Three Tests for Measuring Unjustified Disparate Impacts in Organ Transplantation: The Problem of "Included Variable" Bias

Civil rights statutes often prohibit two distinct types of discrimination, referred to as "disparate treatment" and "disparate impact." Disparate treatment is race-contingent decision making. But even decision making that is not affected by people's race may still produce an unjustified disparate impact. For example, a race-neutral transplantation preference for allografts with partial antigen matches might produce an unjustified disparate impact on African Americans with end-stage renal disease. The transplantation preference might make it harder for African Americans to receive a transplant without significantly increasing the chance of transplant survival. Because disparate impact and disparate treatment claims have distinct elements, they require distinct methods of statistical testing. This article analyzes three different ways of testing unjustified disparate impacts in organ transplantation, which I will call the traditional test, the omitted variable test, and the outcome test. Each of these methods of testing for disparate impact are attuned to the problem of "included variable" bias. Controlling statistically for nonracial variables may actually bias the analysis and mask the existence of unjustified disparate impacts.