Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Risk Assessment topic
No spam. Unsubscribe anytime.
Revalidation shows Commission's risk tool retains moderate predictive power and no clear racial accuracy gap
Summary
Staff presented a statistical revalidation of the Commission's risk assessment that used 2015–2016 sentencing cohorts and a three‑year conviction-based recidivism outcome; the AUC was about 0.65 (previously 0.66) and staff reported no meaningful difference in predictive accuracy for Black and White individuals in the validation sample.
Get email alerts on the Risk Assessment topic
No spam. Unsubscribe anytime.
Commission research staff presented results of a formal revalidation of the Commission’s sentencing-time risk instrument. The staff said the revalidation used people sentenced in 2015–2016 and tracked convictions that resulted from arrests within three years as the outcome measure.
The senior statistician reported that the instrument’s Area Under the Curve (AUC) for the original development sample was 0.66 and 0.65 in the revalidation sample — measures the instrument’s ability to rank-order individuals who did and did not recidivate. Staff said the recidivism rate in both samples was roughly 33 percent overall and that, for given risk scores and categories (low, typical, high), recidivism rates were similar across the original and validation samples.
Staff also presented racial impact analyses that compared the tool’s accuracy for Black and White individuals. The statistician said the AUC values and other predictive metrics were similar by race in the analysis presented and that other racial impact measures in the full written report showed no meaningful disparity in predictive ability in the revalidation sample.
The meeting included a reminder about the tool’s intended use: the instrument is advisory at the time of sentencing and the commission’s approach emphasizes providing judges with a recommendation to obtain additional information (for example, a presentence investigation or a risk‑needs assessment) rather than an instruction to impose a particular sentence.
Ending: Staff said the written revalidation report will be finalized in the fall; the staff recommendation at the meeting was that no substantial revision to the instrument is required based on the current analysis.

