1 Introduction

Intermarriage (exogamy between immigrants and natives) has long been seen as a key indicator of immigrant integration. Previous research has stressed the importance of human capital and marriage market structure for the likelihood of intermarriage (e.g. Alba and Golden 1986; Kalmijn 1998; Furtado 2006; Chiswick and Houseworth 2008; Dribe and Lundh 2008, 2010; Celikaksoy et al. 2010). After controlling for individual characteristics and human capital, however, we previously found large remaining differences in intermarriage propensities between immigrants from different origins which seemed to be related to the degree of cultural dissimilarity between these immigrant groups and Swedish natives, measured by religion and value context in the country of origin (Dribe and Lundh 2008, 2010, 2011). Similarly, cultural dissimilarity seems to have been important for intermarriage of ethnic minorities in Estonia (Van Ham and Tammaru 2011). Other studies have found an earnings premium for some intermarried immigrant men compared to endogamously married ones, which could be related to the spillover of country-specific human capital within marriage in terms of language training and access to native networks (Meng and Gregory 2005; Meng and Meurs 2009; Dribe and Nystedt 2011).

The rather optimistic view that intermarriage is associated with or even promotes immigrant structural integration has been partly contradicted by theories and findings in the union dissolution literature. The effect of exogamy or heterogamy on marital satisfaction and divorce risks has been studied more generally for a range of individual characteristics: for instance, education (Tynes 1990), socioeconomic status (Glenn et al. 1974; Tzeng 1992; Ono 1998; Jalovaara 2003), religion (Lehrer and Chiswick 1993), ethnicity (Jones 1996), and race (Goldstein and Harknett 2006; Bratter and King 2008; Fu and Wolfinger 2009; Zhang and Van Hook 2009). The standard hypothesis in the literature is that exogamous marriages are less stable than endogamous ones. Several American studies on the effects of religious and ethnic exogamy on marital stability also lend moderate support to this idea (Bumpass and Sweet 1972; Becker et al. 1977; Michael 1979; Lehrer 1996). Most of these studies, however, look at descendants of previous immigrants rather than recently arrived immigrants. There has been much less research on the effects of religious or ethnic exogamy on marital stability in a European context. To our knowledge only a few studies exist, such as Finnäs (1997) on linguistic exogamy in Finland, Kalmijn et al. (2005) on religious and nationality exogamy in the Netherlands, and Eeckhout et al. (2011) on ethnic exogamy in Belgium. This apparent lack of knowledge calls for much more research in European contexts, which are often characterized by high immigration rates and poor societal integration of large immigrant groups (Rotte and Stein 2002; Zimmermann 2005). Sweden, with an immigrant population of about 12% of the total population (2005) is no exception to this general pattern (see Gustafsson 2002; Bengtsson et al. 2005).

In this study, we analyze the association between intermarriage and separation risks of native Swedes (Swedish born with two Swedish born parents) and immigrants (foreign born) who came to Sweden as adults. As well as studying the overall connection between intermarriage and union stability, we also look at the importance of gender and value context of the immigrant partner for the risk of separation, controlling for standard demographic and human capital characteristics, and features of the residential location. We use a couple design and event history analysis on longitudinal individual level data for 139 unique immigrant groups and native Swedes in the birth cohorts 1942–1989, and for the period 1990–2005. Owing to the high frequency of non-marital cohabitation among native Swedes we define first partnerships from the time of the first birth regardless of formal marital status, and follow these unions until dissolution or censoring.

2 Theory and Hypotheses

Theoretically, the individual decision to break up a union has been explained by a decline in the gains of being in the union and the presence of tension and external shocks (see, e.g., Becker et al. 1977; Becker 1991; Esser 2002; Blossfeld and Müller 2003). The individual considers the costs and benefits of a potential separation and calculates the net gain. In the literature, two approaches have dominated the discussion of marriage and union disruption; one that has claimed that specialization between spouses, according to comparative advantages in terms of productivity in market work and housework, stabilizes marriage, and that increasing similarity in earnings potentials between spouses leads to higher divorce risks; and another that has seen similarity in characteristics as fundamental for partner selection and marriage stability, and that, consequently, dissimilarity is associated with higher divorce risks. Below we discuss the relationship of these theories to the study of exogamy, value dissimilarity and union disruption.

