Parametric vs. semi-parametric estimation of the male-female wage gap: an application to France [November 2005]

Abstract

We use a semi-parametric method to decompose the difference in male and female wage densities into two parts–one explained by characteristics and one which is attributable to differences in returns to characteristics. We learn substantially more about the gender wage gap in France through this analysis that we do through parametric techniques which we also employ for comparative purposes. In particular, we find that there are no unexplained differences in male and female earnings distributions in the bottom fifth of the data. Occupation and part-time status are the most important determinants of the wage gap for all workers. In the semi-parametric estimates we find that education plays no role in the wage gap once we account for occupation and part-time status.