in private. This can encompass different elements, such as subjective religiosity, which relates directly to the affective component of religion and includes ‘feelings and commitment to religious existence, lifestyles and institutions’ (Guveli 2015, 45), as well as individual religious acts, such as the practice of prayer, which is often not done in public.

Next, religious communal integration acts as an indicator of a person’s connection to their religious community and exposure to religious norms, which commonly prescribe gender hierarchy rather than equality. This dimension of religiosity firstly includes the element of associational involvement or participation, a behavioural type of religiosity often measured by looking at religious attendance. This is thus a public form of religiosity that provides an opportunity for religious socialisation (Alexander and Welzel 2011; Glas, Spierings, and Scheepers 2018; Spierings 2018). Another element is orthopraxy, which denotes the extent to which people comply with religiously prescribed rules. Although this is a personal expression of religiosity, it is one that is very visible in public gatherings. Openly complying or not complying to these rules can affect one’s position within a religious community and perceived degree of integration within it.

Communal religious integration thus acts as an indicator of both integration into one’s religious community and of exposure to religious norms. For this reason – following the logic of socialisation processes – we expect that religious communal integration has a stronger effect on opposition to gender equality than individual religiosity, which allows for more diverse, personal, and deviating interpretations of Islam (Glas, Spierings, and Scheepers 2018).

H1: More religious people are more opposed to gender equality, and this connection will be stronger for communal than for individual religiosity.

2.2. Migration’s impact (1): acculturation

Compared to other Western and non-Western regions, in Europe religion plays an ever-decreasing role (Inglehart 1997; Röder 2014). Moreover, as religiosity often coincides with less egalitarian values (Diehl, Koenig, and Ruckdeschel 2009; Kretschmer 2018), secularisation should go hand in hand with more egalitarian values. According to assimilation theory (Alba and Nee 2003), migrants who end up in a secular host society tend to become more pro-egalitarianism over time and generations, i.e. they dissimilate from their less secular origin society.

Recent studies have stressed that such processes should be understood in light of how the society of origin develops, since the origin context is also not static and many cross-border ties to the origin society remain (Guveli et al. 2017; Guveli, Ganzeboom, Platt et al. 2016; Röder and Mühlau 2014). Considering Turkey as the context of origin, it needs to be noted that it is more Western oriented than many other Muslim-majority countries. It has modernised greatly since the first wave of Turkish migrants arrived in Europe. Yet the country is still home to many devout religious practitioners and despite numerous efforts, patriarchal values have remained upright in Turkey (Engin and Pals 2018). Overall, and compared to European destination countries, Turkey can still be described as a relatively religious and conservative country, as is for instance illustrated by a fifteen percentage point difference in support for gender equality in the years leading up to the survey we use here (Guveli, Ganzeboom, Platt et al. 2016; Spierings 2015). Hence, religiosity and conservative attitudes regarding gender inequality are supposedly both still relatively high.

Meanwhile, among those exposed to secular societies, devoutness to one’s religion is at risk. Adaptation to the host society could thus mean acculturation and dissimilation with regard to religious beliefs, which would cause Muslim migrants in Western Europe to become less religious and therefore more likely to develop gender-equalitarian attitudes, as has been theorised and found before (e.g. Diehl and Koenig 2009; Kogan and Weißmann 2019).

H2a: Migrants will be more supportive of gender equality than non-migrants, which can be explained by a lower level of religiosity among the migrant group.

To further understand the migration–religiosity–gender-equality linkage, it should be considered how the acculturation mechanism develops over generations. Socialisation theory emphasises the role that early-life context is crucial for a person’s attitudes later in life. Primary socialisation concerns the transfer of norms and values through one’s parents or grandparents, whereas secondary socialisation denotes the transfer through institutions such as school or church (see Lubbers, Jaspers, and Ultee 2009). Not surprisingly, several studies show the importance of intergenerational transmission in the maintenance of religiosity and gender equality attitudes among migrant communities (Diehl and Koenig 2009; Diehl, Koenig, and Ruckdeschel 2009; Huschek, de Valk, and Liefbroer 2011; Kretschmer 2018; Spierings 2015).

However, parents are just one piece of the puzzle, as the larger context matter as well. Moreover, the socialisation effect can be expected to be strongest among those who grew up in the origin country, where primary- and secondary socialisation are aligned. Among those who stayed in the origin society, religiosity will be higher on average, leading to stronger opposition towards gender equality. Similarly, among first-generation migrants – who grew up in Turkey – religiosity and opposition to gender equality will also be relatively strong. However, for those who grew up in a more liberal context (generation 1.5 and beyond), the adoption of religious norms is likely to be weaker and consequently their support for gender equality higher. In short, despite efforts of parent to transmit their religiosity, but in line with modest effect found in religiosity studies (Guveli 2015; Kretchmer 2018; Maliepaard and Lubbers 2013), we expect that:

H2b: Migrants have more liberal gender attitudes if they belong to a later migration generation.

2.3. Migration’s impact (2): decoupling

Yet adaptation might not happen with regard to religiosity and attitudes both. Whereas religiosity is often part of a migrant’s social identity, patriarchal gender norms and traditions are not intrinsically part of religion, and thus need not be part of one’s identity’s defining characteristics (Barlas 2002; Glas, Spierings, and Scheepers 2018; Kandiyoti 1988). Hence, migration may lead to a situation in which Muslim migrants remain just as religious as always but start to disagree with or reinterpret certain aspects of their culture, such as gender-related values due to influences from the host country (Glas, Spierings, and Scheepers 2018; Röder 2014). And while complete acculturation is in line with the idea that people assimilate to the host society by adapting to the level of religiousness within Europe (Van Tubergen and Sindradóttir 2011), partial acculturation indicates a changing relationship between religion and migrants’ attitudes (i.e. decoupling).

