The consequences of error: Recidivism prediction and civil-libertarian ratios

Abstract

The utility of recidivistic prediction is limited by the false positive problem: predictions of failure (recidivism) that do not occur. False negatives (predicted successes but observed failures) are also worrisome, and together both types of error can be formally evaluated by what Blumstein, Farrington, and Moitra call the civil-libertarian ratio: the ratio of the subjective cost of a false positive to a false negative. Choice of a recidivistic criterion and selection of a proportion of offenders for criminal justice intervention have implications for the evaluation of the disutility or subjective cost associated with various civil-libertarian ratios. Logistic regression models of four recidivistic criteria are evaluated to demonstrate how base rate (observed failure rate) and selection ratio (proportion selected to fail) affect the disutility associated with a range of civil-libertarian ratios. Use of civil-libertarian ratios by criminal justice policy makers is demonstrated. Predictive utility is relatively difficult to achieve for rare recidivistic events if the decisions involve severe deprivation of liberty (incarceration decisions). Predictive utility is easier to achieve for more common forms of recidivism or in decision contexts where there is less concern for false positives, such as for “intermediate sanctions.”