Partner Choice in Sweden Following a Failed Intermarriage

Abstract

This paper is based on the assumption that divorced and separated individuals bring with them the experience of a failed union which may shape their future choices on the marriage market. It aims to contribute to our knowledge of intermarriage, and social interaction in Sweden in general, by comparing the repartnering choices of immigrants and natives in Sweden who had made what is still considered an atypical choice of entering a native-immigrant union with the partner choices of natives and immigrants whose previous union was endogamous. The empirical analysis in this paper is based on the Swedish register data from the STAR data collection (Sweden over Time: Activities and Relations) and covers the period 1990–2007. All the analyses in the paper include individuals aged 20–55 at the time of union dissolution. The multivariate analysis is based on discrete-time multinomial logistic regression. The results show that for all four groups defined by sex and nativity (native men, native women, immigrant men, and immigrant women), there is a positive association between the previous experience of intermarriage and the likelihood of initiating another intermarriage after union dissolution. Another important finding is that the magnitude of this positive association increases with the degree of social distance between the groups involved in the union. Gender differences are modest among natives and somewhat more pronounced among immigrants.

Electronic supplementary material

The online version of this article (doi:10.1007/s10680-016-9377-1) contains supplementary material, which is available to authorized users.

Introduction

Over the course of the past several decades, the continuous waves of migration to Western Europe from other parts of the world have transformed once relatively homogenous societies into meeting places of different ethnicities and cultures. An inevitable consequence of contemporary migration processes in Europe is an increased occurrence of marriages between natives and immigrants (Lucassen and Laarman 2009; Adserà and Ferrer 2014). Mixed nativity unions have been on rise also in Sweden since the 1970s, and this trend is currently being driven by unions between natives and non-Western immigrants (Haandrikman 2014). One could argue that it is no longer possible to obtain a complete picture of general marital trends in Western societies without looking at mixed nativity marriages. Therefore, it should come as no surprise that there is an increased interest in this type of intermarriage. Another possible reason for this interest may be the belief that intermarriage is an important indicator of social interaction in societies characterized by ethnic, religious, or racial diversity. This view may be traced back to the year 1925 and the early works of Emory S. Bogardus. His social distance scale views the willingness to form a marital union with a member of another social group as the strongest indicator of the absence of social distance between the two groups. Contemporary authors emphasize the intimate nature of marriage and argue that intermarriage may therefore be the strongest test of social boundaries as well as the maximum extent of social integration (Alba and Golden 1986; Choi et al. 2012). Moreover, drawing on important findings from social psychology (e.g., Pettigrew and Tropp 2008, on intergroup contact theory), not only intermarriage is an indicator of social integration, it is also considered an agent producing social integration. In particular, intermarriage leads to more frequent interaction between social groups and indicates, even to those not directly involved in the marriage, that intimate relationships between members of different social groups are possible and acceptable (Kalmijn 1998; Blau et al. 1984; Pagnini and Morgan 1990). Assuming that a marriage with a native person has a positive impact on country-specific skills and social capital, researchers have also recently shown a greater interest in the link between intermarriage and the economic integration of migrants (see review in Furtado and Trejo 2012). However, the enthusiasm regarding the integrative role of intermarriage is somewhat undermined by the empirical evidence of an elevated risk of divorce among those who marry exogamously. This has prompted some researchers to call for more research on intermarriage (Kalmijn 1998; Zhang and Van Hook 2009).

Previous research has helped social scientists develop a deeper understanding of the mechanisms behind the formation and dissolution of intermarriages. This study attempts to go one step further and examines what happens after the dissolution of an intermarriage. The paper departs from the assumption that the experience of a failed union affects future partner choices (Sweeney 1997; Gelissen 2004). It is also assumed that this experience will be all the more important given that, in the Swedish context, a decision to intermarry still constitutes an atypical and, in some cases, more difficult choice than entering an endogamous union. The main research question focuses on examining how the experience of a failed mixed nativity union is associated with subsequent partner choices among natives and immigrants in Sweden.

The general approach to this study has been influenced by the reality that Sweden today—just like most other Western societies—is more diverse than ever and that social boundaries between natives and immigrants vary by immigrant group. As will be seen in following sections, the research study is thus designed in such a way as to reconcile the clarity of presentation with a high degree of diversity in the contemporary Swedish society. For the sake of stylistic variety, the terms intermarriage, native-immigrant union, and mixed nativity union will all be used to denote a union between natives (Swedish-born individuals with two Swedish parents) and the foreign-born individuals living in Sweden.1 Unions between two native individuals as well as unions between two foreign-born individuals from the same immigrant group are referred to as endogamous unions throughout the paper.

The rest of the paper is organized as follows. The mechanisms behind the formation and dissolution of intermarriages are discussed in Sect. 2. The patterns of repartnering, as well as two competing hypotheses, are presented in Sect. 3. Section 4 describes the dataset and methodology. The results of descriptive and multivariate analyses of the relationship between the experience of intermarriage and the subsequent partner choice are presented in Sects. 5 and 6, respectively. Section 7 sets out to analyze to what degree the main conclusions of the paper are robust to the analytical challenges imposed by the diversity of living arrangements in Sweden. Concluding remarks are presented in Sect. 8.

Intermarriage: Patterns of Formation and Dissolution

Contemporary Western societies are characterized by numerous and intersecting social affiliations (Blau et al. 1984), implying that intermarriages may be defined with respect to different dimensions as well. However, it can be argued that, as long as there is at least some social distance between the social groups in question, the major forces behind the formation of intermarriage are similar regardless of whether we refer to intermarriage with respect to ethnicity, nativity, language, race, religion, or even social class. Kalmijn (1998) identifies three major forces of factors shaping the patterns of intermarriage: preferences, marriage markets, and third parties. The first factor refers to the preferences of marriage candidates with respect to norms, values, and lifestyle. In general, the aim is to marry a person who is similar with respect to these characteristics. The role of preferences for partner choice may be constrained by another important factor: the characteristics of marriage markets. In other words, the size of each social group, its geography as well as the structure of individuals’ social networks can substantially increase or reduce the likelihood of meeting a person from a different group. Finally, the probability of the formation of intermarriage may also be influenced by third parties, such as family and friends. Institutions, such as religious communities, can also act as third parties. The role of religious communities may be particularly important in countries where civil marriages are not recognized, or are recognized only under certain circumstances. In the literature on intermarriage, it is typically assumed that third parties, to a lesser or greater extent, are opposed to mixed marriages (Kalmijn 1998; Tolsma et al. 2008).

