P-Values, Priors, and Procedure in Antidiscrimination Law

Abstract

This paper is the second in a series of two papers addressing the influence of priors in antidiscrimination law. The first paper, Hidden Priors, argued that Bayesian probability theory could harmonize two divergent strands of scholarship on the systemic disparate treatment theory of discrimination. This Bayesian approach challenges the conventional use of statistical evidence in systemic discrimination cases. Critically, the Bayesian approach acknowledges the operation of prior information about the likelihood of unlawful discrimination when evaluating statistical evidence. While a handful of scholars have contemplated a transition to a Bayesian antidiscrimination law, there has been no scholarly treatment of the second-order questions raised by the recognition of Bayesian priors. These second-order questions were highlighted at the conclusion of Hidden Priors. They include hard, but not intractable, questions such as: Whose priors matter? How should priors be incorporated into civil litigation procedures? And, what are legitimate sources of priors? The lack of attention to these questions may explain the stubborn refusal of courts, litigants, and employment law scholars to acknowledge the role of priors in systemic discrimination cases. This second article in the series is meant to spur a healthy scholarly discussion of the second-order questions, even if it cannot provide definitive answers to them. This article argues that the procedural devices of our civil litigation system – though imperfect – are actually well-suited to properly allocate responsibility among legislatures, appellate judges, trial judges, and fact-finders for estimating the prior likelihood of discrimination. This article demonstrates that, once acknowledged, priors can be transparently managed by thoughtful implementation of these procedural devices.

Keywords: Bayesian, systemic, discrimination, statistics, priors, Teamsters, Hazelwood, frequentist, transposition