More Careful or Less Marriageable? Parental Divorce, Spouse Selection and Entry into Marriage

Parental divorce is associated with a host of long-term outcomes, some of which are socio-economic (e.g., McLanahan and Sandefur 1994), others psycho-emotional (e.g., Amato 2000), and yet others family demographic (e.g., Wolfinger 2005). Regarding marital behavior, a consistent finding links parental break-up with a heightened risk that one's own marriage dissolves (Wolfinger 2005; Dronkers and Härkönen 2008). Another literature has analyzed the associations between parental divorce and entry into marriage, finding somewhat inconsistent results on whether children of divorce marry at higher or lower rates (Wolfinger 2003a).

Despite the literature on parental divorce and entry into marriage, there is hardly any research on whether parental divorce is related to whom people marry. The few exceptions are the recent studies by Wolfinger (2003b) and Teachman (2004). Wolfinger (2003b) found that children of divorce have a higher likelihood of marrying someone with a similar experience, while Teachman (2004) reported that children of divorce are more likely to marry someone with a low level of education.

In this article, we analyze the effects of parental divorce – we include both divorces and break-ups of parents'' consensual unions – on children's marriage to spouses with different educational qualifications in Finland. Whether one marries and whom one marries can have important consequences on economic and psychological well-being. Due to the positive association between educational attainment and well-being, marital matching according to educational levels has featured widely in the stratification literature (Kalmijn 1998; Blossfeld and Timm 2003). Apart from the clear associations between education, incomes and status, higher education of both partners is now also associated with more stable marriages in Finland (Jalovaara 2003) and in several other countries (Härkönen and Dronkers 2006). Spouse selection behavior can also help understand the now delayed entry into marriage of those with divorced parents (Wolfinger 2003b). It can be that children of divorce do not reject marriage as such, only certain kinds of partners; alternatively, it may be that children of divorce have limited access to partners with preferred characteristics.

The objective of our research is thus to contribute to the understanding of the effects of parental divorce on socioeconomic outcomes and family demographic behavior. To achieve this, in the next section we first draw on the literature on search in the marriage market and spouse selection according to education and on previous research on parental divorce and marital behavior to create hypotheses about the effects of parental divorce on spouse selection. We then test these hypotheses against event-history data on entry into marriage of 44,987 individuals, drawn from the Finnish population registers. Family formation behavior in Finland is similar to that in the other Nordic countries, with near universal cohabitation before marriage (Finnäs 1995; Jalovaara 2012), an early age of leaving the childhood home and late marriage and first childbearing (Andersson and Philipov 2002). Similar to the United States and many other countries, Finnish women are becoming increasingly similar to men in that their higher levels of attained education are now positively associated with entry into marriage (Finnäs 1995; Jalovaara 2012). There is little research on spouse selection behavior in Finland (see however, Pöntinen 1980). As to the effects of parental divorce on family demographic behavior, Dronkers and Härkönen (2008) reported the common finding of an intergenerational transmission of divorce.

Theoretical Background and Hypotheses

Why would parental divorce affect spouse selection by education? To form hypotheses, we combine theories of educational assortative mating with literature on the effects of parental divorce on marriage formation behavior. According to the former, marriage market candidates compete for partners with the highest resources or seek to marry those who are similar to themselves. The latter literature suggests that children of divorce differ either in their preferences or possibilities to marry. Before going to the hypotheses, we review these two strands of literature.

Search and Selection in the Marriage Market

Theories of marriage entry and spouse selection often build on search models similar to those in the job search literature (cf., Oppenheimer 1988; Pollak 2000; Blossfeld and Timm 2003). In such models, marriage entry and spouse selection are inseparable processes occurring in a marriage market with search costs and time limits, imperfect information, a given supply of potential candidates and importantly, two-sided search. A basic assumption in these models is that individuals looking for a marital partner search until they find a “suitable” (though not necessarily the optimal) spouse, who is willing to marry the ego. From this point of view, lower rates of marriage and postponement may occur either because of continued search in hopes of (or trust in) finding a better match (e.g., Boulier and Rosenzweig 1984), or because of rejections from the candidates one finds interesting. This general framework will guide our discussion.

