Estimating the returns to schooling: Implications from a dynamic discrete choice model

This paper assesses the applicability of a dynamic discrete choice model in accounting for the observed ordinary least squares (OLS) and instrumental variable (IV) estimates of the Mincer equation parameter on returns to schooling. A dynamic model of schooling and employment choices is estimated and used to simulate educational attainment, employment history, and wages. Estimations of the Mincer wage equation using simulated data appear to validate the model. Ability selection is found to be the major source of bias in the OLS estimates of schooling returns. Although the IV estimates lie within the support of true returns to schooling if a strong and strictly exogenous instrument is used and if dynamic employment selection is controlled, these conditions may be easily violated in practice.

► A dynamic choice model is estimated and used to simulate labor market outcomes.
► Simulated data are used to estimate the schooling returns in log wage equation.
► Ability selection is the major source of bias in the OLS estimates.
► Under stringent conditions, IV estimates may be unbiased.
► But they may be worse than OLS estimates if any of the conditions is violated.