The Effect of Infancy Radiation Exposure and Its Role as Ability Shifter in Estimating Return to Education

Abstract

The instrumental variables (IV) model is widely used in estimating returns to education. A key, untestable, assumption for the validity of IV is the exclusion restriction. In this paper, I revisit the common schooling instrument based on local college openings to evaluate its validity and estimate heterogeneity in treatment effects. To do this, I use infant radiation exposure in the U.S. in the early 1960s as a measurable shifter that affects the latent ability term, which is assumed to be independent of the IV. Under the IV assumptions, introducing a control function for latent ability should have no effect on the estimated return to schooling. I find that controlling for infancy radiation exposure does not significantly alter the IV estimates. Second, the latent ability shifter can be used to identify heterogeneity in IV treatment effects. I show that the estimated IV treatment effect of schooling on wages decreases sharply with infancy radiation exposure.