2.1 Non-Specialization and Economic Independence

The way that the employment and income of women influence divorce risks has probably been the most frequently discussed theme in the literature. Generally, it has been found that employment and earnings of the woman in a couple have increased the risk of union disruption (South 2001; Blossfeld and Müller 2003; Poortman and Kalmijn 2002; Schoen et al. 2002; Rogers 2004). Theoretically, the basic formulation was made by the New Home Economics (Becker et al. 1977; Becker 1991), which stated that household productivity and joint utility were maximized by the division of labor between the spouses, according to their comparative advantages, normally resulting in men allocating most of their time to paid work and women to housework. As this arrangement was supposedly beneficial for both spouses, union disruption tended to lower utility for both parties. Following this line of thought, the change from a male-breadwinner model to a two-earner model in many Western countries since the 1960s has been seen as a major explanation for increasing divorce rates, because it has lowered the gains to marriage for both spouses (e.g. Becker 1991; Blossfeld and Müller 2003). In a similar way, economic dependency has been seen as a major factor in fostering union commitment and stability (Nock 1995). Greater female independence has been expected to increase the dissolution risk, which is highest when earnings are equal between the partners, because at that point the degree of dependency is at a minimum.

Both the specialization and dependence models have been questioned for not considering that increased total household income, stemming from female labor force participation or higher female wages, could counterbalance the decline in specialization gains and lead to higher, rather than lower, union stability (Greenstein 1990; Oppenheimer 1997; Mofitt 2000). Furthermore, it has been pointed out that female employment could mean different things for men and women (Kalmijn and Poortman 2006). For women, it implies a financially independent position and a feeling of self-esteem that lowers the real and perceived costs of divorce (Cherlin 1979; Oppenheimer 1994), while for men, at least to the extent that they favor traditional gender roles, it is a challenge to their own occupational positions and identities as breadwinners (Parsons 1949; Komarovsky 1962).

2.2 Value Dissimilarity

In accordance with the dominant view that similarity characterizes marriage matches (Kalmijn 1998), one factor of particular importance for union disruption is the dissimilarity between spouses, e.g., with regard to education, religion or ethnicity, because these reflect differences in values, tastes, and communication styles (Glenn et al. 1974; Tynes 1990; Kalmijn 1998; Kalmijn et al. 2004, 2005; Clarkwest 2007, 2010). This perspective could be derived from social-psychological theories implying that similarity between partners is highly valued and positively related to partnership quality (e.g. Antill 1983; Kurdek and Schmitt 1987), as well as from economic models emphasizing consumption complementarities (e.g. shared leisure activities or joint consumption of public goods) as an important gain to marriage (Becker 1991; Stevenson and Wolfers 2007). In dissimilar unions the number of joint activities is reduced, as is the degree of mutual confirmation of values. Such differences also make it more difficult for the parties to understand each other, which is of great importance, especially when the union is under external pressure. Furthermore, mixed unions often enjoy less support from family and social networks, or are subject to group sanctions, because they imply the crossing of group boundaries and the breaking of group norms (Kalmijn et al. 2005; Yancey 2007). In particular, when the relationship is troubled, the lack of support or occurrence of group sanctions should be important. The stronger the social boundary between groups, for instance, for historical reasons (compare Catholics and Protestants), the higher the separation risk for such unions.

According to this dissimilarity perspective, exogamy should increase the risk of separation for both immigrants and natives. Dissimilarity between the partners with regard to values, tastes, and communication could be expected to have a negative effect on the stability of the union, as it may increase tensions and make it harder to deal with difficulties in the partnership. The level of support from family and social networks could also be expected to be lower in exogamous unions, which would further destabilize them. Moreover, a higher degree of value dissimilarity between the partners would reinforce the exogamy effect. Tension in the union could be expected to be stronger when the distance, with respect to values or tastes, between the partners is larger, and the support from family or extended kin correspondingly lower for similar reasons. Hence, we expect separation risks to increase with greater dissimilarity in values between partners.

As an alternative to the dissimilarity perspective it has been maintained that the separation risk for partners from different ethnic backgrounds is a by-product of differences in group norms; in particular, differences in divorce culture in the origin populations. The risk of dissolution of a specific type of mixed union is thus the average of the group-level separation risks of the involved parties. Jones (1994) found empirical support for this hypothesis for immigrant intermarriage in Australia, as well as for ethnic intermarriage in Hawaii (Jones 1996). Similarly, Finnäs (1997) found this kind of convergence to be important in understanding divorce differentials according to language exogamy in Finland.