Homogamy according to education is a well-established empirical finding of long-lasting interest to demographers and stratification researchers (Mare 1991; Kalmijn 1994, 1998; Schwartz and Mare 2005; Rosenfeld 2008). An important explanation of this pattern points to shared institutional contexts – such as schools – where people meet their future spouses. In addition, scholars have sought explanation in the preferences of candidates in the marriage market, and in particular pointed to two forces: competition for marital partners with the highest level of education and a preference to marry someone similar to oneself (Kalmijn 1998).

According to resource competition theories, people prefer to marry someone with as good (economic) resources as possible. With two-sided search, those with high levels of education have better chances of marrying others with high levels of education; as a result, those with the highest levels of education marry each other while those with less education are left to compete among themselves (Kalmijn 1998; Rosenfeld 2005). In the earlier literature, this competition was often regarded as gendered – mainly women competed for economically resourceful husbands, while men valued other traits, such as homemaking skills and looks – but recent studies increasingly argue that this has become more symmetrical as a result of changes in gender roles and increases in women's labor force participation and earning power (Oppenheimer 1988, 1997; Sweeney 2002; Sweeney and Cancian 2004; Schwartz and Mare 2005). Similar results are found in Finland, which has a long history of female labor force participation and where women's educational attainment levels are high (cf., Jalovaara 2012).

Alternatively, instead of seeking to marry as highly educated others as possible, individuals may wish to marry those with educations similar to their own. Reasons for such preference include matching the tastes, values and lifestyles associated with education (Kalmijn 1994,1998). Kalmijn (1994) argued that similarity in these characteristics is more important than competition for economically resourceful partners. Therefore, like marries like – not as an indirect outcome of two-sided competition for high status and resources – but because like prefers like. Further evidence for this matching hypothesis has come from studies reporting positive assortative mating by education even when education has minor economic payoffs for married women (Boulier and Rosenzweig 1985), research on online dating (Skopek, Schulz and Blossfeld 2011) and results showing that homogamy improves marital stability (Blossfeld and Müller 2002). Relevant to our research, Wolfinger (2003b) also showed an increased likelihood that children of divorce marry people with similar experiences.

Parental Divorce and Marital Behavior

Maybe the best-established relationship between parental divorce and own family demographic behavior concerns the intergenerational transmission of divorce (e.g., Amato 1996; Diekmann and Engelhardt 1999; Dronkers and Härkönen 2008; Glenn and Kramer 1987; Wolfinger 1999, 2005). Explanations of the effects of parental divorce – instead of pure selection – that have gained the most support refer to problematic interpersonal skills and behavioral patterns and lower marital commitment found among the children of divorce (Dronkers and Härkönen 2008; Wolfinger 2005). For example, Webster, Orbuch and House (1995) found that children of divorce were more negative about the state and future of their marriages and were more likely to engage in behaviors that undermine marital stability than those from intact families. Similarly, Amato (1996) found a strong association between parental divorce and problematic interpersonal skills. As further evidence for lower marital commitment among children of divorce, Amato and DeBoer (2001) reported how those whose parents ended a low-conflict marriage were particularly likely to divorce. In other words, children of divorce are more likely to be socialized into thinking that divorce is a solution to marital difficulties (see also Glenn and Kramer 1987; Wolfinger 2005; Dronkers and Härkönen 2008).

These traits, attitudes and behaviors are present before marriage. Jacquet and Surra (2001) reported that young women whose parents divorced reported less trust and satisfaction in, and more ambivalence about, their romantic relationships (Prokic and Dronkers 2009). There is also abundant evidence linking parental divorce to higher average incidence of emotional and behavioral problems (e.g., Amato 2000; Amato and Cheadle 2008).

Some studies link parental divorce with more negative views towards marriage (e.g., Axinn and Thornton 1996). However, this result has not been universally supported (e.g., Amato 1988; Trent and South 1992; Tasker and Richards 1994). According to Amato (1988), for example, children of divorce value marriage but are also more aware of its limitations. This may lead them to be more careful in spouse selection.

