Partnership Transitions among Turkish Immigrants and their Descendants in Western Germany Partnership Transitions among Turkish Immigrants and their Descendants in Western Germany* Müşerref Erdoğan, Ayşe Abbasoğlu Özgören** Abstract: Adaptation to host country behaviours encompasses both individual and social change, bringing about rising diversity issues in the host society and societal shifts in the country of origin. This study aims to detect whether Turkish immigrants and their descendants converge towards or diverge from the partnership practices of the native-born population in Western Germany. Specifi cally, transitions from (1) singlehood to the fi rst partnership, (2) singlehood to the fi rst marriage, (3) singlehood to the fi rst cohabitation, (4) cohabitation to marriage and (5) marriage to divorce are investigated. Data from the Panel Analysis of Intimate Relationships and Family Dynamics (pairfam) survey for the period of 2008-2018 are used, applying Kaplan- Meier survival estimations and Cox proportional hazard models. This is the fi rst study that includes natives, immigrants and their descendants simultaneously in an analysis of extensive partnership transitions covering practices of cohabitation, marriage and divorce and to investigate the proportionality assumption in Cox models. We formulate four research hypotheses based on the hypotheses of socialisation, adaptation and the cultural maintenance and segmented assimilation theory. Supporting our fi rst hypothesis, our fi ndings indicate a difference in partnership patterns between both fi rst- and second-generation immigrants and natives, except for the fi nding that second-generation immigrants resemble the native pattern in their transition to the fi rst union (including both cohabitation and marriage). Immigrants and their descendants tend to marry directly and have lower divorce hazard ratios than their native counterparts, while consensual unions are uncommon among Turkish immigrants. As suggested by our second hypothesis, the extent of the divergence varies across partnership transitions. Finally, our results provide support for our third hypothesis rather than the fourth in that partnership transition of Turkish immigrants’ descendants more closely resembles that of fi rst- generation immigrants compared to natives. Comparative Population Studies Vol. 47 (2022): 321-348 (Date of release: 01.08.2022) Federal Institute for Population Research 2022 URL: www.comparativepopulationstudies.de DOI: https://doi.org/10.12765/CPoS-2022-13 URN: urn:nbn:de:bib-cpos-2022-13en0 * This article has a Data Appendix with supplementary material URL: http://www. comparativepopulationstudies.de/index.php/CPoS/article/view/447/360. ** This paper is based on Müşerref Erdoğan’s MA thesis titled “The Union Formation and Dissolution Among Turkish Immigrants and Their Descendants in Germany” at the Hacettepe University Institute of Population Studies (HUIPS), Department of Demography, in 2020. • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören322 Keywords: Union formation · Divorce · Turkish immigrants · Germany · Event history analysis 1 Introduction Labour migrants to Western Germany have lived alongside ethnic Germans and their descendants since the Second World War. While migration infl uxes change the social fabric of cities and neighbourhoods and create diversity, immigrants themselves produce new demographic behaviours and establish a distinct minority culture. Turkish immigrants, together with their descendants, are the most populous immigrant group in Germany (German Federal Statistical Offi ce 2020), and their population has been on the rise since a 1961 labour agreement between the Turkish Government and West Germany. First-generation Turkish immigrants and their descendants may fi nd themselves between two different backgrounds: that of the host country and that of their parents’ country of origin. This difference is also apparent in partnership dynamics, since partnership transitions differ strongly between Türkiye and Germany. In general, adaptation to host country behaviours illustrates both individual and social change. With Turkish migrants affecting ongoing change in the social fabric in Germany, it is vital to study who adapts to what, to what extent and how (if possible). Additionally, as the descendants of Turkish migrants enter young adulthood, it is important and timely to analyse the intergenerational transmission of their partnership preferences. With this background, this paper focuses on partnership formation and dissolution transitions of fi rst- and second- generation Turkish immigrants compared to natives in Western Germany using event history or survival analysis methods, namely the Cox proportional hazard model and Kaplan-Meier method. This study uses data from the Panel Analysis of Intimate Relationships and Family Dynamics (pairfam) survey covering the period of 2008-2018, which is a relatively extensive panel in Germany. Specifi cally, transitions from (1) singlehood to the fi rst partnership, (2) singlehood to fi rst marriage, (3) singlehood to fi rst cohabitation, (4) cohabitation to marriage and (5) marriage to divorce in Western Germany, plus their associations with migration status within the categories of Turkish immigrants, their descendants and Western German natives1 are analysed. Our theoretical approach and research hypotheses are based on various complementary hypotheses that were originally focused on childbearing among migrants (and their descendants) to describe how migration intervenes with fertility dynamics. In forming our hypotheses, we differentiate between fi rst- and second- generation immigrants, as found in Rahnu, Puur, Sakkeus and Klesment (2015). 1 Natives are defi ned as individuals whose mothers and fathers were both born in Western Germany. Partnership Transitions among Turkish Immigrants and their Descendants ... • 323 Recently, many studies have reported on family dynamics among immigrants, thanks to available country-specifi c and cross-national data. These studies tend to focus on immigrants from different origins living in Europe and the US (Andersson et al. 2015; De Valk 2008; De Valk/Liefbroer 2007; González-Ferrer et al. 2014; Hamel et al. 2012; Hannemann/Kulu 2015; Hannemann et al. 2020; Huschek et al. 2010; Kuhnt/Krapf 2020; Milewski/Hamel 2010; Pailhé 2014; Rahnu et al. 2015; Soehl/ Yahirun 2011). Early papers focus on the timing of marriage and use descriptive tools to compare immigrant generations to natives. Later research emphasises differences in intensity of unions, type of union and, to a certain extent, dissolution practices, using multivariate event history analysis techniques. This paper aims to contribute to the existing knowledge in the following ways. First, to the authors’ knowledge, this is the fi rst study to include both fi rst- and second-generation Turkish immigrants in the Western German context in a single analysis with such an extensive partnership transition framework. The analysis covers not only partnership formation and type of union, but also marriage after cohabitation and divorce practices of immigrants. Second, the data used, pairfam, is a specifi c data series used to study intimate relationships and family dynamics. Both marriage histories and cohabitation practices of respondents are recorded separately in the data set. Since pairfam is panel data, it provides data on the family practices of single young generations each year and gives up-to-date information on their status. This feature signifi cantly enriches this paper, since most of the studies use cross-sectional data that do not provide information on partnerships for the younger second generation. Finally, although studies in this fi eld use the Cox model or piece-wise constant model, this paper is the fi rst to discuss the proportionality assumption. In this way, the accuracy and reliability of the standard model are improved. 2 The Context: Partnership Dynamics in Germany and Türkiye Individuals in Germany and Türkiye have signifi cantly different demographic characteristics regarding partnership dynamics and family values. Such variances directly affect the difference of partnership practices between Turkish immigrants and the native population in Germany. This section presents a present-day comparison of selected nuptiality and fertility indicators for Germany and Türkiye, in addition to discussing recent historical trends. In Germany, marriage was historically an almost universal institution and earlier marriages were typical in the “golden age of marriage” in the 1950s and 1960s. During the 1970s, the former FRG2 experienced a pronounced retreat from conjugal family formation compared to the rest of Europe. The connection between sex, marriage and reproduction was relaxed by individualisation trends, advancement in women’s economic positions, the contraception revolution and changes in gender 2 Refers to the territory of the former Federal Republic (FRG before reunifi cation). • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören324 roles. According to the German Federal Statistical Offi ce, the crude marriage rate, which was 11.0 per thousand in 1950, reached its lowest level in 2007 at 4.5. By 2018, the proportion of never-married between 25-45 was 49.1 percent among women and men (German Federal Statistical Offi ce 2020). While marriage has lost its prevalence among all age groups, it has also been postponed. In 1960, the mean age for women at fi rst marriage was slightly under 24. After 1980, it increased, reaching 31 in 2018 (German Federal Statistical Offi ce 2020). This trend is partly because marriage is no longer the primary means of achieving women’s fi nancial stability so that their bargaining power over partner choice and future career lead them to delay marriage formation (Köppen 2011). The same effect reveals itself in the stability of conjugal families. The improvement of women’s economic positions reduces both the benefi ts of marriage and the cost of its dissolution for women (Kalmijn 2007). In 1965, the total divorce rate was around 12 percent, peaking in 2004 with 42 of every 100 marriages on average