The complex ethno-religious mosaic of the USA has inspired social researchers to produce a sizeable and diverse literature on interracial and interreligious marriages in this country (e.g., Glenn 1982; Tucker and Mitchell-Kernan 1990; Kalmijn 1993; Qian 1997; Lehrer 1998; Fu 2001; Sherkat 2004). Mainly as a result of a substantial increase in immigration into the USA over the course of the recent decades, the literature on native-immigrant marriages now also constitutes an important part of the American research on intermarriage (among others, Qian and Lichter 2001; Chiswick and Houseworth 2011; Furtado and Theodoropoulos 2011; Choi et al. 2012). However, these studies frequently combine an analysis of mixed nativity marriage with an analysis of intermarriage with respect to ethnicity or other dimensions of social affiliation. As far as European research is concerned, the increased interest in intermarriage is of more recent date. Some European studies have examined marriage between communities coexisting in the same area for a long period of time (among others, O’Leary and Finnäs 2002 on linguistic intermarriage in Finland and interreligious marriages in Ireland; Kalmijn et al. 2006 on ethno-religious intermarriage in the Netherlands; Smits 2010 on ethnic intermarriage in the former Yugoslavia). However, the European research on intermarriage has mainly been fueled by the ongoing migration flows into Europe. As a consequence, the existing body of research on intermarriage in this part of the world is currently dominated by mixed nativity marriages (see reviews in Kulu and González-Ferrer 2014; Adserà and Ferrer 2014). This is also the case in Sweden, the country under study in this paper. Most European studies examine the characteristics of immigrants who intermarry, whereas studies with a primary focus on natives are scarce, although not entirely absent (e.g., Haandrikman 2014; Medrano et al. 2014).

Apart from a number of context-specific findings in the previous literature, it is possible to identify some contemporary patterns of exogamy among immigrants that can be considered almost universal. One of these frequent findings is that higher-educated immigrants are more likely to marry natives, whereas less educated immigrants are more likely to be in endogamous unions (Lievens 1998 for Belgium, Kalmijn and Van Tubergen 2006 and Van Tubergen and Maas 2007 for the Netherlands, González-Ferrer 2006 for Germany, Qian and Lichter 2007 for the USA, Sánchez-Domínguez et al. 2011 for Spain). The positive relationship between education and exogamy prompted an increased interest in the status exchange hypothesis. This perspective views some mixed nativity marriages as outcomes of a process of exchange of socioeconomic status of a higher-educated immigrant partner for the ethnic status of a lower-educated native partner (Kalmijn and Van Tubergen 2006; Choi et al. 2012). However, the empirical support for the status exchange hypothesis has been mixed. The role of macro-structural factors in general and marriage markets in particular leads to another common finding, namely that the association between group size and exogamy is negative among immigrants and other minority groups (Blau et al. 1982; Alba and Golden 1986; Van Tubergen and Maas 2007). Previous research has also shown that immigrant groups in Western countries differ by their propensity to intermarry, even after controlling for socioeconomic characteristics. This is usually ascribed to the cultural factors, whereby immigrants from countries that are culturally closer to the destination are more likely to intermarry with natives (Van Tubergen and Maas 2007; Hwang et al. 1997). In line with the assimilation perspective (Gordon 1964), the length of exposure to the majority culture also shapes the patterns of intermarriage. Previous studies suggest that immigrants who have stayed longer and those who arrived at a younger age are more likely to intermarry (Muttarak and Heath 2010; Chiswick and Houseworth 2011). Finally, the nature of immigrant–native marriage is also gendered to some degree. Lanzieri (2012) reports that in most European countries native man–immigrant woman unions are more common than immigrant man–native woman unions. It should also be noted that all common patterns of native–immigrant marriages discussed here have also been identified in two Swedish studies by Dribe and Lundh (2008, 2011).

For a better understanding of intermarriages, apart from the mechanisms of their formation, it is also necessary to look at the patterns of their dissolution. From the point of view of homogamy theory (Zhang and Van Hook 2009), the formation of intermarriage in many cases implies overcoming differences in norms, values, and lifestyles, as well as overcoming possible resistance from third parties. However, while these differences may not have constituted an obstacle large enough to prevent the formation of intermarriage, they may nevertheless constitute a difficulty for its subsequent success. Some of the principal sources of conflicts in exogamous unions include differences in childrearing practices, gender role expectations, time orientation, communication style, and family connections (Bustamante et al. 2011). Moreover, and in line with the view that third parties often oppose exogamous choices, it is expected that intermarriages will receive less social support than endogamous unions. This may be particularly problematic for troubled unions and in times of need (Kulu and González-Ferrer 2014). Thus, homogamy theory predicts that, as a result of these differences, mixed marriages are more likely to break up than endogamous marriages. Another prominent theory on the stability of intermarriages (and not mutually exclusive with homogamy theory) is convergence theory (Jones 1996). This theory is based on the evidence that divorce risks vary substantially across countries, as well as on the assumption that partners of different ethnicities inherit the criteria for divorce from their respective ethnic backgrounds. Consequently, the divorce rates of intermarriages should be in between the divorce rates of the ethnicities involved in the relationship.

The empirical evidence tends to lend more support to the view shaped by homogamy theory since most studies, after controlling for socioeconomic characteristics, do indeed identify an elevated risk of divorce in interracial marriages (Bratter and King 2008), in interreligious marriages (Kalmijn et al. 2005), or in marriages between native and immigrants (see Dribe and Lundh 2012 for Sweden; Kalmijn et al. 2005 and Smith et al. 2012 for the Netherlands; Milewski and Kulu 2014 for Germany). Studies analyzing native–immigrant marriages also reveal that the risk of divorce increases if the spouses originate from countries culturally distinct from each other. The findings regarding the increased risk of divorce in exogamous marriages are corroborated by Hohmann-Marriott and Amato (2008). They look at marital conflict rather than at divorce specifically and find that partners in exogamous unions generally reported lower levels of relationship quality than did those in endogamous unions. The support for homogamy theory is dominant, but not universal: in their UK study, Feng et al. (2012) find more support for convergence theory.

Intermarriage and Repartnering

The substantial growth in the general literature on remarriage has mainly been fueled by a sharp increase in divorce rates as well as by an increase in the complexity of family arrangements (Andersson 1998; Wu and Schimmele 2005). In contrast, there is an impression that research focused specifically on assortative mating on the remarriage market remains somewhat scarce, especially when it comes to assortative mating with respect to nativity or ethnicity. Needs, attractiveness, and marriage market are commonly defined as the principal factors determining the decision to enter both the first and subsequent unions (De Graaf and Kalmijn 2003). However, the relative importance of each of these factors may well differ by marriage order. For example, many scholars believe that the desire to have children is on average more important for entering the first union simply because most divorced persons already have children (Dewilde and Uunk 2008). A similar reasoning can be applied when looking at the decision with whom to start a union. Preferences, third parties, and marriage market affect partnering choices irrespective of marriage order. What often does change, however, is the relative salience of the three main forces affecting partnering choice. In particular, the previous research suggests that, at least in some contexts, third parties are less important, whereas marriage markets are a more prominent factor the second time around (Jacobs and Furstenberg 1986; Booth and Edwards 1992; Aguirre et al. 1995; De Graaf and Kalmijn 2003). This difference in the relative importance of factors is usually seen as the driving force behind the predominant evidence that second- and higher-order unions are in general less homogamous than first unions (Dean and Gurak 1978; Aguirre et al. 1995; Gelissen 2004; Fu 2010; Muttarak and Heath 2010).

