The classification is based on statistical estimation of the likelihood of individual failure (revocation and absconding) during the probation supervision period. These estimations are then used to classify individuals as high-, medium-, or low-risk for probation supervision purposes. The analysis uses case history records on offenders admitted to Florida probation between January 1, 1991, and December 31, 1994. While the data are specific to Florida's probationers and policies, the methods appear to be applicable to addressing practical issues associated with risk management in other jurisdictions and program settings. The report includes background information on Florida's probation system, hypothetical examples to illustrate important practical implications of translating individual failure probabilities into a three-level risk classification system, a technical description of mathematical methods by which individual failure probabilities are calculated, and information on work planned by researchers at the Florida Department of Corrections to convert the technical results into a fully automated probationer risk classification system. Appendixes contain additional information on logistic regression models and software for estimating multinomial logit models, along with model coefficients and goodness of fit graphs. Footnotes, tables, and figures