This means religion loses its grip on people’s attitudes in the more secular societies, whereby religion becomes more private and gets a symbolic function that demarcates groups, but that dictates other values less strongly (Diehl and Koenig 2009; Diehl, Koenig, and Ruckdeschel 2009). And as people are confronted with a multitude of opinions through social interaction, democratic signalling (Tétreault, Meyer, and Rizzo 2009), freedom of press (Zakarriya 2014), and more progressive family laws (Htun and Weldon 2015). This causes them to rethink or (un)consciously adjust their stance on gender equality and adapt to a more liberal point of view.

The social-mobility strategy of social-identity theory (SIT; Tajfel and Turner 1979) provides a new explanation for such decoupling processes (cf. Röder 2014). SIT states that people are looking to achieve a positive identity through group membership. However, if the relative status of one’s social group is considered to be inadequate, people are inclined to apply compensatory strategies to change this. One of which is social mobility: people attempt to climb the social ladder by adapting to the norms of the higher-placed social group (Taylor and Moghaddam 1987). Doing so could however come at the cost of abandoning one’s initial social group. We should consider that, as religion has a strong identifying function in the destination society (Diehl and Koenig 2009), abandoning it completely is unlikely. Meanwhile stretching the norms within the religious boundaries is likely. Since gender equality attitudes are less instrumental, they are more likely to be subject to extra-familial socialisation in the destination society. As a consequence, migrants are more willing to adapt to majority attitudes in order to get absorbed into the majority group and enjoy similar privileges. Adjusting one’s attitudes thus facilitates social mobility while one’s personal or (sub)communal integration remains intact.

H3a: The connection between religion and conservative gender attitudes is weaker among those who migrated than among those who did not.

However, among people who grew up in the host society, there is general a disconnect between primary- and secondary socialisation. And in spite of the first generation’s efforts to instil the old values, the influence of the host-country is generally considered quite strong (see Diehl, Koenig, and Ruckdeschel 2009; Röder 2014; Spierings 2015). Yet, evidence also suggests that transmission of religiosity across generations is relatively strong (Diehl and Koenig 2009; Guveli 2015; Kashyap and Lewis 2013), which implies decoupling rather than complete acculturation is likely to strengthen with each migrant generation, given that the primary socialisation of the origin country’s gender norms through parents wanes. Accordingly, Banfi, Gianni, and Giugni (2016) registered that many second-generation immigrants still strongly identify as religious, sometimes even more so than the first generation. However, compared to the first generation, they also feel more connected to the host country. ² Hence we propose:

H3b: Relative to those who stayed in the origin country, decoupling between religiosity and gender attitudes increases with each new generation in destination countries.

2.4. Differential decoupling (2): gender

Lastly, we know that religious practice and identity formation are gendered processes (Avishai, Jafar, and Rinaldo 2015; Khurshid 2015; Prickett 2015), as women are said to adjust their religious contexts more strongly, given that they are limited by the strict norms the most, which therefore and paradoxically provides the motivation for women to re-interpret religion (Glas, Spierings, and Scheepers 2018; Kandiyoti 1988; Prickett 2015). Moreover, besides Röder’s (2014) study introducing decoupling, two other studies on the impact of religiosity on gender equality attitudes among migrants hint at gendered effects, as women are sometimes found to acculturate more strongly (Idema and Phalet 2007; Röder and Mühlau 2014).

These observations on gendered effects resonate with SIT’s compensatory strategies. Since migrant women are the most visible bearers of gender-equality norms in the destination society, it has been argued that they are monitored most and expected to uphold their culture. They may do so via religious behaviour, whilst shifting gender attitudes (Röder and Mühlau 2014). It is also mostly women who are offered more opportunities in the destination society and who gain most from adopting stronger gender equality attitudes (Glas, Spierings, and Scheepers 2018; Prickett 2015; Röder and Mühlau 2014). Renegotiating gender-equality norms provides them with such privileges while while maintaining a safe social position within the migrant communities.

H4a: The decoupling effect between religion and conservative gender attitudes is expected to be stronger among migrant women than among migrant men.

Research with regard to religiosity indicates that the relationship between communal religiosity and gender equality attitudes is stronger for women than for men (Glas, Spierings, and Scheepers 2018). This difference is probably due to the fact that mosque attendance has a different meaning for both genders. Frequency of mosque attendance, for instance, is not necessarily an indication of religiousness, but rather a social event for men (Hassan 2007) and often the result of living in close proximity or simply of the amount of spare time they have. For women, mosque attendance is less a matter of default and so it is a more discerning indicator of piety among them. Moreover, it can be considered less rational for women to attend sermons, as this means they are confronted with norms countering their personal interests.

Allegedly, in the destination countries religious interpretations are under scrutiny more than in the origin society (Glas et al. 2019). Hence, the destination society offers women the perfect opportunity to rethink their religious interpretation. However, a public religious act is also more vulnerable to social pressure than individual, more private religious acts. Given the alternative gender-equality views that women are exposed to in the destination society, women are likely to become more resistant to the conservative messages they are confronted with during these sermons. For these reasons, among women in particular, we expect to observe decoupling between communal integration and gender equality.

H4b: The decoupling effect we expect to find between religion and conservative gender attitudes among migrant women will especially show with regard to communal religious integration.

3. Data and method

3.1. Data: 2000Families

The 2000Families data surveyed Turkish citizens (migrants and non-migrants) who have lived or are descendants of those who lived in the Turkish high-sending regions during the establishment of guest-labour migration towards Western Europe from the 1960s onwards. The data are collected for the regions Acıpayam, Akçaabat, Emirdağ, Kulu, and Sarkisla (Ganzeboom et al. 2015; Guveli, Ganzeboom, Platt et al. 2016). We focus on four of the five regions, as the pilot study in Sarkisla did not include gender-equality items. The 2000Families data uniquely include information on stayer and migrant respondents.

