NCJ Number
184250
Journal
Journal of the American Statistical Association Volume: 94 Issue: 448 Dated: December 1999 Pages: 1045-1052
Date Published
December 1999
Length
8 pages
Annotation
This evaluation of the inmate classification system used by the California Department of Corrections (CDC) to allocated prisoners to different kinds of confinement used a generalized regression discontinuity design and data from 2,746 inmates admitted early in 1994.
Abstract
The research used a simple extension of regression discontinuity designs to involve binary response variables. The researchers chose the start date for data collection because it was the most recent time after which at least 18 months of follow-up data were available to consider inmate misconduct. The sample included virtually all the inmates sentenced under California’s recent three strikes legislation due to the CDC concern that three inmates might become difficult as a result of having little to lose. The study included 734 three-strike inmates. It also included 771 two-strike inmates, who had a prior conviction for a serious felony and were subject to automatic and large sentence-length enhancements. The existing classification scores usefully sort inmates by levels of risk and that placement in the highest security level may well reduce the odds of misconduct compared to placement in the lower three security levels. A large study using an experimental design and 20,000 inmates is now underway to address inmate classification and related issues. Tables and 19 references