Predicting parole outcome via discriminant analysis

Abstract

Much criminological research has dealt with the topic of parole recidivism and the use of prediction devices in correctional settings. It has been suggested that potential predictor candidates need to be tested in different jurisdictions to assess their applicability in various locations. This study examines the relationship between a number of independent variables and parole outcome using a sample of 350 Kansas inmates released on parole between March and September 1979. Parole outcome was determined using a two-year follow-up from the date of parole hearing. Data on independent variables were obtained from a prediction instrument on the research sample completed by the Kansas Adult Authority, as well as information from the Offender Based Statistical Computer Information System (OBSCIS) used in Kansas. Discriminant analysis is utilized to derive discriminant functions which optimally separate those who succeed or fail on parole in terms of the predictor variables. Both practical and theoretical implications are discussed.