All data were collected between 2010 and 2012 and derived by random sampling of 100 Primary Sampling Units (PSU; rural villages or urban streets) in the regions. In each PSU, respondents were sampled via a door-knocking method (skip one door, turn left at the end of the street and right at the end of the next [Ganzeboom et al. 2015; Guveli, Ganzeboom, Platt et al. 2016]). Adult men aged 65 years and above were sampled first, and for each core respondent the whole family tree was charted. Then two of his children were randomly selected, and following those lineages four grandchildren (Ganzeboom et al. 2015). A sampling quota was used to make sure around 20% of the first-generation respondents (i.e. the person of 65–90) did not migrate.

The 2000Families dataset contains the most representative existing data for the guest-workers generation and it is representative of their descendants from the sending regions. The data’s strong point is that they allow a strong comparison between stayers and migrants, which is perfect for this paper’s aims.

Of the respondents from the four regions, 4998 (87%) had valid scores on both gender-equality attitude items (see below). Of these we selected all (Muslim) ³ respondents with valid information on migration status, and of the migrants, we selected those who live or have lived in one of the Western-European countries who attracted labour migrants in the 1960s and 70s. These are also the countries with the largest Turkish communities. Overall, we ended up with 3747 respondents. The four regions are roughly represented by a quarter of the respondents each. Currently, most respondent are still (or again) living in Turkey (61.6%). Of the respondents currently residing in Europe, most live in Germany (47.6%) and Belgium (13.6%). The other countries of residence are France (9.2%), the Netherlands (8.9%), Austria (8.8%), Sweden (6.0%), Denmark (5.1%), Switzerland (0.7%), and the UK (0.1%).

3.2 . Gender equality attitudes

The data include two 5-point Likert scale items on the support for gender equality: ‘A university education is more important for a boy than for a girl’ and ‘On the whole, men make better business executives than women do’. These questions thus focus on the public socio-economic sphere and are rather similar to standard items in the ESS and WVS. The average of the two items is used and rescaled from 0 to 10. A higher score indicates stronger support for gender equality. Table 1 in the results section provides the descriptive statistics of the gender equality attitudes and religiosity variables by migration status.

Table 1. Average religiosity and gender equality attitude by migrant generation.

	Individual religiosity (scale: 0–6)	Religious communal integration (scale: 0–6)	Gender equality attitudes (scale: 0–10)	n
Stayer non-migrants in Turkey	4.48	4.24	7.57	1800
1st generation	4.50	4.49	8.14	610
1.5nd generation	3.97	4.01	8.66	237
2nd generation	3.65	3.99	8.81	308
2.5nd generation	3.48	4.01	8.77	133
3rd generation	3.73	3.73	8.59	177
Return migrants	4.95	4.81	7.06	537
Total	4.38	4.29	7.85	3802

Note: The differences in means between generations were found significant at alpha <0.00 for all three variables.

3.3 . Religiosity

In line with our theoretical discussion, we include two overarching dimensions of religiosity: individual religiosity and religious communal integration. Individual religiosity is measured on a scale from 0 to 6 (less to more religious). The scale is the sum of the scores on two core items on individual religious identification and behaviour, recoded to run from 0 to 3: ‘How important is religion to the way you live your life?’: ‘totally or fairly unimportant’, ‘neither important nor unimportant’, ‘important’, or ‘very important’; and ‘Apart from religious services, how often do you pray (namaz)?’: ‘never’, ‘once a month or less’, ‘weekly to daily’, or ‘five times a day’. Religious communal integration is measured on a similar 0–6 scale, also combining two items: ‘How often do you attend services or go to a place of worship?’: ‘never to hardly ever’, ‘on holy days to monthly’, ‘weekly’, or ‘daily’; and a combined orthopraxy scale asking whether people (a) eat pork and (b) drink alcohol, with the following categories: ‘yes to both’ (0), ‘sometimes’ (1.5), or ‘no’ (3).

Overall, both dimensions show clear variation, although the means are relatively high (Table 1). The differential intergenerational differences between the two variables (see results section) indicate that this distinction provides a more detailed view on religiosity than taking all indicators together. The correlation between the two scales is 0.51, showing some overlap.

3.4 . Migration status

The 2000Families data (Guveli, Ganzeboom, Baykara-Krumme et al. 2016) provide a 7 category migration-status variable. By combining the migration status of the respondent, their parent, and grandparent, we first created five groups: (1) non-migrant or stayer; (2) migrated to Western Europe; (3) born in Europe, but from a parent who migrated from Turkey; (4) born in Europe, from parents born in Europe whose grandfather migrated to Europe; (5) return migrant, migrated from Europe to Turkey (as first- or second-generation migrant from Turkey to Europe). ⁴ Next, we distinguished generations 1.5 and 2.5 indirectly, using additional data available in the 2000Families dataset. Generation 1.5 is defined as those who migrated to Europe, but received partial or complete primary education in Europe. ⁵ Generation 2.5 is defined as having one ‘native’ parent and one second-generation migrant parent. ⁶ Overall, we thus distinguish between stayers, first-generation migrants, 1.5-generation migrants, second-generation migrants, 2.5-generation migrants, third-generation migrants, and return migrants. As a consequence of this subdivide, each individual generation is somewhat smaller which consequently increases the odds of false negatives. However, collapsing groups shows similar and insignificant results too. Meanwhile, theoretical reasons for adding half generations are abundant. Where most relevant, the consequences of our decision are discussed more elaborately in the text (e.g. in section 4.2).

