Keywords: Accuracy, National Institute of Justice, Recidivism Forecasting Challenge, Recidivism Rates, Risk Predictors
Author: Mary Beth Carroll, Rodney Carson, Mike Clark, Adam Cottrell, Jim Georges, Tyler Nelson, Hiwot Tesfaye, Halil Toros, Sree Vuthaluru
Publisher: SAS Institute Inc.
Publication Year: 2022
Summary: The following modeling report attempts to look at racial bias in recidivism and better predict recidivism. This was their entry into the National Institute of Justice (NIJ) sponsored Recidivism Forecasting Challenge to try and improve forecasting recidivism. The goal was to increase accuracy using risk predictors, what they found important variables were age at release, gang affiliation, prior arrest, average days between drug tests and prior convictions.