NCJ Number
149097
Date Published
1993
Length
171 pages
Annotation
This report addresses prediction classification for bail decisions in Harris County, Tex.; how a new point scale was designed for the county; and how that scale performed after implementation.
Abstract
The central issue of the project was whether the existing predictive tool or an empirically derived instrument would provide the consumer courts with greater predictive accuracy in making pretrial release decisions. The project developed a bail classification instrument that used eight predictors out of 40 that were developed from data available through the Justice-Information-Management-System data for the 1990 defendant population. Researchers found the instrument to be substantially more predictive of outcome than the original instrument used in Harris County for more than a decade. Tests for disparate impact on defendants of different racial/ethnic backgrounds or gender showed some differences, but these fell within limits that might be expected from random variation. Statistically removing the influences of race/ethnicity and gender from the classification instrument caused no change in the way the instrument predicted risk. This report concludes that the instrument is performing its intended function well and should be applied widely as a credible information source in making bail decisions. 99 figures