A growing share of children live apart from one of their parents before reaching adulthood. Many policymakers are concerned about the welfare of these children who (partly) grow up in single-parent households. Numerous papers in various social science disciplines document a strong negative empirical association between parental divorce and a wide range of children’s outcomes. This general relationship is highly persistent, leaving the children of divorced parents economically and emotionally worse off, even in adulthood. Most scholars are aware that it is not clear to what degree this relationship is causal (see, e.g., Manski et al 1992, Painter and Levine 2000, Amato 2010, Bhrolcháin 2013, Gähler and Palmtag 2015). A number of confounding factors that provoke parental divorce – for example, emotional stress or parenting disputes – may also be detrimental to children’s outcomes.
In a new paper, we analyse various outcomes for children who experienced parental divorce (Frimmel et al. 2016). To answer the question of whether children are causally affected by parental divorce, exogenous variation in the divorce likelihood is required. The construction of a valid empirical counterfactual is not only necessary for empirical identification, but also essential for identifying the causal channels through which children are affected. If one would use child outcomes emerging from a stable and healthy family background as a benchmark, one would clearly expect a negative effect of divorce, which could operate through multiple channels. Probably a more relevant counterfactual situation is a family background characterised by (at least temporary) parental conflicts. In such a situation, children may even benefit from divorce, if the post-divorce situation is comparably more beneficial than growing up in a two-parent household fraught with conflicts.
Existing evidence is hard to interpret because most of the literature doesn’t sufficiently define the counterfactual situation (which is implicitly presumed in any analysis), nor offer a convincing research design. McLanahan et al. (2013) provide a comprehensive survey of this literature. In our paper, we offer a design-based approach. We suggest exploiting idiosyncratic variation in the extent of sexual integration in fathers’ workplaces within an instrumental variables (IV) approach to establish a causal effect.
McKinnish (2004, 2007) and Svarer (2007) show that individuals who have workplaces with a larger fraction of co-workers of the opposite sex are significantly more likely to divorce later. This empirical finding is in line with the economic model of marriage and divorce (Becker et al. 1977), which stresses imperfect information at the time of marriage and the acquisition of new information while married as key determinants of divorce. In particular, new information regarding alternative outside options – that is, extramarital relationships – is decisive. Sexually-integrated workplaces reduce the cost of extramarital search and allow married individuals to meet alternative mates, which increases the likelihood of divorce. Thus, we aim to identify the causal effect of divorce for the child whose father left the family because he met a new partner at work. We argue that this research design evaluates a realistic divorce scenario and offers a well-balanced relationship between internal and external validity.
To assess the long-run effect of divorce, we analyse children's human capital and demographic outcomes. First, we examine college attendance. In Austria, college attendance implies that this person graduated from a higher secondary school. Second, we check the labour market status (employed; unemployed; out-of-labour force) up to the age of 25 years. Third, we examine children's own family formation behaviour (i.e. fertility and marriage). Finally, we investigate the probability of early mortality (below 25 years of age). Our results show that parental divorce – due to a high level of sexual integration in fathers’ workplaces — has a negative effect on children’s long-term outcomes. Our main findings for human capital outcomes are summarised in Figure 1; our results for demographic outcomes are shown in Figure 2.
Figure 1. The effect of parental divorce on human capital outcomes
Notes: This figure summarises estimation results of the effect of parental divorce on human capital outcomes for boys (dark grey bars) and girls (light grey bars). Labour market outcomes are measured at the age of 25 years. Method of estimation is a Two-Stage Residual Inclusion procedure, which uses the extent of sexual integration in fathers' workplaces as an instrumental variable. Reported estimates are average marginal effects for divorce until age of 18, with standard errors clustered on families in parentheses below. Marginal employment is an employment contract for jobs with a low number of working hours and low pay and covers only accident insurance. This type of employment is, for instance, very common among college students who work while enrolled. LF stands for labour force. Further details are provided in Table 5 of Frimmel et al. (2016)
Figure 2. The effect of parental divorce on demographic outcomes
Notes: This figure summarises estimation results of the effect of parental divorce on demographic capital outcomes for boys (dark grey bars) and girls (light grey bars). Method of estimation is a Two-Stage Residual Inclusion procedure, which uses the extent of sexual integration in fathers' workplaces as an instrumental variable. Reported estimates are average marginal effects for divorce until age of 18, with standard errors clustered on families in parentheses below. Early marriage is a marriage below the age of 20 years. Early mortality is defined as a death below the age of 25 years. Further details are provided in Table 4 of Frimmel et al. (2016).