resulting in divorce. According to the German Federal Statistical Offi ce, in 2019, on average, 32 of every 100 couples divorce after 14.8 years of marriage (German Federal Statistical Offi ce 2020). These substantial changes in living arrangements have been accompanied by the spread of extramarital cohabitation. Although it is not an entirely new form of partnership in Germany, the prevalence of consensual unions, especially at early ages, in the late 20th and early 21st centuries, is apparent. According to the Family and Fertility Survey (FFS), 16 percent of women aged 20-24 and 9 percent of women aged 25-29 were cohabiting in 1996 in Western Germany (Kiernan 2002). A relatively recent report on the household structure of Europe found that 83 percent of German women in their twenties were cohabiting in 2007. This proportion decreased to 48.3 percent among women in a union between 30 to 40 years old (EUROSTAT 2010), so cohabitation has a well-established place in partnership formation trends. Regarding fertility, the total fertility rate in Germany settled to nearly steady levels of about 1.5 after the boosting effect of the post-Second World War period had faded (German Federal Statistical Offi ce 2020). Typically, a low fertility rate is also accompanied by women giving birth at older ages. In 2020, for both married and unmarried women, the average fi rst birth occurred at age 30.2 (German Federal Statistical Offi ce 2020). Table 1 presents selected indicators of partnership behaviours and family values for Germany and Türkiye, revealing signifi cant differences between the two countries. In Türkiye, immigrant marriages occur on average 7 years earlier than that of their German counterparts. The crude marriage rate is slightly higher in Türkiye and the crude divorce rate is lower than that of Germany. According to birth registers, the total fertility rate in both countries in 2020 was below replacement level, at 1.53 in Germany and 1.76 in Türkiye. Offi cial data on cohabitation in Türkiye is lacking, although the World Values Survey sample suggests a proportion of 0.2 percent. For Germany, this share is much higher at 4.7 percent. Values or attitudes related to partnership also differ between Türkiye and Germany. Unmarried or cohabiting couples are not wanted as neighbours in Türkiye by over half the population, whereas such a situation does not exist in Germany. Sex before marriage is seen Partnership Transitions among Turkish Immigrants and their Descendants ... • 325 as totally unjustifi able by 61 percent of the population in Türkiye, while this level is only 5.7 percent in Germany. Moreover, divorce is seen as never justifi able at a proportion fi ve times higher in Türkiye compared to Germany. Although the indicators for Turkish immigrants in Germany are not recent, it is important to mention some of this data. Cohabitation has been a relatively rare phenomenon among immigrants in both older and younger cohorts. Only 22 percent of Turkish immigrants between the ages of 18-29 cohabited at least once in their lives in 2005, while the share of pre-, post- or non-marital cohabitation dropped to 14 percent among males and females aged 45-59 (Naderi 2008). Even among descendants, the conjugal family is the dominant choice; the TIES (The Integration of the European Second Generation) sample showed that 4 out of 5 children of Turkish immigrants in Germany preferred to marry rather than cohabit (Hamel et al. 2012). This scenario is at least partly consistent with the current situation in Türkiye, where marriage is the most widespread family formation. Unlike German endogamous conjugal families, Turkish immigrant marriages are found to be more stable (Milewski/Kulu 2014). Related to this, as many studies suggest, both fi rst- and second-generation Turkish immigrants marry either another immigrant from Türkiye or a Turkish descendant and that appears to be a factor Tab. 1: Selected partnership and fertility indicators, Germany and Türkiye Germany Türkiye Partnership indicators (year: 2020) Mean age at fi rst marriage (male) 34.9 27.9 Mean age at fi rst marriage (female) 32.4 25.1 Crude marriage rate (per thousand) 4.50 5.84 Proportion cohabiting 15.1% - Divorce indicator (year: 2020) Crude divorce rate (per thousand) 1.70 1.62 Fertility indicator (year: 2020) Total fertility rate 1.53 1.76 Selected World Values Survey Indicators (Germany 2017, Türkiye 2018) Living together as married (cohabiting) 4.7% 0.2% Partnership values (Germany 2017, Türkiye 2018) Would not like to have unmarried couples living together as neighbours 0.7% 55.4% Sex before marriage never justifi able 5.7% 61.0% Divorce never justifi able 5.5% 29.1% Source: Mean age at fi rst marriage, Germany: German Federal Statistical Offi ce 2022a; Mean age at fi rst marriage, Türkiye: TURKSTAT 2021a; Crude marriage rate, Germany: German Federal Statistical Offi ce 2022b; Crude marriage rate, Türkiye: TURKSTAT 2021a; Proportion cohabiting, Germany: German Federal Statistical Offi ce 2022c; Crude divorce rate, Germany: German Federal Statistical Offi ce 2022b; Crude divorce rate, Türkiye: TURKSTAT 2021a; Total fertility rate, Germany: German Federal Statistical Offi ce 2020; Total fertility rate, Türkiye: TURKSTAT 2021b; Selected World Values Survey Indicators and partnership values: WVS 2022 • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören326 behind more stable relationships (Constant et al. 2012; Hamel et al. 2012; Naderi 2015). 3 Previous Research and Hypotheses There are various hypotheses used to study immigrant fertility behaviours namely, (childhood) socialisation, adaptation, selection, disruption, interrelation of events and others (Milewski 2010; Tønnessen/Wilson 2020). The fi rst four of these behaviours are the most widely discussed in demography literature. Since fertility behaviours and partnership dynamics are highly interrelated, these four hypotheses are also found to be useful tools when interpreting the partnership dynamics of immigrants (Hannemann/Kulu 2015). Additionally, some phenomena related to these hypotheses may extend beyond fi rst-generation immigrants, so the cultural maintenance hypothesis (Abbasi-Shavazi/McDonald 2000, 2002) and segmented assimilation theory (Portes/Zhou 1993) have provided useful frameworks to study family behaviours of the descendants of migrants (as in Rahnu et al. 2015). The fi rst hypothesis, socialisation, argues that a person’s life-course choices are a product of the prevalent values and norms of their native region (Kulu/Milewski 2007). Therefore, this hypothesis assumes that union trajectories of immigrants are similar to those still living in the country of origin if the immigrant spent their childhood in the sending country. It also posits that experience and exposure to values and norms in the country of origin shapes the life choices of immigrants in the long run and that those choices do not resemble the choices of natives in the destination country. If differences regarding family norms and dynamics in origin and host countries are signifi cant (as between Türkiye and Germany), immigrants will persist with the traditional patterns of their origin country. For instance, fi rst-generation Turkish immigrants typically reject cohabitation in favour of direct marriage, a highly preferred and traditional pattern of family formation in Türkiye (Kuhnt/Krapf 2020; Pailhé 2015). Furthermore, the socialisation hypothesis is not restricted to fi rst- generation immigrants. As mentioned by Rahnu et al. (2015), divergence from the patterns of the host society may extend beyond the fi rst generation. This pattern is referred to as “cultural maintenance” by Abbasi-Shavazi and McDonald (2000, 2002) and it bears similarities to the sub-culture hypothesis. Several studies suggest that Turkish descendants in Europe have diverging partnership patterns from natives wherein they tend to marry earlier and choose direct marriage, rather than cohabitation (De Valk 2008; Hamel et al. 2012; Hannemann et al. 2020; Milewski/ Hamel 2010; Pailhé 2015). Liu and Kulu (2021) have recently found that migrants with Turkish backgrounds in Germany enter marriage and parenthood earlier and in higher proportions than their native counterparts and extra-marital birth is much lower for Turkish Germans. Even in their adolescent years, Turkish immigrants still have stronger preferences for marriage than other immigrant groups and natives (De Valk/Liefbroer 2007). The adaptation hypothesis predicts that immigrant conformation to the demographic, social and economic behaviour of natives increases alongside the Partnership Transitions among Turkish Immigrants and their Descendants ... • 327 duration of their residence in the host country (Hervitz 1985). This approach argues that the effect of mainstream culture is more dominant than childhood experience. Although this is not a sudden transformation, gradually and in medium-term timeframes, immigrants fi nd and defi ne a place within the existing social structure and lifestyle of their adopted culture. Descendants of immigrants also adapt to the host society while being exposed to their family’s past values, norms and behaviours. Meanwhile, the segmented assimilation theory proposed by Portes and Zhou (1993) addressed the confl ict between immigrant families and the social and cultural demands of the host society, as cited in Rahnu et al. (2015). According to segmented assimilation theory, migrant descendants can preserve their community’s culture and values while being successfully economically integrated (Rahnu et al. 2015). Hence, second-generation immigrants exist somewhere between the cultural dynamics of two distinct societies. For instance, in the British case, Hannemann and Kulu (2015) conclude that the second generations of South Asians and Caribbeans maintain the patterns of their parents while making small deviations towards British native society. In Spain, although country-of-origin social patterns seem dominant among immigrants, these preferences are relaxed among second-generation adolescents (González-Ferrer et al. 2014). In the US, Arias (2001) fi nds that the socio-economic gap between immigrants