Employment Discrimination: An Empirical Test of Forward Versus Reverse Regression

Abstract

Using data on faculty salaries at a large university, we follow a methodology suggested by Goldberger to conduct an empirical test of the Conway-Roberts proposition that reverse and not forward regression is the proper way to detect salary discrimination. We obtained 11 different estimates of the amount of salary discrimination. Not only are these estimates statistically inconsistent, but they are also at variance with the estimate derived from the composite reverse regression. Therefore we conclude that, in light of the econometric modeling used, reverse regression does not yield an unbiased estimate of salary discrimination.