Abstract
Numerous studies have shown that parental divorce is associated with an increase in adult children's divorce risk. We extend this literature by assessing how parental divorce on both sides of a couple is related to their partnership dynamics, specifically, whether there is parental divorce homogamy and whether a history of parental divorce for both partners is associated with increased dissolution risks for cohabiting and married unions. We use Finnish Census Panel data on 28,021 cohabiting and marital partnerships to conduct event-history models that follow individuals between ages 18 and 45. Findings show substantial parental divorce homogamy. Children with experience of parental divorce have 13% greater odds of cohabiting with and 17% greater odds of marrying a fellow child of divorcees, compared with those whose parents have not divorced. Moreover, contrary to evidence from the United States and Norway, our findings for Finland support an additive—rather than multiplicative—association between parental divorce homogamy and union dissolution. Parental divorce homogamy increases offspring's union dissolution risk by 20% for cohabitation and 70% for marriage, compared with couples for whom neither partner's parents are divorced. In Finland, the sizes of these associations are notably weaker than in the United States and Norway, likely because cohabitation and separation are more widespread and socially accepted in Finland, and an expansive welfare state buffers the socioeconomic consequences of divorce.
Introduction
In recent decades, divorce, separation, and repartnering have proliferated across many affluent democracies (Thomson 2014). Numerous studies suggest that parental divorce increases the risk of divorce among offspring (e.g., de Graaf and Kalmijn 2006; Diekmann and Schmidheiny 2013; Dronkers and Härkönen 2008; Lyngstad and Jalovaara 2010). The intergenerational transmission of parental divorce, therefore, is one possible driver of rising or persistently high divorce rates (Wagner 2020). Mechanisms linking parents' and offspring's divorce include socioeconomic status transmission, social learning of attitudes and interpersonal behaviors from parents, and a lower threshold of union dissolution when it was observed in parents (Amato 1996; Amato and DeBoer 2001; review also Lyngstad and Jalovaara 2010). Compared with divorce, much less is known about intergenerational transmission of separation when the adult children cohabit.
Union dissolution has numerous consequences for adults and children (Amato 2000; Kalmijn and Leopold 2020). For adults, consequences include new phases of living alone, single and nonresidential parenthood, and the possible formation of new partnerships and stepfamilies. Divorced households are often single-income households that are at an increased risk of poverty (Hübgen 2018; Smock et al. 1999), particularly among those with lower education or weak labor market attachment before or after union dissolution. Comparative research suggests that economic consequences of divorce are tempered by welfare state arrangements (Uunk 2004).
Parental divorce is also related to children's upbringing and family formation. Studies suggest that children of divorced parents tend to grow up, on average, with fewer socioeconomic resources and fewer positive role models for interpersonal behavioral skills (e.g., Amato & Sobolewski 2001; Cherlin et al. 1995), such as compromising and conflict resolution, to maintain long-lasting relationships. Furthermore, they tend to hold more prodivorce attitudes, which could lower thresholds for separation (e.g., Amato 1996).
Most studies focus on simple parent–child dyads to examine the intergenerational transmission of divorce; however, separations are negotiated within couples. Compared with couples in which only one partner's parents divorced, the risk factors for separation and divorce in couples in which both partners experienced parental divorce could accumulate either additively or multiplicatively, depending on the threshold for union dissolution at the societal level and across generations, as well as the intensity of interpersonal conflict (Amato 1996; Wolfinger 2003). In this article, we ask how parental divorce on both sides of a couple is related to union formation and union dissolution risk. First, we examine whether there is parental divorce homogamy—that is, whether children of divorcees are more likely to partner with fellow children of divorcees. Second, we investigate whether and the extent to which the dissolution risk of cohabiting unions and marriages changes when both partners' parents are divorced. This enables a more comprehensive account of the correlates of parental divorce homogamy for offspring's family formation in cohabiting and married unions. We use rich representative register data for Finland that allow us to follow entire coresidential partnership histories between the ages of 18 and 45, and link each cohabiting and married partnership to both partners' parental divorce history, which is usually not possible in available survey data.
The contribution of this study is threefold. First, we account for the entire prior coresidential partnership history from ages 18 to 45 as a pathway through which separation risks are passed from one generation to the next, including parental divorce homogamy in cohabitation and marriage. If children of divorced parents have an elevated likelihood of partnering with each other, parental divorce homogamy is relevant for a larger share of the population. Second, we explore whether the strength of parental divorce homogamy and its association with offspring separation risks are weaker in the Nordic welfare state of Finland than in the United States, the only non-Nordic country for which estimates of this association exist. Third, we present the first study, to our knowledge, that compares these associations for both cohabiting and married couples. Cohabitation has become a common union type that is not only a prelude to marriage. Children of divorced parents may be more likely never to marry (Glenn and Kramer 1987) and may cohabit instead. If parental divorce homogamy is concentrated among cohabiting couples, focusing only on marriage misses a substantial proportion of separations associated with parental divorce.
