Criminal Defense Lawyers

Public Safety Assessment (“PSA”)

The Public Safety Assessment: An Evidence-Based Tool for Pretrial Release Decisions

The pretrial release decision is one of the most consequential in the criminal justice system. Judges must weigh the rights of defendants against public safety concerns when determining whether to release or detain defendants awaiting trial. Historically, judges relied heavily on money bail to manage pretrial release. But money bail has proven to disproportionately impact low-income defendants who cannot afford to post bond.

In response, jurisdictions across the U.S. have adopted evidence-based risk assessment tools to guide pretrial release decisions. These tools evaluate factors like criminal history and age to categorize defendants as low, moderate or high-risk if released pretrial. The Public Safety Assessment (PSA) is one such tool gaining popularity nationwide.

What is the Public Safety Assessment?

The PSA is an actuarial pretrial risk assessment tool developed by the Laura and John Arnold Foundation (now Arnold Ventures) in consultation with criminal justice experts. First implemented in 2012, the PSA aims to provide judges with data-driven, objective information to inform pretrial release decisions. It differs from other risk assessments because it does not use factors like income, education level, or neighborhood that could introduce socioeconomic or racial bias.

The PSA examines nine risk factors to evaluate a defendant’s risk of:

• Failure to appear (FTA) in court
• New criminal activity (NCA) if released pretrial

The nine factors are:

• Age at current arrest
• Current violent offense
• Pending charge at time of offense
• Prior misdemeanor conviction
• Prior felony conviction
• Prior violent conviction
• Prior failure to appear in the past two years
• Prior failure to appear older than two years
• Prior sentence to incarceration

The PSA uses only publicly available data and takes about 5-7 minutes to complete. It generates two risk scores – one for FTA and one for NCA – ranging from 1 (lowest) to 6 (highest). The FTA score determines whether the PSA recommends release on own recognizance or with monitoring. The NCA score indicates whether to flag the defendant as potentially violent if released.

The Need for Evidence-Based Pretrial Practices

Historically, judges made subjective determinations about whether to release or detain defendants pretrial. This resulted in inconsistent decisions and overreliance on financial conditions like cash bail that disadvantage the poor.

Pretrial detention also has significant consequences. Research shows detained defendants have higher conviction rates and longer sentences – likely because they have limited ability to meet with counsel and prepare a defense. Pretrial detention destabilizes employment, housing, and family life. And it exacerbates racial disparities, as Black and Hispanic defendants are detained pretrial more often than white defendants.

Jurisdictions nationwide have worked to reform pretrial practices to be more objective, transparent and data-driven. The federal First Step Act passed in 2018 promotes the use of risk assessment tools to guide pretrial decisions. As of 2019, 42 states used evidence-based pretrial practices statewide.

Why the Public Safety Assessment?

The PSA’s creators designed it to be transparent, racially neutral and empirically validated. They published detailed information on the tool’s development process and made training free for any jurisdiction. This transparency contrasts with some commercial risk assessment tools that do not share information on their proprietary algorithms.

The PSA excludes factors like race, income and zip code that could introduce bias. It relies solely on criminal history data found to predict pretrial misconduct. Automating data collection from court records also reduces human error during risk evaluation.

Perhaps most importantly, the PSA undergoes ongoing independent validation. Multiple studies have analyzed its predictive accuracy, finding it is effective at assessing risk of pretrial misconduct across jurisdictions.

Validating the PSA’s Predictive Accuracy

Validation studies test a tool’s ability to differentiate between high and low-risk defendants in a specific jurisdiction. They ensure statistically significant relationships between the factors and outcomes being predicted.

Initial validation of the PSA in over 300,000 cases across the country found it strongly predicted both FTA and NCA. For example, only 8% of defendants scored 1-2 for FTA failed to appear, versus 46% of those scored 5-6.Many jurisdictions contracting with Arnold Ventures to implement the PSA require an independent validation study. Multiple studies have confirmed the PSA’s accuracy in predicting pretrial misconduct and compliance with court dates in various jurisdictions.

A 2021 study in Fulton County, Georgia found the PSA strongly predicted both FTA and NCA. Only 7% of those scored 1-3 for FTA failed to appear, versus 61% of those scored 4-6. Just 1% of defendants without an NCA flag were rearrested pretrial, versus 16% of those with the flag.

Some studies also assess predictive bias – whether the PSA predicts equally well for all racial groups. A Kentucky study found some differences in predictive accuracy by race, but concluded they were unlikely to significantly impact release recommendations.

PSA Implementation and Impact

Many jurisdictions have implemented the PSA as part of broader pretrial reforms focused on reducing unnecessary detention. Between 2013-2018, over 40 sites adopted the PSA.

Early research on PSA outcomes is promising:

• New Jersey adopted the PSA statewide in 2017 along with eliminating cash bail. From 2015-2017, the pretrial jail population decreased by over 30% with no change in crime rates. Judges followed PSA recommendations over 90% of the time.
• After Mecklenburg County, NC implemented the PSA in 2015, the pretrial release rate rose from 70% to 85% while pretrial crime rates stayed the same. The jail population decreased and more defendants made all court appearances.
• Yakima County, WA saw significant reductions in failure to appear rates, pretrial length of stay and pretrial bed days after implementing the PSA in 2013.

While additional research is needed, these findings suggest jurisdictions using the PSA alongside other reforms safely reduce unnecessary pretrial detention without jeopardizing public safety or court appearance rates.

Concerns and Limitations of Pretrial Risk Assessments

Despite promising evidence, some express concerns about relying on algorithmic risk assessments in criminal justice decisions:

• Tools like the PSA do not eliminate subjective judgment. Judges can override scores. Some argue formulas should not replace judicial discretion.
• Even “race-neutral” factors like criminal history can disproportionately classify minorities as high-risk due to past overpolicing.
• Risk scores may make judges more comfortable detaining defendants labeled high-risk. But no tool can perfectly predict human behavior.
• Some oppose risk-based pretrial systems that allow detention based on potential future actions rather than actual charges.
• Assessing risk does not address root causes of pretrial misconduct like poverty and mental illness.

While these concerns are valid, evidence suggests appropriately implementing and contextualizing pretrial risk information improves fairness. The PSA aims to provide judges with better data to counteract implicit biases that can infect intuition-based decisions. Ongoing training and monitoring helps ensure scores do not unduly influence release decisions.

The Future of Pretrial Risk Assessment

The PSA represents a shift toward data-driven pretrial decision making. But risk assessment is not a panacea. The tool must be implemented alongside reforms like strengthening pretrial services and limiting detention eligibility.

As algorithmic governance expands, transparency and accountability are critical. Arnold Ventures’ commitment to evaluating and improving the PSA is a model. But more research is needed on how predictive tools impact pretrial outcomes for all groups.

With appropriate safeguards and human oversight, actuarial risk assessments like the PSA can make the pretrial system fairer and more effective. But they are only one piece of creating an empirically-grounded, racially equitable pretrial process.