and natives weakens in each successive cohort, thus suggesting that immigrants mirror the practices prevalent in the US. With each succeeding generation, immigrants living in the US opt for dominant patterns, although the level of convergence differs according to their origin (Brown et al. 2008; Landale et al. 2010). Apart from decreasing socio-economic variances, native-immigrant relationships in early age increase the probability of convergence. Huschek, Liefbroer and de Valk (2010) fi nds that Turkish immigrants’ children who have non-coethnic peers will tend to build co-residential unions later than their parents, especially if they have had greater contact with native populations, i.e., if they have a higher proportion of natives in their schools. The adaptation process is also related to the welfare system in the destination country. For instance, in the Swedish case, the state welfare system provides the same opportunity structure and fi nancial support in housing for both immigrants and natives; thus, Turkish immigrants have more freedom to follow the early union formation practices of Swedes or to choose the traditional Turkish practice of marrying early. On the other hand, in Germany, union formation usually occurs after securing access to a paid job and both natives and Turkish second- generation delay partnership formation to later ages (Hamel et al. 2012). For union separation, the patterns and dynamics would be different as traditional family values could impose barriers to separation. The third hypothesis is selectivity, which suggests that migration itself is a selective phenomenon, meaning immigrants are already a selected group with life preferences that may differ from the norm in their country of origin (Hannemann/ Kulu 2015). Their nuptiality preferences, then, are already proximate to the native population in the destination country. This selectivity may originate at the individual level, in their social, cultural and economic capital. Finally, assuming that migration has some psychological, economic and social cost to immigrants, the disruption • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören328 hypothesis envisions that immigrant preferences might be different from those of their country of origin only after they move to their destination (Adserà/Ferrer 2014). In other words, the structure of the marriage market of the receiving country and the available opportunities may not be compatible with those that immigrants are familiar with, or the lack of coethnic partners may result in unplanned deviation from their initial partnership practice preferences. These immigrants may delay the timing of partnership or accelerate the decision to divorce. However, Carlson (1985) suggests that migration has only a temporary effect on the timing of marriages. Single immigrants to Australia delayed marriage compared to their country-of-origin peers. Nevertheless, this effect is weak among successive generations. Similarly, Andersson, Obućina and Scott (2015) fi nd high divorce rates among some people of immigrant origins due to the stressful nature of migration and the changed landscape of the marriage market. This disruption can affect union formation in different ways: while Turkish, Arab, African and Yugoslavian immigrants have elevated risk after migration, some countries have low marriage rates after migration. Based on this theoretical and empirical background, we formulate four hypotheses (H)3: H1: Partnership transitions of fi rst- and second-generation Turkish migrants differ from natives in Western Germany (but converge to patterns prevalent in Türkiye). This hypothesis has roots in the socialisation and cultural maintenance perspectives. The effect of country of origin (Türkiye) is assumed to extend beyond fi rst-generation immigrants. H2: The discrepancy between the natives and Turkish migrants in Western Germany vary across partnership transitions, namely transition to fi rst partnership (including both cohabitation and marriage), transition to fi rst marriage, transition to fi rst cohabitation, transition from cohabitation to marriage and transition from marriage to divorce. Specifi cally, the difference between migrants and natives in transition to fi rst partnership is expected to be small, whereas in transition to cohabitation or marriage, it is expected to be large. These different levels of convergence are due to the fact that fi rst partnership transition may end up in no event occurrence (i.e., remaining single) or an event occurrence that involves either cohabitation or marriage. Any difference of preference of marriage over cohabitation or vice versa between migrant and native groups would cancel each other out in the analysis of transition to fi rst partnership. We also expect high divergence in the transition from cohabitation to marriage and transition from marriage to divorce since family norms in the two contexts (Germany and Türkiye) are quite different from each other. H3a: Partnership transitions of Turkish immigrants’ descendants more closely resemble those of the fi rst generation compared to natives. This proposition has roots in the cultural maintenance hypothesis or segmented assimilation theory. Alternatively: 3 These hypotheses are similar to the ones in Rahnu et al. (2015) for their study in Estonia. Partnership Transitions among Turkish Immigrants and their Descendants ... • 329 H3b: Partnership transitions of Turkish immigrants’ descendants more closely resemble those of natives compared to fi rst generation immigrants. This hypothesis is based on the adaptation hypothesis, since descendants of immigrants are exposed to the host country much longer than their parents. As we’ve seen, a single hypothesis is not enough to understand the position of immigrants and their descendants compared to natives. This paper mainly uses the frameworks of socialisation, adaptation, cultural maintenance and segmented assimilation approaches. Considering disruption and selectivity hypotheses are not as relevant for second-generation immigrants, we exclude these two from our formulated hypotheses. Moreover, due to a lack of information regarding the behaviours and characteristics of natives in the country of origin (Türkiye), we cannot directly test for disruption or selectivity hypotheses for even fi rst-generation immigrants. However, we elaborate on all approaches in our interpretation of the results. 4 Data and Methodology 4.1 Data The analysis in this study is based on the Panel Analysis of Intimate Relationships and Family Dynamics, known as the German Family Panel (pairfam), release 10.0, which is funded by the German Research Foundation (DFG) as a long-term project (Brüderl et al. 2019a).4 We use the fi rst ten waves of pairfam covering the period of 2008 to 2018. This paper uses anchor data of the respondents themselves consisting of three birth cohorts namely those born between 1991-93, 1981-83 and 1971-73. Each birth cohort consists of a nearly equal number of respondents: 4,338 for the youngest, 4,010 for the middle and 4,054 for the oldest cohort (Brüderl et al. 2019b). Our analytical sample includes only those respondents whose parents were both born in Western Germany, i.e., natives and those who have a Turkish background living there. After removing all other origins from the data, the sample size is 7,560. Of this population, 6,950 form the fi rst analysis group in this research, namely natives, which is defi ned as those respondents whose parents were both born in Western Germany. Turkish second-generation individuals, meaning those born in Western Germany but with at least one parent born in Türkiye, correspond to 370 respondents. The remaining 240 anchors are fi rst-generation Turkish immigrants born in Türkiye with migration experience. Among natives, 2 respondents are dropped since their gender information is missing. Table 2 shows the percentage distribution of respondents according to migration status and transition types and outcomes. The second generation is the youngest group and the proportion that remain never partnered is high. As expected, this 4 It was initiated in 2008 and coordinated by Josef Brüderl, Sonja Drobnič, Karsten Hank, Bernhard Nauck, Franz J. Neyer and Sabine Walper. • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören330 T a b . 2 : P e rc e n t d is tr ib u ti o n o f tr an si ti o n t y p e s an d o u tc o m e s o f re sp o n d e n ts , p a ir fa m fi r st 1 0 w av e s (2 0 0 8 -2 01 8 ), a n a ly si s p e ri o d : 19 8 5 -2 01 8 F ir st t ra n si ti o n t y p e M ig ra ti o n s ta tu s R e m a in e d S in g le S in g le  C o h a b it a ti o n S in g le  M a rr ia g e T o ta l N u m b e r N a ti v e 3 9 .7 5 6 .3 4 .0 10 0 .0 6 ,9 4 8 1 st G e n e ra ti o n 15 .8 2 2 .5 61 .7 10 0 .0 2 4 0 2 n d G e n e ra ti o n 5 8 .1 18 .4 2 3 .5 10 0 .0 3 7 0 T o ta l 3 9 .8 5 3 .3 6 .8 10 0 .0 7, 5 5 8 F ir st c o h a b it a ti o n o u tc o m e s M ig ra ti o n s ta tu s M a rr ia g e S e p a ra ti o n R e m a in e d C o h a b it in g D e a th o f P a rt n e r T o ta l N u m b e r N a ti v e 4 7. 1 3 0 .6 2 2 .2 0 .1 10 0 .0 3 ,9 11 1s t G e n e ra ti o n 74 .1 16 .7 9 .2 0 .0 10 0 .0 5 4 2 n d G e n e ra ti o n 5 5 .9 3 0 .9 11 .8 1. 5 10 0 .0 6 8 T o ta l 4 7. 6 3 0 .5 2 1. 