These findings lead us to expect that children of divorce have lower rates of marriage, either because they are less willing to marry or they have a harder time attracting candidates willing to marry them. However, there are also reasons to expect that children of divorce have a higher rate of marrying. Children of divorce generally “grow up faster” (Weiss 1979) and get romantically involved and enter co-residential unions at an earlier age (Cherlin, Kiernan and Chase-Lansdale 1995; Kiernan and Hobcraft 1997; Wolfinger 2005). Their home environments can be economically and socially unpleasant and they might even have “inner neediness leading them to seek out romantic involvement.”(Wolfinger 2003a:340)

The evidence has been somewhat contradictory, some studies showing higher marriage rates, others showing lower marriage rates and yet others finding no difference between the children of divorce and those who grew up with both parents. Wolfinger (2003a) furthermore reported that this relationship has changed over time in the United States: in earlier cohorts, children of divorce were marrying at higher rates, whereas in the later ones they were marrying at lower rates (with the exception of a higher entry rate to teenage marriage). One possible explanation to this finding is the increasingly acceptable alternative of cohabitation, which children of divorce may be especially willing to engage in (ibid.; Axinn and Thornton 1996; Ongaro and Mazzuco 2009). Given the comparatively weak economic effects of divorce (Aassve et al. 2007) and all but universal premarital cohabitation in Finland, we expect that parental divorce is related to lower rates of entering marriage in Finland (Hypothesis 1).

Furthermore, we expect that these lower rates of entry into marriage are not universal, in the sense that we expect that children of divorce have a lower rate of marrying others with certain characteristics, rather than having a lower overall marriage rate. Next, we present our hypotheses of these processes. Following our search theoretical framework and empirical approach, all our hypotheses refer to the rates of marrying someone with given characteristics at a given timepoint, relative to remaining single (and potentially continuing search).

Parental Divorce and Spouse Selection: Hypotheses

We draw from findings that children of divorce have, on average, characteristics – such as interpersonal skill deficits, less trust and lower commitment in partnerships, or socioeconomic family backgrounds (Ermisch, Francesconi and Siedler 2006) – that can make them less marriageable. In this scenario, they can be disadvantaged in a competition for highly educated partners, who may (and can) exercise more discretion in spouse selection and avoid those from divorced backgrounds (Wolfinger 2005). This, in turn may help explain Wolfinger's (2003b) finding that college graduates were less likely to marry children of divorce. This would lead us to expect that children of divorce have lower rates of marrying people with high education – independently of their own level of education (Hypothesis 2).¹

This hypothesis stresses the other as the decision maker – that preferred candidates are unwilling to marry children of divorce. Those with divorced parents are obviously active decision makers themselves. Our discussion of the higher awareness of the fragility and other limitations of marriage held by children of divorce may lead them to be more careful in choosing their mates; instead of rejecting marriage as such, they might be more prone to reject only some kinds of candidates. Or in other words, their generally lower willingness to enter marriage can be trumped by a particularly fitting potential partner. If matching according to similar characteristics drives spouse selection, children of divorce might avoid candidates who differ from themselves – and thus have lower rates of entering an educationally heterogamous marriage (Hypothesis 3).² Alternatively, in the case of competition for economic resources in the marriage market, this may show up as a general tendency for children of divorce to avoid those with lower educational qualifications (Hypothesis 4).

Data and Variables

Our data come from the Finnish Census Panel, provided by Statistics Finland. The data are formed from national population registries and are of very high quality. They are based on a sample drawn from the 1970 census, and are extended to cover all members of the household. Each family member is followed to next censuses, and new family members are included in the follow-up. Our sample consists of 44,987 persons born in Finland from 1970 through 1980. We have data on their demographic and socio-economic living conditions from the years 1970, 1975, 1980, 1985 and 1987, after which we have annual records until 2007. Therefore, while we need to resort to the quintannual data to construct the measure on parental divorce, we can use annual measures to construct event-history data for entry into first marriage starting from age 18 to the maximum observed age of 37.³

Our dependent variable indicates whether the individual was married or not at a given age, and if yes, the level of educational attainment of the spouse. We focus on marriages instead of cohabitations. As mentioned, premarital cohabitation is nearly universal in Finland (Jalovaara 2012), and of the marrying couples in this study 81 percent were observed to cohabit at the end of the year prior to marriage. One could thus analyze entry into cohabitation as the “first step” in family formation. However, cohabitating couples are very heterogeneous, particularly at the beginning of cohabitation, ranging from “dating with shared housing” to more serious and committed couples, and our data do not distinguish between these levels of commitment. The couples are also still in the process of obtaining their educations, which makes determining their (future) educational level difficult. Furthermore, marriage continues to send a stronger signal of mutual commitment and a common future. We nonetheless did ancillary analyses for entry into cohabitation and comment on the results in the discussion section. We excluded marriages to partners who were more than 15 years older or younger than the index person, and in the event-history analyses these marriages were censored.