For both sexes, we find a substantially lower level of educational attainment – parental divorce reduces college attendance by about 9 to 10 percentage points. The effects on family formation, labour market, and health outcomes differ by sex. In the case of boys, we find little effect on their fertility or marriage behaviour. However, we find a higher likelihood of early mortality and worse labour market outcomes at 25 years of age. Figure 3 shows the pattern over time, and reveals that the negative employment effect kicks in at 22 years. In the case of girls, we find strong effects on their fertility behaviour. Parental divorce increases the likelihood of a pregnancy in the teenage years up to the early twenties. Most of these additional children are born out-of-wedlock; we find only very small treatment effects on the likelihood of early marriage. Regarding labour market outcomes there is some evidence for an increased employment probability for girls in their early twenties, which dissipates over time (see Figure 3). This effect could be a direct consequence of teenage motherhood, which may initiate early entry into the labour market.
Figure 3. The effect of parental divorce on employment over time
Notes: This figure summarises employment effects due to parental divorce for boys (upper panel) and girls (lower panel), estimated at different ages of the child separately. The empirical specification is equivalent to those of our standard empirical model presented in Tables 3 to 5 of Frimmel et al. (2016).
While the external validity of an analysis is, in general, hard to assess, our approach provides us with a treatment effect at a margin of broad interest. Our estimates inform us about the consequences of divorce in situations where the separation was triggered by the father meeting a new partner at work. We consider this type of divorce to be a realistic scenario and preventable, in principle. A small increase in the cost of divorce or in the benefit of the existing marriage — for instance, due to a change in divorce legislation or in the social approval of divorce — may avert some of these divorces. In contrast, divorces that result from more severe shocks (such as domestic violence) can and should not be averted.
Our findings are also consistent with expectations about possible pathways. After divorce children typically grow up in female-headed households, since maternal sole custody is the dominant arrangement. These households have lower incomes, tend to live in worse neighbourhoods, have fewer and weaker male role models, and access to smaller social networks. Moreover, affected children may suffer from the separation with their father, parental hostility, and residential and school dislocation (Painter and Levine 2000).
Our results also imply that the negative consequences of parental divorce on children’s long-term outcomes should ideally not only be internalised by parents, but also by policymakers who design policies affecting parents’ incentives to divorce, or programmes which support children from disrupted families.
Amato, P R (2010) “Research on divorce: Continuing trends and new developments”, Journal of Marriage and Family, 72(3): 650-666.
Becker, G S, E M Landes and R T Michael (1977) “An economic analysis of marital instability”, Journal of Political Economy, 85(6): 1141-1187.
Bhrolcháin, M N (2013) “Divorce effects' and causality in the social sciences”, European Sociological Review, 17(1): 33-57.
Frimmel, W, M Halla and R Winter-Ebmer (2016) “How does parental divorce affect children’s long-term outcomes?”, CEPR, Discussion Paper No11339.
Gähler, M and E Palmtag (2015) “Parental divorce, psychological well-being and educational attainment: Changed experience, unchanged effect among Swedes born 1892-1991”, Social Indicators Research, 123(2): 601-623.
Manski, C F, G D Sandefur, S McLanahan and D Powers (1992) “Alternative estimates of the effect of family structure during adolescence on high school graduation”, Journal of the American Statistical Association, 87(417): 25-37.
McKinnish, T G (2004) “Occupation, sex-integration, and divorce”, American Economic Review, Papers and Proceedings, 94(2): 322-325.
McKinnish, T G (2007) “Sexually integrated workplaces and divorce: Another form of on-the-job search”, Journal of Human Resources, 42(2): 331-352.
McLanahan, S, L Tach and D Schneider (2013) “The causal effects of father absence”, Annual Review of Sociology, 39: 399-427.
Painter, G and D I Levine (2000) “Family structure and youths' outcomes: Which correlations are causal?”, Journal of Human Resources, 35(3): 524-549.
Svarer, M (2007) “Working late: Do workplace sex ratios affect partnership formation and dissolution?”, Journal of Human Resources, 42(3): 582-595.