Constructing a rehabilitation index to assess the relationship between prison environments and reoffending

Abstract Correctional rehabilitation policy typically targets individual risk, with little attention to institutional risk. Using administrative data on 11,302 adult men released from 45 prisons in one Australian jurisdiction, we estimate multiple-membership multilevel logistic models that time-weight exposure across facilities and benchmark prisons with a risk-adjusted rehabilitation index. Prisons differed significantly in reimprisonment rates after adjusting for individual risk profiles. Contextual features and in-prison experiences were associated with outcomes. Including people’s in-prison experiences made the model better at distinguishing who returned to custody and explained much of the remaining difference between prisons. Findings challenge person-only explanations of reoffending and show that institutional context matters. The approach offers a learning-oriented model of prison accountability and points to practical ways to improve rehabilitative performance.