Unobserved Worker Ability, Firm Heterogeneity, and the Returns to Schooling and Training

Abstract

It is well known that unobserved heterogeneity across workers and firms seriously impacts the computation of the determinants of individual earnings in standard human capital earnings functions. Following the tradition of AKM (Abowd, Kramarz, and Margolis, 1999), this paper offers an alternative way of controlling unknown worker and firm heterogeneity by taking full advantage of a matched employee-employer dataset based on two key Portuguese micro databases. Our modelling strategy assumes that the gap between individual and firm average wages, unexplained by differences in observable characteristics, gives the extent to which the unobserved ability of a given individual deviates from the unobserved worker average ability in the firm. This methodology has, in particular, the advantage of not relying exclusively on information on job switchers to identify worker and firm effects, thus avoiding any bias arising from endogenous worker mobility. Another important aspect of our treatment is that it allows the estimation of worker effects without risk of contamination from firm effects. To test our modelling we use an original 2-year longitudinal LEED dataset, comprising of more than 400 thousand workers and 1,500 firms in each year. We focus on two separate sets of individuals (i.e. stayers and switchers) and provide a variety of robustness tests, including replication of the original AKM methodology. After controlling worker and firm effects, our results show that the acquisition of schooling, labor market experience, and training, inter al., pays off. Moreover, we do find evidence of a large bias in standard OLS return rates to typical covariates. Evidence from Monte Carlo simulation and bootstrapping also shows that our estimated rates of return to human capital do not seem to be sensitive to changes in various assumptions. Our study does provide therefore further evidence that a wide set of individual and firm characteristics is crucial to understanding the true role of human capital variables in labor markets.