Testing 'Discrimination Plus' Cases

Abstract

This paper develops an estimation methodology to test the hypothesis of discrimination appropriate to a particular class of employment discrimination cases, referred hereto as ?discrimination plus? cases, involving more than one legally protected employee characteristic. In discrimination plus cases, we need to compare subgroups of employees based on the charges brought forward. To do so, one may use a reduced sample consisting of the subgroups in question or use the full sample with appropriate interaction terms. However, between these two alternatives, the tests based on the full sample are more reliable because the full sample estimates the parameters of our interest more precisely. If we use the reduced sample instead, we are more likely to find ourselves drawing a wrong conclusion. From a litigation point of view, an erroneous conclusion may have a profound impact.