This paper differs from previous related research in two aspects. First, while the aforementioned scholars sought to explore the link between remarriage and heterogamy with respect to characteristics such as education or race, this is one of the first studies whose main focus is on unions between natives and immigrants. Second, most of the previous literature has focused on comparing the average levels of marital heterogamy in the first- and higher-order unions. This paper has a stronger focus on individual marital trajectories and is one of the first studies seeking to establish the relationship between previous and subsequent partner choices. Hence, the primary independent variable of interest is the partner’s origin in the previous union. Two competing hypothesis can be introduced at this stage. The proneness hypothesis predicts that personal preferences, marriage markets, social networks, and third parties will act in such a way that the individuals who experienced a breakup of an exogamous union are more likely to enter another exogamous union the next time around, compared to the divorced and separated whose previous union was endogamous. So, one should expect to find a positive association between the exogamy in the previous and in the subsequent union since some natives and immigrants are more prone to outmarry than other members of their groups, whether for the first time or when starting a higher-order union. An alternative view is that the failure of the previous union affects the preferences regarding partner characteristics in the subsequent union. This perspective emphasizes the process of learning through divorce (Whyte 1990; Gelissen 2004) and the possibility that the divorced individuals “apply their new knowledge in a second marriage—a marriage which is more homogamous and in which greater value-sharing leads to less conflict” (Dean and Gurak 1978). The learning hypothesis considers these factors and assumes that a substantial share of those with an experience of a failed exogamous union will view their previous choice of partner as a mistake. Moreover, in the context of this research, this hypothesis predicts that these individuals will have an above-average preference for endogamous unions the next time around. This should result in a negative association between the previous experience of exogamy and the likelihood of entering another exogamous union the next time around.

Data and Methodology

The empirical analysis in this paper is based on Swedish register data from the STAR data collection (Sweden over Time: Activities and Relations) and covers the period from 1990 to 2007. This compilation of datasets makes it possible to obtain detailed longitudinal information on civil status, household composition, partner information, and a number of other socio-demographic characteristics that may help disentangle partnering choices after a union disruption. The affiliation to immigrant groups is based on the country of birth, as there is no information available on ethnicity, religion, or mother tongue. In the Swedish register data, it is possible to identify married couples and cohabitants with children in common. On the other hand, cohabitants with no common children cannot be identified and are displayed as two separate entities in the registers. This is a somewhat unfortunate feature of the data. However, since childbearing and marriage are often considered the principal indicators of commitment to the union (Vikat et al. 1999; Lyngstad and Jalovaara 2010), this paper should be understood as an analysis of transitions from one committed union to another.

In this paper, cohabitants with at least one common child are treated the same way as married couples. In other words, cohabitation dissolution is treated as equal to divorce, while cohabitation formation with a common child after a union dissolution is equivalent to remarriage. The event that constitutes the start of a new union is either the act of marriage or the birth of a child, whichever comes first. In a section below, there is a discussion on whether classifying marriages and non-marital unions with common children as being equal affects the general conclusions of the paper. An individual enters the study if he or she experiences a union dissolution in 1990 or after and is between 20 and 55 years of age at the time of the union dissolution. Individuals are followed from the union dissolution until repartnering, or until they are censored because of emigration, death, or at the end of the study period in 2007. Since the research question is motivated by failure of the previous union, individuals at risk of entering a new union due to the death of a partner are not included into the analysis. This also applies to individuals who repartner with the same person they previously divorced or separated from. All descriptive and multivariate analyses are done separately for four groups defined by sex and nativity: native men, native women, immigrant men, and immigrant women. Natives are defined as Swedish-born individuals with two Swedish-born parents, while the foreign-born (first-generation immigrants) are referred to simply as immigrants. The definition of natives in this study implies that an overwhelming majority of native–immigrant unions are also interethnic unions, which one should bear in mind when interpreting the results of the analysis. The number of individuals included in the study is as follows: 265,364 native men, 278,641 native women, 66,977 immigrant men, and 70,584 immigrant women.

Looking at the cultural dimensions developed by Inglehart (1997), Sweden scores very high on secular-rational values (opposite to traditional values), just as it scores very high on self-expression values (opposite to survival values). Dribe and Lundh (2012) show that the most prevalent among native–immigrant unions in Sweden are those in which the immigrant partner stems from a value context similar to that in Sweden. Their results also show that these marriages are more stable than other mixed nativity unions in Sweden. It is therefore necessary to account for “degrees of otherness” to some extent and introduce a classification that goes beyond immigrant–native or monoethnic–interethnic dichotomies. Yet there is no particular reason why we should believe that the structure of social boundaries from the perspective of a native person perfectly corresponds to that from the perspective of a foreign-born person. This is why the sets of competing events for natives and immigrants on the repartnering market are not identical in this paper. The model assumes that following union dissolution each native person on the repartnering market is at risk of experiencing one of four competing events (i.e., entering one of four types of union)2:

• new union with a native person

• new union with an immigrant from a Western country (persons born in the EU-15, the USA, Canada, Australia, or New Zealand)

• new union with a non-Western immigrant

• new union with a second-generation immigrant

It should be noted that adult second-generation immigrants in Sweden are fairly similar to natives. First, many of them have one Swedish-born parent. Second, a large majority of those who do not were born to parents from earlier immigrant cohorts, who mainly originate in Nordic and other West European countries.

Turning to partner choices among the foreign-born, the most similar group for an immigrant consists of his or her co-ethnics, whereas the most numerous group on the marriage market is made up of natives. However, in order to obtain a more fine-grained picture of remarriage trends among the foreign-born, it would also be interesting to explore which immigrants form unions with non-co-ethnic immigrants. While this type of union should undoubtedly be considered an interethnic marriage, it may have some distinct features compared to a marriage between a native and an immigrant, at least in the context in which the native community constitutes a clear majority in the Swedish society. To illustrate, in such a context two individuals from different immigrant communities may be more similar to each other in terms of language skills, country-specific knowledge, or social capital than they are to a typical native person. This reasoning should especially apply to non-Western immigrants, who constitute a majority of the total foreign-born population in Sweden, as well as among the foreign-born at risk of repartnering. Thus, the competing risks for foreign-born persons are:

• new union with a native-born person

• new union with a co-ethnic immigrant (also simply referred to as co-ethnic)3

• new union with other immigrant (also referred to as non-co-ethnic immigrant)

• new union with a second-generation immigrant

The multivariate analysis is based on event history models of competing risks or, more precisely, discrete-time multinomial logistic regression. The dependent variable takes the value 1, 2, 3, or 4 if the individual at risk of repartnering starts one of four types of union during year t; otherwise, it is equal to 0. The explanatory variable of major interest is the origin of the previous partner in the union, with the classifications by origin corresponding to those used when identifying competing risks. A number of other potentially relevant time-varying and time-invariant variables are used in the model. The value of time-varying variables refers to December 31 of the year t − 1. The individual time-varying variables used in the model are:

• Years since union dissolution (and its squared term)

• Categorical variable children, with three categories: (1) childless, (2) has children and all children live apart, and (3) has children and lives with at least one of them

• Education level, with four categories: (1) less than secondary education, (2) secondary education, (3) more than secondary education, and (4) unknown (only for the foreign-born)

• Disposable household yearly income, adjusted by household size using the square root scale and expressed in quintiles

• Categorical variable activity, constructed on the basis of the primary source of income of the individual, with five categories: (1) employed, (2) unemployed, (3) welfare recipient, (4) in education, and (5) a heterogeneous residual category involving all other individuals

The analysis also incorporates a number of contextual time-varying indicators. One of these is the size of municipality of residence and is categorized as: (1) less than 10,000 inhabitants, (2) 10,000–50,000 inhabitants, (3) 50,001–100,000 inhabitants, (4) 100,001–200,000 inhabitants, and (5) more than 200,000 inhabitants. Three sets of municipality-level marriage market indicators are introduced into the model in order to control for the opportunities on the marriage market. All marriage market indicators refer to the population comprised of the individuals of the opposite sex who are identified as unpartnered in the registers and are 18–65 years of age. The first set of indicators describes the presence of other social groups in the municipality of residence, whereby the definition of social groups corresponds to that used to denote the type of previous union. So, when analyzing natives, the model includes the following variables: (1) percentage of Western immigrants, (2) percentage of non-Western immigrants, and (3) percentage of second-generation immigrants. When looking at immigrants, the three variables are the following: (1) percentage of natives, (2) percentage of non-co-ethnic immigrants, and (3) percentage of second-generation immigrants. All regressions also include the interaction terms between the type of previous union and municipality-level share of the group to which the former partner belongs.4 The educational structure of the municipality is captured through shares of individuals with less than secondary education, secondary education, and more than secondary education, respectively. The age structure of the marriage market is controlled for by shares of individuals in the municipality who belong to each of these age groups: (1) 18–29, (2) 30–39, (3) 40–49, and (4) 50–65.5 All marriage market indicators are centered around the mean for each of four groups defined by sex and nativity. A separate categorical variable controls for possible period effects, with periods being defined as follows: (1) before 1996, (2) 1996–2001, and (3) after 2001. The period effects are included in order to capture the effects of the general socioeconomic climate on the process of family formation. Besides, they control for possible changes in general attitudes toward intermarriage among both natives and immigrants.

The model also controls for some additional characteristics of the previous union and takes into account the previous findings that the formation of an intermarriage may be a by-product of assortative mating with respect to characteristics other than origin. For instance, Kolk (2012) shows that exogamous and native endogamous marriages also differ with respect to age heterogamy. Therefore, the following time-invariant variables are also introduced into the model:

• Age at union dissolution, and its squared term

• Dummy variable, which takes value 1 if the previous union was marriage

• Dummy variable, which takes value 1 if the previous union was a second- or higher-order union (or, put another way, if the individual is at the second- or higher-order repartnering risk)6

• Categorical variable age heterogamy in the previous marriage, with five categories: (1) older by more than 3 years, (2) older by up to 3 years, (3) same age, (4) younger by up to 3 years, and (5) younger by more than 3 years

• Categorical variable educational heterogamy in the previous marriage, with four categories: (1) less educated than partner, (2) same level of education as partner, (3) more educated than partner, and (4) unknown

The immigrant models also control for years since migration at union dissolution, with the following categories: (1) less than 5 years, (2) 6–10 years, (3) 11–15 years, and (4) more than 15 years. Finally, the geographical origin of the foreign-born is controlled for through a number of dummy variables based on a somewhat broader geographical definition than when defining immigrant groups.7 Included are the dummies for the following countries or regions of birth: (1) Nordic countries, (2) other Western countries, (3) former Yugoslavia, (4) Poland, (5) the rest of Europe, (6) Iran, (7) Turkey, (8) Thailand and other Southeast Asian countries, (9) Latin America, and (10) residual heterogeneous group comprising all other immigrants.

Descriptive Findings

The first step in the analysis of repartnering is to classify the formerly married and the formerly cohabiting natives and immigrants by their previous choice of partner, and then observe with whom those who repartner by the end of 2007 do so. Figure 1 shows repartnering choices of native men and women. Regardless of the origin of the previous partner, native men and women who repartner most frequently do so with a native partner. As far as gender differences are concerned, the tendency to choose a native partner the next time around is somewhat more pronounced among women than men. Previous experiences do seem to play a certain role, as the share of those who repartner with a native is the highest among those whose previous union was endogamous (78 % of men and 84 % of women), while it is clearly the lowest among natives who previously lived with non-Western immigrants (46 % of men and 63 % of women). When the natives who were previously with Western immigrants are compared with those who were previously with other immigrants, the share of those who form a new union with a native is clearly higher among the former. This pattern may be due to marriage market characteristics, but it may also indicate that preferences for a partner from a more distant culture, once established, are more persistent. Repartnering choices among natives who were previously partnered with natives differ very little from those of natives whose previous partner was a second-generation immigrant.

Fig. 1.

Repartnering choices among natives. Source: Swedish register data

Repartnering choices are more heterogeneous among immigrants than natives, both across types of previous unions and between sexes. Considering that natives clearly outnumber immigrants on the marriage market, a higher degree of heterogeneity among immigrants should come as no surprise. Concerning repartnering choices of immigrants who previously lived with natives, Fig. 2 reveals some remarkable gender patterns: whereas 70 % of women from this group who repartner will make the same choice the next time around, a majority of men (around 59 %) will form a new union with a foreign-born or second-generation immigrant woman. Most immigrant men and women who repartner after a failed union with a co-ethnic will also make an endogamous choice the next time around (70 % of men and 58 % of women). Interestingly, among men whose former partner is a co-ethnic, starting the subsequent union with a woman from another immigrant group is more prevalent than forming a new union with a native woman. This is a noteworthy result given that immigrants from other origins, just like natives, are not co-ethnics, while they are a significantly smaller group than natives on the marriage market. The most heterogeneous pattern of repartnering choices has been found among foreign-born who were previously with a non-co-ethnic immigrant. This is particularly the case among women, for whom repartnering with natives, co-ethnics, and other immigrants are almost equally common choices the next time around (with respective shares being 34, 28, and 32 %).8

Fig. 2.

Repartnering choices among immigrants. Source: Swedish register data

The patterns of repartnering can be observed from a different angle by applying the cumulative incidence approach proposed by Coviello and Boggess (2004). Figure 3 shows the cumulative incidence of entering different types of union by type of previous union among natives, whereby types of new union are represented as competing risks.9 For native men, the risk of starting any of the four types of union is the highest for those who were previously in the same type of union. This is also the case for women, except when looking at the risk of starting a new union with a native man, which is slightly higher for women formerly partnered with a second-generation immigrant. There seems to be a sizeable difference with respect to the magnitudes of relative risks of entering the four types of union. For instance, previous experiences imply only a relatively moderate degree of difference in terms of risk of entering a union with a native person. On the other hand, the cumulative incidence of a new union with a non-Western immigrant 10 years after union dissolution is around five (among men) to seven (among women) times higher for natives who already had that experience, as compared to natives who were previously in an endogamous union.

