Comparison and combination of clinical and statistical predictions of recidivism among adult offenders

Abstract

Clinical and statistical predictions of 6 categories of recidivism among 198 adult male felony probationers (average age 27.99 yrs) were compared and combined, both before and after correcting for the restricted range of the predictor variables. The statistical composite consistently outperformed decision makers for the undifferentiated recidivism criteria of arrest and conviction. The opposite state of affairs was observed to exist for all 3 indices of violent recidivism but only after correction for restriction of range. The combination of both variables tended not to produce significant increases in criterion variance compared to the superior predictor alone, the major exception being incarceration for any offense. Decision makers performed best when forecasting violent criminal conduct that resulted in incarceration, and this contingency table was subjected to a detailed analysis of the outcomes of prediction. This included 2 methods of estimating the probable probationary performance of 141 members of the initial sample who were sentenced to prison, and consideration of the utilities associated with false positive and false negative errors. Partially because of the seeming impossibility of achieving perfect prediction, it is concluded that there is a need for decision-making guidelines that include only those criminological variables that can serve the dual purpose of predicting recidivism and indicating which offenders are the most deserving of imprisonment as punishment. (19 ref) (PsycInfo Database Record (c) 2025 APA, all rights reserved)