2.3 Gender Differences

There are clear gender implications for a study of the influence on separation risks of exogamy and value dissimilarity, because union dissolution tends to mean different things to men and women, and as gender roles differ between groups from different cultural contexts. Several studies of Western countries have indicated more positive attitudes towards union dissolution among women than among men, and that women more often initiate separations (Booth and White 1980; Pettit and Bloom 1984). In a study based on a retrospective life-course survey in the Netherlands, Kalmijn and Poortman (2006) found that women were twice as likely as men to initiate divorce, and that standard determinants of union disruption worked more strongly through “her” divorce than through “his” divorce. Also the stabilizing effect of a child in the household was larger for men than for women.

Traditional gender roles have been characterized by the dominant position of the man as the official representative of the household and the breadwinner, whereas the woman has played a more passive and subordinate role, concentrating on domestic duties and work within the household. Sweden has been depicted as a forerunner in terms of gender equality, both with regard to institutional support for working parents and factual female labor force participation (see Bernhardt 1992; Sainsbury 1996; Duvander 2000; Bernhardt and Goldscheider 2007; Stanfors 2007). The male breadwinner norm has for a long time been replaced by a dual-earner model. In the period after 1990, male employment rates have been between 75 and 85%, and female employment rates between 70 and 80%. While the vast majority of men have been working full-time, however, a large proportion of women, about 40%, have worked part-time (Stanfors 2007; Lundh 2010). Time-use studies have shown that Swedish women did more unpaid housework instead, and as a consequence the total numbers of hours used for paid and unpaid work was quite equal for men and women (Rydenstam 2003; Dribe and Stanfors 2009; see also Bittman and Wajcman 2000). Thus, the dual-earner model in Sweden has been based on women often working part-time and taking a larger responsibility for domestic work, in particular for childcare and childrearing (Ahrne and Roman 1997). On the other hand, family benefits, public childcare services and rather flexible working practices have made Swedish women expecting a continuous working life without longer career interruptions (Corman 2000).

Since many immigrant groups differ from Swedish norms when it comes to female employment, division of housework and gender roles, intermarriage can be expected to have different implications for men and women, not least depending on whether it is the man or the woman who is native. A union between a native Swedish woman and an immigrant man could be assumed to be less stable than a native man/immigrant woman union. In the former, the immigrant man often has more traditional attitudes about female employment and the sharing of household tasks, which would likely clash with the native woman’s attitudes, and hence create tension and disagreement. Furthermore, since women usually are more prone to initiate dissolution, this kind of mixed union can be expected to be more vulnerable to separation than a similar native man/immigrant woman union.

Moreover, as the dual-earner model in Sweden usually implies that men work full-time while women work part-time during childhood years, and do a larger share of housework, the value dissimilarity between a native man and an immigrant woman concerning female employment and household division of labor can be expected to be less destabilizing, as the difference in ideal situations for each party is smaller than in a woman native/man immigrant union.

In addition, some Swedish men prefer more traditional gender roles than average within the population, and seek women from countries with traditional values in terms of male authority and household division of labor. These unions will have a lower degree of value dissimilarity than is indicated by country averages. Commercial marriage agencies, non-commercial websites, and travel have increasingly made this sort of matchmaking easier (Niedomysl et al. 2010). Women in such unions will often be highly dependent on their husbands for both income and contact with society, which together with the high degree of similarity in values would promote union stability.

In conclusion, the interaction of gender and value context can be expected to imply a stronger exogamy effect on union dissolution for native woman/immigrant man unions than for native man/immigrant woman unions. For the same reasons, the impact of the degree of value dissimilarity on union dissolution should be greater for the former type of union.

2.4 Hypotheses

Based on the preceding discussion we posit the following three hypotheses:

  1. 1.

    Intermarriage increases the risk of union dissolution for both immigrants and natives compared to endogamous unions (the exogamy hypothesis).

  2. 2.

    Greater dissimilarity in value context between the partners reinforces the exogamy effect (the value dissimilarity hypothesis).

  3. 3.

    The exogamy and value dissimilarity effects are greater for native woman/immigrant man unions than for native man/immigrant woman unions (the gender difference hypothesis).