Background
Previous Research
Numerous studies suggest that parental divorce increases offspring's divorce risk and that intergenerational divorce transmission is stronger in some contexts than in others (e.g., de Graaf and Kalmijn 2006; Diekmann and Schmidheiny 2013; Dronkers and Härkönen 2008). For example, transmission is weaker in countries where divorce among the parent generation is more common (Dronkers and Härkönen 2008) and less socially stigmatized (Kalmijn and Uunk 2007). A meta-analysis of 120 European divorce studies showed that lower divorce barriers in a society are associated with weaker intergenerational divorce transmission (Wagner and Weiß 2006). As divorce proliferates, its intergenerational transmission might weaken, but it becomes relevant for a larger number of individuals who have experienced parental divorce. While there is a sizeable literature on intergenerational divorce transmission in simpleparent–child dyads, the couple level in which both partners' have experienced parental divorce has received much less research attention.
To our knowledge, only three studies to date have examined how parental divorce on both sides of a couple is associated with offspring's divorce risk (Amato 1996; Storksen et al. 2007; Wolfinger 2003). They all support up to three times higher divorce risks when both spouses' parents are divorced, compared with those couples in which the parents of neither partner are divorced. This suggests a multiplicative—rather than an additive—association of parental divorce on both sides of the couple. For the United States, Amato (1996) attributes this largely to a compound effect of poor interpersonal skills of both spouses that accumulate, causing conflict. Using different data for more recent cohorts in the United States, Wolfinger (2003) similarly finds a three times higher likelihood of divorce for couples in which both partners' parents are divorced than for those in which neither are. In addition, children of divorced parents are 31% more likely than others to marry a fellow child of divorcees. Wolfinger (2003) concludes that parental divorce homogamy multiplicatively compounds the risk of offspring divorce.
The findings might be specific to the United States, where marriage is particularly idealized and culturally loaded, despite high divorce and remarriage rates (Cherlin 2004; Lewis and Kreider 2015; Sharp and Ganong 2011). Further, the socioeconomic consequences of divorce are severe in the United States, especially for women, because of the limited welfare provisions for single mothers and persistent gender gaps in employment and wages (Cherlin 2010). Compared with the estimates for the United States, those reported for Norway by Storksen et al. (2007) are slightly smaller for both parental divorce homogamy and the divorce risk when both spouses' parents are divorced.
The Finnish Context
Finland is an interesting context in which to study intergenerational divorce transmission and parental divorce homogamy for several reasons. First, it is a forerunner of changes in partnership dynamics associated with the second demographic transition (Guzzo 2014; Lesthaege 2010). The average age at first marriage has increased: in 1982 it was about 25 for both men and women, and in 2017 it was 32 for women and 34 for men (Official Statistics of Finland 2015, 2018). The average age at divorce was 41 for women and 43 for men in 2017 (Official Statistics of Finland 2018). Cohabitation is widely accepted: it is common for younger adults to cohabit for long periods before they marry (if they do marry) (Jalovaara and Fasang 2020). Nonetheless, marriage remains important as a signal of the highest commitment, and cohabitations dissolve at a much higher rate than marriages (Jalovaara and Kulu 2018).
Second, Finland is a gender-egalitarian country where women's employment rates are high. Further, the Finnish welfare state provides generous, largely universal, and individualized services and income transfers that buffer the socioeconomic consequences of union dissolution (Hakovirta 2011). Thus, poverty is not as strongly transmitted by parental divorce as in liberal and less generous welfare states, such as the United States. The association between parental divorce and offspring's union dissolution is likely stronger in countries where divorce barriers are higher and divorce has more severe socioeconomic consequences. Thus, we expect weaker parental divorce homogamy and weaker associations of parental divorce homogamy with offspring's union dissolution in Finland than in the United States (Amato 1996; Wolfinger 2003).
Extending Mechanisms of Intergenerational Transmission of Union Dissolutionto Parental Divorce Homogamy
Commonly discussed mechanisms of transmitting union dissolution (and union formation) from parents to children include socioeconomic status transmission, socialization, and a genetic component (Fasang and Raab 2014). These mechanisms partly play out and are reinforced by partnership histories preceding union dissolution (Lyngstad and Jalovaara 2010), marrying, partner selection, and cohabitation. Union dissolution is negotiated within couples. Mechanisms of transmission between parents and children could either multiplicatively or additively compound when both partners have experienced parental divorce. To extend the foregoing mechanisms to parental divorce homogamy, we assume that separation crucially depends on (1) the frequency and intensity of interpersonal conflicts and (2) both partners' thresholds for divorce at given levels of conflict. Factors that increase the frequency and intensity of interpersonal conflict are more likely to multiplicatively compound separation risks associated with parental divorce homogamy (Amato 1996; Wolfinger 2003). Conflicts accumulate and easily trigger further conflict multiplicatively without compensating forces. In contrast, factors that merely change thresholds for divorce but do not increase conflict are more likely to additively compound among couples in which both partners experienced parental divorce than among those in which only one partner did so. Lower thresholds for divorce refer to lower levels of commitment to the relationship and a lack of love or positive perspective on the relationship (e.g., Amato and Previti 2003; Amato and Rogers 1997; Clarke-Stewart and Brentano 2006). Under these circumstances, partners will more readily see separation as an acceptable option at given levels of conflict.
