Public safety assessment: Predictive utility and differential prediction by race in Kentucky

Abstract

Research Summary: We assess the predictive validity and differential prediction by race of one pretrial risk assessment, the Public Safety Assessment (PSA). The PSA was developed with support from the Laura and John Arnold Foundation (LJAF) to reduce the burden placed on vulnerable populations at the front end of the criminal justice system. The growing and disparate use of incarceration is one of the most pressing social issues facing the United States. The implementation of risk assessments has provided fuel for both sides of the reform debate with proponents arguing that the use of these assessments offers a policy mechanism to alleviate populations and bias. Risk assessment critics, however, argue that the use of the assessments exacerbates bias and does not improve decision-making. By examining a statewide data set from Kentucky (N = 164,597), we found the PSA to have predictive validity measures in line with what are generally accepted within the criminal justice field. The differences we found indicate the PSA scores for failure to appear (FTA) are moderated by race, but these differences do not lead to disparate impact. Policy Implications: We point to data limitations and the need for localized risk assessment studies, and we emphasize that risk assessments are decision-making tools that require ongoing refinement. Risk assessment developers, opponents, and proponents would do better to focus on the reality of risk assessments as probabilistic models. The results of these assessments cannot predict with certainty, and they are not inherently biased. Rather, criminologists and policy makers need to understand the uncertainty that comes with any predictive model.