Intergenerational Transmission of Education in Japan: Nonparametric Bounds Analysis with Multiple Treatments

Abstract

This paper investigates the intergenerational effects of education in Japan using a nonparametric bounds approach. The educational levels of parents are considered key factors in explaining children's educational success. Nevertheless, the literature has not reached consensus on the causal effects of parents' education on their child's schooling. This is because both parents' and the child's schooling depend on unobserved heterogeneity. Moreover, the strong positive correlation of the mother's and father's schooling makes it difficult to separate the effects of each parent's schooling, making it unclear how to control spousal schooling in the analysis. Therefore, this paper estimates a set of semi-ordered vectors of both parents' schooling as an application of the nonparametric bounds method with multiple treatments. It thus derives bounds depending on relatively weak semi-monotonicity assumptions on treatment response, selection, and instrumental variables. A combination of these assumptions provides informative bounds on the average treatment effect of both parents' education on their child's schooling. The main results show that the tightest lower bounds suggest the positive causal effects of parents' schooling, but the tightest upper bounds on the effects are lower than the point estimates that rely on the assumptions of an exogenous selection for parents' schooling. These results suggest that simple regressions overestimate the true causal effect of parents' education.