Fig. 3.

Cumulative incidence of forming different types of union by type of previous union, natives. Note n.u. new union. Source: Swedish register data

Some similar patterns emerge among the foreign-born. Most importantly, as Fig. 4 demonstrates, the risk of any of the four types of union is the highest for immigrant men and women whose previous union was of the same type. The risk of starting an endogamous union is the lowest for immigrants previously partnered with natives, while those previously partnered with a co-ethnic are the least likely to repartner with a native. Gender differences are modestly pronounced, except when looking at relative risks of repartnering with a co-ethnic. In particular, the previous experience of endogamous union for entering endogamous union the next time around is more important for women. The next section reveals to what degree the results presented in Figs. 3 and 4 are sensitive to the introduction of additional covariates.

Fig. 4.

Cumulative incidence of forming different types of union by type of previous union, immigrants. Note n.u. new union. Source: Swedish register data

Multivariate Analysis Results

The main results of the multivariate analysis for natives are shown in Table 1. As explained in Sect. 4, the model also contains an interaction term between the type of previous union and mean-centered characteristics of local marriage markets. Therefore, the coefficients in Table 1 show the impact of the previous partner’s origin on the type of subsequent union for natives with an average (or, typical) exposure to other groups. Controlling for observable individual and contextual characteristics does not substantially affect the conclusions drawn from the previous section. One modification relative to the results of the cumulative incidence analysis is that now there are no exceptions to the pattern that the risk of entering any of the four types of union is the highest for those who were previously in the same type of union (i.e., it is now also the case among native women when it comes to the risk of repartnering with a native man).

Table 1.

Association between partner’s origin in the previous and the subsequent union, natives (discrete-time multinomial logit; relative risk ratios; base outcome: without a partner or cohabiting with no common children)

	Previous union with (ref.: native)	New union with
		Native	Western immigrant	Non-Western immigrant	Second-generation immigrant
Native men	Native	1.00	1.00	1.00	1.00
	Western immigrant	0.89***	1.61***	1.30***	0.97
	Non-Western immigrant	0.54***	0.92	4.04***	0.64***
	Second-generation immigrant	0.89***	0.99	1.07	1.04
Native women	Native	1.00	1.00	1.00	1.00
	Western immigrant	0.90***	1.40***	1.78***	1.00
	Non-Western immigrant	0.63***	0.90	5.02***	0.74***
	Second-generation immigrant	0.90***	1.04	0.98	1.06

Source: Swedish register data. Control variables shown in Table 4 in Appendix

* p < 0.10; ** p < 0.05; *** p < 0.01

So, in other words, the analysis lends support to the proneness hypothesis. However, the magnitude of the coefficients varies considerably. If it is assumed that the social distance is the smallest between natives and second-generation immigrants and the greatest between natives and non-Western immigrants, then the importance of the type of the previous union on the formation of the same type of union when repartnering increases with the social distance between the groups involved in the union. To illustrate, when comparing two native individuals with an average exposure to other social groups, the risk of repartnering with a non-Western immigrant is about four to five times higher for a native whose previous union was with another non-Western immigrant than for a native who was previously in an endogamous union. The analogous associations are much lower when it comes to the risk of repartnering with a Western immigrant (the corresponding relative risk ratios are 1.61 for men and 1.40 for women) or a second-generation immigrant (1.04 and 1.06 for men and women, respectively, both not statistically significant). On a similar note, the stronger the social boundary that was crossed when the previous union was formed, the lower the likelihood of entering an endogamous union the next time around; the risk of starting a new union with a native is smallest among natives who were previously partnered with a non-Western immigrant. Another general pattern emerging from this analysis is that coefficients for native men and women are fairly similar. The type of previous union has a similar impact for repartnering with natives as it does for repartnering with second-generation immigrants. This is in line with the findings in the descriptive analysis.

Although the main concern of this study is the impact of previous experience of intermarriage on subsequent partner choice, it may be interesting to briefly discuss how some other socio-demographic individual and contextual characteristics operate on the repartnering market. These results are shown in Table 4 in Appendix. The role of structural factors for native men and women manifests itself in the expected way as the size of each group increases the likelihood of repartnering with a person from that group. In general, interactions of group size and the type of previous union indicate that, all else equal, the likelihood of entering the same type of intermarriage the next time around increases somewhat in areas with a lower-than-average presence of the former partner’s group, and vice versa. This suggests that the previous experience of intermarriage is more important in contexts where the decision to intermarry is less typical. Yet, considering the size of the effect and the actual shares of Western and non-Western immigrants in Swedish municipalities, the inclusion of interaction terms does not substantially affect the conclusions emerging from Table 1. The risk of repartnering decreases with time since union dissolution and age, although age matters less for men when it comes to the risk of repartnering with a Western or non-Western woman. Natives living with one or more children are less likely to start a new union compared to other separated or divorced natives. Highly educated and employed natives, as well as those in the highest income quantile, are also most likely to repartner. Women entering a new union with a non-Western immigrant constitute a noteworthy exception here. For women, income and education do not affect the likelihood of repartnering with a non-Westerner, while being employed actually decreases the risk of entering this type of union compared to women identified as unemployed, inactive, or enrolled in education. Natives whose previous union was a marriage as well as those who experienced more than one union dissolution are more likely to repartner.

Table 4.

Discrete-time multinomial logit model of repartnering, natives (base outcome: without a partner or cohabiting with no common children)

