Risk, Race & Recidivism: Predictive Bias and Disparate Impact

Abstract:      

One way to unwind mass incarceration without compromising public safety is to use risk assessment instruments in sentencing and corrections. These instruments figure prominently in current reforms, but controversy has begun to swirl around their use. The principal concern is that benefits in crime control will be offset by costs in social justice — a disparate and adverse effect on racial minorities and the poor. Based on a sample of 34,794 federal offenders, we empirically examine the relationships among race (Black vs. White), actuarial risk assessment (the Post Conviction Risk Assessment [PCRA]), and future arrest (for any/violent crime). First, application of well-established principles of psychological science revealed little evidence of test bias for the PCRA — the instrument strongly predicts arrest for both Black and White offenders and a given score has essentially the same meaning--i.e., same probability of recidivism — across groups. Second, Black offenders obtain higher average scores on the PCRA than White offenders (d=.34; 13.5% non-overlap in groups’ scores). Although groups’ scores largely overlap, some applications of the PCRA could create disparate impact — which is defined by moral rather than empirical criteria. Third, most (66%) of the racial difference in PCRA scores is attributable to criminal history — which strongly predicts recidivism for both groups and is embedded in sentencing guidelines. Finally, criminal history is not a proxy for race — instead, criminal history mediates the otherwise weak relationship between race and future violent arrest. Data may be more helpful than rhetoric, if the goal is to improve practice at this opportune moment in history.