Our dependent variable has five categories: not married and four categories of completed education. We use a modified version of the internationally comparable CASMIN scheme (Braun and Müller 1997). This scheme differentiates not only between levels of education, but also types of education. This makes sense in an educational system such as the Finnish one, where students are tracked into different paths around age 16 (general/academic and vocational).

We differentiate between four levels of education of the respondent and the spouse. All Finnish children take nine years of compulsory education (Level 1: Compulsory schooling), after which they can choose either the vocational track (Level 2a + b: Intermediate secondary) or the academic high school track (Level 2c: Higher secondary), which is the normal route to university education. High school graduates can continue either to further secondary, usually vocational, education (Level 2c: Higher secondary) or tertiary education either in universities or polytechnic universities (Level 3ab: Tertiary).

These four educational groups are also used to measure the educational attainment levels of mothers. Mothers with only compulsory schooling (Level 1) are a somewhat heterogeneous group, with some having only 6 years (kansakoulu), while others had 9 to 12 years of schooling (alakoulu 4-6 years plus keskikoulu 5-6 years), which was the track to high school and further education.

Experience of parental divorce before age 17 is our main explanatory variable. We include information both from divorces and separations, but we excluded those born to unpartnered mothers.⁴

We capture the age pattern of entry into marriage by including dummy variables for each age in which the individual was at risk of first marriage into our models, with the modal age of marriage at 28 as our reference category.⁵ Our control variables can be divided into two broadly defined groups: social background variables and educational attainment and life-course variables. The background variables are year of birth, gender, mother's education at age 16 and mother's age at birth.⁶ Women enter marriage faster than men, and there can be gender differences in spouse selection. Parental background is related both to the experience of divorce (Härkönen and Dronkers 2006) and to marital behavior (Finnäs 1995). The same can hold for the mother's age at birth of the child: younger mothers have a higher risk of marital dissolution and age of the mother can affect outcomes in adulthood (e.g., Kalmijn and Kraaykamp 2004). We also include a measure of cohabitation as the time in years (linear and squared) one cohabits. Unfortunately, we lack information about the possible parental divorce or separation of the partners.

In addition to the four-class education variable, we include a dummy variable indicating whether one is still in education and a linear measure of the years after finishing education (zero if in education). These account for the facts that marriage is less likely while in education (Blossfeld and Huinink 1991) and that the longer people have been out of education, the higher their chances of meeting marital candidates with different educational levels (for example at work) (Blossfeld and Timm 2003).

To analyze entry into first marriage, we transformed the data into discrete-time event history form with age as our time unit (Yamaguchi 1991). The discrete-time specification enables us to better handle time-varying covariates and ties in the data. Due to the large number of cases (44,987 persons; 521,910 person-years), we do not experience a notable loss in efficiency due to the discrete-time form of the data.

Table 1 presents descriptive information on some of the variables by parental divorce. Approximately 20 percent of our cases experienced parental divorce during their childhood. Those from broken homes had, on average, lower levels of own, mother's and spouse's education (if married), and they were born, on average, a month later. Their mothers were younger at the time of their births. Children of divorce were less likely to be married, and if they married, they married at a somewhat older age than those from intact families. The partners had lower levels of education than the index persons in both groups. This is because partner information is, of course, available only for those who married, and because the younger and better educated cohorts had less time to marry (and were in general marrying later). The observed partners were also, on average, approximately one year older than the index persons (not shown in Table 1). This was due to the same reasons as the educational difference, and also due to the fact that women (who marry older spouses and at a younger age) were more likely to have married in our cohort data. However, there was no difference in this age gap between children of divorce and those from intact families.

Method

Event-history analysis is commonly used to analyze transitions from one discrete state to another, in our case, entry into first marriage. Given that this method takes into account the time to the event, it is well-suited for our purposes since our theoretical framework stresses search in the marriage market, where people can extend or have to extend search for a spouse.

x are individual-level covariates, such as educational attainment and experience of parental divorce, which we would find from a multinominal logistic regression model for competing risks event-history data. The conditional model differs from these more familiar models because of the alternative-specific covariate z, which is often interpreted as giving the “value” of alternative k for the individual. In our case, z is the educational level of the spouse, conditional on that of the ego (for example, the same level of education). For example, an odds ratio above one for the homogamy parameter demonstrates that people have a higher probability of entering marriage with someone with the same level of education. Similar models are used in social mobility research (Breen 1994; Dessens et al. 2003), but have not, to our knowledge, been used in research on entry into marriage.

