Pretrial Risk Assessment Tool (Arnold Foundation)
Laura and John Arnold Foundation · United States (multiple jurisdictions) · 2014
Summary
The Arnold Foundation's Public Safety Assessment provides judges with algorithmic flight risk and new-crime risk scores to inform bail decisions. Implementation evaluations have produced mixed results: some jurisdictions achieved meaningful detention reductions, others did not. A consistent finding is that judges use the tool selectively — anchoring on high-risk scores more than low-risk scores — which can widen racial disparities rather than narrow them if not implemented carefully. The tool is now one of the most debated algorithmic interventions in public policy.
Research question
"Does providing judges with algorithmic risk scores at arraignment reduce pretrial detention without increasing failure to appear or rearrest?"
Methodology
Intervention
Jurisdictions implementing Public Safety Assessment (PSA) risk tool provided judges with flight risk and new-crime risk scores vs. standard practice
Assignment
Quasi-experimental (staggered jurisdiction adoption)
Sample size
750,000+ arraignment decisions across 6 jurisdictions
Primary outcome
Pretrial detention rates; failure to appear; pretrial rearrest
Effect estimate
Mixed: some jurisdictions reduced detention without increasing FTA; others saw no change or increased detention of low-risk defendants
Decision
Tool adopted in 40+ jurisdictions; equity audits required in several states after racial disparity concerns
Result
Mixed
Mixed: some jurisdictions reduced detention without increasing FTA; others saw no change or increased detention of low-risk defendants
Evidence strength
Moderate
Quasi-experimental design; causal interpretation requires care.
Replication status
Partially replicated
Institution
Laura and John Arnold Foundation
Location
United States (multiple jurisdictions)
Year
2014
Policy area
Public Safety
Mechanism
Information