3.5 . Control variables

The data further allow us to control for seven core confounding variables of gender equality attitudes; i.e. sex (‘1’ being female); age (18 and up), which also captures hierarchically ordered cohort differences ⁷ ; education (the common educational-level metric included as an interval-like variable); labour-market participation (employed/working, unemployed, in education, homemaker, retired, other); occupational status (ISCO classification of their current or last job), ⁸ whereby the missing data were mean imputed; ⁹ marital status (single, partnered, other; and the number of children). Descriptive information of all control variables by migration status is given in Appendix 1.

3.6 . Models

To take clustering of the data into account, we estimated standard linear multilevel models, whereby the respondents are nested in families. This estimation prevents false positive conclusions (type-2 errors) due to deflated p-values. Respondents are not nested in countries as this in itself would capture the migration-status effect. The models were tested for multicollinearity, of which no indication was found.

4. Results

Before turning to the explanatory analyses, Table 1 presents a descriptive overview of respondents’ religiosity and gender equality attitudes. These data show four core patterns: (1) first-generation migrants on average hold more gender-equal attitudes than stayers (t = 4.723; p = 0.00 – here it should be recalled that the first-generation sample includes relatively more men than the stayer sample), but they are not less religious. They are even considerably more religious when we look at communal religiosity (t = 5.153; p = 0.00); (2) generation 1.5 and beyond hold more gender-equal attitudes (t = 4.591; p = 0.00) and are less religious than the first generation (individual religiosity: t = −12.163; p = 0.00; communal religiosity: t = −10.313; p = 0.00); (3) between these later migration generations, the differences are rather minimal; and (4) return migrants are the most religious (individual religiosity: t = 11.357; p = 0.00; communal religiosity: t = 10.288; p = 0.00) and gender conservative (t = −6.726; p = 0.00) of all groups, which has been ascribed to a retention-selection backlash effect (Spierings 2015). But in this case, particularly regarding gender attitudes, this could also be partly attributed to the return-migration sample containing relatively more men. Overall, this suggests that migration first leads to a liberalisation of gender equality attitudes, but in the long run is linked to both secularisation and the liberalisation of gender equality attitudes. The multivariate analyses below will further reveal whether this is indeed the case.

4.1. Acculturation via secularisation

Table 2 presents how gender equality attitudes differ between the migrant-status groups, as well as what the effects of religious communal integration and individual religiosity are on attitudes about gender equality, all controlled for socio-economic and demographic factors. For the control variables, the Table 2 shows either no effects or effects in line with what we know from the literature (e.g. Glas et al. 2019; Inglehart 1997; Röder 2014; Spierings 2018): young people (or people from later birth cohorts) are more pro gender equality than older people, as are those who are employed, those with a higher educational degree, and women for whom this has been registered across Europe and Muslim majority countries (Davis and Greenstein 2009; Glas et al. 2019; Norris and Inglehart 2012; Spierings 2018).

Table 2. Multilevel regression analysis of influences of migration status and religiosity on support for gender equality.

	Model 1		Model 2		Model 3
	B	SE	B	SE	B	SE
Intercept	8.717***	0.305	7.641***	0.309	8.489***	0.372
Migration
Migrant status(ref = stayers)
1st generation migrant			1.014***	0.135	0.980***	0.135
1.5 generation migrant			0.985***	0.190	0.926***	0.190
2nd generation migrant			0.812***	0.177	0.771***	0.177
2.5 generation migrant			0.668**	0.252	0.631*	0.252
3rd generation migrant			0.698**	0.220	0.604**	0.220
Re-migrated(from EU to Tky)			0.339*	0.152	0.327*	0.152
Religiosity
Communal	−0.124*	0.052			−0.128*	0.051
Individual	−0.121**	0.043			−0.079^#	0.043
Control variables
Age (in years)	−0.017**	0.006	−0.019**	0.006	−0.017**	0.006
Gender (1 = female)	0.645***	0.122	0.706***	0.116	0.620***	0.121
Education (1–11)	0.101***	0.022	0.092***	0.218	0.086***	0.022
Employment (ref = employed)
Unemployed	−0.207	0.223	−0.189	0.222	−0.200	0.221
In education	−0.484**	0.183	−0.455*	0.183	−0.432*	0.183
Housemaker	−0.394**	0.149	−0.255^#	0.149	−0.232	0.150
Retired	−0.109	0.185	−0.169	0.187	−0.079	0.188
Other	−0.292	0.233	−0.222	0.232	−0.237	0.232
Occupational status (ISCO88)	0.038	0.024	0.008	0.024	0.012	0.024
Marital Status (ref = never married)
Married (incl. registered partnership)	0.176	0.150	0.095	0.151	0.111	0.151
Other	−0.033	0.252	−0.103	0.252	−0.148	0.253
Number of children	−0.012	0.038	−0.047	0.038	−0.038	0.038
N	3802		3802		3802
−2 log liklihood	18238.534		18184.346		18176.023
BIC	18255.012		18200.823		18192.498
Residual	6.396***		6.337***		6.331***
Variance(intercept)	0.712**		0.668***		0.642***

Note: Significant two-tailed results ^#alpha < 0.10, *alpha < 0.05, **alpha < 0.01; ***alpha < 0.001 n1 = 3746; n2 = 1235.

Model 2 shows that effects for migrant status are present after controlling for these socio-economic and demographic factors. Compared to Turks who have not left Turkey, all groups of (re-)migrants hold more pro-gender equality attitudes. Moreover, the coefficients in Model 2 seem to indicate that first-generation migrants are the most supportive of gender equality, and that this support declines somewhat with each generation; however, this decline is not statistically significant ¹⁰ (cf. Table 1).