Given the size of our data, even weak effects become easily significant at the conventional levels. Therefore, we report significance levels at 1 percent and .1 percent. The estimates from these models are descriptive in the sense that they cannot be interpreted as causal effects. Instead, they show the difference between children of divorce and those who grew up in intact families, controlling for other family background characteristics, and education and other life course variables of the individual. Therefore, our estimates do not distinguish between the effects of possible family conflicts that lead to divorce from the divorce itself. However, both processes should affect (albeit maybe differently) the marital behavior of the offspring.⁷

Results

Parental Divorce, Entry into Marriage and Education of the Spouse

We begin our analyses with descriptive analyses of marriage rates by parental divorce, both overall and according to the educational level of the partner. Firstly, Figure 1 shows the cumulative probability of entry into marriage. While in our sample almost 60 percent of those from intact families had married by age 37, approximately 50 percent of the children of divorce had done so. This result gives preliminary support for our first hypothesis.

Figure 2 shows the cumulative incidence rates (Coviello and Boggess 2004) of marrying spouses with different levels of education by parental divorce. These figures clearly demonstrate how parental divorce is not only associated with lower overall rates of marriage as shown by Figure 1, but even more clearly with lower rates of marrying people with higher education. Children of divorce are more likely to marry someone with only compulsory education, and there are no differences in marrying partners with intermediate secondary qualifications. However, those with divorced parents have clearly lower rates of marrying partners with higher secondary education and in particular, partners with a tertiary degree. These different rates remain throughout the ages covered so that whereas 17 percent of those from intact families had married someone holding a tertiary degree by age 37, only 10 percent of those with divorced parents had done so. The corresponding figures for marrying a partner with higher secondary education were 14 percent and 10 percent, respectively. These results support our second hypothesis and suggest that the lower marriage rates among the children of divorce are actually lower rates of marrying certain people, in this case, those with higher educational qualifications.

But do these results remain after controlling for additional covariates? Table 2, column 1 presents the results from a (binomial) discrete-time, event-history model of entry into marriage and corresponds to the descriptive analysis in Figure 1. As can be seen from the estimate for parental divorce, children of divorce have lower marriage rates even after adjusting for the control variables. Our estimate of a 15 percent lower marriage rate is, in fact, nearly identical to Wolfinger's (2003a) estimate of the effects of parental divorce on entry into marriage in later periods in the United States, and similar to recent findings from Norway (Storksen et al. 2007). The other estimates show that men, those who were born in later cohorts, born to mothers without higher education, who were in education, who had a lower level of education and who had a longer time since graduation have lower rates of marriage relative to the reference groups. The duration of cohabitation increases the odds of entry into marriage, but with a concave slope. Regarding age, our dummy variable specification shows a hump-shaped hazard of entry into marriage, which peaks around ages 26-28 (not shown) (cf., Jalovaara 2012).

We next turn to the main finding of the descriptive analysis: the lower marriage rates to better educated partners among children of divorce. The results from the competing-risks, discrete-time event-history analysis are presented in columns 2 to 5 of Table 2. Our results continue to support the second hypothesis: parental divorce is associated less with lower marriage rates as such than with lower rates of marrying others with higher levels of education. According to these estimates, the higher marriage rates to partners with compulsory education only are not statistically significant. However, children of divorce have lower rates of marrying others with post-compulsory education. Children of divorce have a 13 percent lower rate of marrying others with intermediate secondary education, compared to those from intact families. The differences in marriage rates grow with the educational level of the spouse: children of divorce have an average 20 percent lower rate of marrying those with higher secondary education and a 28 percent lower rate of marrying someone with tertiary education.