9 0 .1 10 0 .0 4 ,0 3 3 F ir st m a rr ia g e o u tc o m e s M ig ra ti o n s ta tu s R e m a in e d M a rr ie d D iv o rc e D e a th o f a p a rt n e r T o ta l N u m b e r N a ti v e 8 4 .2 15 .3 0 .6 10 0 .0 2 ,1 2 2 1s t G e n e ra ti o n 8 6 .7 12 .8 0 .5 10 0 .0 18 8 2 n d G e n e ra ti o n 8 9 .6 10 .4 0 .0 10 0 .0 12 5 T o ta l 8 4 .6 14 .8 0 .5 10 0 .0 2 ,4 3 5 S o u rc e : A u th o rs ’ c al cu la ti o n s b as e d o n p ai rf am d at a Partnership Transitions among Turkish Immigrants and their Descendants ... • 331 proportion is the lowest among fi rst-generation Turkish immigrants, since they are older and are more likely to form a union compared to second generation immigrants. Cohabitation levels represented in Table 2 do not represent the current status of respondents. Rather, they present the percentage of respondents whose fi rst transition was from singlehood to cohabitation in the data set. In many cases, cohabitation is a fi rst step for couples towards forming a marriage or it may lead to a dissolution of union through separation. Approximately 78 percent of the pairfam sample chose to marry or separate as opposed to maintaining extramarital cohabitation. While 74 percent of cohabiting Turkish fi rst-generation immigrants proceed to marriage, this proportion decreases to 47 percent among German- descendant cohabiting unions. This table may also roughly imply that those still in cohabiting unions also tend to proceed with one of the events in the future, since only 881 out of 4,033 respondents are still cohabiting. Relatively lower shares of marriages end up in divorce among both migrants and natives. The sample compositions of our event history setting are represented for each transition in Table A.1, A.2 and A.3, where the number and percentage distribution of person- months and events are shown. Our event history setting is explained in more detail in the following section. 4.2 Method We analysed fi ve partnerships transitions including the transitions 1) from singlehood to the fi rst union, 2) from singlehood to direct marriage, 3) from singlehood to cohabitation, 4) of dissolution of cohabiting unions through marriage and 5) divorce of ever-married couples by employing Cox proportional hazard models. Figure 1 below shows the number of respondents analysed in each transition.5 Regardless of the type of partnership, all unions are defi ned as co-residential unions lasting at least six months. The risk period starts with age 15 for the fi rst and second transitions. Censoring is applied at the last interview date to those respondents who do not form a union. The date of cohabitation and marriage (ever-married respondents) are the onset of the risk period for third and last transition models, respectively. Censored cases are those that have not experienced any dissolution at the last interview they participated in or on the date of the death of a partner. As a modelling strategy, in our Cox proportional hazard models, we fi rst control for sex and birth cohort. Model 2 adds education to Model 1. Finally, the last model controls for pregnancy and parity status of respondents. Cohabitation outcomes and divorce events have additional controls on age at union formation and type of union. The use of a proportionality assumption that suggests a constant hazard ratio over time is critical for proper implementation of a Cox model and required for accurate results. In this study, we use Schoenfeld residuals to monitor the effects 5 Excluded cases are those respondents who have their fi rst union before age 15 in the transition to the fi rst partnership formation (38 cases). The other excluded observations in the transition to marriage after cohabitation and divorce are those cases that event occurs at the same time at the beginning of the risk period (36 cases). • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören332 of covariate changes over time. In the case of variables violating proportionality assumptions in the standard Cox model, the hazard ratios reported should be taken as average hazard ratios. In such instances, we also provide results from the stratifi ed and extended Cox model, which are reported in the Appendix (see Table A.4, A.5, A.6 and A.7). In addition, we carried out robustness checks for both sexes for each transition using restricted samples. We analysed (1) transition to fi rst union excluding the fi rst generation immigrants who were married or cohabiting before migration, (2) transition to fi rst cohabitation excluding fi rst generation immigrants who were cohabiting before migration, (3) transition to fi rst marriage excluding fi rst generation immigrants who were married before migration, (4) transition to marriage after cohabitation excluding fi rst generation immigrants who experienced the transition before migration and (5) transition to divorce excluding fi rst generation immigrants who divorced before migration. We present them as robustness checks of standard Cox models in the Appendix tables A.9, A.10, A.11, A.12 and A.13. The multivariate results in the main text are reported without dividing analysis by sex, but separate analyses by males and females can be found in the Appendix (see from Table A.14 to A.23). In total, we include eight explanatory variables in the models, both time-fi xed and time-varying. These include: Migration Status; migration status of the respondent is a time-fi xed categorical variable with three groups: “natives” whose parents were born in Western Germany, “fi rst-generation Turkish immigrants” born in Türkiye and “second-generation Turkish immigrants” born in Western Germany with at least one parent born in Türkiye. Natives are the reference category in our multivariate analysis. Fig. 1: Number of union transitions of pairfam data, 2008-2018 S in g le 7 5 2 0 7 5 5 8 - 3 8 (e x cl u d e d b e fo re s ta rt ) Stay Single 2982 Cohabitation 4027 Separation 1225 Stay cohabited/death of partner 884 Marriage 1918 Marriage 511 Ever-married 2429 2435 - 6 (excluded before start) Stay married/death of partner 2072 Divorce 357 Source: Authors’ design based on pairfam data Partnership Transitions among Turkish Immigrants and their Descendants ... • 333 Sex of the respondent: This variable is used to analyse any different patterns related to gender in the formation and/or separation of a partnership. This is a time- fi xed covariate having two levels; “male” (reference category) and “female”. Birth Cohort: There are three categories, those born between 1971-1973, 1981-1983 and 1991-1993. This variable is constant over time and aims to refl ect the variation of demographic behaviours across birth cohorts since they have completely different social, economic and demographic realities. Education of respondents: This is a time-varying variable that focuses on the highest level of education completed. It is produced by using years of schooling and operationalised by giving reference to the 1997 International Standard Classifi cation of Education (ISCED-97). There is an assumption that education starts at the age of 6 and continues without any break until reported completion. The fi rst level refers to 6 completed years of schooling and is categorised as “no or primary education”. While computing this reference category, those graduating from primary school and no education at all are combined since only 8 cases report no education before the event of interest occurs. “Lower secondary”, “upper secondary” and “tertiary” are the other categories used to investigate the effect of education on timing and prevalence of event. Each education level, respectively, refers to completed 10, 13 and 16 or more years of schooling. A similar classifi cation is done in ISCED-11 as presented in Appendix 1.b of Liu and Kulu (2021). Their classifi cation presents the cut-off points for years of schooling as 10.5, 13.5 and 16, where they also control for age when last observed in school. Our thresholds are close to theirs but we only control for years of schooling. Parity: This variable is operationalised as time-varying and computed by using the birth history generated dataset provided by pairfam called “biochild”. All biological births are those of the respondent. It has three categories: childless (reference category) and 1+ (having 1 or more children). The logic behind this division is that there is a limited number of cases reporting more than two children before forming or dissolving any union (see Table A.1, A.2 and A.3). Pregnancy: Pregnancy is a time-varying variable that is adjusted as 7 months before the date of birth of the children and refers to two categories, not pregnant (reference category) and pregnant. For males, it is the pregnancy status of the partner. Age at union formation: This is a categorical variable grouped under fi ve levels among those forming a partnership with the age of 15. Age at union can take following values; 15-18, 19-22, 23-26, 27-30, 31+. Those respondents who formed their union between age 19-22 is taken as the reference category. For the transition to marriage after cohabitation, this variable refers to the age at fi rst cohabitation. For the divorce model, age at union formation refers to the age at fi rst union, regardless of type of partnership. Type of union: Type of union is included in the divorce of ever-married couples to shed light on whether the type of fi rst partnership explains the divorce patterns of respondents. Direct marriage and pre-marital cohabitation (reference category) are the two defi ned levels for this variable. • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören334 5 Results 5.1 Median Age and Duration for Partnership Trajectories Below, Table 3 provides descriptive results, namely median age and duration for partnership trajectories from Kaplan-Meier survival estimates. At the age of 25, half of the respondents have formed their fi rst partnership in pairfam data. Earlier union formation is more evident among fi rst-generation immigrants compared to descendants and natives. On the other hand, the timing of fi rst partnership is similar for natives and second-generation immigrants. This pattern persists across the three groups when the analysis is divided into female and male. In line with previous studies, women tend to form any type of union earlier than men. The patterns of immigrant generations resemble each other once the type of union is considered. This harmony is highly prominent in the transition to fi rst cohabitation, whereas descendants show an obvious preference for direct marriage compared to their fi rst-generation counterparts. Since less than half of Tab. 3: Median Age and Duration for Partnership Trajectories Both Men Women First Union Native 25.25 26.83 23.66 1st Generation 21.66 23.00 20.25 2nd Generation 24.50 26.75 22.66 Total 25.08 26.75 23.50 Direct Marriage Native - - - 1st Generation 22.83 25.00 21.00 2nd Generation 28.58 28.58 26.66 Total - - - Cohabitation Native 25.58 27.16 23.92 1st Generation 32.00 30.67 34.25 2nd Generation 30.00 33.00 29.83 Total 25.75 27.33 24.08 Marriage after cohabitation Duration since cohabitation (year) Native 4.42 4.42 4.42 1st Generation 0.75 2.17 0.58 2nd Generation 3.08 2.25 3.83 Total 4.42 4.42 4.33 Note: The event of divorce is not shown in the table. Since less than half of the risk population get divorced, no median age can be calculated for this transition for any of the sub-groups. Source: Authors’ calculations based on pairfam data Partnership Transitions among Turkish Immigrants and their Descendants ... • 335 the risk population chooses direct marriage, no median age can be calculated for this transition for the total population. It should be noted that the median age at fi rst cohabitation is about 5 years higher among Turkish immigrants compared to natives. This is an interesting fi nding and may be related to the selectivity of this migrant group who prefer to cohabit. It is a fi nding that invites analysis in future studies. Half of the cohabiting partners get married sooner than 4.4 years after cohabitation. The fi rst-generation Turkish immigrants are the fastest group to advance their relationship into marriage, with an average of 50 percent getting married within 9 months. On the other hand, while male descendants more or less follow the patterns of fi rst-generation immigrants, female second-generation Turkish immigrants prolong their period of cohabitation before getting married. Natives, regardless of gender, seem to keep the duration of cohabitation at about 4.4 years before they form a conjugal family. 