one might expect, given the determining power often attributed to educational change (Mensch, Singh, and Casterline, 2005). Furthermore, a regression analysis of the amount of intercohort change in early marriage that might be expected to follow from the intercohort change in educational attainment in 39 DHS countries reveals that, in 15 countries, the expected change exceeds the observed change. That is, the magnitude of the decline in early marriage between cohorts is less than would be expected given the increase in schooling (see Mensch, Singh, and Casterline, 2005, for a detailed description of the methodology and findings).12 Indeed, in about half of these 15 countries, the probability of early marriage actually increases between cohorts despite the increase in schooling.

The pattern in the majority of countries, however, is that the percentage marrying at early ages declined from the older to the younger cohort, and this observed decline exceeds the expected decline. The regional differences are considerable. In Latin America and the Caribbean, the expected decline in early marriage following from increased schooling far exceeds the actual decline. In many of these countries, of course, the probability of early marriage has not changed. Perhaps there is a threshold beyond which increased schooling is not associated with a change in age of marriage. By contrast to Latin America, for about two-thirds of sub-Saharan African countries, half or more of the decline in early marriage can be linked to increased schooling. In sum, the rise in schooling hardly appears to be the entire story, although in sub-Saharan Africa a substantial fraction of the reduction in early marriage is associated with the expansion in education.

Links with Women’s Labor Force Participation

Not only is increased schooling widely believed to contribute to the delay in marriage among young women, but also access to wage employment is frequently cited in discussions of rising age at marriage (Mathur, Greene, and Malhotra, 2003). It seems logical that there are greater opportunity costs associated with marriage for young women who are in the paid labor force. Indeed, a daughter’s enhanced income-earning potential is ar-


Logit regressions were estimated for each cohort. Then the coefficients from one cohort were applied to the other cohort to calculate a predicted logit of early marriage for each woman, which was then transformed into a predicted probability. The mean of these probabilities is the expected proportion marrying early, and the expected proportion minus the observed is the expected change in early marriage due to schooling change. This analysis was conducted only in the 39 countries in which the DHS interviewed all women, not just ever-married women. This exclusion effectively eliminates South-central and South-eastern Asia and the Middle East from this analysis.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement