Improved Statistical Methods for the Calculation of Damages in Discrimination Lawsuits
56 Pages Posted: 12 Dec 2017
Date Written: November 14, 2016
Abstract
This paper develops a new method to calculate individual-specific damage payments in discrimination lawsuits using empirical Bayes techniques and a simple random coefficients model. The method yields payments that can be mathematically proven to be more accurate than existing statistically-based approaches, as measured by the mean squared error. This method also provides a natural justification for zero payments to members of the non-injured class and substantially reduces the bias caused by the legal restriction on negative payments. We empirically demonstrate our method in the context of mortgage pricing decisions using detailed loan-level data from a large subprime lender.
Keywords: Empirical Bayes, Discrimination, Class Action
JEL Classification: K13, J15, J71
Suggested Citation: Suggested Citation