Strategies for improving offender risk prediction

Abstract

There is a considerable debate in the field with respect to what is the best strategy for
improving the predictive accuracy of offender risk assessments. Among these differences of
opinion are opposing viewpoints regarding static and dynamic risk factors, the mathematics of
computing scores (e.g., adding/removing items, weighting items), and the use of offender
responsivity factors (e.g., gender). There are also techniques used in other behavioral science
fields (e.g., subgroup norming) that have yet to be explored on offender populations. This study
examines the ability of three approaches (criminal history, selected items, subgroup weighted) in
improving the predictive accuracy of the Level of Service Inventory- Revised (LSI-R) on the
same sample of offenders. This study did not find that any of these methods were able to produce
a statistically significant improvement over the standard LSI-R assessment score alone.
Implications of these findings and recommendations for future research are discussed.
Keywords: Risk assessment, Level of Service Inventory-Revised, LSI-R, community corrections,
correctional policy, subgroup norming