This brings us to the Hypothesis 1, which stated that (especially communal) religiosity links to less supportive gender equality attitudes. Model 3 shows a significant negative effect of communal integration and a marginally significant effect of individual religiosity. The direction of both effects indicates that more religious people are on average more opposed to gender equality. As expected, the coefficient of religious communal integration is substantially larger and more strongly significant than that of individual religiosity (b = −0.13 and b = −0.08 respectively). At the same time, these two coefficients do not significantly differ for the current sample, as the standard errors show. ¹¹ Lastly, while the two religiosity variables are not multicollinear in the model, we also ran the analyses with only one variable at the time (see Appendix 2). Here, too, both showed a negative coefficient (communal still the largest but without a significant difference), and the individual level of religiosity indicator becomes more clearly statistically significant. In sum, religiosity relates to stronger opposition towards gender equality, but this effect is not significantly larger for communal- than for individual religiosity.

Continuing to acculturation (H2a and H2b), we expected to find that the difference between non-migrants and migrants is due to lower levels of religiosity, and that this mechanism becomes more apparent with every new migrant generation. If this is the case, then the impact of migration, which we see in Model 2, should be substantially reduced by the inclusion of the religiosity measures (Model 3). The results of migration in Model 2 in comparison to Model 3 tells us that this difference decreases slightly (roughly 5% of the coefficients in Model 2) after the inclusion of religiosity, but not much more for later generations. So even though we find decreases in religiosity by migration (see Table 1 and Appendix 2), these changes do not structurally coincide with an increase in gender-equal attitudes. Overall, these analyses suggest rather minimal acculturation of gender equality attitudes via secularization and so no support for hypotheses 2a and 2b.

The fact that this specific acculturation process is rather weak seems to be due to the fact that, after controlling for socio-economic factors, gender equality attitudes do not rise over generations. Overall, though there might be acculturation when it comes to gender equality attitudes (see also Table 1), for instance via education, our findings show that it is not clearly linked to the secularisation process.

4.2. Decoupling religiosity and gender equality

Although acculturation is not taking place, decoupling still might. To test Hypotheses 3a and 3b, which state that the link between religiosity and gender equality is weaker among migrants and weakens by generation, we need to look at the interactions between religiosity and migrant generation as detailed in Table 3. One would expect to see a decreasing effect with every generation. However, Table 3 shows no effect of individual religiosity at all. A significant main effect of communal religiosity of −0.15 does show. This indicates that among stayers, religiosity leads to a decrease in support for gender equality attitudes. Though the interaction effect is insignificant, these effects of religiosity are less strong among migrants. One might argue that the group sizes explain this, but we also find no statistically significant interaction effect when we group first- to third-generation migrant together (Appendix 3, Model 1), or when running the decoupling model on migrants only and with a linear generations variable (Appendix 3, Model 2). At the same time, it should be noted that if we change the reference group so that the main coefficient holds for migrants, the influence of communal religiosity is insignificant (Appendix 3, Model 1). In other words, among non-migrants we do find significant negative relationships between religiosity and gender equality attitudes, which we did not find among migrants. This does lend some support for the idea of decoupling, but only for communal religiosity.

Table 3. Multilevel regression analysis of how migration background influences the religiosity-support for gender equality linkage.

	Model 1: interaction model
	B	SE	B	SE	B	SE
			Communal religiosity		Individual Religiosity
Religiosity (main effects)			−.147*	.069	−.082	.059
	Main effects generations		Communal religiosity * generation		Individual religiosity * generation
stayers	ref		ref		ref
1st generation migrant	.715	.597	.057	.141	.002	.118
1.5 generation migrant	−.190	.809	.229	.201	.049	.165
2nd generation migrant	−.355	.818	.023	.203	.284	.171
2.5 generation migrant	.618	1.130	.294	.332	−.335	.267
3rd generation migrant	1.035	.875	.219	.233	−.337	.212
Re-migrated (from EU to Tky)	.668	.577	−.077	.132	.005	.126
N	3802
−2 log liklihood	18185.376
BIC	18201.845
Residual	6.334***
Variance(intercept)	0.639***

Note: The model are controlled for age, gender, education, employment status, occupational status, marital status, and number of children.

Significant two-tailed results ^#alpha < 0.10, *alpha < 0.05, **alpha < 0.01; ***alpha < 0.001 n1 = 3746; n2 = 1235.

In sum, shifts in religiosity and gender equality seem to happen independently (see above), and our data suggest that decoupling is indeed taking place among Turkish people in Western Europe; particularly when comparing migrants to non-migrants (slightly supporting H3a but not H3b). Yet, the effects are limited and suggestive rather than irrefutable. Taken together with the analyses on acculturation via secularisation, it seems safe to conclude that the linkages between migration, religiosity and gender equality are subject to some acculturation as well as decoupling processes, leading to less strongly carved out effects of each individual process.

4.3. Gendered decoupling

One of the reasons for our mixed results might be that the decoupling effects differ for men and women (H4). Table 4 presents the core results of the same analyses presented in Tables 2 and 3, but split by gender. Although H4 focusses on decoupling and not on acculturation, the results on the latter are provided in the top half of Table 4 for a comprehensive picture. These models show that different dimensions of religiosity have different effects on women than on men. Communal religious integration has a particularly negative effect on gender equality attitudes among women, especially of the first generation. Meanwhile, individual religiosity has a clearly negative effect on men, while it has no effect among women. Again, we see some reduction of the migration effects, but mostly among men. Indicating that the acculturation effects only happens among men and only with regard to individual religiosity.

Table 4. Multilevel regression analyses of migration status, religiosity and gender equality linkage per gender.