In the following, we provide an extension of the proposed mechanisms of intergenerational divorce transmission to parental divorce homogamy. Specifically, we distinguish whether observed risk factors, such as interpersonal conflict, act as multipliers when observed in both partners (multiplicative effect) or whether they are more likely to simply operate additively (additive effect) (Figure 1). More generally, an additive effect could be found in societies and among generations in which union dissolution is more socially acceptable and common, because the intergenerational transmission of divorce could be weaker and the role of divorce different. Because we cannot test all mechanisms discussed directly with our data, we focus on their implications for effect size in the added separation risks associated with parental divorce homogamy, which we can determine with high accuracy and reliability. We can directly account for indicators of status transmission and include extensive information on partnership histories preceding offspring divorce.
Concerning status transmission, low socioeconomic status is associated with higher separation risks (Amato 2010; Jalovaara 2001; Kulu 2014). Because socioeconomic status is transmitted from one generation to the next, divorce transmission can arise as a by-product. Elevated stress, conflict, and separation due to economic hardship have been found for married and cohabiting couples (Halliday Hardie and Lucas 2010), and low-income couples are less likely to ever marry and more likely to cohabit compared with couples with higher income. If parental divorce homogamy is disproportionately associated with both partners experiencing economic disadvantage, the stress related to economic hardship is likely to increase the risk of union dissolution multiplicatively because of heightened interpersonal conflicts. Associations between dual low incomes or joblessness with separation are likely weaker in more generous welfare states, such as Finland, which mitigate immediate economic hardship and related interpersonal conflict, than in liberal and conservative welfare countries (Hansen 2005).
Socialization and social learning refer to the norms and values about desirable and appropriate family lives that children learn in early childhood (Fasang and Raab 2014). Children who have experienced parental divorce generally stigmatize separation less than children whose parents have not divorced (Dronkers and Härkönen 2008). More positive attitudes toward divorce among both partners in a couple will lower thresholds for separation but seem unlikely to increase interpersonal conflicts. We, therefore, expect a mere additive increase in separation risks because of both partners having more favorable attitudes to divorce and more quickly agreeing on separation at a given level of conflict. Prodivorce attitudes were already widespread for our study cohorts—particularly the younger generation—in Finland. We expect that both partners holding more prodivorce attitudes additively lowers thresholds for divorce and that this is particularly salient for our study cohorts in Finland.
Socialization and social learning also occur with regard to relationship skills that are observed in the parental home (Wolfinger 2003). If union dissolution is triggered by weak interpersonal skills between partners (typically, the ability to communicate and compromise), children of divorced parents may adopt the same behavioral patterns that then undermine the stability of their relationships (Amato 1996; Storksen et al. 2007). A transmission of weak relationship skills would increase the frequency and intensity of interpersonal conflicts and, therefore, lead to a multiplicative increase in couples' separation risk when both partners' parents are divorced, as compared with when only one partner's parents are divorced. Prodivorce attitudes are related to country-specific cultural and institutional factors. Social learning of poor interpersonal relationship skills could multiplicatively accumulate conflict, particularly when marriage and romantic love are strongly culturally idealized, such as in the United States, where disappointment in not reaching these ideals could trigger more conflict than in a country context such as Finland.
A genetic component can also contribute to intergenerational transmission of union dissolution. Shared genetic factors between siblings account for some of the intergenerational transmissions of divorce (McGue and Lykken 1992). The exact biological and psychological mechanisms linking parental and offspring's divorce largely remain a black box. In any case, genetic and health-related factors—for example, the inheritance of certain personality traits—seem more likely to affect interpersonal conflict (multiplicative) than to affect separation thresholds (additive effects). Genetic effects on union dynamics could be stronger in liberal contexts such as Finland where social norms are less restrictive (see Kohler et al. (2002) for fertility).
The foregoing mechanisms partly play out over the partnership histories preceding divorce. Several factors are associated with elevated divorce risks and are more likely for children of divorced parents: marrying at an early age (Kulu 2014; Lehrer 2008; Moore and Waite 1981), never marrying (Glenn and Kramer 1987; Storksen et al. 2007), and having divorced previously (Amato 2010). Moreover, separated individuals are more likely to separate again. In this study, we focus on two aspects of relationship histories preceding divorce: (1) parental divorce homogamy as part of assortative mating, and (2) its role for separating from cohabiting relationships, not only marriages.
Parental divorce homogamy (Storksen et al. 2007; Wolfinger 2003) is part of assortative mating. It can result from emotional closeness of sharing the experience of parental divorce or arise as a by-product of assortative mating. The strength of assortative mating differs across countries, likely leading to country-specific associations between parental divorce homogamy and offspring's separation risk. The stronger assortative mating is and the more negatively selected couples with parental divorce homogamy are, the stronger the association between parental divorce homogamy and offspring's union dissolution will be. Assortative mating on socioeconomic characteristics and their association with family behavior are stronger in such high-inequality contexts as the United States than in such low-inequality contexts as Finland (Schwartz 2013). We expect weaker and socioeconomically less negatively selected parental divorce homogamy and, therefore, weaker associations with offspring dissolution risk in Finland than in the United States.
