NCJ Number
111433
Date Published
1986
Length
61 pages
Annotation
A research study conducted by two Michigan agencies resulted in the development of a proposed prediction and classification instrument for use in classifying probation cases into high-risk, medium-risk, and low-risk groups.
Abstract
The research aimed to develop a model that would be easy to use and that would have a high level of accuracy in predicting the probability of recidivism as a function of criminological variables. Research teams from the Michigan Department of Corrections and the Michigan Office of Management and Information Systems identified the variables for analysis. The research emphasized the use of multiple linear regression analysis to examine 22 variables. Several sets of analyses were performed in the effort to develop a satisfactory prediction instrument. The recommended prediction equation is a linear additive model that is easy to understand, interpret, and use. It uses the following variables: age at first arrest, prior employment record, length of employment, total number of juvenile arrests, presence or absence of a substance abuse problem, and outcome on prior probation. This equation was converted to a prediction table for use in classification. Figure, tables, 8 references, and appended methodological information.