5.2 Union Formation Timing and type of fi rst partnerships are profoundly correlated with the type of economic, social and demographic factors that an individual was born into. This section presents multivariate analyses (Cox proportional hazard models) to investigate the position of fi rst- and second-generation Turkish immigrants compared to natives. The analysis is based on respondents’ entry into their fi rst partnership, regardless of the type of union. Respondents are at risk at the beginning of age 15 and followed until they form the fi rst union or until censoring occurs at the last interview they participated in before forming a partnership. Table 4 shows the results from the Cox proportional hazard model for the transition to fi rst union formations of direct marriage or cohabitation. In general, comparisons of immigrant generations and natives reveal compelling differences between the two groups. While fi rst-generation Turkish immigrants have a 79 percent higher risk to form a union than the native population, descendants and natives are almost alike. These patterns persist when we control for respondent’s educational level. Even though, in general, those pursuing higher education delay union formation longer than those respondents having a lower level of educational attainment, in our model, education does not explain differences among immigrants and natives. First-generation immigrants still have a higher hazard ratio than their children when compared to natives. In the full model, hazard ratios stay almost the same. The following regression models investigate whether and in which direction natives and immigrant generations differ in terms of their preferences for the type of fi rst entry into a union. The results assert systemic differences between immigrant generations and natives. The most important fi nding of this model is that the type of union is the primary force that generates and reinforces the contrast between German natives and Turkish immigrants. Table 5, the hazard of cohabitation for both sexes, illustrates dramatically different behaviours of migrant groups and natives specifi cally when controlled for sex and birth, education, and parity and pregnancy. As seen on the standard model, among cohabiting unions, • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören336 T a b . 4: R e la ti v e h a za rd o f fi rs t u n io n f o rm a ti o n M 1 M 2 M 3 V a ri a b le B e ta p -v a lu e C I B e ta p -v a lu e C I B e ta p -v a lu e C I N a ti v e 1. 0 0 . . . 1. 0 0 . . . 1. 0 0 . . . 1s t G e n e ra ti o n 1. 7 9 0 .0 0 1. 5 7 2 .0 9 1. 6 5 0 .0 0 1. 4 4 1. 9 4 1. 6 3 0 .0 0 1. 4 0 1. 9 0 2 n d G e n e ra ti o n 1. 11 0 .2 1 0 .9 4 1. 3 0 1. 0 6 0 .4 9 0 .9 0 1. 2 5 1. 0 6 0 .4 5 0 .9 0 1. 2 5 N o te : M o d e l 1 c o n tr o lle d fo r se x a n d b ir th c o h o rt . M o d e l 2 a d d it io n al ly c o n tr o lle d fo r e d u ca ti o n al le v e l. M o d e l 3 a d d it io n al ly c o n tr o lle d fo r p ar it y a n d p re g n an cy . S o u rc e : A u th o rs ’ c al cu la ti o n s b as e d o n p ai rf am d at a M 1 M 2 M 3 V a ri a b le B e ta p -v a lu e C I B e ta p -v a lu e C I B e ta p -v a lu e C I N a ti v e 1. 0 0 . . . 1. 0 0 . . . 1. 0 0 . . . 1 st G e n e ra ti o n 0 .5 4 0 .0 0 0 .4 1 0 .7 1 0 .5 0 0 .0 0 0 .3 8 0 .6 6 0 .5 0 0 .0 0 0 .3 8 0 .6 5 2 n d G e n e ra ti o n 0 .5 1 0 .0 0 0 .4 0 0 .6 5 0 .4 9 0 .0 0 0 .3 8 0 .6 2 0 .4 9 0 .0 0 0 .3 8 0 .6 2 T a b . 5: R e la ti v e h a za rd o f fi rs t co h a b it a ti o n N o te : M o d e l 1 c o n tr o lle d fo r se x a n d b ir th c o h o rt . M o d e l 2 a d d it io n al ly c o n tr o lle d fo r e d u ca ti o n al le v e l. M o d e l 3 a d d it io n al ly c o n tr o lle d fo r p ar it y a n d p re g n an cy . S o u rc e : A u th o rs ’ c al cu la ti o n s b as e d o n p ai rf am d at a Partnership Transitions among Turkish Immigrants and their Descendants ... • 337 the patterns of descendants largely resemble those of fi rst-generation immigrants. The risk level predicts that all Turkish immigrant generations produce, adapt and sustain compatible behavioural trends. Controlling for educational level, parity and pregnancy does not change hazard ratios signifi cantly. According to the third model, the fi rst and second generation have 50 percent and 49 percent lower risk of cohabitation than natives, respectively. This is not surprising since most native respondents prefer cohabiting with their partners, while only one-fi fth of Turkish immigrants form cohabiting unions as their fi rst partnership (see Table 2). The marriage model stands out as the most distinctive in the analyses, since this is rarely preferred by natives in the analysed age groups. Additionally, fi rst partnership formation highly coincides with marriage among fi rst-generation Turkish immigrants (see Table 2). Table 6 shows that while immigrant generations have a higher intensity of direct marriage, their preferences are not as consistent once they are in cohabitation. Turkish immigrants have a 16.1 times higher hazard of marrying than that of natives. This risk level decreases to 11.6 times among descendants. When educational level, parity and pregnancy status of respondents are controlled, the intensity of direct marriage decreases among immigrants. In the fi nal model, fi rst-generation Turkish immigrants have 13.7 times higher and second generation have 11.0 times higher marriage intensity compared to natives. When we look at other covariates (Appendix Table A.4, A.5 and A.6), the analysis shows that women enter into the fi rst union earlier and have a signifi cantly higher risk of direct marriage or cohabitation than their male counterparts. In line with the gradual retreat from marriage and increasing popularity of cohabiting unions over time, pathways to partnership formation vary across birth cohorts. The intensity of direct marriage is evident among older generations, whereas the younger generation prefers cohabiting unions as their fi rst partnership. Pregnancy is another important variable increasing the hazard of direct marriage. 5.3 Marriage after Cohabitation Until this point in the study, the results indicate that cohabitation is a highly preferred practice among native respondents but still rare across Turkish immigrant generations in Western Germany. However, there are further questions to be answered about those respondents who cohabit in the fi rst place, such as “what comes after cohabitation?” Once a cohabiting partnership is formed, couples may prefer either to proceed to marriage or to end the relationship through separation. Secondly, duration of cohabitation indicates whether cohabitation is perceived as a long-term form of relationship or a waiting room for couples still indecisive about their future together. In this respect, natives and Turkish immigrant generations may opt for a different path on what to do with the cohabiting union. Table 7 presents the intensity of marriage after cohabitation. First-generation Turkish immigrants have 2.10 times higher risk of marriage than their native counterparts on average. Although the risk level is slightly lower among descendants, the intensity of marriage is 2.02 times higher than that of natives. Including education in the model does not change risk levels much in the transition • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören338 M 1 M 2 M 3 V a ri a b le B e ta p -v a lu e C I B e ta p -v a lu e C I B e ta p -v a lu e C I N a ti v e 1. 0 0 . . . 1. 0 0 . . . 1. 0 0 . . . 1 st G e n e ra ti o n 16 .1 0 0 .0 0 13 .1 3 19 .7 7 13 .9 7 0 .0 0 11 .1 1 17 .5 7 13 .7 4 0 .0 0 10 .9 1 17 .3 1 2 n d G e n e ra ti o n 11 .5 6 0 .0 0 9 .0 4 14 .8 7 10 .8 2 0 .0 0 8 .4 3 13 .9 0 10 .9 5 0 .0 0 8 .5 3 14 .0 6 T a b . 6: R e la ti v e h a za rd o f fi rs t m a rr ia g e N o te : M o d e l 1 c o n tr o lle d fo r se x a n d b ir th c o h o rt . M o d e l 2 a d d it io n al ly c o n tr o lle d fo r e d u ca ti o n al le v e l. M o d e l 3 a d d it io n al ly c o n tr o lle d fo r p ar it y a n d p re g n an cy . S o u rc e : A u th o rs ’ c al cu la ti o n s b as e d o n p ai rf am d at a T a b . 7: R e la ti v e h a za rd o f m a rr ia g e a ft e r co h a b it a ti o n M 1 M 2 M 3 V a ri a b le B e ta p -v a lu e C I B e ta p -v a lu e C I B e ta p -v a lu e C I N a ti v e 1. 0 0 . . . 1. 0 0 . . . 1. 0 0 . . . 1s t G e n e ra ti o n 2 .1 0 0 .0 0 1. 5 3 2 .9 0 2 .0 5 0 .0 0 1. 4 7 2 .8 6 1. 8 2 0 .0 0 1. 3 1 2 .5 4 2 n d G e n e ra ti o n 2 .0 2 0 .0 0 1. 4 6 2 .8 0 1. 9 3 0 .0 0 1. 3 9 2 .6 8 1. 6 5 0 .0 0 1. 