Long-lasting cohabitations are increasingly widespread in countries in advanced stages of the second demographic transition. If children of divorced parents tend to cohabit rather than marry (Storksen et al. 2007), parental divorce might contribute to many more separations of coresidential unions than are captured in divorce transmission only. Despite the growing popularity of cohabitation, marriage and cohabitation differ in many respects. In Finland, most marriages start with cohabitation (Jalovaara 2012), and couples that continue to cohabit often remain in lower socioeconomic positions than those that eventually marry (Jalovaara 2013; Jalovaara and Kulu 2018).
Normative barriers are lower to dissolve cohabitations than divorce. Even inScandinavian countries, married couples report higher commitment and higher levels of relationship quality than cohabiters (Wiik et al. 2009). Married couples benefitfrom stronger social support and experience higher social pressure to stay together. Marriage is legally binding, and its dissolution requires formal divorce procedures, while cohabiting relationships end by (simply) moving apart. Married couples are also more likely to have children and jointly own property, which further operate as barriers to union dissolution (Jalovaara 2013; Jalovaara and Andersson 2018). Despite these differences, previous research has shown that the antecedents of union dissolution in Finland are similar for cohabitation and marriage, although socio-economic resources are somewhat more important for marriages (Jalovaara 2013).
Because marriage is usually preceded by cohabitation, any parental divorce homogamy found in marriages is also likely in cohabitations. If children of divorcees are more likely to never marry and cohabit instead (Glenn and Kramer 1987; Storksen et al. 2007), parental divorce homogamy could be even stronger among cohabiting couples that eventually separate and do not marry. If lower thresholds for divorce rather than elevated interpersonal conflicts drive the association between parental divorce homogamy with offspring union dissolution in Finland, associations are likely weaker for separation from cohabitation than from marriage. Parental divorce homogamy would then arguably contribute less to already low thresholds for separation and an already elevated separation risk in cohabiting couples. In contrast, parental divorce homogamy might be more relevant in lowering overall higher barriers and thresholds for divorce.
Finally, the mechanisms might operate in gender-specific ways when only the man or only the woman has experienced parental divorce. Note that gender differ-ences in effect sizes would not distort our general framework of additive and multiplicative associations of dissolution risk with parental divorce homogamy relative to only one partner experiencing parental divorce. But the theoretical reasoning above easily extends to gender differences based on (1) women's higher likelihood of initiating the (emotional and bureaucratic) process of union dissolution (Hewitt 2009; Hewitt et al. 2006; Sayer et al. 2011) and (2) gendered norms and socioeconomiccorrelates of divorce (Pessin 2018). Both accounts suggest a stronger association when the woman's parents are divorced than when the man's parents are. In Finland, about 70% of divorce applications are filed by women (Kontula 2013). Gendered initiation of separation from cohabitation is unknown. Women experiencing parental divorce might, therefore, be more relevant in lowering thresholds for initiating divorce and separation. Finland represents a fairly gender-egalitarian context with high levels of full-time female employment and generous individualized universal benefits that lower dependence on family members. But the gender earnings gap remains substantial, especially among married men and women (Jalovaara and Fasang 2020), which suggests that women on average do lose more economically from divorce than men.
Hypotheses
Our core research interest is to estimate the dissolution risk in cohabiting and married unions for couples in which both partners' parents are divorced, compared with couples in which neither or only one of the partner's parents are divorced (see Lundberg et al. 2020).
Hypothesis 1: Children of divorced parents are more likely than others to form cohabiting and married unions with a fellow child of divorced parents (parental divorce homogamy hypothesis).
Hypothesis 2: Children of divorced parents have a higher union dissolution risk in both cohabitation and marriage than those without divorced parents (general parental divorce hypothesis).
Hypothesis 3: Couples in which both partners experienced parental divorce have a multiplicatively higher risk of union dissolution in both cohabitation and marriage than those in which neither partner experienced parental divorce (dual parental divorce hypothesis, see Figure 1).
Hypothesis 4: Parental divorce on both sides of a couple increases dissolution risk more in marriage than in cohabitation (union type hypothesis).
Data and Methods
Data
We used high-quality Finnish register data, The Finnish Growth Environment Panel (FinGEP), which is based on a 10% sample of individuals living permanently in Finland in 1980. The data structure for one example case is displayed in Figure 2. First, the index-persons (“Parents”) were linked to all their biological children (our focal “Individuals”). Second, “Individuals” were linked to each of their opposite-sex,1 coresidential, either cohabiting or married partners (“Partner 1,” “Partner 2,” “Partner 3”), and each partner was linked to their parents (“Partner's parents”).
