Returns to Skills in Europe – Same or Different? The Empirical Importance of the Systems of Regressions Approach

Abstract

We estimate the Mincer equations for a set of European countries. The variability of parameters, describing the impact of years of schooling and the experience to the wages, was obtained by application of the system of Seemingly Unrelated Regression Equations (SURE). The differences between parameters were tested given two alternative stochastic assumptions. In the first model, no contemporaneous correlations between error terms in the system is imposed. This may be related to the standard country regression approach. In the second approach the unrestricted covariance matrix is considered, making error terms stochastically dependent. The contemporaneous correlations of error terms in the SURE system were empirically supported. Also, rich parameterisation of covariance matrix of contemporaneous relations reduced statistical uncertainty about differences in parameters describing return on education effect. Consequently, the country heterogeneity of return on education, which seems intuitively correct, was obtained in the system of regressions with complex stochastic structure.