	Native men—new union with				Native women—new union with
	Native	West. imm.	Non-West imm.	Second-gen. imm.	Native	West. imm.	Non-West imm.	Second-gen. imm.
Age at union dissolution	0.90***	0.99	1.00	0.92***	0.88***	0.92***	0.92***	0.92***
Age at union dissolution squared	1.00***	1.00	1.00*	1.00	1.00***	1.00	1.00	1.00***
Education level (ref.: primary or less)
Secondary	1.08***	1.05	1.05	1.09***	1.10***	1.01	0.83***	1.11***
More than secondary	1.32***	1.43***	1.28***	1.27***	1.28***	1.29***	0.93	1.27***
Household income (ref.: first quintile)
Second quintile	1.05***	0.93	0.77***	0.93**	1.00	0.86***	0.93	1.02
Third quintile	1.17***	1.02	0.85***	1.03	0.93***	0.89*	0.97	0.93*
Fourth quintile	1.38***	1.36***	1.01	1.25***	0.97*	0.95	0.87*	0.99
Fifth quintile	1.87***	1.80***	1.29***	1.61***	1.23***	1.20**	0.93	1.31***
Main source of income (ref.: employed)
Unemployed	0.68***	0.99	0.90	0.78***	0.84***	1.15*	1.26***	0.97
Student	0.46***	0.56***	0.63***	0.62***	0.46***	0.81*	1.18*	0.58***
Welfare recipient	0.75***	0.79	0.79*	0.60***	0.63***	0.79**	0.98	0.70***
Other	0.66***	0.89*	1.03	0.71***	0.78***	1.14**	1.27***	0.84***
Children (ref.: childless)
Lives apart from children	0.89***	0.85***	0.75***	0.96	1.19***	1.06	1.06	1.12**
Lives with children	0.76***	0.69***	0.69***	0.86***	0.80***	0.69***	0.77***	0.83***
Years since union dissolution	0.99***	0.94***	0.92***	0.99	1.00	0.97*	0.87***	1.01
Years since union dissolution squared	1.00***	1.00	1.00	0.99***	0.99***	1.00***	1.00	0.99***
Previous partner (ref.: native)
Western immigrant	0.89***	1.61***	1.30***	0.97	0.90***	1.40***	1.78***	1.00
Non-Western immigrant	0.54***	0.92	4.04***	0.64***	0.63***	0.90	5.02***	0.74***
Second-generation immigrant	0.89***	0.99	1.07	1.04	0.90***	1.04	0.98	1.06
Western immigrants in municipality (%)	1.01***	1.06***	1.04***	1.02***	1.03***	1.08***	1.03**	1.03***
Non-West. immigrants in municipality (%)	0.98***	0.99	1.02***	0.98***	0.97***	0.98***	1.02***	0.98***
Second-gen. immigrants in municipality (%)	0.95***	1.02	0.95***	1.07***	0.97***	1.01	0.98	1.08***
Interaction Western immigrants	0.99***	0.97***	1.00	0.98*	0.98**	0.98	1.01	0.98
Interaction non-Western immigrants	0.98***	0.98	0.98**	0.99	0.98***	0.97*	0.98**	0.98**
Interaction Second-gen. immigrants	1.00	0.97*	1.00	0.98**	1.00	0.95***	1.00	0.98**
Previous union marriage	1.48***	1.47***	1.43***	1.49***	1.56***	1.62***	1.50***	1.57***
Second- or higher-order risk	1.41***	1.46***	1.55***	1.36***	1.45***	1.87***	1.75***	1.61***

N (men): 265,364; person-years (men): 1,950,076; N (women): 278,641; person-years (women): 2,103,752; * p < 0.10; ** p < 0.05; *** p < 0.01. The model also controls for period effects, educational and age heterogamy in the previous union as well as for the size, educational and age structure of marriage markets. These coefficients and standard errors are not reported for the sake of space. Source: Swedish register data

Multivariate results for immigrants, shown in Table 2, also support the proneness hypothesis. Just as is the case among natives, the risk of entering any of the four types of union is the highest for those who were previously in the same type of union. As compared to immigrants with an experience of a failed endogamous union, immigrants who were previously in a native–immigrant union are slightly more than twice as likely to choose a native partner the next time around. The coefficients for immigrant men and women do not differ markedly, except when it comes to the risk of starting a new union with a co-ethnic. In particular, whereas the type of previous union matters only to a certain degree among foreign-born men, the previous union with a native implies a substantially lower risk of repartnering with a co-ethnic for women. Given the heterogeneity of the immigrant population in Sweden, the link between social distance, previous union, and subsequent union can best be observed if additional analyses are performed including non-Western immigrants only. The coefficients in the lower half of Table 2 demonstrate another important similarity with the results reported for natives: the likelihood of entering the same type of exogamous union the next time around increases with the social distance between the groups involved in that type of union. Non-Western men and women with an experience of a failed native–immigrant union and with an average exposure to other social groups are about two and half times more likely to repartner with another native as compared to non-Westerners whose previous union was endogamous. The corresponding associations are lower when it comes to repartnering with a non-co-ethnic immigrant.

Table 2.

Association between partner’s origin in the previous and the subsequent union, immigrants (discrete-time multinomial logit; relative risk ratios; base outcome: without a partner or cohabiting with no common children)

	Previous union with (ref.: co-ethnic)	New union with
		Native	Co-ethnic immigrant	Non-co-ethnic immigrant	Second-generation immigrant
Immigrant men	Co-ethnic immigrant	1.00	1.00	1.00	1.00
	Native	2.24***	0.83***	0.83**	1.89***
	Non-co-ethnic immigrant	1.22***	0.89***	1.95***	1.32**
	Second-generation immigrant	1.88***	0.83***	0.92	2.03***
Immigrant women	Co-ethnic immigrant	1.00	1.00	1.00	1.00
	Native	2.22***	0.35***	0.85**	1.53***
	Non-co-ethnic immigrant	1.29***	0.64***	1.90***	1.18
	Second-generation immigrant	1.93***	0.47***	0.87	1.75***
Immigrant men, non-Western	Co-ethnic immigrant	1.00	1.00	1.00	1.00
	Native	2.65***	0.89***	0.86**	2.42***
	Non-co-ethnic immigrant	1.42***	0.93**	1.82***	1.47**
	Second-generation immigrant	2.36***	0.86***	0.93	2.83***
Immigrant women, non-Western	Co-ethnic immigrant	1.00	1.00	1.00	1.00
	Native	2.51***	0.36***	0.85**	1.80***
	Non-co-ethnic immigrant	1.46***	0.72***	1.71***	1.48***
	Second-generation immigrant	2.28***	0.47***	0.89	2.16***

Source: Swedish register data. Control variables shown in Table 5 in Appendix

* p < 0.10; ** p < 0.05; *** p < 0.01

Table 5 in Appendix shows more detailed results for immigrants. The effect of group size on the local marriage market works in the expected direction: the larger the exposure to each group, the higher the likelihood of repartnering with a member of that group. There are also certain similarities with results for natives when it comes to the effect of interactions. The likelihood of repeating the same choice following the failure of union with a native is somewhat lower in the areas with an above-average presence of natives. The signs of interaction terms for other types of union indicate a similar effect, but these coefficients are generally not significant. Younger immigrant women are more likely to repartner, irrespective of the type of the subsequent union. On the other hand, younger men are more likely to repartner with a native or second-generation immigrant, whereas age is less important when it comes to repartnering with immigrants, either from own or from other groups. The risk of repartnering is the lowest among immigrant men and women who live with one or more children. A more favorable socioeconomic status, as measured by income, labor market status, and education, increases the risk of repartnering with a native and second-generation immigrant among both sexes, whereas it decreases the risk of endogamous repartnering among women. In line with the assimilation perspective, having lived in Sweden longer makes it more likely to repartner with a native or second-generation immigrant, and less likely to repartner with a co-ethnic or other immigrant (the effect is stronger for endogamous unions). In general, the effects of group size, socioeconomic status, and duration of stay at destination largely resemble common patterns identified in previous studies on the formation of first mixed unions. As is the case among natives, immigrants are more likely to repartner if their previous union was a marriage or if they experienced two or more union dissolutions.10

Table 5.

