Effect of conditional release on violent and general recidivism: A causal inference study

Abstract

Objectives: To study the effect of Conditional Release (C.R.) on recidivism. To compare this effect along different recidivism risk levels, to evaluate whether risk-assessment-based policies that prioritize people in lower risk categories for release maximally reduce recidivism. Methods: We use a dataset of 22,726 incarcerated persons released from 87 prison centers in Spain. We apply multiple causal inference methods including Propensity Score Matching (PSM), Inverse Propensity score Weighting (IPW), and Augmented Inverse Propensity Weighting (AIPW) to determine Average Treatment Effect (ATE) of C.R. on recidivism. Results: Granting C.R. significantly reduces violent and general recidivism risks. Conclusions: The results suggest that C.R. can promote a safe and supervised return to the community while protecting public safety. ATEs obtained through causal inference methods suggest that granting C.R. exclusively to low-risk inmates does not lead to the maximum reduction of recidivism, and hence we propose alternatives to be studied further.