Author(s)

Zejun Jiang, Ke Lyu

Corresponding Author

Zejun Jiang

Abstract

Previous researches indicated a positive causal effect of education on earnings. In these studies, a widely acknowledged point is the endogeneity of education which causes bias for estimate. One common method to deal with the endogenous problem is to introduce instrumental variables. Angrist and Krueger employ the quarter of birth as the instrument, however, subsequent researches point out that the quarter of birth is a weak instrument because of weak relationship with education leading to implausible results. We still use the quarter of birth as instrument and data of Angrist and Krueger’s research in our supplementary work to focus on how to check the existence of weak instrument problem and then obtain the weak-identification-robust inference in return to schooling. The first finding is that the quarter of birth may not bring about weak instrument problem at the 30% bias level. The second is that we gain the credible confidence interval for the estimate of our interest under three weak-identification-robust test, namely the Conditional Likelihood Ratio (CLR) test, Anderson and Rubin (AR) test, and Kleibergen (K) test. Furthermore, we want to get data from a tracking survey of the same people to ensure that the return to schooling changes for individuals during the whole life.

Keywords

Return to schooling; Quarter of birth; Instrumental variable; Endogeneity