19 2 .2 9 N o te : M o d e l 1 c o n tr o lle d f o r se x , b ir th c o h o rt a n d a g e a t u n io n f o rm at io n . M o d e l 2 a d d it io n al ly c o n tr o lle d f o r e d u ca ti o n al le v e l. M o d e l 3 a d d it io n al ly c o n tr o lle d f o r p ar it y a n d p re g n an cy . S o u rc e : A u th o rs ’ c al cu la ti o n s b as e d o n p ai rf am d at a Partnership Transitions among Turkish Immigrants and their Descendants ... • 339 to marriage after cohabitation. The last model, which additionally controls for parity and pregnancy, indicates that marriage intensity is 1.8 times higher among fi rst generation Turkish immigrants compared to natives, while this relative ratio is 1.65 among descendants. Controlling for pregnancy is important as almost 20 percent of cohabited respondents are pregnant before marrying (see Appendix Table A.7 for results related to covariates). 5.4 Divorce Divorce is the fi nal transition examined in our analyses. Respondents are followed from the time of marriage – either directly or after a cohabitation period – until a divorce event. Censoring occurs at the date of a partner’s death or interview date for ongoing marriages. Before that, it may be good to remember that natives in Western Germany prefer to delay marriage to later ages and, in most cases, they prefer a period of cohabitation. In addition, fi rst-generation Turkish immigrants are less prone to practice divorce and descendants are still too young to proceed to higher-order unions, since the share of divorced individuals are lower among ever- married Turkish immigrants (Table 2). Table 8 presents the estimates for the divorce of ever-married partnerships. The results show signifi cant differences among natives and Turkish immigrants. The hazard of divorce among the fi rst generation is lower than their native counterparts where descendants follow a similar pattern with the fi rst-generation immigrants. The intensity of divorce slightly decreases when the educational composition of respondents is considered. In the last model, the hazard of divorce reaches approximately one third of the natives for both fi rst- and second-generation Turkish immigrants. The hazard of divorce among the fi rst generation is 66 percent lower than their native counterparts. Descendants follow a similar pattern, wherein the intensity of divorce is 0.36 of that of natives. With regard to the association of explanatory covariates; increasing levels of education do not seem to be highly correlated with a decreasing risk of union formation including all transitions (as in Hannemann/Kulu 2015; Pailhé 2015). Instead, regardless of the type of union, educational attainment may explain the relative delay in partnership formation. That is, highly educated people do not necessarily reject marriage or cohabitation, but because they spend longer periods in education, union formation occurs later in life than lesser-educated respondents (Kalmijn 2007). Thus, the aforementioned differences in partnership trajectories are not infl uenced by the human capital that education provides. Pregnancy in all transitions except divorce increases the intensity of partnership formation. The effect of pregnancy and parity are consistent with the fi ndings of dissolution literature: Couples are less likely to end a marriage with the existence of children (Choi et al. 2020; Dribe/Lundh 2012; Kaplan//Herbst-Debby 2017; Milewski/ Kulu 2014). Of particular note, direct marriage is found to have an increased risk of divorce compared to previously cohabiting couples. This fi nding contrasts many studies in the literature and invites further investigation (see Appendix Table A.8). • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören340 T a b . 8: R e la ti v e h a za rd o f d iv o rc e o f e v e r- m a rr ie d M 1 M 2 M 3 V a ri a b le B e ta p -v a lu e C I B e ta p -v a lu e C I B e ta p -v a lu e C I N a ti v e 1. 0 0 . . . 1. 0 0 . . . 1. 0 0 . . . 1 st G e n e ra ti o n 0 .5 1 0 .0 0 0 .3 3 0 .7 8 0 .4 3 0 .0 0 0 .2 7 0 .6 8 0 .3 4 0 .0 0 0 .2 1 0 .5 6 2 n d G e n e ra ti o n 0 .5 0 0 .0 2 0 .2 8 0 .9 0 0 .4 5 0 .0 1 0 .2 5 0 .8 1 0 .3 6 0 .0 0 0 .1 9 0 .6 7 N o te : M o d e l 1 c o n tr o lle d f o r se x , b ir th c o h o rt a n d a g e a t u n io n f o rm at io n . M o d e l 2 a d d it io n al ly c o n tr o lle d f o r e d u ca ti o n al le v e l. M o d e l 3 a d d it io n al ly c o n tr o lle d f o r p ar it y, p re g n an cy a n d t y p e o f u n io n . S o u rc e : A u th o rs ’ c al cu la ti o n s b as e d o n p ai rf am d at a Partnership Transitions among Turkish Immigrants and their Descendants ... • 341 5.5 Robustness Checks Our robustness check results are presented in Appendix Tables A.9, A.10, A.11, A.12 and A.13. According to these results, our fi ndings are robust to sample restrictions, i.e., exclusion of fi rst-generation immigrants who experienced the event of interest before migration. One difference observed is the decline in hazard ratios of fi rst- generation immigrants related to overall union formation, transition to cohabitation and transition to marriage. A decline in the hazard ratio of divorce also takes place with the restricted sample, although the difference is smaller. These fi ndings can be explained by the fact that marriage or cohabitation can take place as a reason for migration due to the need for family unifi cation. Some individuals may tend to enter a union earlier to migrate to Western Germany afterwards. Hence, excluding these cases may have lowered the hazard ratios of fi rst immigrants. However, the main conclusions derived do not change when analyses are applied to the restricted samples. 6 Discussion and Conclusion This study has investigated union formation and dissolution practices of fi rst- and second-generation Turkish immigrants compared to natives in Western Germany. It addressed how fi rst-generation Turkish immigrants and their descendants respond to the native patterns and it analysed the extent of convergence or divergence between generations. We also formulated four hypotheses, with two being alternatives to each other, based on established theoretical approaches. Our fi ndings indicate that both the timing and incidence of union formation comply among descendants of Turkish immigrants and natives in Western Germany, while fi rst-generation immigrants pursue a traditional early union formation pattern. First- generation immigrants also have a signifi cantly higher risk of a partnership than both natives and their descendants. Descendants mostly share practices of the fi rst generation when it comes to cohabitation, but neither fi rst- nor second-generation groups prefer cohabiting unions or postponing marriage to their early 30s. On the other hand, marriage practices of descendants indicate a relative divergence from fi rst-generation immigrants. Traditional early direct marriage ceases, Turkish descendants signifi cantly delay the decision to form a conjugal family and they have slightly less risk than fi rst-generation Turkish immigrants. Despite these fi ndings, however, the data cannot be interpreted as cohabitation replacing marriage preference for Turkish immigrants in Western Germany. The dominance of the marriage institution is partly related to the partner choice of Turkish immigrants. Studies show that national homogamy is highly typical among Turkish minorities in Europe (Constant et al. 2012; Hannemann et al. 2018; Milewski/ Hamel 2010; Soehl/Yahirun 2011). Choosing a Turkish partner may intensify the risk of marriage and reproduce the traditional pattern. However, children of Turkish immigrants are exposed to mainstream structure and culture from childhood onwards. Contact with non-coethnic peers may postpone entry into the fi rst union • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören342 among second-generation Turkish immigrants in Germany (Huschek et al. 2010). Furthermore, these immigrants mostly form a conjugal family with another Turkish descendant in Germany rather than with a spouse from Türkiye (Gonzalez-Ferrer 2006; Hamel et al. 2012). Overall, these results suggest that children of immigrants socialise into their family’s norms and values instead of mirroring native patterns in their transitions to fi rst marriage and cohabitation. As opposed to natives, marriage is still dominant over cohabitation as a fi rst choice among Turkish immigrants in Germany. One of the aims of this paper was to examine cohabitation outcomes among natives and immigrants. As descriptive results show, cohabitation is rare among Turkish immigrants, while natives prefer it over direct marriage. This behaviour is not surprising since consensual union is a culturally disapproved phenomenon among Turkish people. Even though adolescent Turkish immigrants see cohabitation as an alternative partnership formation, in reality, they abstain from consensual unions due to strong parental infl uence and low social acceptance (De Valk/Liefbroer 2007). Therefore, Turkish descendants still mostly reject the premarital sexual activity of females in the European context (Hamel et al. 2012), as also suggested by the indicators in Section 2. At fi rst glance, the fi ndings predict that immigrants and natives are two distinct groups in their preferences for type of cohabiting union. Based on the ideal types of cohabitation conceptualised by Heuveline and Timberlake (2004), natives treat this type of partnership as “must-have stage” or “trial marriage”. On average, German- descendant respondents decide to form a conjugal family within 4.42 years after cohabitation. By contrast, and in line with Naderi’s (2008) fi nding, among Turkish immigrants, the fi rst generation tends to formalise their union within less than a year. The critical point here is that while three-fourths of fi rst-generation people marry their fi rst cohabiting partners, less than half of their descendants formalise their fi rst cohabiting union. Second-generation Turkish immigrants also prefer separating from a cohabiting partner more than natives (see Table 2). In addition, female descendants stay in cohabiting unions signifi cantly longer than their parents and male counterparts. This may hint at a transformation regarding cohabitation among second-generation immigrants. In other words, the marginality of cohabiting unions