تفاوت غیر قابل تطبیق؟ ازدواج های قومی و طلاق در هلند، 1995-2008

Abstract This study uses population data of the Netherlands (municipality registers) between 1995 and 2008 to describe and explain the occurrence of divorce among recently newlywed interethnic and mono-ethnic couples (N = 116,745). In line with homogamy theory, divorce risks are higher for interethnic couples, in particular if the spouses were born and raised in countries that are culturally distant from each other. In addition, the effect of cultural distance is smaller for second generation immigrants than for first generation immigrants. There is no evidence for a higher risk of divorce among Black–White marriages. In line with convergence theory, results show that the higher the divorce propensity in the wife’s origin country, the higher the divorce risk of a couple is.

Introduction Research on marriage between members of different ethnic groups can be traced back to the roots of sociology as it refers to one of its core problems, namely the problem of cohesion. Ethnic intermarriage, also known as ethnic exogamy, indicates strong links between members of different ethnic groups within society and is therefore considered to be an important indicator of the social integration of ethnic groups (Gordon, 1964, Kalmijn, 1998 and Monden and Smits, 2005). From this perspective, it is important to consider divorce of ethnic intermarriages because ethnic exogamy does not have the same value for societal cohesion if a large proportion of these marriages end in divorce (Zhang and Hook, 2009). Generally, studies showed that interethnic couples are more likely to divorce than mono-ethnic couples in the United States (Fu, 2006, Jones, 1996 and Zhang and Hook, 2009), Australia (Jones, 1994) and the Netherlands (Janssen, 2002 and Kalmijn et al., 2005). Nevertheless, exceptions to this pattern have been found. Several specific types of interethnic couples appeared to have more stable marriages in comparison to mono-ethnic marriages (Jones, 1996, Schwertfeger, 1982 and Zhang and Hook, 2009). In this study, we aim to describe and explain the occurrence of divorce among interethnic couples in comparison to mono-ethnic couples. Therefore, we address the general research question: to what extent and why does the ethnicity of partners affect the risk to divorce? We aim to contribute to existing literature in four ways. First, theoretically: we test old and new hypotheses from existing theory to broaden our understanding of interethnic divorce patterns. Scholars have relied on homogamy theory and convergence theory to guide their research (Kalmijn et al., 2005 and Zhang and Hook, 2009) and we extend previous research by arguing that effects proposed by the theories should be less strong for second generation immigrants. Second, conceptually: we study divorce patterns of (almost) every national origin group in the Netherlands, including immigrants as well as the native Dutch. In other studies, in particular in the United States, pan ethnic identity measures of ethnicity are often used. In these studies, immigrants are classified into large groups, such as Asians, Hispanics, Blacks and Whites. As a consequence, interethnic marriage (and divorce) within such panethnicities go unnoticed. In a pan ethnic categorization, for example, Blacks from Suriname (Dutch speaking post-colonial immigrants) would be classified together with Blacks from Somalia (non-Dutch speaking refugees), although these groups differ in terms of language, religion, and traditional norms and values. Hence, our national-origin concept of ethnicity captures the more fine-grained group boundaries in the Netherlands in comparison to pan ethnic concepts of ethnicity. We distinguish between 124 national origin groups, allowing us to examine patterns of divorce risk among various combinations of national origin groups. We provide descriptive information about divorce risks among and between the most important ethnic groups in the Netherlands. Our hypotheses, however, are about interethnic divorce within and between all 124 groups in general instead of between only the largest groups in the Netherlands. Third, methodologically: we contribute to existing literature by using longitudinal data on the entire Dutch population. Previous research has been mostly of static nature, thereby relying on cross-sectional data to examine the risk of divorce. A major problem with these static data is that they start with a selective sample of (surviving) marriages. The longitudinal design of the data we use circumvents this problem because we start our observations with newlyweds. We follow couples over a 13-year period and use event history techniques to examine the annual risk of divorce while correcting for two-way clustering in our data on the husband’s and wife’s origin country. Fourth, descriptively: little is known about recent divorce rates of interethnic and mono-ethnic couples in the Netherlands. Most Dutch studies relied on data until 1999 (Janssen, 2002 and Kalmijn et al., 2005). This study contributes to the literature by providing up-to-date descriptive information on the linkage between national origin and the risk of divorce in the Netherlands, by covering the period between 1995 and 2008.

