Separating risk in education from heterogeneity: a semiparametric approach

Returns to education are variable within the same educational group. If uncertain payoffs are a concern for individuals when selecting education, wage variance is the resultant of unobserved heterogeneity and pure uncertainty. The first element is known to the individual, but unknown to the researcher. If individuals exploit private information to select their level of education the variance observed in the data will overestimate the real magnitude of education uncertainty and the impact that risk has on educational decisions. In this paper we will apply a "semi-non parametric" estimator of an unknown density to tackle selectivity issues. This method does not rely on the joint normality of errors in the selection and primary equation and is robust to misspecification of the residual distribution. Results suggest that pure risk tend to increase with education. Private
information accounts for a very minor share of total wage variance observed in the data.