Retornos à educação no Brasil: uma abordagem de pseudo painel e viés de seleção amostral

Abstract

The goal of this work is to analyze the returns to education in the hourly wages in Brazil for the period of 2011 to 2015, last five years of PNAD data. Even thought the relationship between years of education and wages has been long studied, the accuracy of the results from traditional models like the Ordinary Least Squares are to be questioned due to the necessity to check the presence of group heterogeneity and sample bias. The hypothesis tested in this work is that there is significant effect from non observable factors that leads to biased estimators if not taken into account. By estimating a pseudo panel, through the creation of cohorts, it is seen that the group heterogeneity leads to underestimation of the returns to education from the WLS model. The sample bias also affects the magnitude of the coefficients, however, the two different models used to capture and correct the estimators present different results. The two step model reports smaller returns in comparison to the WLS benchmark, while the Maximum Likelihood model had mixed results.