Improving Risk to Reoffend Predictions in Florida Juvenile Justice

• In Event: Predicting Future Delinquency and Improving Youth Outcomes

Abstract

Criminologists have a compelling interest in predicting the risk of reoffending for offenders in the criminal justice system. Varying risk assessment instruments have been used toward that goal, but their levels of predictive accuracy often fall short of reasonable expectations (Baird, 2009). Indeed, as some have noted, recent decades may have witnessed little improvement in predictive accuracy (Austin, 2006; Caudy, Durso, & Taxman, 2013). These patterns prompt research into what advances should be pursued. The present study reports results from one such effort. We examine data from a large sample of offenders in Florida who completed juvenile justice dispositions from 2010-15; reoffending was tracked for a 12-month period after completion of services. We compare the accuracy of existing practices to what was achieved when various procedures were used to improve predictive accuracy. This included alterations of such things as which criminogenic risk components were considered, how such components were scored, and what modeling strategies were used to predict reoffending. We describe key findings that emerged from the analysis and conclude by examining those findings in the context of current dialogue on best practices for predicting the risk of reoffending.

Authors

• Carter Hay, Florida State University

• Jennifer Copp, Florida State University

• Brian Stults, Florida State University

• Brae Campion Young, Florida State University

• Tiffaney Tomlinson, Florida State University