Bias Testing of the Public Safety Assessment: Error Rate Balance Between Whites and Blacks for New Arrests

Abstract

This study is an evaluation of predictive bias by race when using a pretrial risk assessment. We use data from Kentucky from July 2013 through December 2014 (n = 164,597) to evaluate differences across five error measures. Pretrial risk assessments were developed to respond to growing awareness about how intuitive processes were leading to disparate impact for people of color and the poor. This study contributes data to the debate about the merits of pretrial risk assessments. The crux of this debate is the lack of standards about how to assess predictive bias. Our analyses demonstrate that a more nuanced understanding of pretrial outcomes and decision making are needed before fully understanding the strengths and weaknesses of pretrial risk assessments.