In testing these hypotheses it is vital to control for variables measuring specialization and dependency, as these factors can be assumed to differ a great deal between endogamous and exogamous unions, and also according to value dissimilarity. Moreover, to control for the possibility of a convergence in divorce cultures, comparisons of exogamy effects on union dissolution must be made with both native endogamous unions and immigrant endogamous ones.

3 Data and Methods

We use data from the Swedish population registers maintained by Statistics Sweden. Information from a number of different registers, such as the multigenerational register, the income register, the education register, the migration register, the death register and the register of the total population (RTB), has been linked. We study all couples consisting of natives and immigrants belonging to the birth cohorts 1942–1989. Natives are defined as people born in Sweden with two Swedish born parents, while immigrants are foreign born with foreign born parents who came to Sweden after the age of 15. The reason for the latter restriction is that immigrants who came in early childhood can be expected to be quite different in terms of values and attitudes towards host country norms and behavior than those who arrived in adolescence or early adulthood. Individuals in the sample are followed from entering their first partnership until separation, or until they are censored because of emigration, death, or the end of the study period in 2005. Only individuals entering their first partnership (having a first birth) after 1989 are included, because it is only after this date that we have full information on income, education, and non-marital cohabitation with common children. This means that about 90% of the individuals in the sample are born in the period from the mid-1950s to the mid-1970s.

We study formal marriage as well as non-marital cohabitation, and follow such unions from first birth onwards. The reason for choosing this design is that the frequency of cohabitation differs greatly between immigrants and natives. In our sample, about 50% of endogamously partnered natives have their first birth in non-marital cohabitation, while the corresponding figure for endogamous immigrants is only 8% (for intermarried the proportion is 30%), which means that looking only at the formally married would introduce a sample selection bias which could, potentially, distort the results. In the registers information about non-marital cohabitation is restricted to couples with common children (based on information about residence at the house unit level; fastighet), which means that it is impossible to study all cohabiting unions. This is not a serious limitation, however, because it seems most reasonable to limit the analysis to serious partnerships and not include all early cohabitation.

It is important to treat marriage and cohabitation in the same way because of the big differences between immigrants and natives in the frequency of non-marital cohabitation. To analyze formal marriages from the time of marriage, but cohabitation only from the first birth, would, on average, inflate the duration of marriage for the groups most inclined to marry (immigrants), which could seriously distort the analysis. Therefore, we define partnerships as having a common child, regardless of being legally married or not. Owing to the high frequency of non-marital cohabitation in Sweden, both before and after the first birth (e.g. Kiernan 2004), having a common child is a more relevant indicator of a serious partnership than only being legally married; the latter group being more selective. We limit the analysis to first partnerships, i.e. partnerships in which it is the first child for both spouses. This design allows equal treatment of formal marriages and non-marital cohabitation, which is crucial for the comparison between immigrants and natives. In the analysis estimations of the formally married only are also presented for comparison.

We have information on country of birth for a total of 141 different immigrant groups. In the analysis we exclude the small number of immigrants belonging to seven residuary groups with less than 100 individuals from each country in the sample, because it is impossible to decide whether or not the union is endogamous. We also exclude immigrants from Cuba and Myanmar because we lack information on value context. This leaves 139 individual countries of origin in the analysis. In total, we have about 2.8 million observations for 400,000 couples. Corresponding to these observations, we have about 63,000 separations (see Table 1).

Table 1 Means and distributions of the variables
Full size table

As many of the variables in the registers are available annually, a dataset is constructed with one observation per year for each couple. We use discrete-time event history analysis, estimating a series of complementary log–log models with duration since first birth as a control variable (see Rabe-Hesketh and Skrondal 2008, pp. 356–360). The dependent variable is binary, indicating whether or not the union is dissolved, by formal divorce or split-up of cohabitation, during the year of observation. All time-varying covariates refer to January 1 of the observation year. The variable of main interest is partnership type, which is defined based on country-of-origin endogamy and the value context of each country. Endogamous unions are those where the man and the woman are born in the same country, and exogamous those where they originate from different countries. This classification is made using the 139 different countries of origin. For the endogamous and the intermarried couples (exogamous including a native Swede) we also distinguish between different value contexts, while exogamy between two immigrants from different countries of origin are included as a comparison without identifying value dissimilarity (exogamy immigrant-immigrant).

