Text

theses

With an increasing number of offenders released to the community from prisons each year, prisoner reentry has become an important area of focus for criminal justice practitioners and policymakers. Parole supervision was originally established as a means of offender reintegration, but high rates of re-offending present a challenge to discretionary release to parole supervision. As a result, the most important function of parole supervision has shifted to the protection of public safety, leading to the development of risk prediction instruments used to identify offender risk to the community. While early forms of risk assessment used limited practices of professional judgment to predict offender risk, more contemporary risk assessment instruments use individual factors known to affect recidivism to determine offenders who pose the greatest risk to society once they are released from prison. The present study examined the reliability and validity of one such instrument, the Correctional Offender Management for Profiling Alternative Sanctions (COMPAS) tool, using a diverse sample of male offenders released to parole supervision in New York City. Specifically, this study assessed the psychometric properties of the COMPAS using correlations and internal consistency estimates to assess reliability. The predictive validity of the COMPAS composite risk scores on several recidivism outcomes was assessed using AUC and RIOC analyses. Additionally, this dissertation study compared the predictive validity of the two composite COMPAS risk scores with the predictive validity of risk scores calculating using several static factors from each offender’s computerized case history (CCH) known as the DCJS any risk and risk for violence scores on a re-arrest for any crime and a re-arrest for a violent crime. Results from this study indicate that the COMPAS is a reliable risk/needs instrument with this sample. While the COMPAS composite recidivism risk and history of noncompliance scores achieve moderate predictive validity on their corresponding outcomes, the COMPAS composite violent recidivism risk score does not. When the COMPAS composite recidivism risk score is compared to the DCJS any risk score on validity in predicting re-arrest, the DCJS risk score, based solely on static variables, out-performs the COMPAS. The implications of this research for parole practice and policy, both locally in the State of New York and more broadly, as well as directions for future research, are also discussed.

ETD_6049

Ph.D.

Includes bibliographical references

Includes vita

by Bryn A. Herrschaft

doi:10.7282/T38C9XX2

ETD doctoral

The author owns the copyright to this work.