Labor Market Discrimination in the US: A Quantile Regression Approach

In this study I apply quantile regression techniques to the well-known Oaxaca coefficient of discrimination. This methodology provides different coefficients for different quantiles of the conditional wage distribution and is more informative than the technique based on OLS regression, which concentrates only on the conditional mean of the wage distribution and provides only an average estimate. Results, using data from the 1999 Current Population Survey, show that the part of the wage gap not explained by differences in skills between the genders is higher at the upper quantiles of the conditional wage distribution, suggesting that ``discrimination" against women increases as we move from low to high quantiles. That is, more than twenty years after Oaxaca's (1973) results, we notice that women and men are much more alike in terms of labor market experience, education and attachment to the labor force. Moreover, the data show that in the last twenty years female workers increased their share in predominantly male occupations and decreased their participation in female occupations. During this period, the wage gap actually decreased, however from the little wage gap remaining, almost nothing can be explained by differences in observable skills, and women at higher paying jobs are the most affected by this unexplained wage differential