	MEN		WOMEN
	Model 2	Model 3	Model 2	Model 3
DIRECT EFFECTS: ACCULTURATION
Religiosity
Communal		−.080		−.195*
Individual		−.166**		.016
Migrant status
stayers	Ref	ref	ref	ref
1st generation migrant	.946***	.888***	1.115***	1.136***
1.5 generation migrant	.931***	.854**	1.112***	1.116***
2nd generation migrant	1.013***	.959***	.584*	.592*
2.5 generation migrant	.795*	.742*	.652^#	.665*
3rd generation migrant	.870**	.766*	.516	.464
Re-migrated (from EU to Tky)	.350^#	.346^#	.332	.302
INTERACTION EFFECTS: DECOUPLING	MEN		WOMEN
	Communal	Individual	Communal	Individual
Religiosity (main effects)	−.069	−.222*	−.216*	.035
Interaction with migration status
* stayers
* 1st generation migrant	−.032	.150	.209	−.141
* 1.5 generation migrant	.192	.387	.036	−.225
* 2nd generation migrant	−.008	.293	−.083	.247
* 2.5 generation migrant	.175	−.397	.795	−.360
* 3rd generation migrant	.093	−.304	.157	−.204
* Re-migrated (from EU to Tky)	−.095	.035	−.255	.123

Note: all models are controlled for age, gender, education, employment status, occupational status, marital status, and number of children; the interaction model include the main effects of migrant status generation too.

Significant two-tailed results ^#alpha < 0.10, *alpha < 0.05, **alpha < 0.01; ***alpha < 0.001 n1 = 3746; n2 = 1235.

Regarding the decoupling mechanism shown at the bottom half of Table 4, we see that the negative main effects of individual religiosity on gender equality attitudes among men and of communal religiosity on gender equality attitudes among women are both significant. This indicates that the effects discussed above hold among the subgroup of stayers. Again, we find no statistically significant interaction effects. But similar to our across-the-board finding, if we combine all migrant groups and model the impact of religiosity on gender equality attitudes, the impact of individual religiosity is smaller and insignificant among male migrants (Appendix 4). In turn, among female migrants, the impact of communal religiosity decreases and turns insignificant (Appendix 4). This means we again find an indication of decoupling, or at least no clear effect of religiosity on gender equality attitudes within the migrant groups, whereas this effect is present among comparable non-migrant stayers. Since we do not find the decoupling effect to be stronger for women than for men, Hypothesis 4a is rejected. Nevertheless, the decoupling effect does seem gendered, as it only shows with regard to communal religiosity for women and with regard to individual religiosity for men. The former is in line with Hypothesis 4b, which stated that the norms that women are exposed to through communal integration can especially be reinvented in a society with more liberal viewpoints. The implications of this finding are discussed in greater detail in the discussion section.

5. Discussion

There is much public discussion about Islamic religiosity and gender equality attitudes among migrants in Europe. Scientifically, the predominant view is that acculturation into the destination society leads to dissimilation from the origin society, suggesting inter- and intra-generational secularisation that leads to more progressive gender equality attitudes (e.g. Diehl and Koenig 2009; Kogan and Weißmann 2019). However, it is increasingly argued that this tripartite relationship is rather complex (Kogan, Fong, and Reitz 2019) and not well understood. Building on empirical finding of others (e.g Diehl, Koenig, and Ruckdeschel 2009), Röder (2014) was one of the first to present an alternative theoretical connection. She concluded that acculturation is taking place, but is accompanied by decoupling: migration causes a weakening of the influence of religion on gender equality attitudes among Muslim migrants. In this study, we built on her work by investigating these linkages in greater detail and for the largest group of Muslim migrants in Europe (i.e. Turks). This study has thus shed additional light on the complexity of acculturation processes regarding gender equality attitudes (e.g. Spierings 2015), and assessed the tenability of Röder’s results by focusing on a group with strong religious and ethnic infrastructures in the receiving countries (cf. Bird, Saalfeld, and Wüst 2011; Cesari 2004).

Theoretically, we applied the acculturation and decoupling logic by linking migration to secularisation (and socialisation) and religious re-interpretation (decoupling), which in turn can translate to the adoption of more progressive gender equality attitudes. We moreover refined the existing reasoning by integrating insights on how different dimensions of religiosity have different and gendered impacts (see Glas, Spierings, and Scheepers 2018) and by theorising the decoupling process more clearly within the existing frame of social identity theory (Tajfel and Turner 1979; Taylor and Moghaddam 1987). Empirically, these theories make implicit assumptions about stability in the country of origin, which we could explicitly test by taking an origin perspective (see Guveli, Ganzeboom, Platt et al. 2016; Guveli et al. 2017).

Both the theoretical and empirical refinement turned out to be highly relevant. While the overall models showed that religiosity has a negative impact on gender equality attitudes, a closer inspection shows that individual religiosity and religious communal integration have different and diverging effects on acculturation and decoupling. In line with research by Glas, Spierings, and Scheepers (2018) on citizens (non-migrants) in the Middle East, we found that communal religiosity has the clearest negative impact on women, whereas individual religiosity has such an effect on men. This means that the connection between religiosity and gender equality attitudes is highly gendered and not very straightforward (see also Idema and Phalet 2007; Röder and Mühlau 2014). Religious communal integration is not necessarily an indication of religious conservatism for men to begin with; attitudes and behaviour are already decoupled among those who did not migrate. Consequentially, migration appears to have little additional influence. ¹² Yet individual religiosity is still coupled for men in the origin country, and our results for men with regard to individual religiosity show both signs of acculturation and decoupling. Migrant men are religiously conservative to some extent, but this is separated from their patriarchal attitudes. Women meanwhile show a similar decoupling pattern without changing their religious habits regarding religious communal integration.

