PSA Research

PSA Research

The PSA: A Re-Validation and Assessment of Predictive Utility and Differential Prediction by Race and Gender in Kentucky

Authors: Matthew DeMichele, Peter Baumgartner, Michael Wenger, Kelle Barrick, Megan Comfort & Shilpi Misra

The authors use a statewide dataset from Kentucky of more than 164,000 cases to test three elements of the PSA. First, they assess the overall predictive validity of the PSA. Second, DeMichele et al. assess differential validity and predictive bias between black and white defendants. Third, they conduct the same analyses to determine accuracy and assess if bias exists between male and female defendants.

Key Findings:

•  Overall, the PSA predicts at a level that is considered statistically “good” by researchers in the criminal justice field.

•  Defendants with a New Violent Criminal Activity (NVCA) flag are 3 times more likely to commit an NVCA than those who are not flagged.

•  The PSA predicts equally well for black and white defendants for New Criminal Activity (NCA) and NVCA.

•     The PSA is free of predictive gender bias for Failure to Appear (FTA) and NCA.

Full Study: PDF, SSRN

What do Criminal Justice Professionals Think About Risk Assessment at Pretrial?

Authors: Matthew DeMichele, Peter Baumgartner, Kelle Barrick, Megan Comfort, Samuel Scaggs & Shilpi Misra

The authors describe findings from surveys (72% response rate) with judges, prosecutors, defenders, and pretrial staff in 30 jurisdictions using the Public Safety Assessment (PSA). The surveys assessed stakeholders’ experiences applying the PSA in their decision making; training and technical assistance received to use the PSA; views about racial disparity and the PSA; and perceptions of strengths and weakness of the PSA. DeMichele et al. frame local jurisdictions as courtroom communities in which criminal justice actors share decision-making responsibilities, but these actors have unique perceptions of risk assessment based on their role in the local criminal justice system. Stakeholder views and use of the PSA are critical because substantial discretion exists in whether and how they implement recommendations.

Key Findings:

•  79 percent of judges surveyed report that the PSA’s recommendation “always” or “often” informs their decision making.

•  61 percent of all respondents report they “often” agree with the PSA’s recommendation.

•  69 percent of all respondents report they consider it a strength that the PSA is research-based.

Full Study: PDFSSRN

The Intuitive-Override Model: Nudging Judges Toward Pretrial Risk Assessment Instruments

Authors: Matthew DeMichele, Megan Comfort, Shilpi Misra, Kelle Barrick & Peter Baumgartner

The authors identify and examine the various factors that influence judicial decision making about pretrial release decision. DeMichele et al. conducted in-depth interviews with judges in a diverse set of courts that were using the Public Safety Assessment (PSA) as a decision making took for pretrial release. Interview questions gained insight into how judges define and assess risk, as well as how they perceive the potential for bias and disparate impacts for communities of color in the use of risk assessments. The analysis is informed by Guthrie and colleagues’ intuitive-override model and suggest that risk assessment instruments can help judges engage both intuitive and deliberative models of decision making.

Key Findings:

•  Judges try to reconcile the actuarial aspect of the PSA with their experience-based inclination to learn about defendants’ lives.

•  Judges develop their own roster of information they seek when considering a case, and there is variability in the factors they value.

•  Judges viewed the PSA “one tool among many.”

Full Study: PDF, SSRN