To derive all married and cohabiting partnerships, we selected a subsample of index-persons' daughters born between 1969 and 1973, who we followed from ages 18 to 41–45 (i.e., between the years 1987 and 2014). In each year, we matched (adult) daughters with their cohabiting or marital partner, if they had one. If we used both sons and daughters, we would have inflated our sample by including some relationships twice. We, therefore, derived all relationships by reconstructing the daughters' relationship histories. Robustness checks using only the sons as index-persons yielded qualitatively the same results, but a somewhat lower case number of couples, as sons entered unions at a higher age. Since 1987, Finnish registers contain information about the place of residence down to the specific apartment, thereby enabling the linkage of opposite-sex individuals to coresidential couples, even when they are unmarried and childless (see Jalovaara and Kulu 2018). The partners can then be linked to their biological parents.
Unions were followed from their start until (if relevant) their dissolution. We considered all coresidential partnerships that women had between 18 and 41–45. Cohabiting couples entered the analysis when they started to cohabit (move in together), and married couples entered when they married. In both cases, right-censoring occurred after emigration, a partner's death, or age 41–45 (depending on cohort). For cohabitations, entry into marriage was an additional right censor. The final sample included 28,021 cohabiting or married couples, who contributed 284,802 total couple-years at risk of union dissolution. Married couples contributed 219,935 couple-years at risk and 4,305 divorces, whereas cohabiting couples contributed 66,499 couple-years at risk and 10,897 separations. In contrast to survey data, register data do not suffer from nonresponse or memory bias, enabling a reliable and representative linkage of couples and both partners' parental divorce. We excluded 1.9% of the cases because there was no information on the parents of both partners; in almost all of these cases, the parents were born abroad.
Our main predictor—parental divorce—was operationalized as follows: 0 = “not divorced,” 1 = “woman's parents divorced,” 2 = “man's parents divorced,” and 3 = “both partners’ parents divorced.” The category “not divorced” includes still married parents and widowed parents. We excluded persons who were born to single mothers because, in such cases, information on the father is often missing and our analysis required information on both parents. After this restriction, less than 1% of the parents of the remaining sample-persons (6,078 cases) were never-married or did not have information on both parents; these were excluded from the analysis. We included all marriages of biological parents, regardless of whether parents married each other before or after the child was born.
Following the literature, we controlled for both partners' education (a time-varying variable, henceforth indicated by “tv”), both partners' parents' education, age at union formation, union order (tv), partnership duration (time elapsed since entry into either the cohabiting or the married partnership, tv), birth cohort, and age of the youngest child in the family (tv) (Table 1). Parents' education is time-constant and given by the dominance principle (i.e., the highest observed maternal or paternal education). Parental education was categorized into primary, secondary (including vocational and general tracks), and tertiary education (those with a bachelor's degree or higher). Both partners' education was coded as time-varying as the highest degree attained at each observation point. Their education was categorized into primary, secondary (including vocational and general tracks), lower tertiary (bachelor's degree), and higher tertiary (master's degree or higher).
In line with previous research, cohabiting couples in our sample dissolved their relationships more often than did married couples (e.g., Jalovaara 2013). Partners' and their parents' education levels were lower among cohabiters than among married partners. The mean age of moving in together was slightly lower among married couples. Cohabiters had fewer children and more often had experienced parental divorce.
Methods
To assess parental divorce homogamy (Hypothesis 1), we followed previous research to first estimate logistic regression models on the probability of entering a cohabiting or married relationship with a woman with divorced parents (Storksen 2007; Wolfinger 2003). The central independent variable was whether the man also has divorced parents. To assess the association between parental divorce and offspring's dissolution risk (Hypotheses 2, 3, and 4), we used piecewise constant exponential event-history models that divided the time axis into one-year intervals (Blossfeld et al. 2009). The baseline hazard was assumed to be constant within each one-year interval, but could vary flexibly without assuming any specific functional form between intervals.
For some women, we observed multiple partnerships, as separation is a potentially recurring event. Women who separate multiple times might differ from women who do not on the basis of unobserved factors. Furthermore, previous separations tend to increase the risk of future ones. We followed two strategies to account for potential bias due to unobserved factors and recurring separations. First, we ran the entire analysis considering only the first cohabitation and the first marriage (analysis available from authors). The results remained substantively the same. Second, we ran all models including a “frailty” term, a woman-level random effect that controlled for the time-invariant unmeasured characteristics of a woman (or unobserved heterogeneity), which could influence the hazard of union dissolution for any of her partnerships—for example, personality traits or interpersonal behavior. We tested both gamma- and inverse Gaussian-distributed shared frailty, but the results were similar. Table A1 in the online appendix presents gamma-distributed shared frailty, which is widely used in the lit-erature because it has a flexible shape and is analytically tractable (Gutierrez 2002). The results remained substantively the same with or without frailty. We, therefore, conclude that unobserved time-invariant characteristics and recurring separations do not seem to bias our estimates of separation risks to a substantive degree. In the following, we present models without the frailty term. Stata 15.1 was used for the data analysis.
Results
Do Children of Divorced Parents Partner More With Each Other?
