Can risk-based release policies improve pretrial outcomes, compared to unstructured judicial decisions?
We analyze how officers and judges currently decide who to keep in jail before trial, then model the impact of a policy that would release defendants classified as low risk by the Pretrial Risk Assessment (PTRA).
Context of pretrial reform
Nearly 500,000 unconvicted individuals are held in U.S. jails each year while awaiting trial—often because they cannot afford bail or because decision-makers default to detention. Risk Assessment Instruments (RAIs) offer an evidence-based alternative by estimating each defendant’s likelihood of failure to appear and reoffending, thereby identifying who can safely await trial in the community. However, concerns about algorithmic bias and uneven judicial adoption limit their real-world impact.
Comparing unstructured judgment with a hypothetical presumptive low risk release policy
Our first study of nearly 150,000 federal defendants (Skeem, Montoya & Lowenkamp, 2022) revealed that officers recommend detention for Black defendants at higher rates than White defendants—largely due to excessive emphasis on criminal history. In a second study, we estimated how much defendants’ outcomes would improve if judges’ actual pretrial decisions were replaced by a hypothetical policy that released defendants who fell below a high-risk score on the Pretrial Risk Assessment instrument (PTRA).
Presumptive low-risk release pretrial policies have been recommended because judges can—and often do—ignore risk-based recommendations. Under such a policy, defendants classified as low risk by an RAI would be released unless the judge identified specific circumstances that warranted detention and met a standard for departure. By limiting incarceration to those who pose a genuine public-safety threat, this approach aims to curb unnecessary detention.
Key findings
- Boost successful releases by more than one-third, while increasing failures to appear or reoffending by less than 1%,
- Save $3.5 billion in avoidable detention costs, and
- Advance racial equity, reducing detention for Black defendants by 42% versus 33% for White defendants.
Implications & next steps
These findings strongly support piloting a risk-based presumptive release policy that melds established categorical release frameworks with validated risk assessments.
In the shorter term, two targeted improvements could deliver immediate impact:
- “Right-size” criminal-history weighting. Ensure that prior arrests and convictions inform release decisions without overwhelming other factors.
- Elevate validated risk scores. Pretrial reports and other decision-support tools should present risk estimates in clear, concise formats so practitioners can interpret, trust, and act on them.
Although the federal Pretrial Risk Assessment is well validated, only 15 percent of districts currently include its scores in pretrial reports. Consistently integrating these scores—and refining how they’re communicated and displayed—could increase their influence on judicial decisions and thereby improve pretrial outcomes.
To bridge this gap, we’re partnering with magistrate judges on pilot projects to identify the most effective ways to present criminal history and risk information. The goal is to co-design report templates and digital interfaces that make risk estimates useful to, and valued by practitioners.
Partners & funding
Jennifer Skeem led this project, in collaboration with Lina Montoya and Christopher Lowenkamp. This project was made possible through our partnership with Pretrial and Probation Services (though affiliation does not imply endorsement).