NCJ Number
95816
Date Published
1984
Length
74 pages
Annotation
A method of predicting violence and recidivism is demonstrated with data from a study of recidivism among a sample of 1,000 offenders released from Iowa prisons from 1976 through 1980 and monitored for 4 years. Implications are drawn for sentencing and parole policy.
Abstract
The efficiency of predictive instruments is discussed, and the Mean Cost Rating (MCR) is identified as the instrument of choice for measuring predictive efficiency in situations where the criterion variable is dichotomous. Historical levels of predictive efficiency are reviewed, and the development of the Iowa Offender Risk Assessment Scoring System is chronicled. To refine the model, a new data base, in the form of the 1,000-offender sample, was generated. Several criterion measures of recidivism were defined for purposes of model refinement, including a simple dichotomous variable indicating whether the offender received a new charge for a violent felony, a weighted measure of all new violent felony charges, and a dichotomous measure indicating a new prison sentence for violent, property, and drug crimes. Factors that separated recidivists from nonrecidivists were identified and separated into Class II factors (exhibiting the strongest relationship to recidivism criteria) and Class I factors (exhibiting a lesser relationship). Three criterion measures were used to test the predictive validity of the 1980, 1983, and 1984 versions of the Iowa Risk Assessment Model. For all criteria, the 1984 version results in the highest MCR values. Construction, validation, and sample results are provided. Study results suggest that high-rate offenders can be accurately identified with factors oriented more to the just deserts philosophy of sentencing than has been the case. Forty-three references and approximately 50 tables are included.