In analyses not reported here we used likelihood ratio tests to check for three types of interactions: between parental divorce and sex, parental divorce and own education and parental divorce and the age dummies. Firstly, parental divorce may have different effects on boys and girls. For example, Ermisch, Francesconi and Pevalin (2004) found that growing up in a non-intact family had greater effects on boys'' than girls'' educational and psychological outcomes. Furthermore, many theories of entry into marriage hold that it is particularly women who compete for the status and resources of the husband (cf., Oppenheimer 1988). We did not find any evidence that parental divorce affected spouse selection differently for women than for men. Secondly, if the effects of parental divorce on entry into marriage by the educational level of the spouse vary by own educational attainment, simply adjusting for the differences in own educational attainment levels by parental divorce may not be sufficient. These interactions were neither jointly nor separately significant. However, there is an exception in which the effect of parental divorce seems to depend on own educational level, which will be discussed in the next section. Finally, we tested whether the effect of parental divorce varies according to age. Wolfinger (2003a) found that children of divorce were more likely to marry young (below age 20), but in more recent cohorts less likely to marry in their 20s or at older ages. We did not find that the effect of parental divorce varied by age.

Summing up, both our descriptive and our regression results support the second hypothesis which predicted that children of divorce have lower rates of entering marriage with people who have higher educational qualifications. This result could not be explained by different family backgrounds nor by the lower educational credentials of the children of divorce. In fact, the effect of parental divorce was overall rather insensitive to inclusion of the control variables. In particular, the negative effect on marrying graduates with a degree from tertiary institutions remained very robust.

Parental Divorce and Homogamy/Heterogamy

The previous analyses showed broad support for our first and second hypotheses: children of divorce have lower marriage rates in general, and in particular to partners with higher educational qualifications. These hypotheses were motivated by the theory that search in the marriage market is primarily driven by a preference to marry someone with the highest possible (economic) resources.

These analyses did not, however, take into account the possibility that parental divorce might also affect the rates of entering educationally homogamous or heterogamous marriages (Hypothesis 3). Our analyses that apply conditional logit models for discrete-time event-history data take this into account.

We fitted four separate models (A to D) with homogamy/heterogamy parameters and their interactions with parental divorce. Table 3 shows the results for these models. The figures in the first four rows of the table show the effects of parental divorce on marrying spouses with different levels of education, just as in Table 2. The next rows show how parental divorce affects different types of homogamous or heterogamous marriage. In Model A, we analyzed the effects of parental divorce on marrying upward, that is, a spouse with higher education than oneself. The relevant parameter estimate shows no effect. Neither do we find effects of parental divorce on downward marriage (Model B), or on marrying homogamously (Model C). Finally, we analyze whether parental divorce affects homogamous marriage depending on one's educational attainment level, following earlier findings of particularly high levels of homogamy among the highly educated (e.g., Mare 1991; Schwartz and Mare 2005; Rosenfeld 2008).

In figures not shown here, we too find particularly high rates of homogamous marriage among the highly educated (tertiary education approximately doubles the odds of homogamy). But our interest is mainly in the effects of parental divorce. Results from Model D show that although parental divorce does not affect rates of homogamous marriage among most, it does further increase the rates of homogamous marriage among the tertiary educated.⁸ This result was not predicted by our hypotheses. It is worth noting that in Model D, the baseline effect of parental divorce on marrying highly educated partners became more negative, compared to the estimates of the competing risks model shown in Table 2. Parental divorce lowers the rate of marrying highly educated partners by more than 40 percent among those with less than tertiary education. However, the higher rates of homogamy partly offset this negative effect of parental divorce among those who themselves have tertiary degrees. Therefore, the negative effects of parental divorce on marrying highly educated spouses are weaker among those with tertiary degrees.

Discussion and Conclusions

Parental divorce can have various long-term consequences. In this research, we used Finnish population register data to analyze whether parental divorce is associated with whom – in terms of educational attainment – people marry. Despite several studies on the effects of parental divorce on own marriage and divorce (e.g., Wolfinger 2003a; 2005; Dronkers and Härkönen 2008), there is very little research on parental divorce and spouse selection behavior. Whether and whom one marries can, nevertheless, have major effects on one's well-being. Education is a key socioeconomic resource, and spouse selection according to educational attainment has an important effect on socioeconomic (Schwartz and Mare 2005; Ermisch, Francesconi and Siedler 2006) and family outcomes (e.g., Jalovaara 2003; Härkönen and Dronkers 2006). We used a search theoretical framework combined with theories of assortative mating by education and of the effects of parental divorce on marital behavior to formulate hypotheses of how parental divorce affects rates of marrying spouses with different levels of educational attainment.

