"PA-PVR" or "Projecting Violent Re-Offending in a Parole Population: Developing a Real-Time Forecasting Procedure to Inform Parole Decision-Making"

The purpose of this research is to develop and evaluate a forecasting model for use by the Pennsylvania Board of Probation and Parole (PBPP) in parole supervision. Forecasting procedures will be developed to allow the PBPP to have formal forecasts of the chances that an offender is likely to re-offend violently while under supervision. The forecasting instrument will be incorporated into PBPP decision making and evaluated as to effectiveness in reducing violent recidivism among offenders under parole supervision. Data will be collected from the PBPP for all individuals released on parole in 2007 to comprise a cohort of at least 10,000 individuals released from the Pennsylvania prison system. Forecasting models will be developed and tested using the 2007 cohort using machine/statistical learning procedures. Models will be evaluated for accuracy and then implemented by PBPP in making parole supervision decisions. A process, outcome, and impact evaluation will be conducted on approximately 2,000 individuals released on parole using a regression discontinuity design to estimate the effects of violent crime risk forecasts for a period of 2 years of following release from prison. ca/ncf