NCJ Number
161627
Journal
Journal of Research in Crime and Delinquency Volume: 33 Issue: 1 Dated: special issue (February 1996) Pages: 49-69
Date Published
1996
Length
21 pages
Annotation
Risk dimensions used in guideline systems have been implicated as contributing to racial and gender disproportionalities in U.S. prison and jail populations.
Abstract
Developers of some guideline systems dealt with invidious risk predictors by purposely ignoring them, and this resulted in misspecification and did not eliminate their effects. Eliminating all variables associated with suspect factors would greatly attenuate power and render practical decisionmaking tools useless. The authors show that one risk prediction device forming the basis of an operational guideline system correlates with race and gender and that control factors are not correlated with the final model even after second-order policy controls are implemented. Further, although the suggested approach is agnostic with respect to the nature of policy controls, these will have considerable practical importance. The authors demonstrate that unbiased models can be estimated with little appreciable loss of predictive utility. The nature and extent of statistical and ethical problems associated with risk components of guideline systems are described, along with criticisms generally resulting from the fact that race and gender are correlated with common outcomes predicted by criminal justice functionaries. 44 references, 8 notes, and 6 tables