4. Results 4.1. Descriptive findings As a first exploration, we will present descriptive findings. Mono-ethnic native couples have the lowest divorce risk (86% are still married after the 13th year of marriage), followed by mono-ethnic immigrant couples (82%), native-immigrant couples (80%) and interethnic immigrant-immigrant couples (73%), respectively (Fig. 1). Survival curves for four types of marriages in the Netherlands, 1995–2008. Fig. 1. Survival curves for four types of marriages in the Netherlands, 1995–2008. Figure options More specifically, Table 2 illustrates divorce patterns in more detail. It shows the percentage of divorced couples within 10 years of marriage for natives, the largest five immigrant groups, and all other countries subsumed in continental categories.3 Only the marriages celebrated in 1995, 1996 and 1997 are included in Table 2 because only these marriages are observed for at least 10 years. Table 2. Percentages of divorced couples within 10 years for couples married between 1995 and 1998 in the Netherlands. Mono-ethnic marriages a Interethnic marriages with natives % Divorced n Marriages % Divorced n Marriages Dutch 12.2 186,732 Indonesian 12.5 295 17.2 3310 Antillean 16.2 294 22.6 393 Surinamese 27.2 1821 22.6 1453 Turkish 21.1 1414 27.6 392 Moroccan 15.7 1528 42.4 455 Other American (North and South) 21.2 33 17.3 925 Other Asian 12.1 1253 16.8 1091 Other European 11.6 902 14.6 5083 Other African 22.2 378 37.0 750 Total 12.5 194,468 19.9 15,060 Data: Municipality registers (GBA) for 1995–2008, Statistics Netherlands (own calculations). a The subsumed categories include mono-ethnic marriages as well as interethnic marriages. However, only 3.6% of all couples is an interethnic immigrant couple. Hence, most marriages in these categories are mono-ethnic marriages. Table options Focusing on the mono-ethnic marriages, it can be seen that the marriages of the native Dutch are most stable (12.2% divorced), followed by the Indonesians (12.5%), Moroccans (15.7%), Netherlands Antilles (16.2%), Turkish (21.1%) and lastly the Surinamese (27.2%). Table 2 also shows that the finding in Fig. 1 holds for most of the national origin groups: interethnic couples are more likely to divorce than mono-ethnic couples. There are, however, deviations from this pattern. For example, Surinamese-Dutch couples have a divorce rate (22.6%) that is in between the divorce rate of mono-ethnic Dutch couples (12.2%) and mono-ethnic Surinamese couples (27.2%). 4.2. Hypotheses testing We tested our hypotheses using logistic regressions, of which the results are presented in Table 3. Model 1 estimates the effect of exogamy and all control variables, in Model 2 all independent variables except for immigrant group size are added and in Model 3 we include interaction effects concerning couples with at least one second generation immigrant partner. Models 4 and 5 are based on sub samples. In Model 4, we examine ethnic intermarriages between immigrants and include the effect of group size. In Model 5, we replicated Model 4 for native-immigrant marriages, taking the group size and crude divorce rate of the immigrant partner for a couple. Table 3. Odds ratios from multiway clustering event history models of divorce in the Netherlands, 1995–2008. Model 1 Model 2 Model 3 Model 4 Model 5 Couple level 2nd Generation partner 1.05 1.31⁎⁎⁎ 0.92 1.24 Origin country level Divorce rate husband 0.89 0.85 0.82 Divorce rate wife 1.09⁎⁎ 1.13⁎⁎ 1.12 Divorce rate immigrant partner 1.01 Group size husband 1.02 Group size wife 1.12⁎⁎ Group size immigrant partner 1.00 Combination level Exogamy 1.56⁎⁎⁎ 1.21⁎ 1.17 Religious difference 1.27⁎ 1.30⁎ 1.41⁎⁎ 1.17⁎⁎⁎ GDP difference 1.01 1.01 0.97 1.03⁎⁎⁎ Color difference 1.00 1.00 1.01⁎⁎ 1.00 Interaction effects 2nd gen⁎ religious difference 0.75⁎⁎ 0.77⁎ 0.87 2nd gen⁎ divorce rate husband 1.19 1.34 2nd gen⁎ divorce rate wife 0.96 0.82⁎ 2nd gen⁎ divorce rate immigrant p. 1.13 Control variables Age at marriage wife 0.93⁎⁎⁎ 0.93⁎⁎⁎ 0.93⁎⁎⁎ 0.95⁎⁎⁎ 0.93⁎⁎⁎ Child (<6 years) 0.47⁎⁎⁎ 0.47⁎⁎⁎ 0.47⁎⁎⁎ 0.67⁎⁎⁎ 0.52⁎⁎⁎ Registered partnership 1.69⁎⁎⁎ 1.68⁎⁎⁎ 1.69⁎⁎⁎ 0.77 1.63⁎⁎⁎ Age heterogamy – h > 5 years older 1.05 1.04 1.04 0.92 0.98 – h − 2–5 years Ref. Ref. Ref. Ref. Ref. – h > 2 years younger 1.62⁎⁎⁎ 1.63⁎⁎⁎ 1.60⁎⁎⁎ 1.91⁎⁎⁎ 1.51⁎⁎ Marital duration 2.00⁎⁎⁎ 2.01⁎⁎⁎ 2.01⁎⁎⁎ 2.52⁎⁎⁎ 2.07⁎⁎⁎ Marital duration ⁎ marital duration 0.95⁎⁎⁎ 0.94⁎⁎⁎ 0.95⁎⁎⁎ 0.92⁎⁎⁎ 0.94⁎⁎⁎ Model characteristics χ2 (df a) 7459(6) 63,897(10) 51,652(12) 16,144(14) 289,069(11) N total observations 957,719 957,719 957,719 29,969 324,813 N couple level 116,745 116,745 116,745 4235 40,960 N origin country level 124 124 124 123 124 N combination level 1250 1250 1250 1028 239 Notes: All continuous variables are centered around their mean. 2nd gen = 2nd generation immigrant. Data: Municipality registers (GBA) for 1995–2008, statistics Netherlands (own calculations). ⁎ p < 0.05 (one-tailed). ⁎⁎ p < 0.01 (one-tailed). ⁎⁎⁎ p < 0.001 (one-tailed). a Due to invalid (inadmissible) covariance matrices in the multiway clustering specification, we used the proposed adjustment of Cameron and Trivedi (2005) to correct for this. A byproduct of this adjustment is a reduction in the degrees of freedom. Table options Model 1 shows that the odds for interethnic couples