Predicting troublemaking offenders: A discriminant analysis

Violence, attempted escapes, thefts, persistent rule violations, and other extreme forms of maladaptive prison behavior have created problems for correctional authorities since congregate confinement became the watchword in American corrections. The evolution of prisoner classification programs has aided prison administrators in their attempts to recognize potential behavioral problems but the subjective nature of a number of state classification models has resulted in a continuous wave of litigation and fiscal constraints. The objective of this paper is to provide a prediction model to assist in the classification of adult prisoners. Data on all members of the Georgia adult, male prison population from 1971 to 1978 were used as a baseline. A sample was chosen from inmates in long-term disciplinary segregation whose files contained all necessary demographic, psychometric and judicial variables. A discriminant analysis of these two groups identified 14 variables which correctly assigned inmates in 85% of the cases. The predictive variables include both demographic and personality characteristics.