Gender Effects in Actuarial Risk Assessment: An Item Response Theory Psychometric Study of the LS/CMI

Abstract

Actuarial tools play an important role in correctional and risk management systems as they are widely used to assess potential recidivism. In psychometric studies of the predictive value of these tools, it is, rightly or wrongly, common to find all items and all components being given the same weight such that they contribute equally to the case management of individuals and to the determination of criminal recidivism risk. Item Response Theory (IRT) allows for psychometric analysis permitting to evaluate the quality of the items in actuarial tools, such as the LS/CMI. As such, IRT methods can improve our understanding of an instrument’s psychometric properties well beyond what is already known from traditional approaches. This paper draws on IRT to explore the predictive evidence for the LS/CMI across inmates’ reported gender. The sample consisted of male (n = 1200) and female (n = 1148) inmates serving a custodial sentence. The analyses suggest that the predictive evidence for the LS/CMI is strongly related to the discrimination parameter of the items and varies considerably by gender. In conclusion, this study contributes to the understanding of criminal recidivism as a function of gender and questions the interpretation of the total score (from the section General Risk/Need Factors) generated by the LS/CMI.

Keywords:

• Actuarial tools
• discrimination index
• female inmates
• item characteristic curve
• item response theory
• male inmates
• predictive validity
• psychometric research
• recidivism