seems to erode gradually for the female second generation. Another possible explanation for this might be that, as Pailhé (2015) suggests in the context of France, Turkish second-generation female immigrants are a selected group that have individual characteristics that distinguish them from their male descendants and the fi rst generation. Note that this study is unable to control for the origin of the partner. This characteristic might also be the motive behind longer cohabitation of female descendants. Another aim of this study was to analyse the divorce practices of Turkish immigrants as opposed to natives. As the second demographic transition indicates, divorce has an upward trend and marriage is weaker among the youngest generation who also have higher intensity than older generations in deciding to end a marriage. In line with other studies (Milewski/Kulu 2014), both female and male Turkish immigrants have signifi cantly more stable marriages than natives. An explanation Partnership Transitions among Turkish Immigrants and their Descendants ... • 343 for this practice may again be found in the partner choice of Turkish immigrants and in the low divorce rate within the Turkish population. Native-immigrant marriages are found to be most fragile in any context due to socio-cultural distance between partners (Choi et al. 2020; Kaplan/Herbst-Debby 2017; Milewski/Kulu 2014; Zhang/ Van Hook 2009). Since the fi rst and second-generation immigrants form a union with a coethnic partner, contradicting social factors that can weaken the marriage are not common. Thus, fi rst-generation Turkish immigrants also transmit their preferences regarding the dissolution, which justifi es the socialisation hypothesis. It should be noted that in our analytical sample, the number of divorce events among fi rst- and second-generation Turkish immigrants were low and the results should be interpreted considering this situation. Regarding our four hypotheses, our fi rst hypothesis, which expects difference in partnership patterns between migrants and natives, mostly holds true, except for the fi nding that second-generation immigrants resemble the native pattern in transition to the fi rst union (including both cohabitation and marriage). In other words, the socialisation hypothesis for fi rst-generation immigrants appears to hold, whereas it is not extended towards second generation immigrants in terms of fi rst union formation (including both marriage and cohabitation). Nevertheless, the type of union reveals the actual distinction between natives and immigrants. When type of union is considered, our fi rst hypothesis holds for descendants of immigrants, too, supporting the cultural maintenance perspective. As suggested by our second hypothesis, the extent of the difference varies across partnership transitions. First- and second-generation immigrants have an about 50 percent lower hazard of cohabitation compared to natives, whereas the hazard of getting married is 13.7 and 11.0 times higher among fi rst- and second- generation immigrants compared to natives, respectively. Marriage intensity following cohabitation is also higher among Turkish immigrants, where the difference is greater for fi rst-generation immigrants. Finally, the divorce hazard is 66 percent and 64 percent lower among fi rst- and second-generation immigrants compared to natives. Overall, our results provide support for our third hypothesis (H3a) rather than the fourth related to second-generation immigrants, in that partnership transition of Turkish immigrants’ descendants more closely resembles those of fi rst-generation immigrants compared to natives. This provides support for the cultural maintenance hypothesis or segmented assimilation theory. Hence, the adaptation hypothesis for descendants of immigrants (hypothesis H3b) does not appear to hold, except in the transition to fi rst union (covering both cohabitation and marriage). The study at hand is not free of limitations. First, this investigation was unable to control for the selection and disruption hypotheses or their effects on immigrants’ union trajectories. Addressing whether fi rst-generation immigrants are selected groups that have different characteristics than the Turkish population requires comparable data that belongs to the society of origin. Available data sources on Turkish partnership practices lack cohabitation histories, which preclude comparing type of fi rst union. With some imagination, however, it is relatively fair to say that fi rst-generation Turkish immigrants are not selected groups. In pairfam data, • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören344 90 percent of Turkish fi rst-generation immigrant respondents were born between 1971-1983, corresponding to the 35-47 age group at the last wave of pairfam in 2018. The descriptive results show that the median age at fi rst marriage is 21 for the fi rst- generation Turkish female immigrants. According to the 2018 Turkey Demographic and Health Survey (2018 TDHS), the average median age at fi rst marriage is 21 among women aged between 35-49 (Hacettepe University Institute of Population Studies 2019). Therefore, it seems that Turkish society and Turkish immigrants have similar patterns in terms of the timing of fi rst marriage. A further limitation of this study is that the disruption hypothesis is not controlled in the multivariate analysis. Therefore, there is a question of whether immigrants might be affected by the immigration process and, as a result, struggle to adapt to the new partnership market in Western Germany. Hence, the disruption stemming from immigration itself results in a decision to postpone union formation. However, when the median age at fi rst union is compared, the age gap between those starting a co-residential union before and after migration is minimal, respectively 21 and 21.33 (see Appendix Table A.24). These descriptive results roughly suggest almost no disruption. Additionally, the data had some limitations when controlling for other socio- economic or demographic factors that may contribute to the convergence process of immigrants. The employment history of immigrants, the religious affi liation of respondents, partner choice and educational attainment of father and mother are examples of factors that could not be included in our analysis. A high proportion of missing data and a lack of retrospective information on these covariates have led the authors to renounce using them as an explanatory variable in multivariate models in order to keep the analyses as accurate as possible. Furthermore, parental infl uence, especially the impact on partner choice, should be considered extensively in analyses of future Turkish generations. Most surveys do not ask about “religious partnerships”, which may be observed among fi rst-generation Turkish migrants. The meaning and social acceptance of unregistered religious unions and cohabitation are different. Therefore, future studies may shed light on transformations across generations by illuminating the details of unoffi cial co-residential unions. Another issue for further study is the moderating effect of education. As Krapf and Wolf (2015) fi nd, highly educated Turkish women of the second generation do not differ signifi cantly in their fi rst birth risks from natives. This could also apply to partnership transitions. Finally, our fi ndings of the delayed cohabitation among Turkish immigrants in pairfam data and direct marriage increasing risk of divorce compared to previously cohabiting couples stand out as issues to be analysed further. References Abbasi-Shavazi, Mohammad Jalal; McDonald, Peter 2000: Fertility and multiculturalism: Immigrant fertility in Australia, 1977-1991. In: International Migration Review 34,1: 215- 242. https://doi.org/10.2307/2676018 Partnership Transitions among Turkish Immigrants and their Descendants ... • 345 Abbasi-Shavazi, Mohammad Jalal; McDonald, Peter 2002: A comparison of fertility patterns of European immigrants in Australia with those in the countries of origin. In: Genus 58,1: 53-76. Adserà, Alícia; Ferrer, Ana 2014: Immigrants and Demography: Marriage, Divorce, and Fertility. IZA Discussion Paper No. 7982. Andersson, Gunnar; Obucina, Ognjen; Scott, Kirk 2015: Marriage and divorce of immigrants and descendants of immigrants in Sweden. In: Demographic Research 33,2: 31-64. https://doi.org/10.4054/DemRes.2015.33.2 Arias, Elizabeth 2001: Change in Nuptiality Patterns among Cuban Americans: Evidence of Cultural and Structural Assimilation? In Source: The International Migration Review 35,2: 525-556. Brown, Susan L.; Van Hook, Jennifer; Glick, Jennifer E. 2008: Generational Differences in Cohabitation and Marriage in the US. In: Population Research and Policy Review 27,5: 531-550. https://doi.org/10.1007/s11113-008-9088-3 Brüderl, Josef et al. 2019a: The German Family Penal (pairfam). https://doi.org/10.4232/pairfam.5678.10.0.0. Brüderl, Josef et al. 2019b: The German Family Panel: Study Design and Cumulated Field Report (Waves 1 to 10) [https://www.pairfam.de/fi leadmin/user_upload/ redakteur/publis/technical_papers/TP01_Cumulated_Field_Report_ _pairfam_2019. pdf, 19.03.2020]. Carlson, Elwood D. 1985: The impact of international migration upon the timing of marriage and childbearing. In: Demography 22,1: 61-72. https://doi.org/10.2307/2060986 Choi, Yool; Kim, Doo-Sub; Ryu, Jungkyun 2020: Marital dissolution of transnational couples in South Korea. In: Journal of Ethnic and Migration Studies 46,14: 3014-3039. https://doi.org/10.1080/1369183X.2019.1585021 Constant, Amelie; Nottmeyer, Olga; Zimmerman, Klaus 2012: Cultural Integration in Germany. In: Algan, Yann et al. (Eds.): Cultural Integration of Immigrants in Europe. Oxford University Press: 69-124. De Valk, Helga 2008: Union and family Formation. In: Crul, Maurice; Heering, Liesbeth (Eds.): The position of the Turkish and Moroccan second generation in Amsterdam and Rotterdam: the TIES study in the Netherlands. [https://www.imiscoe.org/publications/ library/2-imiscoe-research-series/39-the-position-of-the-turkish-and-moroccan- second-generation-in-amsterdam-and-rotterdam, 14.06.2019]. De Valk, Helga; Liefbroer, Aart