Research Summary

Pretrial reform is vital, as nearly 500,000 un-convicted people are held in U.S. jails while awaiting trial. Risk assessment instruments (RAIs) offer data-driven identification of defendants who are unlikely to reoffend and can safely be released, but face criticism from both ends of the political spectrum—for potentially perpetuating racial bias or endangering public safety. Moreover, inconsistent application by judges undermines the effectiveness of RAIs.

Using data on 146,841 federal defendants, we apply causal analyses with machine learning to estimate the effects of a policy that presumptively releases defendants classified as relatively low-risk by the Pretrial Risk Assessment Instrument (PTRA). Compared to magistrate judges’ status quo decisions, this approach would reduce pretrial detention by 34.2% and increase successful community releases by 31.8%, with only a 1.6% rise in public safety risk. Black defendants would experience greater benefits (39.0% detention reduction versus 27.3% for White defendants). Detention costs savings would be approximately $3.5 billion.  

Policy Implications

 Although the PTRA outperforms unstructured judgment, it should strongly guide rather than replace judicial discretion. The challenge is to structure judgment in pretrial decision-making.

As an initial step, the federal system should expand inclusion of PTRA risk estimates in pretrial reports beyond the current 15% of districts, implementing standardized reporting formats to ensure clear communication and consistent application. For greater impact, policymakers should consider a risk-based presumptive release framework where defendants below a specified risk threshold are released unless magistrate judges identify specific statutory factors justifying detention. This approach addresses inconsistent RAI application while preserving judicial authority. 

Implementation will require investment in trust-building and practice fidelity, but structuring pretrial decisions around validated risk measures promises to enhance fairness, reduce costs, and maintain public safety—outcomes with broad political appeal.