Discrete-time multinomial logit model of repartnering, immigrants (base outcome: without a partner or cohabiting with no common children)

	Immigrant men—new union with				Immigrant women—new union with
	Native	Co-ethnic imm.	Non-co-ethn. imm.	Second-gen. imm.	Native	Co-ethnic imm.	Non-co-ethn. imm.	Second-gen. imm.
Age at union dissolution	0.87***	1.01	0.99	0.92***	0.94***	0.95***	0.94***	0.97
Age at union dissolution squared	1.00***	1.00	1.00	1.00	1.00***	1.00	1.00	1.00**
Education level (ref.: primary or less)
Secondary	1.21***	0.98	0.97	1.05	1.19***	0.90***	0.94	1.34***
More than secondary	1.65***	0.99	1.05	1.22*	1.53***	0.74***	1.07	1.62***
Unknown	1.37***	0.94	0.84	0.82	0.78***	0.65***	0.67***	0.73
Household income (ref.: first quintile)
Second quintile	1.12**	1.14***	0.93	1.10	1.04	0.91**	1.06	0.94
Third quintile	1.20***	1.22***	1.08	1.12	0.97	0.82***	0.91	0.89
Fourth quintile	1.41***	1.20***	1.13*	1.20	1.03	0.71***	0.83**	0.88
Fifth quintile	2.03***	1.12*	1.41***	1.59***	1.12*	0.72***	0.84	0.89
Main source of income (ref.: employed)
Unemployed	0.74***	0.97	0.92	0.93	0.75***	1.06	1.02	0.82
Student	0.48***	0.76***	0.84***	0.66***	0.29***	0.77***	0.85**	0.38***
Welfare recipient	0.73***	0.68***	0.75**	0.66***	0.67***	0.71***	0.89	0.68***
Other	0.61***	0.61***	0.72***	0.64***	0.65***	0.71***	0.86***	0.71***
Children (ref.: childless)
Lives apart from children	0.94	0.74***	0.86***	0.83**	1.08*	0.85***	0.91	1.02
Lives with children	0.79***	0.88***	0.79***	0.73**	0.76***	0.71***	0.76***	0.78***
YSM at union dissolution (ref.: up to 5)
6–10 years	0.99	0.81***	1.07	0.90	1.01	0.72***	1.00	1.14
11–15 years	1.03	0.67***	0.91	1.12	1.04	0.53***	0.90	1.17
More than 15 years	1.24***	0.42***	0.81***	1.36***	1.26***	0.39***	0.70***	1.39***
Years since union dissolution	1.11***	1.04***	1.08***	1.07**	1.05***	1.01	1.01	1.06*
Years since union dissolution squared	0.99***	0.99***	0.99***	0.99***	0.99***	0.99***	0.99***	0.99***
Previous partner (ref.: co-ethnic immigrant)
Native	2.24***	0.83***	0.83***	1.89***	2.22***	0.35***	0.85***	1.53***
Non-co-ethnic immigrant	1.22***	0.89***	1.95***	1.32**	1.29***	0.64***	1.90***	1.18
Second-generation immigrant	1.88***	0.83***	0.92	2.03***	1.93***	0.47***	0.87	1.75***
Natives in municipality (%)	1.09***	0.97***	1.07***	1.04**	1.04***	0.96***	1.02	1.03*
Other immigrants in municipality (%)	1.04***	0.96***	1.07***	1.02	1.00	0.96***	1.03***	1.00
Second-gen. immigrants in municipality (%)	1.05**	0.94***	1.03	1.12***	1.02	0.94***	1.00	1.13***
Interaction natives	0.98***	1.00	0.98***	1.00	0.98***	1.01*	0.98**	0.99
Interaction non-co-ethnic immigrants	1.00	1.00	0.99	0.97*	0.99	0.97***	0.99	1.02
Interaction Second-gen. immigrants	1.06***	1.00	1.05	0.97	1.03*	1.09***	1.05	1.00
Previous union marriage	1.32***	1.44***	1.50***	1.27***	1.23***	1.37***	1.30***	1.38***
Second- or higher-order risk	1.32***	1.25***	1.35***	1.55***	1.23***	1.52***	1.49***	1.04

N (men): 66,977; person-years (men): 417,668; N (women): 70,584; person-years (women): 481,301; * p < 0.10; ** p < 0.05; *** p < 0.01. The model also controls for period effects, geographical origin, educational and age heterogamy in the previous union as well as for the size, educational and age structure of marriage markets. These coefficients and standard errors are not reported for the sake of space. Source: Swedish register data

Robustness Analysis

The analysis in this paper has revolved around the idea that a couple needs to be either married or have at least one common child in order to be considered as living in a committed union. Nevertheless, this approach also implies that, among other things, childless marriages and childbearing cohabitations have been treated in the same way in the analysis, even though the nature of these unions may be different. So, it cannot be ruled out that the approach adopted in this paper masks a certain degree of heterogeneity among those at risk of repartnering. In order to test whether the patterns of repartnering vary with the definition of union commitment employed, two alternative approaches have been considered. In the first approach, the focus is on a more traditional definition of union, so that only transitions from one marriage to another marriage are of interest. In other words, only divorced individuals are at risk of repartnering, while repartnering only takes place if the person at risk marries again. The second alternative scenario emphasizes childbearing and only considers transitions from one childbearing union (whether cohabitation or marriage) to another childbearing union. These additional analyses show that, regarding the association between the degree of exogamy in the previous and the subsequent union, the results obtained in the additional analyses (see Tables S3–S6 in the Online Supplement) are very similar to those presented in Tables 1 and 2. The differences in coefficients range from marginal to moderate and the main conclusions of the paper remain unaffected. The general patterns also remain unaltered when the analysis only considers individuals who are at their first repartnering risk (Tables S7 and S8 in the Online Supplement).

These results show that the main conclusions from Sect. 6 are robust to different definitions of what constitutes a committed union. Moreover, the analysis that focuses on childbearing unions only can be used to address the issue of “marriages of convenience” (i.e., unions formed with the primary goal of obtaining a residence permit or citizenship in Sweden). Of course, there is no way to identify these unions in registers, but it is safe to assume that a vast majority of these couples will not have a common child. Therefore, the similarity of the results in Sect. 6 with those reported in Tables S5 and S6 strongly indicates that the existence of marriages of convenience does not affect the main conclusions on the propensity for intermarriage among natives and immigrants in Sweden.

Discussion

The aim of this paper was to contribute to our knowledge about intermarriage and the patterns of social interaction in Sweden by analyzing the relationship between the experience of intermarriage and subsequent partnering choices. The results show that for all four groups defined by sex and nativity, the risk of entering any of the four types of union is the highest for individuals who were previously in the same type of union. This means that the study lends support to the proneness hypothesis, as there is a positive association between the previous experience of intermarriage and the likelihood of starting another intermarriage after union dissolution. Another important finding is that the magnitude of this positive association increases with the degree of social distance between the groups involved in the union. It can be argued that descriptive and multivariate analyses complement each other well in this paper, primarily because the interpretation of some results may depend on the perspective one chooses to adopt. One of the most intriguing results of the descriptive analysis concerns immigrant men who previously lived in a union with a native woman. On the one hand, men from this group are clearly the most likely to form a union with another native woman the next time around. On the other hand, most of them do not do so—almost 60 % of foreign-born men who repartner after having been with a native woman make a different choice the next time around. Keeping in mind that the number of native women on the marriage market clearly exceeds that of women from other groups, this result, more than any other in this paper, may indicate traces of a behavior corresponding to the learning hypothesis.