Table 2 shows logistic regression models of the probability of entering into cohabitation or marriage with a fellow child of divorcees to assess parental divorce homogamy (Hypothesis 1). The estimates support considerable parental divorce homogamy, which is slightly stronger in marriages than in cohabitations. Net of controls, the odds ratios for women whose parents are divorced of being in a union with a man whose parents are divorced are 1.17 (95% CI, 1.14–1.20) for married couples and 1.13 (95% CI, 1.08–1.17) for cohabiting couples. Parental divorce appears to influence assortative mating, making parental divorce homogamy a relevant feature on the population level.
Is Parental Divorce Related to Cohabitation and Marriage Separation Risk?
Figure 3 shows Kaplan-Meier survival curves for separation from cohabitation (top) and marriage (bottom) by parental divorce. In line with the previous literature, cohabitations dissolve at a higher rate and more quickly than marriages. In addition, differences in separation risks by parental divorce are more pronounced for marriages than for cohabitations. Couples in which both partners experienced parental divorce show the highest divorce risks, followed by couples in which only one partner experienced parental divorce. The survival estimates are thus consistent with Hypotheses 2 (general parental divorce hypothesis), 3 (dual parental divorce hypothesis), and 4 (union type hypothesis).
Table 3 shows exponential piecewise constant models for dissolution risks for cohabitation and marriage. In line with Hypotheses 2 and 3, among both cohabiting and married couples, the risk of separation is higher for couples in which both partners' parents are divorced than for those in which just one or neither partner had experienced parental divorce (model 0). Parental divorce on both sides of the couple increases the risk of union dissolution more among married couples than among cohabiting ones (Hypothesis 4). For married couples, when one of the partners' parents are divorced compared with neither, we observe a 28%–35% increase in the risk of divorce (see Table A2 in the online appendix). If both partners' parents are divorced, the risk of offspring's divorce increases additively to 70%. For cohabiting couples, the difference in the association between parental divorce and their own risk of union dissolution is much lower: an 8%–13% increased risk for union dissolution if one of the partners' parents are divorced and an additive increase of 20% if bothpartners' parents are divorced. When all control variables are included (full model in Table 3), the differences between parental divorce status slightly diminished, especially for married couples, compared with the raw associations. Including theeducation of parents and offspring leads to the largest reduction in the size of the association between own and parental divorce, which remains notable, suggesting that some but not all of the transmission of divorce can be attributed to status transmission.
The increase in separation risk associated with parental divorce on both sides of the couple, compared with on one side, is almost twice as high, supporting an additive and not a multiplicative association in Finland. We further tested whether there is any indication of a multiplicative association on separation risk owing to dual parental divorce by including an interaction term between the men's and women's parental divorce. The interaction term proved to be close to zero for both cohabitations and marriages (see Table A3 in the online appendix). Contrary to previous studies in which the size of the estimates for the probability of own divorce was three times larger for couples whose parents had both divorced, pointing toward a multiplicative association (Amato 1996; Storksen et al. 2007; Wolfinger 2003), our study finds strong evidence of a merely additive increase in separation risk when parents on both sides of a couple are divorced.
For the first time, we show that the association between parental divorce and offspring separation risk differs between cohabitation and marriage. Parental divorce increases separation risk more for married couples than for cohabiting couples, thus supporting Hypothesis 4. Although our data do not allow us to clearly disentangle the mechanisms driving the heterogeneous association for dual parental divorce for cohabitation and marriage, we are confident that birth cohort, offspring, and parental education (status transmission), as well as age at union formation, union order, and the age of the youngest child in the family (preceding relationship history), do not account for these differences to a large extent.
Robustness Checks
In addition to the analyses reported in detail above, we also performed a series of robustness checks to further substantiate our results. First, we replicated all analyses using men as sample index-persons. The results from the multivariate analyses (see Tables A4 and A5 in the online appendix) are highly consistent with those discussed earlier. The only notable difference is in the descriptive distributions:2 individuals in the sample based on male index-persons were slightly older when they entered unions (cohabitation, 26.7 vs. 25.0; marriage, 25.4 vs. 23.8) than in the women's sample, and we observe slightly fewer unions of men than of women when keeping the same age-bracket as the observation period (cohabitation, 68,889 couple-years vs. 70,460 couple-years; marriage, 197,987 couple-years vs. 216,525 couple-years). Second, we conducted all analyses using a different indicator for parental background: the results did not change when we used ISEI (international socioeconomic index of occupational status) instead of education.3
Discussion and Conclusions
This study aimed to assess the role of parental divorce homogamy in couple's partnership dynamics. Specifically, we examined parental divorce homogamy in partner selection, and the dissolution of cohabiting and married unions. To our knowledge, only three previous studies—two in the United States (Amato 1996; Wolfinger 2003) and one in Norway (Storksen et al. 2007), and all among older cohorts—have examined consequences of parental divorce homogamy for marital stability. Our study is the first to consider individuals' entire history of coresidential partnerships and to investigate the differences between cohabitations and marriages.