Value context is measured at the country-of-origin level, and not at the individual level, because this information is not available, either in the registers or anywhere else. Individual level measures would have been more precise, not least since immigration is selective, and immigrants to a varying degree deviate from the average population of the home country. Nonetheless, the value context of the country of origin reflects the basic differences between immigrant groups and the native population, which do not disappear quickly after immigration. Thus, although not being ideal measures of individual values, they will be valid indicators of group level differences in value context.

We use two indexes developed by Inglehart (1997) to measure values along two dimensions. The traditional–secular/rational value dimension reflects the difference between societies in which religion is very important, and those in which it is not. More specifically, traditional values imply an emphasis on parent–child ties, authority, high levels of national pride and a nationalistic outlook, traditional family values and absolute standards, as well as the rejection of abortion, euthanasia, suicide and divorce. Societies closer to the secular/rational pole emphasize the opposite values. The survival–self-expression dimension reflects the difference between societies in which great emphasis is put on economic and personal security, and societies in which survival is taken more or less for granted, and values on subjective well-being, self-expression, and quality of life are more important. Societies near the survival pole emphasize the priority of economic and physical security over subjective well-being, self-expression, and quality of life. Linked to this are little support for gender equality, low levels of political engagement or trust in other people, and negative attitudes towards homosexuals, foreigners, and people with AIDS or criminal records. Societies near the self-expression pole emphasize the opposite values (see Inglehart and Welzel 2005).

The indexes are constructed with data from the World Values Survey, which was undertaken in five waves 1981, 1990, 1995, 2000, and 2006. Respondents in about 100 countries answered questions on beliefs and values in at least one of the survey waves. The indexes are constructed from the answers to ten questions that were similar in all surveys and were highly correlated with the answers to a broader set of questions (see Inglehart and Welzel 2005). The index scales on both dimensions range from about −2 to +2. Sweden is an outlier on both dimensions, close to the poles of secular/rational values (similar to Germany) and self-expression values (similar to the Netherlands and Australia), respectively.

We use the index scores of these two dimensions for countries where information is available from the survey of 1995 or surveys close to 1995 (www.worldvaluessurvey.org). In total, 88 out of our 139 countries of origin have information on the indexes from at least one survey, and 60% of these from the 1995 survey. For the remaining 51 countries we interpolate using information for adjacent, similar countries. We checked the robustness of this procedure previously and found that it did not bias the results in any way (Dribe and Lundh 2011).

To construct a couple variable measuring the value context we aggregate values into four groups. The first group, with values closest to Sweden, are countries with positive index values in both dimensions, the second group has a positive survival-self-expression index and a negative traditional-secular/rational index, the third group has a negative survival-self-expression index and a positive traditional-secular/rational index, and the fourth group has negative index values in both dimensions, thus being furthest away from Sweden in terms of values.

Since immigrants and natives differ greatly with respect to the prevalence of non-marital cohabitation, as was pointed out previously, and because cohabitation is associated with much higher dissolution rates, exogamy effects will also differ substantially depending on whether or not union type is controlled for. In a model without union type, exogamy effects will be underestimated because part of the effect will be concealed by differences in cohabitation. On the other hand, a model controlling for marital status will most likely overestimate the association between exogamy and union dissolution due to an overrepresentation of unstable couples among exogamously married compared to native–endogamous married couples. Thus, part of the exogamy effect estimated might be due to a selection of unstable couples into marriage among the intermarried couples, while equally unstable couples in the native–endogamous group cohabit to a larger extent. Consequently, we estimate one model with a control for union type (time-varying) and one model without a control. The true associations between exogamy and union dissolution can be expected to be somewhere between these estimates.

We also control for a number of variables commonly included in divorce models to capture specialization/dependence, as well as demographic and geographic factors (see, e.g., Bumpass and Sweet 1972; Tzeng and Mare 1995; Ono 1998; Jalovaara 2001, 2003; Schoen et al. 2002; Henz and Jonsson 2003; Kalmijn 2003; Goldstein and Harknett 2006; Lyngstad and Jalovaara 2010): man’s education, woman’s relative education; man’s age; age difference between the spouses; man’s income (from employment, self-employment, unemployment benefits, sickness insurance, and pre-retirement benefits); woman’s share of family income; and the age of the youngest child. We also include type of municipality and the county-level unemployment rate to capture level differences in union dissolution according to urbanity and regional economic fluctuations (cf. the discussion in Lyngstad 2011). In alternative specifications we estimated models with county-specific fixed effects, which yielded substantively identical results for all variables (results not shown).

