Generality and Variation Across Time and Space
Discussions of “diverging destinies” are similar to research on the broader patterns of family change comprising the second demographic transition in recognizing that cross-national differences in the pace and magnitude of family bifurcation may reflect differences in policy, demography, or social context. Scholars have focused on (a) evaluating the extent to which the general pattern of family bifurcation highlighted by McLanahan (2004) is observed across countries, (b) identifying family behaviors and countries for which the general pattern does not appear to hold, and (c) attempting to identify contextual factors that contribute to observed cross-national differences in the degree of bifurcation (and its change over time).
A number of cross-national studies have examined change over time in educational differences in family behavior. This work has examined the educational gradient in marriage timing and patterns of union formation (Kalmijn 2013; Perelli-Harris and Amos-Lyons 2014; Rendall et al. 2003, 2005, 2009, 2010), union dissolution (Harkonen and Dronkers 2006; Kalmijn 2013; Matysiak et al. 2014), childbearing among unpartnered women (Perelli-Harris et al. 2010), and early childbearing (Raymo et al. 2015). The results of these studies generally demonstrate a pattern of bifurcation in family behaviors. For example, Perelli-Harris and colleagues (2010) find that childbearing within cohabitation is more common at lower levels of educational attainment across Europe and Raymo and colleagues (2015) show that the negative educational gradient in early childbearing has become stronger over time in many countries and has become weaker in none.
In addition to these multi-country studies, a good deal of research has examined educational gradients in family behavior and their change over time in specific countries. Examples include documentation of the emergence or strengthening of a negative educational gradient in divorce in Korea (Park and Raymo 2013), the Netherlands (De Graaf and Kalmijn 2006), Sweden (Hoem 1997), and the U.K. (Chan and Halpin 2005), research showing an increase in the association between low education of mothers and children’s experience of parental separation in Sweden (Kennedy and Thomson 2010), and evidence of an emerging negative educational gradient in the likelihood of never marrying in the U.S. (Goldstein and Kenney 2001). McLanahan and Jacobsen’s (2015) summary of this research concludes that findings are generally consistent with the claim that growing differentials in family behavior are a universal characteristic of low-fertility countries.
However, it is certainly not the case that all findings are consistent with a pattern of diverging destinies. In some cases, the educational gradient in family behaviors associated with reduced resources is positive. For example, divorce is more common among highly educated women in Italy and other southern European countries (Harkonen and Dronkers 2006). In other cases, educational differences in family behavior are small (Matysiak et al. 2013) or are explained by country-level factors (Perelli-Harris and Amos-Lyons 2014). Finally, there are other studies that document a strong educational gradient in the family behavior of interest but find little evidence of change over time in educational differences. For example, Raymo et al. (2015) found no evidence of statistically meaningful change in the negative educational gradient in early childbirth in ten of the twenty countries they examined. Overall, existing empirical evidence is generally consistent with the pattern described by McLanahan (2004), but there is also an abundance of evidence pointing to the need for caution in evaluating the generality of a pattern of diverging destinies. Much of this cautionary evidence highlights the importance of contextual factors in shaping socioeconomic differences in family behavior and their change over time.
Our own reading of the cross-national evidence suggests the importance of four contextual factors. First, it seems safe to say that the U.S. is an extreme case, with educational divergence in family behavior much more pronounced than in most other countries. For example, Perelli-Harris and Amos-Lyons’ (2014) examination of educational differences in partnership trajectories across 14 countries found that the U.S. fits nicely with the expectations of “diverging destinies”—with trajectories characterized by union dissolution and long-term cohabitation concentrated at lower levels of education—but evidence for other countries is more mixed. Similarly, Carlson et al. (2014) found that the negative educational gradient in children’s exposure to unpartnered-parent family structure is far more pronounced in the U.S. than in the other 14 countries they studied. In addition to evidence that the U.S. is somewhat of an outlier, cross-national studies point to the importance of differences in public policy, the level of economic inequality, and the degree of gender inequality in shaping cross-national differences in pace and nature of socioeconomic bifurcation in family behavior.