The clearest result of our competing-risks, discrete-time event-history analysis is that children of divorce have lower rates of marrying others with high levels of education, and with a university degree in particular. This finding is not due to lower levels of education or the social background of the children of divorce. This result is in line with our second hypothesis, which predicted that children of divorce carry with them characteristics – such as lower interpersonal skills and lower commitment to intimate relationships – that make them disadvantaged in the marriage market. Because of this, marital candidates with higher levels of education – who, according to the resource competition theory of search in the marriage market have more demand and can therefore continue search for more suitable spouses – may be less likely to accept marriage proposals from children of divorce.

Our results are not in line with our alternative hypotheses, which predicted that children of divorce are more careful in the marriage market due to their first-hand experience of the fragility of marriage. In such a case, we expected that children of divorce would be less likely to marry someone with low levels of education or different levels of education than themselves (such partner candidates could be seen as less suitable than those with high or the same levels of education). We did, however, find one unexpected result: children of divorce with a tertiary degree have an increased tendency of marrying homogamously. As a result, tertiary education partly compensates for the negative effects of parental divorce in access to highly educated partners, and the gap between children of divorce and those from intact families becomes smaller. One possible explanation is that highly educated children of divorce place particular emphasis on homogamous marriage, for example as a more secure way to guarantee benefits from those similar tastes and lifestyles that are often found among those with corresponding levels of schooling. This can be facilitated by the marriage market for the highly educated, in which the importance of educational institutions as the main context for meeting and choosing partners is more important than other, non-educational, characteristics such as family-history (including parental divorce) and ethnic background (Qian 1997; Kalmijn 1998). Either or both of these mechanisms can then result in an increased tendency for homogamy among children of divorce with tertiary education compared with children of divorce with less education (see Qian (1997) on ethnic intermarriage and educational homogamy).

Our results give detailed information on recent results showing that children of divorce on average postpone or even forego marriage (Wolfinger 2003a), a result that we also replicated (in line with our first hypothesis). Instead of having lower rates of marriage as such, children of divorce have lower rates of marriage only to certain groups, which in our case were defined by educational attainment. The result suggests that one of the intergenerational consequences of divorce is lower access to highly educated spouses, especially for those without high education themselves. This can have important effects on the economic level of living and marital life of the children of divorce. These results are among the first to document a link between parental divorce and characteristics of one's spouse.

Despite the importance of our findings, they carry some limitations. Our focus on marriage – as compared to all unions, including cohabitations – may miss an important part of current family formation processes particularly in a country such as Finland. While this critique is partly valid, marriage continues to send a stronger signal of commitment to a relationship. Many consensual unions begin as “extended dating,” and a large share or first cohabitations are dissolved within the first years (Jalovaara 2011). They cannot thus be seen as equal to marriage when analyzing union formations. In additional analyses (results available upon request) we ran the same models for entry into first cohabitation. In general, parental divorce played a smaller role in entries into cohabitation. Children of divorce even had higher rates of starting a consensual union with someone with the lowest level of education. Different results for entry into marriage and cohabitation have been found in other studies of the effects of parental divorce (e.g., Cherlin, Kiernan and Chase-Lansdale 1995; Kiernan and Hobcraft 1997; Wolfinger 2005). The differences to our main results can reflect lower conversion rates into marriage or higher break-ups from cohabitations. They support other findings of the role of resources in entry into marriage and cohabitation (Jalovaara 2012) and the relatively lower rates of forming marriages compared to consensual unions among the children of divorce (Wolfinger 2005).

We are not able to control for more than a limited set of family background variables. Our estimates cannot, therefore, be given a causal interpretation. One consequence of this is that we cannot distinguish between parental divorce as an event (parents breaking up) from parental divorce as a process (including the possible conflicts preceding and following the breakup). However, since its intergenerational effects are a function of both, we nevertheless consider our results a valuable addition to the existing knowledge.⁷

A final limitation concerns our limited ability to suggest specific factors – such as psychological or behavioral traits – that could explain our results. Neither do we observe the family background of the spouse. If children of divorce are prone to marry each other (Wolfinger 2003b) and if they have lower average rates of education, this may lead to lower marriage rates to better educated spouses as an indirect effect of family background homogamy. This possible explanation would – in contrast to our interpretation of our results – point to the children of divorce as the active decision-makers that shape their marriage patterns. We leave exploration of this and other relevant explanations for further research that will hopefully aim to replicate our results in other societies. But no matter what the explanation, our main conclusion remains. Parental divorce can have longstanding effects, even in Finland where it is common, socially accepted and has relatively weak economic consequences.

Notes

References