Tools for Safe and Smart Bail System Changes: Pretrial Assessments
This is part of the “Tools for Safe and Smart Bail System Changes” series. Read the introduction and view other posts here.
What are they?
Pretrial risk assessments are data-driven algorithms used to determine an individual’s likelihood of absconding or committing a crime while awaiting trial. Their purpose is to produce more accurate and consistent pretrial determinations by turning educated guesses into science-based predictions.
Several types of pretrial risk assessments exist, but they generally use historical data patterns to group people into different risk categories and predict outcomes. These predictions are not individualized; rather, assessments compare one person’s history to others who have previously succeeded or failed on pretrial release. They may use “static” factors (unchanging facts), “dynamic” factors (which capture attitude, behaviors and needs), or both.
The table below lists some of the better known, most widely used pretrial risk assessments and the factors they use to predict outcomes. (Download the table.)
Of course, pretrial risk assessments cannot predict success with complete accuracy. They are not meant to replace professional judgment or judicial discretion; rather, they provide additional information for a judge to consider.
Pretrial assessments can also facilitate release decisions before a defendant appears in front of a judge. They can supplement or replace a bail or bond schedule (a predetermined amount a defendant can pay before their first court appearance) or guide other delegated release authority, wherein the court authorizes another agency to make release decisions. When used in this manner, local decision-makers set policies and procedures to guide the interpretation of pretrial assessments.
What concerns do they address?
The U.S. Supreme Court has repeatedly held that pretrial detention should be the exception, not the rule. Judges, however—who are subject to natural human error—often choose to detain out of fear that an individual will commit a violent crime upon release. National news coverage of communities experiencing these tragic outcomes reinforce this fear. Consequently, more than 80 percent of individuals in county jails are being detained prior to conviction for a crime.
Public safety is a priority for most communities, and determining which individuals are safe for pretrial release is essential to maintaining it. By providing judges with a clearer understanding of an individual’s risk level, pretrial risk assessments can help prevent high-risk individuals from being released and ensure that low-risk individuals are not needlessly detained. Judges can also use them to help determine individualized nonmonetary conditions that support court attendance and address safety concerns.
Pretrial risk assessments can also help address racial disparities within the criminal justice system. Unfortunately, pretrial determinations have disproportionately impacted communities of color. While several components may contribute to this disparity, research shows that racial stereotypes and unconscious bias in judicial decisions could be the reason more people of color are detained prior to trial. Objective assessment tools could mitigate implicit and explicit discrimination from the bench.
Finally, the cost of pretrial incarceration is massive, with experts estimating a $13.6 billion annual cost to taxpayers. Pretrial risk assessments can save jails money and resources by expediting the release of low-level, low-risk offenders. They can also transform bail schedules into more effective tools. A traditional bail schedule assigns a dollar amount to each offense, allowing a defendant to pay that amount to be released from jail before their first court appearance—regardless of their flight or safety risk. Delegated release and pretrial risk assessment would ensure a more individualized approach.
Ultimately, pretrial risk assessments allow for more accurate, less biased release and detention decisions that maximize resources, improve fairness and increase public safety.
Are they effective?
Pretrial assessments can provide an empirical and objective review of a defendant’s ability to succeed on pretrial release. Proponents argue that more informed decisions on pretrial release can reduce unnecessary pretrial detentions and mitigate public safety risk. In fact, research shows that judges inappropriately release about half of the highest-risk defendants. Experts predict that using an objective assessment could result in a 24.7 percent reduction in failures to appear and an 11.1 percent reduction in rearrests, with no change in pretrial detention rates. It could also improve consistency between jurisdictions.
However, to be effective, these assessments must be as accurate as possible. To this end, pretrial assessments are validated using historical data. To compare the validity of predictive findings, most validation studies use “area under the curve” (AUC), which calculates the probability that an individual who was unsuccessful before trial would have been given a higher risk classification with a particular risk assessment tool. When interpreting AUC data for risk assessments, the following performance indicators are commonly used: Poor (less than 0.55), Fair (0.55-0.63), Good (0.64-0.70) and Excellent (0.71-1.00).
Validation studies support the accuracy of the six pretrial assessments described above. (Download the table.)
Thus, identifying the appropriate release of low-risk defendants not only frees up jail space, saving taxpayers money and prioritizing resources for violent offenders, but it can also reduce recidivism. Limiting the collateral consequences of detention, like loss of employment or housing, renders low-level offenders less likely to commit future crimes. Further, predicting pretrial outcomes more accurately ensures that resources are prioritized for high-risk offenders—thereby increasing public safety.
What challenges or concerns do they present?
Researchers have challenged whether assessments are any more predictive than untrained humans. One study concluded that untrained individuals given random samples describing a defendant’s sex, age, current charge, degree of offense and criminal history accurately predicted recidivism rates at an average of 64 to 67 percent. Not only is this rate of accuracy equal to or greater than that of pretrial assessments, significantly fewer factors were considered when making these predictions. This study did not include failures to appear or violent criminal activity.
Research on judges overriding assessment recommendations is also troubling. Because predictions often fail to account for interventions that might reduce risk—such as court date reminders, transportation assistance or check-ins with a pretrial services agency—judges are much more likely to override a recommendation for release than for detention. This can lead to increased and unnecessary detentions.
Further, while some experts promote pretrial assessments as tools to reduce racial disparities in pretrial detention, opponents have raised concerns about their potential biases. Critics caution that because predictions rely on historical data tainted by systemic racism within the criminal justice system, these assessments may inadvertently produce less accurate outcomes for traditionally marginalized communities. Because assessment tools largely ignore race, they are unable to counter or correct this issue.
Lastly, jurisdictions may have concerns regarding the cost of implementing pretrial risk assessments. While there are upfront costs to consider, studies show that assessments can actually save communities millions of dollars in the long run.
Conclusion
Releasing someone from jail before trial is a calculated risk based on the need to balance public safety with the individual’s right to be presumed innocent. Though pretrial risk assessments are not perfect, they can be a useful tool in this calculation. On the front end, they can expedite and individualize recommendations for release; in court, they can supplement judicial discretion. It is important, however, to continue evaluating these tools for accuracy and inequity and to improve them accordingly.