4 Results

Table 2 displays relative risks from the multivariate estimates. Four different models are estimated. All models control at least for duration since first birth (partnership duration) and time period. Model 1 shows the crude differences in dissolution between various kinds of endogamous and exogamous unions. Model 2 includes all additional control variables except union type, which is added in Model 3. Model 4 is a comparison model only for the formally married. Looking briefly at the control variables, it is interesting to note that our results neither support specialization theory nor the dependency explanation.

Table 2 Regression estimates (relative risks, RR) of union dissolution
Full size table

Comparing Model 1 with Model 2 it seems clear that adding the control variables does not change the association between exogamy and union dissolution a great deal when comparing to native endogamous unions, while the associations gets much stronger when comparing to immigrant endogamous unions in the same value context. This implies that differences in education, income, residence etc. conceal the lower dissolution risks of endogamous immigrants.

Looking at the full models (Models 2–4) we find that endogamous immigrant couples have lower dissolution risks than native endogamous couples in Model 2, where union type is not controlled for, but higher dissolution risks when union type is controlled for or when only looking at formally married. Clearly, the lower dissolution risks in Model 2 are to a large extent explained by the relative unimportance of non-marital cohabitation among immigrants, as previously discussed. When controlling for union type, or when looking only at the formally married, they have higher, and not lower, risks of dissolution. As has already been explained, this difference is partly due to higher dissolution in cohabiting unions than in married ones, and partly to an over-representation of unstable couples in formal marriages among immigrants compared to natives.

Turning to exogamy, all forms of exogamous unions have higher dissolution risks than endogamous natives, with the possible exception of man-native unions with women from Value group 1 (most similar to Sweden). As expected the magnitude of the estimates differ between specifications, but overall the results offer clear support for the exogamy hypothesis. Intermarried unions are up to 155% more likely to split up than endogamously partnered natives, and in no specification are exogamous unions more stable than endogamous native ones. We find the smallest differences when union type is not controlled for (Model 2), while the estimates in Models 3 and 4 are more similar. In both these specifications the exogamy effects are substantial.

In all specifications there are pronounced differences in the magnitude of the effects between different exogamy combinations. The general pattern is that value contexts most dissimilar from Sweden have higher dissolution risks. For example, while intermarried unions in Value group 1 (most similar to Sweden) have 10–38% higher dissolution risks than endogamous native unions in Model 3, the corresponding range for Value group 4 (most dissimilar to Sweden) is 61–155%. This clearly supports the value dissimilarity hypothesis.

The magnitude of both the exogamy effect and the dissimilarity effect also differs considerably by gender. Among the intermarried, effects are generally bigger for intermarried couples where the woman is born in Sweden, than in similar couples where the man is Swedish born. Looking at Model 3, dissolutions risks are 25–58% higher in woman-native unions than in man-native unions in the same value context. This pattern is clearly consistent with the gender difference hypothesis. In the man-native unions the highest dissolution risks are in Value groups 3 and 4, while in the woman-native unions the effect is more linear, with the highest dissolution risk in the most dissimilar group. The fact that the gap between the two types of unions tends to increase with greater value dissimilarity is also consistent with the gender difference hypothesis.

We now turn to a comparison between intermarried unions and endogamous immigrant unions. It is quite clear that intermarried unions in all cases have higher dissolution risks than endogamous unions from the same value context as the immigrant spouse. Overall, the exogamy effects are somewhat larger in Model 2 than in Model 3. For man-native intermarriage the dissolution risks are 22–67% higher than in the corresponding endogamous unions if we look at Model 3. In Model 2 the relative risks are 51–94% higher. For woman-native exogamy the higher dissolution risks range from 67 to 112% in Model 3 (119–198% in Model 2). In all cases Value group 3 shows the largest estimates, but there is no clear ordering of the magnitudes according to the degree of value dissimilarity. Thus, while there is considerable support for the exogamy hypothesis also when comparing to endogamous immigrant unions, there is no clear support for the value dissimilarity hypothesis. The gender differences are of course the same as when comparing with native endogamy.

