Much of the research on diverging destinies has focused on the growing concentration of early childbearing among women with lower levels of education. Because marriage and childbearing remain tightly linked in Japan, we examine trends over time in educational differences in the likelihood of marrying by age 22. This particular age is an admittedly arbitrary threshold, but it is relevant in that it corresponds to the age at which university graduates complete schooling and marriage before this age is relatively uncommon in Japan. According to census tabulations, the proportion of 22-year-old women who had ever married was 0.19 in 1980 and 0.09 in 2010.
Because we employ a similar analytical strategy for most of the analyses presented below, we describe our approach in some detail here and note modifications in subsequent analyses. We begin by defining the outcome of interest as a 0-1 variable—marrying before one’s 23rd birthday in this case. We then estimate a logistic regression model for this outcome as a function of women’s age at the time of the survey, her educational attainment, her birth cohort, and the interaction between educational attainment and birth cohort. Because our interest is in the description of trends over time in educational differences, and not in evaluating explanations for those trends, we do not include any other covariates in the models. Based on results from this descriptive model, we calculate the predicted probabilities of marrying by age 22 for each combination of cohort and education. We then present these predicted probabilities graphically to provide both a visual and statistical representation of evidence to evaluate the emergence of diverging destinies. We restrict our analyses to women who were at least 23 years old when interviewed and present predicted probabilities for women age 30-34 when interviewed.
Figure 3.1 shows the predicted probability of marriage by age 22 across birth cohorts and educational attainment (here, and in subsequent figures, JHS refers to junior high school, HS refers to high school, VOC refers to vocational school, JC refers to junior college, and UNI refers to four-year college or more). At first glance, this figure suggests little support for a pattern of diverging destinies. The absolute difference between women with lower levels of education and those with higher levels was highest for the 1940s cohort and lowest for the 1970s cohort, reflecting the steady decline in early marriage for women of all educational levels except university graduates (for whom the probability has always been very low). A substantial proportion of women who did not complete high school continue to experience early marriage (35 % of the 1970s cohort married by age 22), but we believe it more informative to focus on comparisons of high school graduates with more highly educated women. As will be clear from subsequent figures, the small and increasingly select group of junior high school graduates are engaging in a much different pattern of family behavior than all other groups of women. This pattern is generally consistent with patterns of diverging destinies and merits attention, but we consider this group to be too small to serve as a meaningful comparison group.
Looking first at differences between high school graduates (our preferred reference group of less-educated women) and university graduates, we see that the relative difference has increased over time as the proportion in the latter group who experience marriage by age 22 has shrunk to nearly zero. The empty circle markers indicate that the difference between university graduates and high school graduates is statistically larger than in the first cohort (1940-49) and the solid circle marker
Fig. 3.1 Predicted probability of marriage by age 22, by education and birth cohort
indicates that the difference with high school graduates is larger than in both the first and the previous cohort. The odds ratio for early marriage among university graduates relative to high school graduates declined from 0.14 in the earliest cohort to 0.07 in the 1960s and 1970s cohorts. So, while we do not see any obvious “fanning out” in the probability of early marriage for high school and university graduates, there is some evidence that the relative likelihood of early parenthood has grown across cohorts. This pattern is consistent with expectations articulated in the literature on diverging destinies.
An important focus in our analyses is on women in the two “intermediate” educational groups (vocational school graduates and junior college graduates). Motivated by the growing body of research documenting the increasing similarity of women with some college and high school graduates in the U.S., we are interested in assessing the extent to which vocational school and junior college graduates increasingly look more like women with less education or women with more education. We are also interested in assessing the extent to which these two groups increasingly differ from (or resemble) each other. This focus on the middle of the educational distribution will not only shed light on the relevance of patterns of diverging destinies in Japan, but will also provide insight into the nature of rapid shifts in women’s educational outcomes. In particular, we are interested in assessing evidence of two plausible scenarios. In one, the rise in four-year education results in an increasingly selective group of women in the intermediate educational categories who increasingly look like their less-educated counterparts. This is what appears to be happening in the U.S. In another, the shrinking pool of junior college graduates consists of women with relatively strong family orientation whose family behavior resembles women with four-year degrees and increasingly differs from vocational school graduates whose family behavior more closely resembles high school graduates. So, the second scenario is one in which family bifurcation is occurring among women with intermediate levels of education.
Figure 3.1 clearly shows a growing divergence between high school graduates and the two intermediate educational groups. In all cases, the difference in the predicted probability of early marriage for these groups (relative to high school graduates) is significantly larger than in the first cohort and this gap appears to be growing over time (although the change across cohorts is only statistically significant in one case). With respect to early marriage, these two groups increasingly resemble university graduates and continue to resemble each other. It is worth noting that change appears to be particularly pronounced for the 1960s birth cohort —growth in the gap with high school graduates was statistically different from zero in this cohort for both junior college and university graduates. That is, the difference in the probability of early marriage between these women and high school graduates was not only larger than in the 1940s birth cohort, but also larger than in the preceding (1950s) birth cohort. The odds ratio of early marriage for vocational school graduates relative to high school graduates declined from 0.75 in the 1940s cohort to 0.49 in the 1960s cohort. The corresponding decline for junior college graduates (relative to high school graduates) was from 0.52 to 0.25.