It makes sense that women are more influenced by communal integration, as the sermons emphasise their subordinate role as women. However, in a migration context, the impact of attendance decreases as attending women are also exposed to more liberal viewpoints giving them more freedom, leading to a reinterpretation of their religious beliefs. The fact that this causes decoupling rather than acculturation indicates that mosque attendance and obeying religious rules (orthopraxy) are still important features of migrant women’s lives, most likely due to their social function. In terms of the societal consequences of our findings, this indicates that for women at least the cultural distance between Western and Muslim norms does become smaller with time, regardless of religiosity.

Returning to the theoretical basis of our results, the descriptive statistics showed that non-migrants are more religious and less supportive of gender equality than first-generation migrants, and first-generation migrants are more religious and less supportive of gender equality than later-generation migrants. In other words, acculturation and secularisation are taking place. However, for gender attitudes we found far less proof of acculturation via secularisation, which was our focus. Only among men did we find somewhat, yet minor support for this claim. This suggests that migrants may undergo various acculturation processes through different cultural dimensions simultaneously and that the linkages between these dimensions differ for men and women, depending on self-interest and gendered practices (Glas, Spierings, and Scheepers 2018). The results also indicate that socialisation processes take place beyond the origin country and among those who migrated. Although this is not significantly different for younger generations than older ones, and in our study only found among men.

Furthermore, the decoupling mechanism should not be contrasted with these acculturation and secularisation processes, as our results indicate both are taking place simultaneously (leading to more complex and seemingly blurry results). The interaction terms on the religiosity–gender equality attitudes linkage among migrants and non-migrants (in Turkey) were insignificant, there were some hints that decoupling was taking place. In the separate investigation of men and women we found some evidence for decoupling, albeit on different religious dimensions. In short, the decoupling framework can be developed more strongly by taking into account the incentives for, and resources of, re-interpretation and socialisation processes. This could partly be explained through social identity theory, as our results provide indications that both men and women apply the social mobility strategy (regardless of generation).

Our study thus provides some support for Röder’s (2014) and others’ findings (cf. Banfi, Gianni, and Giugni 2016; Diehl and Koenig 2009; Diehl, Koenig, and Ruckdeschel 2009). Most importantly, our comparison between migrants’ and non-migrant counterparts helps to attribute the found differences to migration. Moreover, our focus on Turks presented a less likely case (see introduction); nevertheless, we found important indications of decoupling processes. Lastly, we only found the acculturation effect for men, stressing the importance of a gendered perspective.

While this study has provided new insights, it also confronted us with future challenges. First, migration evidently rests on self-selection, and thus migrants possess specific, unobservable characteristics different from non-migrants. Still, the survey we rely on is unique in sampling comparable non-migrant and migrant groups as much as possible, which proxies the before/after approach. Moreover, the likely selectivity will centre on economic traits as the respondents came from high-sending labour migration regions (Guveli, Ganzeboom, Platt et al. 2016; Guveli et al. 2017). It is unlikely that the decision to migration was based on religiosity or gender equality attitudes. If anything, selectivity on cultural issues plays a role in re-migration: those not feeling at home in the relatively more liberal Western Europe are more likely to return (Spierings 2015). Moreover, after control for other factors, of all migrant groups, return migrants are most dissimilar in their gender equality attitudes from the 1st generation migrants and least open to gender equality of all migrant groups. However, in the absence of panel data, causal claims should be made with care. Future projects might set out to collect such data in currently high-sending regions.

Second, though selecting Muslim migrants from a community with a large religious and ethnic infrastructure in the receiving country did make for a clean design, it also implies context-dependent effects by host societies. In this study, we simply could not systematically compare these potential differences. Similarly, distinguishing generations and sexes resulted in relatively small subsamples; future research might take this into account, for instance by oversampling subgroups or by starting from the result that the main division is found between non-migrants, first- generation migrants, and 1.5-generation and up.

Lastly, our study is unique in comparing non-migrants and emigrants, which led to new information on potential decoupling processes. A next step is to collect similar data in a panel design, which could focus on over time developments. For instance, the indication of decoupling that we found could technically be partly caused by no change among emigrants and recoupling among non-migrants due to societal and political changes in Turkey. This still means that migration impacts the religiosity-gender equality attitudes linkage, but it does have different implication for theory- and policy development.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Also included: France, Denmark, Sweden, Switzerland, and the UK.

2 Comparing first- and second-generation immigrants in The Netherlands, Beek and Fleischmann found no substantial difference for religiosity’s impact. This variation in results calls for more theoretical and empirical work on decoupling.

3 In the 2000Families data almost all respondents are Muslim; fewer than 1.5% said to be other or non-religious (including atheist), or gave no answer.

4 Over 90% of this group consists of first-generation return migrants, not second-generation migrants moving to Turkey, for whom the label ‘return migrant’ could be debated. Given the group it is rather small, we kept that label though.

5 We selected those who migrated age >11. We did not use age at migration more generally, as there are some inconsistencies and missing scores in the data, and because it fits our theoretical reasoning less.

6 The nationality of the second parent is unknown. The language spoken at home served as proxy; if that was mostly/only that of the destination country, we considered this generation 2.5 (a ‘native’ partner most likely does not speak Turkish). We expect few false positives (people we coded 2.5 actually being second generation), but we cannot exclude false negatives as strongly (people still coded as second generation belonging to generation 2.5).

7 Core models are rerun with age’s quadratic term and with dummies for every 10-year age group to control for non-linear cohort effects. We found no substantive change in our core results, and the age/cohort effect was roughly ordered hierarchically.

8 If absent, the ISCO code of the first job was taken.

9 Mean imputation was used. Running a model with a dummy for the cases with imputed scores does not change the effects of the variables of interest (see Appendix 2). Still, given limitations in operationalizing occupational status, one should be careful in drawing strong conclusion on occupation’s impact itself.

10 Tested by changing the reference category. The migration groups do not differ statistically from each other on gender equality attitudes.