Turning back to the main result, what are the causal mechanisms behind the individual proneness to intermarriage among some natives and immigrants? One interpretation is that the same preferences that previously led to exogamy persist and are still at work when repartnering choices are made. However, as Blau et al. (1984) argue, this does not necessarily mean that people who enter exogamous unions have a preference for exogamy as such; it is more likely that these individuals have preferences for certain values and lifestyles more prevalent in other social groups. In some contexts, especially when there is a large social distance between groups, an above-average preference for other social groups may simply be a result of a lower-than-average level of prejudices or aversion toward these groups. One should not rule out the possibility that some results in this paper are also shaped by preferences with regard to physical appearance, which in some social contexts may be important for explaining the formation of native–immigrant marriages. Because it takes two to marry, the preferences of potential partners on the marriage market are also relevant when explaining the positive association between the previous experience of intermarriage and the likelihood of entering intermarriage when forming a subsequent union. For instance, an immigrant formerly married to a native is more likely to be proficient in Swedish, and thus more attractive to prospective native partners, as compared to an immigrant whose previous union was endogamous. Nevertheless, preferences are most likely not the only explanation for the main result of interest. Work, school, family, family networks, and other meeting settings often play an important role in the process of partner choice (Kalmijn and Flap 2001). The structure of these settings (which cannot be fully captured by the municipality-level indicators) is often very little affected by divorce or separation, which may contribute to the positive association between the previous experience of intermarriage and starting another intermarriage. As for the role of third parties in shaping the results, it is most likely marginal for natives. The role of third parties for immigrants should not be entirely discarded, but if the results of other European studies (Van Zantvliet et al. 2014) have some validity for Sweden, then one should expect this role to be at most modest. Finally, the results on the proneness to intermarriage may also be influenced by the nature of the learning process. First, the increased risk does not mean that all intermarriages that break up do so due to cultural differences. Second, even if this were the case, a dissolution of an intermarriage may negatively affect the attitude toward intermarriage not only among the former partners, but also among individuals in their surroundings who have not experienced an intermarriage, but are also at risk of repartnering themselves.

Gender differences are somewhat more pronounced among immigrants. In spite of some notable differences in descriptive statistics, in the multivariate setting the coefficients for native men and women are remarkably similar. In contrast, substantial differences emerge among the foreign-born concerning the risk of repartnering with a co-ethnic partner. When a comparison is made with immigrants who left an endogamous union, immigrant men who have experienced an intermarriage in Sweden are only somewhat less likely to enter an endogamous union the next time around, whereas foreign-born women with the same experience are substantially less likely to repartner with a co-ethnic. Differences in gender norms between origin and destination could play an important role when explaining this result. Sweden ranks very high in terms of gender equality, which is not the case with many countries with immigrant populations in Sweden. Previous research has indicated that, due to contextual and institutional differences in gender norms, immigrant women on average benefit more from migration to Sweden than immigrant men (Darvishpour 2002). Applying the same reasoning, it may well be the case that intermarriage too is more beneficial for women (e.g., in terms of intra-household bargaining power). This would mean that, when confronted with the choice between exogamy and endogamy, immigrant women value the gain in intra-household power more than they are afraid of marital conflicts due to cultural differences. Another possible explanation is that, for immigrant women, entering a union with a native person implies a more radical cutting of ties with the community of origin than for men, which in itself limits the opportunities to meet potential co-ethnic partners following the divorce or separation.

As discussed in previous sections, this is a study on committed unions, i.e., marriages and childbearing cohabitations. Sweden is among the countries with the highest prevalence of cohabitation, and where this type of living arrangement is practically indistinguishable from marriage in terms of social acceptance (Duvander 1999; Kiernan 2001; Andersson 2004). Why then would one believe that, for instance, married Swedish couples without common children are on average more committed than cohabitants without common children? It can be argued that the same level of social acceptance does not imply the absence of differences with respect to commitment between two types of living arrangement. In a context with very weak (or entirely absent) social pressure to start a marriage, a decision to do so can signal the partners’ desire to upgrade their level of mutual commitment. Also, Swedish law still distinguishes between cohabitations and marriages, usually in such a way that less committed couples have little incentive to enter a marriage (Agell and Brattström 2008; Perelli-Harris and Gassen 2012; Holland 2012). The empirical evidence supports the view that marriage matters also in Sweden as an indicator of commitment (Andersson and Philipov 2002; Wiik et al. 2009). Having this in mind, the lack of information on cohabitants without common children in the STAR data should not pose a threat to the validity of results presented here. However, having access to this information would have enabled some additional analyses. Among other things, it would have been possible to construct more precise marriage market indicators. Also, register data lack information usually obtained through survey-type questions, such as self-declared ethnicity, individuals’ social networks, or which of two partners initiated the dissolution of the previous union. This information would have helped in identifying more precisely the causal paths behind the main findings. Nevertheless, there are still appealing topics that can be addressed with register data. In that regard, future research should build on the findings in this paper and explore how assortative mating with respect to education and origin interact to shape the patterns of repartnering.

If we depart from the view that social integration is a two-way process including both natives and immigrants, how do the findings in this paper contribute to the debate on the link between intermarriage and social integration? It does not seem that an elevated risk of divorce or separation for native–immigrant unions in Sweden substantially undermines the importance of intermarriage as an agent of social integration, since a substantial share of those who experience the failure of an intermarriage will subsequently start another intermarriage. What might be a cause for less optimism in this debate is the low share of those who choose an exogamous union following a failure of an endogamous union. If natives with an experience of a union with a non-Western immigrant are four to five times more likely to repartner with another non-Western immigrant than natives who were previously in an endogamous union, and if this association is indeed to be ascribed to preferences (other than those related to physical appearance) or to social networks, then this may suggest that social integration in Sweden is taking place only partially—at least when looking through the prism of intermarriage.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Appendix

See Tables 3, 4, and 5.

Table 3.

Classification of immigrant groups

Immigrant group	Countries/regions of birth
1	Finland
2	Denmark, Iceland
3	Norway
4	Bosnia and Herzegovina
5	The rest of the former Yugoslavia
6	Poland
7	UK, Ireland
8	Germany
9	Southern Europe (Spain, Italy, Greece, Portugal, Malta)
10	Estonia, Latvia, Lithuania
11	The rest of the former Soviet Union
12	Albania, Bulgaria, Romania, Hungary, Czech Republic, Slovakia
13	Other European countries
14	USA, Canada
15	Central America, Guyana, Surinam
16	Chile
17	The rest of Latin America
18	Djibouti, Eritrea, Ethiopia, Somalia
19	Algeria, Egypt, Libya, Morocco, Tunisia
20	Middle Eastern countries (excl. Egypt, Iraq and Iran)
21	Africa (excl. northern Africa)
22	Iran
23	Iraq
24	Turkey
25	Japan, China, and other East Asian countries
26	Thailand, Vietnam, and other South-east Asian countries
27	Other Asian countries
28	Australia, New Zealand

Source: Swedish register data