Finally, as a sensitivity test, we estimate a model for all unions, also including second-generation immigrants as natives (results not shown). Generally speaking, the exogamy effects are smaller than in Model 3, which also seems reasonable given that some of the exogamous unions include a native with a shared cultural heritage, which tends to lower the degree of dissimilarity and thus the dissolution risks.

5 Discussion

Overall, our results provide strong support for the exogamy hypothesis, that mixed unions face higher dissolution risks than endogamous unions. This higher dissolution risk for exogamous couples is present regardless of whether we compare with native endogamous unions or immigrant endogamous unions, which means that the exogamy effects are not simple by-products of convergence in separation probabilities between different partnership cultures. As the major determinants of union dissolution are controlled for in the analysis, the observed exogamy effects are not a result of heterogeneities in age, education, income or family context. Instead, our interpretation is that the lower stability among exogamous unions is a result of dissimilarities in values, attitudes, communication styles, and traditions, which lead to difficulties and disagreements in the partnership, or lower levels of support from, and interaction with, social networks and kin.

If this interpretation is correct, it also seems logical to expect that the greater the dissimilarity the greater the dissolution effects. Our results also give some support for this value dissimilarity hypothesis, but mainly when comparing with native endogamous couples. Greater distance in value context between the partners is associated with higher risks of union dissolution.

The effects on union dissolution of intermarriage and value context dissimilarity are also quite different for men and women. In general, the effects are stronger for native woman/immigrant man unions than for unions between native men and immigrant women, and this effect increases with greater value dissimilarity. These results are consistent with the gender difference hypothesis, and stress the gender dimension of value dissimilarity and of the bargaining power of men and women in unions.

The clash between traditional and modern values on gender roles is sharper in native woman/immigrant man unions. On average, native women have rational and self-expressive values, including a positive attitude towards female employment and gender equality, whereas many immigrant men have more traditional family values. If the male immigrant opposes female employment and the sharing of household tasks, this could lead to tensions in the union because it runs against the values of the native woman and her expectations of a career and financial independence. In comparison, the tension in similar native man/immigrant woman unions is much lower because it is the woman who represents the traditional and survival values, and the man the rational and self-expressive values. This implies that the role of the man is not much different from what it would be in an endogamous native union in terms of time spent on market work, although he would spend less time on housework. Thus, his occupational identity is not threatened by the traditional and survival values of the partner.

The bargaining power of the woman is also different in the two types of unions. In native man/immigrant woman unions the wife has the double disadvantage of being female, according to traditional values subordinated to the husband, and being foreign, probably lacking native language proficiency, networks, and labor market attachment. Consequently, their bargaining power is less than that of immigrant men with native partners. Also native women have more bargaining power than immigrant women because modern values include ideals of gender equality and the possibility of breaking up unwanted unions. The weak position of immigrant women in relation to their partners makes it more difficult to initiate union dissolution.

As is the case with most other research in this field, our results cannot immediately be interpreted in a causal manner. Although theory predicts a causal effect from exogamy and value context dissimilarity to union dissolution, and even though the empirical results appear to be consistent with the theoretical predictions, we cannot rule out that selection effects in partner choice could be important in explaining the patterns observed. More liberal-minded, or progressive, individuals—immigrants as well as natives—can be expected to be more likely to cross boundaries and enter into exogamous unions, and at the same time they can be expected to have less prejudice against separation and divorce, which could account for some of the associations between exogamy and union instability.

Returning to the issue of immigrant integration, there is mounting evidence that intermarriage with natives is a crucial aspect of the integration process in most Western countries. It is clear, for example, that the propensity to intermarry with natives is greatest among immigrants who are more similar in terms of values to the majority population (Dribe and Lundh 2011). There are also strong indications, however, that intermarriage has a causal impact on the earnings of immigrant men (Meng and Gregory 2005; Meng and Meurs 2009), and that the effects are greatest for the groups with the lowest intermarriage rates, i.e. the most dissimilar immigrants (Dribe and Nystedt 2011). Thus, marrying a native seems to promote economic integration of the most disadvantaged immigrant groups. This study shows that there is also a potential downside to this process, in that entering a union with a native increases dissolution risks, especially for those of highly dissimilar origins in terms of values. In short, our findings indicate the complexity of intermarriage and its association with immigrant integration, and that analyses in the field should consider the aspects of value differences and gender, as well as their interaction.