Professor of Justice Administration, Texas Southern University
- Characteristics such as location, jurisdiction, minority status and offense severity matter more than a simple overall risk score given to female offenders by risk assessment models.
- It is important to examine a model’s predictive ability by also paying attention to its predictive error, including false positives and negatives.
- The real-life circumstances of minority female offenders must be taken into account when determining the predictive utility of risk assessment algorithms.
In the article, “Gender, Race/Ethnicity and Prediction: Risk in Behavioral Assessment,” Henderson and his co-authors provide new evidence about the predictive accuracy of risk assessment instruments for minority female offenders. Utilizing risk scores from the Wisconsin Risk Needs Assessment Instrument and other contextual variables (such as location and offense severity), this analysis examines probation outcomes across a sample of female minority probationers. Their findings identified location and minority status as obstacles to equitable predictions by risk assessment models. Given the unique importance of achieving predictive equity, examining error holds special appeal to those policymakers who have been charged with allocating millions towards these instruments.