StatRec – Performance, validation and preservability of a static risk prediction instrument.

Abstract

StatRec is a prediction instrument for recidivism making use of only a limited number of static factors. In this paper, we discuss and update the StatRec scale for four-year reconviction and evaluate its predictive performance over several dimensions: time, region and non-random subsamples. Additionally, using criminal file data we investigate to what extend adding dynamic factors improves the predictive performance. The predictive performance of the scale proves to be relatively stable over time and comparable to the England and Wales’ OGRS-scales. The precision of the estimated probabilities is the only indicator that decreases slightly as time passes. Though the scale does not make use of dynamic and situational factors related to the risk of re-offending, criminal file analyses show that adding these enhance the predictive power only slightly. The scale could play a role in the validation of more specified dynamic risk instruments.