C. 2007: Parental Infl uence On Union Formation Preferences Among Turkish, Moroccan, and Dutch Adolescents in the Netherlands. In: Journal of Cross-Cultural Psychology 38,4: 487-505. https://doi.org/10.1177/0022022107302316 Dribe, Martin; Lundh, Christer 2012: Intermarriage, Value Context and Union Dissolution: Sweden 1990-2005. In: European Journal of Population 28: 139-158. https://doi.org/10.1007/s10680-011-9253-y Eurostat 2010: Household Structure in the EU. https://doi.org/10.2785/56366 German Federal Statistical Offi ce 2020: Population [https://www.destatis.de/DE/ Themen/Gesellschaft-Umwelt/Bevoelkerung/_inhalt.html, 17.05.2020]. German Federal Statistical Offi ce 2022a: Marriages and average age at marriage of singles – 2020 [https://www.destatis.de/EN/Themes/Society-Environment/ Population/Marriages-Divorces-Life-Partnerships/ Tables/marriages-average-age- marriage.html, 22.01.2022]. • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören346 German Federal Statistical Offi ce 2022b: Marriages, divorces (time series) – 2020 [https://www.destatis.de/EN/Themes/Society-Environment/Population/Marriages- Divorces-Life-Partnerships/Tables/lrbev06.html#fussnote-3-269872, 22.01.2022]. German Federal Statistical Offi ce 2022c: Couples with and without children – 2020 [https://www.destatis.de/EN/Themes/Society-Environment/Population/Households- Families/Tables/couples.html, 22.01.2022]. González-Ferrer, Amparo 2006: Who Do Immigrants Marry? Partner Choice Among Single Immigrants in Germany. In: European Sociological Review 22,2: 171-185. https://doi.org/10.1093/esr/jci050 González-Ferrer, Amparo et al. 2014: Partnership Formation and Dissolution in Spain. In: Families and Socities 13. Hacettepe University Institute of Population Studies 2019: Turkey Demograhic and Health Survey [www.hips.hacettepe.edu.tr, 21.05.2020]. Hamel, Christelle et al. 2012: Union Formation and Partner Choice. In: Crul, Maurice; Scheider, Jens; Leile, Frans (Eds.): The European Second Generation Compared. Amsterdam University Press: 225-284. https://doi.org/10.1515/9789048516926 Hannemann, Tina; Kulu, Hill 2015: Union formation and dissolution among immigrants and their descendants in the United Kingdom. In: Demographic Research 33: 273-312. https://doi.org/10.4054/DemRes.2015.33.10 Hannemann, Tina et al. 2018: Co-ethnic marriage versus intermarriage among immigrants and their descendants: A comparison across seven European countries using event-history analysis. In: Demographic Research 39,1: 487-524. https://doi.org/10.4054/DemRes.2018.39.17 Hannemann, Tina et al. 2020: Partnership dynamics among immigrants and their descendants in four European countries. In: Population, Space and Place 26,5: e2315. https://doi.org/10.1002/psp.2315 Hervitz, Hugo M. 1985: Selectivity, Adaptation, or Disruption? A Comparison of Alternative Hypotheses on the Effects of Migration on Fertility: The Case of Brazil. In: International Migration Review 19,2: 293-317. https://doi.org/10.1177/019791838501900205 Heuveline, Patrick; Timberlake, Jeffrey M. 2004: The role of cohabitation in family formation: The United States in comparative perspective. In: Journal of Marriage and Family 66,5: 1214-1230. https://doi.org/10.1111/j.0022-2445.2004.00088.x Huschek, Doreen; de Valk, Helga A. G.; Liefbroer, Aart C. 2010: Timing of fi rst union among second-generation Turks in Europe: The role of parents, peers and institutional context. In: Demographic Research 22: 473-504. https://doi.org/10.4054/DemRes.2010.22.16 Kalmijn, Matthijs 2007: Explaining cross-national differences in marriage, cohabitation, and divorce in Europe, 1990-2000. In: Population Studies 61,3: 243-263. https://doi.org/10.1080/00324720701571806 Kaplan, Amit; Herbst-Debby, Anat 2017: Mixed-ethnicity marriages and marital dissolution in Israel. In: Journal of Israeli History 36,2: 291-312. https://doi.org/10.1080/13531042.2018.1549833 Kiernan, Kathleen 2002: The State of European Unions: An Analysis of Partnership Formation and Dissolution. In: Macura, Miroslav; Beets, Gijs (Eds.): Dynamics of Fertility and Partnership in Europe: insights and lessons from comparative research. New York: United Nations: 57-76. Köppen, Katja 2011: Marriage and Cohabitation in western Germany and France. Rostock: Max Planck Institute for Demographic Research (MPIDR). Partnership Transitions among Turkish Immigrants and their Descendants ... • 347 Krapf, Sandra; Wolf, Katharina 2015: Persisting Differences or Adaptation to German Fertility Patterns? First and Second Birth Behavior of the 1.5 and Second Generation Turkish Migrants in Germany. In: KZfSS Kölner Zeitschrift für Soziologie und Sozialpsychologie 67,1: 137-164. https://doi.org/10.1007/s11577-015-0331-8 Kuhnt, Anne-Kristin; Krapf, Sandra 2020: Partnership Living Arrangements of Immigrants and Natives in Germany. In: Frontiers in Sociology 5, 86. https://doi.org/10.3389/FSOC.2020.538977 Kulu, Hill; Milewski, Nadja 2007: Family change and migration in the life course: An introduction. In: Demographic Research 17: 567-590. https://doi.org/10.4054/DemRes.2007.17.19 Landale, Nancy S.; Schoen, Robert; Daniels, Kimberly 2010: Early Family Formation Among White, Black, and Mexican American Women. In: Journal of Family Issues 31,4: 445-474. https://doi.org/10.1177/0192513X09342847 Liu, Chia; Kulu, Hill 2021: First comes marriage or fi rst comes carriage? Family trajectories for immigrants in Germany. MigrantLife Working Paper 4. University of St. Andrews. Milewski, Nadja; Hamel, Christelle 2010: Union Formation and Partner Choice in a Transnational Context: The Case of Descendants of Turkish Immigrants in France. In: International Migration Review 44,3: 615-658. https://doi.org/10.1111/j.1747-7379.2010.00820.x Milewski, Nadja; Kulu, Hill 2014: Mixed Marriages in Germany: A High Risk of Divorce for Immigrant-Native Couples / Mariages mixtes en Allemagne: un risque de divorce élevé pour les couples immigrant(e) et natif(ve) d’Allemagne. In: European Journal of Population 30,1: 89-113. https://doi.org/10.1007/s10680-013-9298-1 Naderi, Robert 2008: BiB – GGS – Publications – Ehen und nichteheliche Lebensgemein- schaften im Lebensverlauf von Deutschen und türkischen Staatsbürgern in Deutsch- land. In: Zeitschrift Für Bevölkerungswıssenschaft 33,3-4: 433-448. Naderi, Robert 2015: Kinderzahl und Migrationshintergrund Ein Vergleich zwischen Frauen türkischer Herkunft mit oder ohne eigene Wanderungserfahrung sowie Frauen ohne Migrationshintergrund in Westdeutschland. In: Zeitschrift Für Familienforschung, 27,3: 322-342. Pailhé, Ariane 2014: Union Formation and Dissolution Among Immigrants and Their Descendants in the French Context. In: Families and Socities 13: 39-51. Pailhé, Ariane 2015: Partnership dynamics across generations of immigration in France: Structural vs. cultural factors. In: Demographic Research 33: 451-498. https://doi.org/10.4054/DemRes.2015.33.16 Portes, Alejandro; Zhou, Min 1993: The new second generation: Segmented assimilation and its variants. In: The Annals of the American Academy of Political and Social Science 530,1: 74-96. https://doi.org/10.1177/0002716293530001006 Rahnu, Leen et al. 2015: Partnership dynamics among migrants and their descendants in Estonia. In: Demographic Research 32: 1519-1566. https://doi.org/10.4054/DemRes.2015.32.56 Soehl, Thomas; Yahirun, Jenjira 2011: Timing of union formation and partner choice in immigrant societies: The United States and Germany. In: Advances in Life Course Research 16,4: 205-216. https://doi.org/10.1016/j.alcr.2011.09.004 Statistisches Bundesamt 2020: Statistisches Jahrbuch 2019 [https://www.destatis.de/ DE/ Themen/Querschnitt /Jahrbuch/jb-bevoelkerung.pdf?_ _blob=publicationFile, 17.05.2020]. • Müşerref Erdoğan, Ayşe Abbasoğlu Özgören348 Tønnessen, Marianne; Wilson, Ben 2020: Visualising Immigrant Fertility – Profi les of Childbearing and their Implications for Migration Research. In: Journal of International Migration and Integration. https://doi.org/10.1007/s12134-020-00762-5 TURKSTAT 2021a: Press Release No. 37211 [https://data.tuik.gov.tr/Bulten/ Index?p=Evlenme-ve-Bosanma-Istatistikleri-2020-37211, 22.01.2022]. TURKSTAT 2021b: Press Release No. 37229 [https://data.tuik.gov.tr/Bulten/ Index?p=Dogum-Istatistikleri-2020-37229, 22.01.2022]. WVS (World Values Survey) 2022: WVS Wave 7: 2017-2020 [https://www. worldvaluessurvey.org/WVSOnline.jsp, 22.01.2022]. Zhang, Yuanting; Van Hook, Jennifer 2009: Marital Dissolution Among Interracial Couples. In: Journal of Marriage and the Family 71,1: 95-107. https://doi.org/10.1111/j.1741-3737.2008.00582.x Date of submission: 28.07.2021 Date of acceptance: 12.05.2022 Müşerref Erdoğan (). Groningen, Netherlands. E-mail: erdoganmuserref93@gmail.com Dr. Ayşe Abbasoğlu Özgören. Hacettepe University Institute of Population Studies. Ankara, Türkiye. E-mail: ayse.ozgoren@hacettepe.edu.tr URL: https://avesis.hacettepe.edu.tr/ayse.ozgoren Published by Federal Institute for Population Research (BiB) D-65180 Wiesbaden / Germany Managing Publisher Dr. Nikola Sander 2022 Editor Prof. Frans Willekens Managing Editor Dr. Ralina Panova Dr. Katrin Schiefer Editorial Assistant Beatriz Feiler-Fuchs Wiebke Hamann Layout Beatriz Feiler-Fuchs E-mail: cpos@bib.bund.de Scientifi c Advisory Board Karsten Hank (Cologne) Ridhi Kashyap (Oxford) Michaela Kreyenfeld (Berlin) Natalie Nitsche (Rostock) Zsolt Spéder (Budapest) Alyson van Raalte (Rostock) Rainer Wehrhahn (Kiel) Comparative Population Studies www.comparativepopulationstudies.de ISSN: 1869-8980 (Print) – 1869-8999 (Internet) Board of Reviewers Bruno Arpino (Barcelona) Kieron Barclay (Rostock) Laura Bernardi (Lausanne) Gabriele Doblhammer (Rostock) Anette Eva Fasang (Berlin) Michael Feldhaus (Oldenburg) Tomas Frejka (Sanibel) Alexia Fürnkranz-Prskawetz (Vienna) Birgit Glorius (Chemnitz) Fanny Janssen (Groningen) Frank Kalter (Mannheim) Stefanie Kley (Hamburg) Bernhard Köppen (Koblenz) Anne-Kristin Kuhnt (Duisburg) Hill Kulu (St Andrews) Nadja Milewski (Wiesbaden) Roland Rau (Rostock) Thorsten Schneider (Leipzig) Tomas Sobotka (Vienna) Jeroen J. A. Spijker (Barcelona) Heike Trappe (Rostock) Helga de Valk (The Hague) Sergi Vidal (Barcelona) Michael Wagner (Cologne)