The role of public policy is suggested by McLanahan’s (2004) focus on increasing means-testing and reductions in levels of public income support as an explanation for family bifurcation in the U.S. The argument is that educational differentials (and growth therein) should be less pronounced in countries with a strong public safety net because more generous, universal welfare policies reduce differences across educational categories in the costs or benefits associated with engaging in specific family behaviors. Broad support for this posited relationship can be found in a series of papers in which Rendall and colleagues examine socioeconomic differences in the timing of childbearing in a number of European countries (Rendall et al. 2003, 2005, 2009, 2010). Their analyses show an increasing concentration of relatively early childbearing among women with lower levels of education and occupational status in countries with less generous, means-tested welfare policies (i.e., English-speaking countries and Southern Europe) but declining educational differentials in age at first birth in countries with more generous, universal welfare policies (Northern and Western Europe). Similarly, the multilevel models estimated by Harkonen and Dronkers (2006) show that welfare generosity is associated with a smaller educational gradient in divorce.
Another contextual factor that appears to be related to the degree of family bifurcation is the degree of economic inequality. As noted above, the reciprocal linkages between inequality and family bifurcation have been emphasized in research on the U.S. (McLanahan and Percheski 2008). Here, the argument is that high levels of income inequality make it difficult for those at the bottom of the socioeconomic distribution to generate and maintain the economic resources necessary for marriage and for marital stability. Consistent with this emphasis, Kalmijn (2013) finds that measures of educational inequality (educational differences in unemployment and poverty) are associated with a stronger negative educational gradient in marital dissolution for men (but not for women).
One of the most intriguing findings from the cross-national research summarized above is that the negative educational gradient in family behaviors associated with reduced resources is stronger in more gender-inegalitarian countries. As discussed in the next chapter, this is a central focus in our expectations regarding educational differences in family behavior in Japan. Here, the basic argument is that policies or social structures facilitating work-family balance reduce the opportunity costs of family transitions that compete with employment (e.g., early childbearing, single parenthood). Stated differently, the greater the difficulty that women face in working while raising children, the more likely women with the highest earnings potential will be to engage in family behaviors that facilitate continued employment (e.g., delayed marriage, delayed childbearing, marriage to high-earning men, stable marriage). Evidence consistent with this emphasis includes relatively weak educational gradients in Northern European countries and stronger educational gradients in relatively gender-inegalitarian countries in Southern Europe (e.g., Perelli-Harris et al. 2010; Rendall et al. 2010). More direct evidence comes from multilevel models that include country-specific indicators of gender equality or support for work-family balance. For example, Kalmijn (2013) finds that highly educated women are more likely to be married and less likely to be divorced in more gender-egalitarian countries.
Most of the cross-national research on educational differences in family behavior is limited by its lack of attention to differences across countries in the nature and meaning of various family outcomes. For example, analyses of childbearing within cohabitation (e.g., Perelli-Harris et al. 2010) and analyses of patterns of union formation (e.g., Perelli-Harris and Amos-Lyons 2014) are complicated by the fact that the duration, nature, and selectivity of cohabiting unions are qualitatively different across countries. In the U.S., nonmarital childbearing is often to unpartnered mothers or to mothers in relatively unstable partnerships whereas most unmarried mothers in western and northern European countries are in stable, marriage-like cohabiting unions (Heuveline and Timberlake 2004; Kiernan 2001). Similarly, efforts to understand cross-national differences in early childbearing (Raymo et al. 2015; Rendall et al. 2009, 2010) are complicated by the fact that the meaning of “early” varies across both time and space. While these studies do make reference to these issues, it is not easy to incorporate such contextual specificity within a comparative analytical framework. One insight from this work is that the prevalence of a given family behavior may be a good predictor of the direction and strength of the educational gradient. When new behaviors (i.e., those associated with the second demographic transition) are relatively uncommon, they are more likely to be experienced by “innovators” who tend to be more highly educated. As the formerly nonnormative behavior becomes more common and the social and economic costs of engaging in the behavior decline, other factors such as those listed above become more relevant and the positive gradient diminishes, disappears, and ultimately a negative gradient appears. This general argument has been made in several studies (Harkonen and Dronkers 2006; Matysiak et al. 2014; Park and Raymo 2013; Raymo et al. 2015).