Separate Effects of Sibling Gender and Family Size on Educational Achievements - Methods and First Evidence from Population Birth Registry

Son-preferring parents tend to continue to have babies until a son's birth. After deciding the set of children, the parents with resource constraints may divert family sources from daughters to a son. Thus, the presence of a son, relative to a daughter, have 2 distinct effects on his sister's educational out- comes; the direct effect while holding constant family size and the indirect effect through decreasing family size. Previous estimates of the direct effect take family size as an exogenous and predetermined covariate, and assume the size is endogenous and dependent on the sex composition of early-born siblings. We show that even if child gender and family size are both exogenous, use of an instrument for family size is required to isolate the direct effect from the main effects of family size. Using a large and unique administrative data from Taiwan, we demonstrate how Instrumental-Variable Methods resolve both prob- lems of endogeneity and causal dependence of an important covariate (family size) on treatment status (sibling sex). Furthermore, we minimize the incident of sex-selective abortion by restricting our birth data on cohorts prior to abor- tion legalization and prior to prevalent practice of prenatal sex determination. Using the occurrence of twining to instrument for family size conditional on birthweights, our IV estimates show a strong direct effect of a male sibling, relative to a female, on women's college attainment, if the women were born in the earliest year of our data, 1978. After 1978, both effects of sibling gender and family size are almost zero.