We add to the literature in three ways. First, we show that parental divorce on one or both sides of a couple not only elevates offspring's divorce risk but also drives parental divorce homogamy and additively compounds separation risk in both cohabitation and marriage. We were able to include all coresidential partnerships regardless of marital status, which is rarely possible given that data on cohabitations and parents of cohabiting partners over long periods are usually unavailable. The mechanisms transmitting union dissolution, therefore, also operate and are reinforced along different stages of the partnership history preceding union dissolution. Because children of divorced parents have an increased likelihood of partnering with each other, the elevated separation risk when both partners’ parents are divorced is relevant for a larger share of the population. Therefore, it is important to study intergenerational divorce transmission on the couple level, instead of focusing on the parent–child dyad of one partner only.
Second, together with previous studies, our findings contextualize the role of parental divorce homogamy for offspring's family formation and union dissolution. In Finland, a forerunner of the second demographic transition, cohabitation and separation are common, widely socially accepted, and not economically disastrous, especially for the younger cohorts in our data. Wolfinger (2003) found that, in the United States, parental divorce increased the likelihood of choosing a partner whose parents are also divorced by 58% without controls and by 31% when including a broad set of controls. In contrast, in our analyses, which included a relatively narrow set of controls, we found just a 13% and a 17% increase in the likelihood of choosing a partner whose parents are also divorced for cohabitation and marriage, respectively. The size of the coefficient for parental divorce homogamy in marriage in Finland roughly corresponds to half of that estimated for the United States. Thus, our findings are in line with weaker assortative mating and less-negative selection into parental divorcehomogamy in low-inequality contexts, such as Finland, compared with high-inequality contexts, such as the United States. Moreover, contrary to previous studies, we found a merely additive (double)—and not multiplicative (triple)—increase in the coefficient for parental divorce on both sides of a couple for their separation risk from both cohabitations and marriages.
Previous studies have interpreted the multiplicative association as deriving from poor interpersonal skills that cumulatively spiral into conflict and separation. While we could not directly measure interpersonal skills, our findings for Finland do not support such an interpretation. Rather, our findings are consistent with the interpretation that dual parental divorce primarily lowers thresholds for divorce for both partners, rather than cumulatively increasing interpersonal conflict. This is in line with the assumption that the transmission of prodivorce attitudes and lower thresholds for divorce when having observed it in one's own parents, as well as weaker and less negatively selected parental divorce homogamy, is a salient mechanism connecting parental divorce homogamy to offspring's separation in Finland. Associations between both partners’ parental divorce and offspring's union dissolution are likely stronger in countries where cohabitation and separation are more stigmatized, marriage is idealized and symbolically loaded, the welfare state does not effectively buffer its socioeconomic consequences, and assortative mating on parental divorce is stronger and more negatively selected. Unfortunately, given the data at hand, we cannot clearly specify the relative importance of each of these factors in contributing to the lower associations found between parental divorce homogamy and offspring's union dissolution in Finland. This remains an important task for future research.
Third, to our knowledge, we present the first study that compares the role of parental divorce from both sides of a couple on dissolution risks in cohabiting and married relationships. The examination of cohabitation is gaining importance as it proliferates as a substitute—not only a prelude—for marriage and is associated with lower socioeconomic standing and higher baseline union dissolution risks (Jalovaara 2013). If dual parental divorce is concentrated among cohabiting couples, elevating their separation risk, this could additionally reinforce cohabiting couples’ socioeconomic disadvantages relative to married couples. However, this is not supported by our findings. Instead, we show that both parental divorce homogamy and divorce transmission from one or both sides of a couple are stronger in marriages than in cohabitations. It seems that the same event—divorce from marriage—is more strongly transmitted across generations. The stronger normative signal of commitment attached to marriage likely creates a higher threshold for divorce—in addition to the legal and economic burdens associated with divorce—than for dissolving a cohabiting union. Further, having observed one's parents’ divorce might encourage offspring to divorce rather than stay in an unsatisfactory partnership. Separation risks from cohabitation are generally higher, and other factors appear to be more influential than parental divorce, although parental divorce notably increases separation risks for cohabiting unions. Indeed, we would miss crucial information about the separation of coresidential unions due to dual parental divorce if we focused only on offspring's divorce from marriage, as has been done in previous studies (Amato 1996; Storksen et al. 2007; Wolfinger 2003). More generally, the lower separation risk from cohabitation than from marriage associated with parental divorce homogamy is consistent with our interpretation of its additive rather than multiplicative effects in Finland: parental divorce seems to primarily lower thresholds for separation—which are generally higher in marriage than cohabitation—rather than increase interpersonal conflict. Heightened interpersonal conflict associated with parental divorce homogamy could be expected to multiplicatively increase separation risks from both marriage and cohabitation, which is not supported by our estimates for Finland.
Our findings must be interpreted in light of several limitations. First, while the register data allowed us to include representative information on parental divorce on both sides of the couple and to reconstruct entire partnership histories, including all cohabiting unions on a yearly basis over a long period, the data contain limited information to disentangle potential mechanisms that drive the intergenerational transmission of divorce. In contrast to survey data, register data do not include information on prodivorce attitudes and values, interpersonal behavior, or relationship quality. Next to the mechanisms considered so far, recent evidence (Gager et al. 2016) suggests that it is not the parental divorce per se (i.e., change in family structure) that increases offspring dissolution risk, but rather parental conflict and the poor relationship quality that preceded the divorce. Gager and colleagues (2016) show that parental conflicts increase offspring separation risks irrespective of parental divorce. With register data, we could not test these mechanisms directly, but we did provide some insights by theoretically considering which mechanisms would plausibly lead to additive or multiplicative effects of parental divorce homogamy. We can assess population-level effect sizes with particularly high precision and reliability with the register data and hope that our theoretical considerations—while partly speculative—will be useful in informing future research.
Second, we could not statistically compare our data on the weaker association between parental divorce homogamy on own-union dissolution risk in Finland to previous studies or other countries, which would have required merging the respective national samples. Still, if the U.S. and Norwegian data are reasonably representative, the weaker association found in our study strongly suggests that intergenerational divorce transmission is weaker at various stages of family formation than for the older cohorts studied in the United States and Norway. This weaker association for Finland relative to the United States is in line with the cross-nationally comparative divorce literature suggesting that intergenerational continuity in union dissolution is lower in contexts where separation is more widespread, is less socially stigmatized, and has less severe socioeconomic consequences (Dronkers and Härkönen 2008; Kalmijn and Uunk 2007). Against this backdrop, the relatively strong association found by Storksen et al. (2007) for Norway is somewhat puzzling. One reason for different findings for Finland may be that our data represent younger cohorts among whom parental divorce is more common. This might explain why we found additive associations when previous studies in Norway found multiplicative ones. This would suggest an important role of normative change and the relative prevalence of divorce, in addition to structural conditions of separation entrenched in welfare state institutions. Another reason might be that our data are representative of the Finnish population, whereas the Norwegian data are for only one county. Harmonized cross-national longitudinal data that account for changes in family constellations of partners would allow for directly testing differences in the strength of associations with parental divorce homogamy, as well as estimating country and cohort interaction effects.
Third, our information on parental relationship histories was limited. We were unable to include parental separation from cohabitation, as cohabitations can be reliably identified in the Finnish registers only from 1987 onward. It is possible that offspring's separation risks from cohabitation are more strongly associated with parental separation than with divorce, if indeed transmission of the same demographic event is particularly strong. But cohabitations were less widespread among the parent generation and, therefore, possibly played only a minor role compared with younger generations. Moreover, our data do not allow us to precisely know the children's age when their parents divorced. Previous studies suggest that parental divorce early in the child's life is particularly consequential for later life outcomes (Amato 1996). Finally, because of our research design, we could observe only relatively early separations and divorces before ages 41–45. For Norway, Storksen et al. (2007) show that the association of parental divorce with offspring's divorce was highest within the first 10 years of offspring's marriages. These are arguably well-covered in our data, given an average first marriage age of about 32 and average age of divorce of 41 in Finland (Official Statistics of Finland 2015, 2018). Yet, the associations found in our study might differ for later-life and higher order union dissolutions. Future research is needed to assess how the timing of single or dual parental divorce in a child's life matters for their family formation and how these associations vary across countries. Finally, the results presented in this article should be interpreted as associations, and not causal relationships.
We conclude that not considering how parental divorce on both sides of a couple affects their broader family formation processes risks underestimating the consequences of parental divorce for demographic behavior and associated socioeconomic outcomes. Parental divorce homogamy additively increases separation risk to a greater extent in marriages than in cohabitations—even in a generous welfare state with liberal family values, such as Finland. Analyzing multigenerational dynamics of wider kinship and in-law networks as determinants of demographic behavior and socioeconomic outcomes (Kailaheimo-Lönnqvist et al. 2019; Mare 2011), as well as systematically assessing their cross-national variation, remains an important task for future research. Studies on the intergenerational transmission of family formation more broadly, including the timing and sequencing of union dynamics and fertility (Fasang and Raab 2014), would benefit from considering both partners’ parents’ family formation. Because family formation is negotiated within couples, both partners’ parents likely matter for all of these processes but are often relegated to unexplained components in regression analyses. For example, parental divorce homogamy could contribute to rising divorce rates (Wagner 2020), and parental divorce is linked with a lower probability of becoming a parent (Jalovaara 2013). Thus, parental divorce homogamy and its growing frequency might also reduce fertility.
Acknowledgments
This research was funded by the Academy of Finland (decisions 321264 for the NEFER project and 320162 for the INVEST Flagship), ECSR-Society (research exchange grant), University of Turku, Freie Universität Berlin, WZB Berlin Social Science Center, and Humboldt University of Berlin. Emanuela Struffolino's research was partly funded by the Support Network for Interdisciplinary Social Policy Research (FIS) of the German Federal Ministry of Labor and Social Affairs. We thank Statistics Finland for the permission (TK-53-507-12) to use their data. Warm thanks to those who have provided comments in the writing workshop in WZB 2020.
Notes
We do not study same-sex unions because the register data do not allow us to distinguish cohabiting couples from roommates, such as students who share a living facility to reduce expenses, which would be a serious problem in these age-groups.
Results available from the first author.
Results available from the first author.