Incorporating Dynamic Measures of Risk and Desistence into Risk Assessment Instruments for Community Supervision in Georgia

• In Event: Innovations in Corrections: Findings and Lessons Learned from Using Technology in the Field

Abstract

Actuarial risk assessments of rearrest built on static factors such as criminal history have been used in community supervision since the early 20th century. Newer generations of instruments have sought to incorporate more dynamic factors. However, models that rely on static measures or that privilege only risk factors, may struggle to produce varying predictions over time. This study leverages the findings from focus groups with community supervision officers to inform predictive models of rearrest. To model rearrest, we follow a cohort of 87,000 probationers and parolees in Georgia between 2016 and 2019 and employ survival models and time-specific classification models that use both protective and risk factors. We also test the utility of measuring time- and charge-specific criminal history. Qualitative findings indicate that risk scores derived from mostly static factors were less useful in assessing risk over the course of supervision. Incorporating both protective and risk factors in models resulted in greater accuracy compared to static models or those omitting protective factors. Furthermore, models that used only recent criminal history did not compromise accuracy. This study provides support for including dynamic measures that are positively and negatively associated with the risk of rearrest while on supervision.

Authors

• Christopher Inkpen, RTI International

• Pamela K. Lattimore, RTI International

• Sharon Johnson, Applied Research Services, Inc.

• Emily Burtch, RTI International