Predicting wages from transformed wage equations

Abstract

The log-linear specification of wage equations is commonly used in the applied economics literature. Predictions of the retransformed wage are not straightforward as the standard approach (naive estimator) produces biased predictions. Duan (1983) proposes a smearing estimator which overcomes this problem, and is easy to implement. In the paper we predict wages using these two estimators and highlight the difference in the predictions they generate.