to divorce are 56% higher compared to mono-ethnic couples (eB = 1.56). Thus, controlling for age at marriage of the wife, having a child younger than 6 years at home, having a registered partnership, age heterogamy, and marital duration, we find evidence that partners from different origin countries have a higher divorce risk than partners from the same origin country (hypothesis 1). Without these controls, the odds to divorce are 41% higher for interethnic couples (model not shown). The odds ratio of exogamy becomes insignificant in Model 2 and Model 3, indicating that explanatory variables added in those models partly explain the higher divorce risk of interethnic couples. It appears that the odds to divorce for couples who have a different predominant religion in their origin country are 27% higher in comparison to couples who have the same predominant religion in their origin country (Model 2). Religious differences appear to be important for the risk of divorce among marriages between immigrant spouses (Model 4), and for marriages between immigrants and natives (Model 5). Our other measure of cultural distance (i.e., the difference in the GDP per capita of the country of origin of the husband and wife) is not significant in the analysis of the entire population (Models 2 and 3) and the analysis of immigrant marriages (Model 4). However, when the difference in the GDP per capita from an immigrant spouse and the native Dutch spouse is larger, the risk of divorce is higher (Model 5). More specifically, for immigrant-native couples, the odds to divorce increase with 37% (1.0310.70) with every standard deviation increase of GDP difference in the wife’s and husband’s origin country (SD = 10.70; Table 1). The effect also holds in a model without any interaction effects (eB = 1.03; model not shown). The results therefore largely confirm the hypothesis on cultural distance. We ran additional analyses using Hofstede’s (1980) cultural index. We found that the odds to divorce increase with 16% with every standard deviation the wife’s and husband’s origin country differ on Hofstede’s measure of individualism (model not shown). Again, this supports our hypothesis that a larger cultural distance between spouses’ origin countries increases the risk to divorce (hypothesis 2). Hypothesis 3 was about the role of color differences. Interestingly, it appears that in the analysis involving native Dutch (Models 1–3 and Model 5), there is no evidence for increased risk when spouses differ in skin color. When looking at marriages between immigrants only, we do find a rather large effect of color differences (Model 4). The odds to divorce for interethnic immigrant couples increase with 39% (1.0132.96) with one standard deviation increase in the value of color difference (SD = 32.96; Table 1). Hence, the color line plays a significant role for marriages between immigrants, but it does not affect the risk for divorce when one spouse is Dutch. Hypothesis 4 stated that the risk of divorce increases with the size of the immigrant group. Model 4 shows for interethnic immigrant couples that the odds to divorce increase with 25% (1.122.00) with a one standard deviation increase of the natural logarithm of the wife’s immigrant group size (SD = 2.00; Table 1). We find no evidence, however, that the immigrant group size of the husband has a positive effect. For native-immigrant couples, we do not find evidence either. Thus, we find partial support for hypothesis four. Are the effects of cultural distance smaller among the second generation (hypothesis 5)? For reasons of multicollinearity, we could not test the interaction between immigrant generation and GDP difference. With respect to our other measure of cultural distance – religion – results are in line with expectations. Thus, the odds to divorce for couples without a second generation partner increase with 30% (eB = 1.30) if the predominant religion in the partners’ origin country differs, whereas for couples with at least one second generation immigrant partner, the odds to divorce increase with only 2.5% (1.30 * 0.75) if the predominant religion the partners’ origin country differs (model 3). Hypothesis 5 is confirmed. In line with convergence theory, it was hypothesized that the divorce rate in the country of origin positively affects the odds of divorce (hypotheses 6a and b). We find that with a one standard deviation increase of the crude divorce rate of the wife’s origin country (SD = 1.01, Table 1), the odds to divorce increase with 9% (Model 2). This is a rather small effect. We do not find our hypothesized positive effect of the husband’s origin country divorce rate. As a sensitivity check, we estimated Model 2 on the basis of only those couples of whom the crude divorce rates in the wife’s and husband’s country of origin is reported to the UNSD. The effects of the crude divorce rates are of similar magnitude (View the MathML sourceedivorce rate origin country husbandB=0.94;edivorce rate origin country wifeB=1.09∗; model not shown). This leads us to the conclusion that the results are robust. Thus, the data only modestly support convergence theory. Finally, we find no clear patterns of interaction between immigrant generation and the divorce rate in the country of origin (hypothesis 7).