Simulated Recidivism Risk Assessment: A New Way to Assess Criminogenic Needs?

• In Event: In God We Trust. All Others Must Bring Data

Abstract

Risk assessment involves predicting who will recidivate, while needs assessment involves predicting who will benefit from certain types of interventions. Because the performance of contemporary risk and needs assessment instruments is typically evaluated by examining how well they predict recidivism, existing research has not demonstrated that current instruments perform well in assessing criminogenic needs. In this study, we explore whether simulated recidivism risk assessment can be used to not only improve how we assess criminogenic needs, but also how we evaluate the performance of needs assessment instruments. Simulated risk assessment entails examining the potential impact of correctional interventions on recidivism risk for individual prisoners. Our dataset is the same one used to develop and validate the Minnesota Screening Tool Assessing Recidivism Risk 2.0 (MnSTARR 2.0), the recidivism risk assessment the Minnesota Department of Corrections has used since November 2017. In addition to containing criminal history, demographic, and institutional measures (e.g., prison misconduct, gang involvement, etc.) on more than 40,000 males and females released from prison, the dataset includes 10 items related to participation in prison programming found to be effective in reducing recidivism. Using a variety of statistical and machine learning algorithms on training and test sets, we estimate baseline prediction models. Next, we estimate models that simulate participation across the 10 interventions. Comparing the differences in predicted probabilities from the baseline and simulation models, we examine whether these differences (i.e., the simulated impact of participating in an intervention) successfully predict desistance. We propose that simulated risk assessment, when combined with traditional needs assessment, may be a promising strategy to help identify the most beneficial interventions for individuals within correctional systems.

Authors

• Grant Duwe, Minnesota Department of Corrections

• KiDeuk Kim, Urban Institute