Predictive Validity of the Joint Risk Matrix With Juvenile Offenders

Abstract

Actuarial risk assessment instruments promise to increase decision-making accuracy and equity in settings such as the juvenile justice system, but both aims presume high levels of predictive validity. Prior research suggests that the predictive validity of some juvenile justice risk assessment instruments differs across gender and race/ethnicity. The Joint Risk Matrix (JRM) described herein is an instrument developed to increase the predictive validity of risk assessment for the diverse populations served by the nation’s juvenile courts. The predictive validity of the JRM was estimated on a sample of 536 court-involved juveniles. The instrument demonstrated acceptable levels of validity across all juveniles (AUC = .710). Gender-based differences were explained by gendered patterns of referral to out-of-home placements. Differences by race/ethnicity were reduced compared with previous reports. The findings suggest that risk assessment can be improved by including measures related to the behavior and demeanor of offenders and the cooperation of their parents or caretakers.