This paper investigates the effects of health on wages by controlling for a number of problems: first, the unobservable genetic endowment may cause an omitted variable bias; second, using a self-reported health variable could induce measurement error; third, the issue of reverse causality arises; and fourth, panel attrition driven by the endogenous decision to participate in the labour market may result in inconsistent estimation. By using recently developed methods, I control for all of the above issues in one framework. The results show that good health raises wages for both women and men. I find the health variable to suffer from measurement error. In the male sample, applying OLS or 2SLS, instead of methods accounting for selection and individual heterogeneity, causes an upward bias in the health coefficient.
Health wages fixed effects sample selection instrumental variables