Income’s Role in Explaining Black-White Differences in the Educational Gradient in Health: Evidence from the NLSY79 and G-Computation

Michael H. Esposito, University of Washington

Although it is well-established that the association among health and education is less strong for Non-Hispanic U.S Blacks than it is for Non-Hispanic U.S Whites, little empirical work has been produced to explain why said racial difference exists. The aim of this paper then, is to clarify the role of one of the more popular explanations of why the health returns to education vary by race: income. In this endeavor, we employ a combination of rich data from the National Longitudinal Study of Youth 1979, and G-Computation, a technique which allows us to quantify the role income plays in maintaining Black-White differences in educational gradients while avoiding post-treatment bias. In addition, to alleviate the model specification concerns that come with a G-Computation approach, we make use of a nonparametric machine-learning algorithm (Bayesian Additive Regression Trees) to estimate the regression models necessary to the G-Computation process.