11 The standard errors form the basis of the 95% confidence intervals. For communal religiosity the interval ranges from −0.03 to −0.23, for individual religiosity from 0.01 to −0.16, this is too large an overlap to conclude with acceptable certainty that communal religiosity has a stronger effect in the larger population.

12 Unless strong Europeanization processes took place regarding religion and gender equality in Turkey, leading to similar processes taking place among migrants and non-migrants. While this might be partly the case, it is unlikely to fully explain these findings given the conservative resurgence in Turkey, particularly in the areas the respondents come from and previously documented developments, for instance in gender equality attitudes (see Spierings 2015).

Appendices

Appendix 1. Mean of control variables by migrations status.

Table

	Stayer non-migrants in Turkey	1st generation	1.5nd generation	2nd generation	2.5nd generation	3rd generation	Return migrants
Age (in years)	37.21	49.33	38.80	24.00	21.81	33.12	60.11
Gender
Male	0.53	0.73	0.55	0.55	0.49	0.56	0.84
Female	0.47	0.26	0.45	0.45	0.51	0.44	0.16
Education (1–11)	4.71	3.76	5.69	6.79	7.14	6.59	3.26
Employment
Employment	0.42	0.50	0.66	0.52	0.38	0.73	0.24
Unemployed	0.04	0.03	0.06	0.07	0.08	0.09	0.03
In education	0.11	0.02	0.04	0.28	0.41	0.02	0.01
Housemaker	0.29	0.14	0.14	0.09	0.07	0.14	0.09
Retired	0.10	0.29	0.03	0.00	0.00	0.01	0.61
Other	0.04	0.03	0.08	0.05	0.06	0.01	0.03
Occupational status (ISCO88)	6.16	7.08	6.07	6.10	6.17	5.85	6.62
Marital Status
Never married	0.28	0.04	0.12	0.62	0.83	0.17	0.06
Married (incl. registered partnership)	0.68	0.90	0.80	0.36	0.17	0.76	0.86
Other	0.03	0.06	0.08	0.03	0.00	0.07	0.08
Number of children	1.88	3.26	2.28	0.46	0.13	1.65	3.83

Appendix 2. Core coefficients in additional analyses of Table 2.

Table

	Model 3 without occupational status	Model 3 with dummy for imputed cases occupational status	Model 3 with communal religiosity, without individual religiosity	Model 3 with individual religiosity, without communal religiosity	Individual religiosity as dependent variable	Communal religiosity as dependent variable
	B	B	B	B	B	B
Migration
Migrant status (ref = stayers)
1st generation migrant	0.990***	0.912***	0.997***	0.979***	−0.260***	−0.115*
1.5 generation migrant	0.929***	0.879***	0.953***	0.929***	−0.437***	−0.245***
2nd generation migrant	0.773***	0.748***	0.797***	0.763***	−0.357***	−0.079
2.5 generation migrant	0.632*	0.604*	0.670**	0.606*	−0.478***	−0.014
3rd generation migrant	0.605**	0.565*	0.631**	0.627**	−0.586***	−0.415***
Re-migrated (from EU to Tky)	0.329*	0.290^#	0.328*	0.333*	−0.017	−0.050
Religiosity
Communal	−0.127*	−0.129*	−0.170***
Individual	−0.078^#	−0.078^#		−0.127**

Note: Models are specified exactly the same as Model 3 in Table 2, i.e. including the same control variables, levels, N etc.

Significant two-tailed results ^#alpha < 0.10, *alpha < 0.05, **alpha < 0.01; ***alpha < 0.001.

Appendix 3. Core coefficients in additional analyses of Table 3.

Table

	Model 1: Group all migrant generations together			Model 2: Linearising generations variable (1st Gen = 0; 3rd Gen = 4; on migrants only)
	Main effects	Interaction with communal	Interaction with individual	Main effects	Interaction with communal	Interaction with individual
Migration
Migrant status
Stayers	−0.282	−0.116	−0.020
1st generation migrant	Ref.
1.5 generation migrant	Ref.
2nd generation migrant	Ref.
2.5 generation migrant	Ref.
3rd generation migrant	Ref.
Re-migrated (from EU to Tky)	0.363	−0.194	−0.014
Migrant generation linear				0.050	0.015	−0.039
Religiosity
Communal	−0.034			−0.072
Individual	−0.067			−0.028

Note: Models are specified exactly the same as in Table 3, i.e. including the same control variables, levels, N etc.

Significant two-tailed results ^#alpha < 0.10, *alpha < 0.05, **alpha < 0.01; ***alpha < 0.001.

Appendix 4. Core coefficients in additional analyses of Table 4.

Table

INTERACTION EFFECTS: DECOUPLING	MEN			WOMEN
	N per subgroup	Communal	Individual	N per subgroup	Communal	Individual
Religiosity (main effects)		−0.045	−0.094		−0.056	−0.008
Interaction with migration status
* stayers	951	−0.021	−0.128	849	−0.159	0.025
* 1st generation migrant	445	Ref.	Ref.	165	Ref.	Ref.
* 1.5 generation migrant	131	Ref.	Ref.	106	Ref.	Ref.
* 2nd generation migrant	170	Ref.	Ref.	138	Ref.	Ref.
* 2.5 generation migrant	63	Ref.	Ref.	70	Ref.	Ref.
* 3rd generation migrant	97	Ref.	Ref.	80	Ref.	Ref.
* Re-migrated (from EU to Tky)	450	−0.117	−0.092	87	−0.423	0.156

Note: Models are specified exactly the same as in Table 4, i.e. including the same control variables, levels, N etc.

Significant two-tailed results ^#alpha < 0.10, *alpha < 0.05, **alpha < 0.01; ***alpha < 0.001.