Unwarranted Disparity in High-Stakes Decisions: Race Measurement and Policy Responses

Studies of racial discrimination often condition on endogenous measures of race or on earlier decisions that might themselves be affected by discrimination. We develop
quasi-experimental tools for estimating the impact of racial misclassification on measures of unwarranted disparity, and for designing policy responses to unwarranted disparity that account for discrimination in earlier decisions. We apply these tools to the setting of child protective services (CPS), where previous work in our context has found that Black children are placed into foster care at higher rates than white children with identical potential for future maltreatment. CPS investigators misclassify 8 to 9 percent of Black and white children relative to their self-reported race, and this misclassification obscures around 24 percent of unwarranted disparity in foster care placement decisions. Policies that use algorithmic recommendations to eliminate total unwarranted disparity in placement rates are also meaningfully affected by earlier discrimination in CPS call screening.