Evidence on Gender Wage Discrimination in Portugal: parametric and semi-parametric approaches

Abstract

In this paper we use two alternative approaches to study the extent of gender wage discrimination in Portugal. Both methods involve the estimation of wage equations for males and females and the Blinder [1973] and Oaxaca [1973] decomposition. However, to take into account possible sample selection bias, we consider both parametric and semi-parametric methods. First, we consider a parametric approach that relies on distributional assumptions about the distribution of the error terms in the model (Vella (1992, 1998) and Wooldridge (1998)). Within this approach, if the distributional assumption is not satisfied, the parameters’ estimates may be inconsistent. Secondly, we apply Li and Wooldridge [2002] semi-parametric estimator, which does not assume any known distribution on the joint distribution of the errors of the wage equation and of the sample selection equation; the distribution has an unknown form and is estimated through non-parametric kernel techniques.We employ micro data for Portugal from the European Community Household Panel (ECHP). The results from both approaches provide evidence in favour of the existence of gender wage discrimination in Portugal. However, the extent of labour market discrimination decreases when sample selection bias corrections are taken into account.