This paper investigates the impact of education on income distribution of Brazilian states and regions, using a semi parametric method, following Dinardo, Fortin and Lemieux (1996) and dataset from the PNAD 1999. Contrafactual densities were constructed weighting the distribution of the poorest region/state (Northeast / Ceará) by the profile of education in the richer one (Southeast / São Paulo). Results: between 12% and 36% of the difference in income is explained by the educational differences; the weighting by education increased by about 55% the average income in counterfactuals; the counterfactual income of the Northeast amounts to 93% of the average Brazilian income; the higher the percentile considered income, the greater the contribution of the difference in schooling for the difference in income; and the income dispersion of the poorest regions increases when they provide the level of schooling of the richest regions, while the wage profile of the region is kept constant.
income distribution; education; semiparametric method
