By Hena Rafiq and Emily Audet
From October 2020 to June 2022, the Harvard Kennedy School Government Performance Lab (GPL) led a pilot program in Harris County, Texas, to transform how judges received data from the pretrial services agency. Under the pilot, judges reduced the intensity of supervision requirements for over 2,200 defendants released into the community while awaiting trial. Even as supervision orders were lowered for thousands of people, rates of re-arrest and compliance with bond conditions did not change. What started as a select group of six judges initially participating in the pilot more than doubled to 13 judges – one-third of the criminal court’s bench – and at the end of the pilot, an additional seven judges were on a waitlist to join.
The reductions, or “step-downs,” to conditions of pretrial release issued by judges through the pilot freed up capacity for Harris County Pretrial Services staff, who had to administer and monitor these orders, and for defendants awaiting trial. For instance, before a step-down, a defendant required to take semimonthly drug tests would be randomly assigned two testing days a month and would find out the morning of that they needed to show up in-person to take a test. This system left people scrambling to arrange all the necessary logistics – like arranging child care, finding transportation, and taking time off from work – in order to comply. A step-down could eliminate a person’s drug testing requirement, barring a change in their case, substantially reducing the demands on their time and making their schedule more predictable. GPL Project Leader Hena Rafiq, who spearheaded the pilot, described another way step-downs could reduce burdens on people awaiting trial: “People assigned alcohol monitors had to pay for their device each month, so removing the device requirement for even a few months of their release could save them hundreds of dollars."
Hena Rafiq recently spoke to an audience of over 80 criminal justice stakeholders, including judges and pretrial agency staff, during the Safety and Justice Challenge Network Meeting. The gathering brought together representatives from the 57 jurisdictions participating in the challenge, which aims to safely reduce the local jail population by half by 2025. Rafiq presented alongside Natalie Michailides, Director of Harris County Pretrial Services; Amy Dezember and Kevin Kuehmeier, Senior Associates at Justice System Partners; and David Asma, Lead Investigator at the Public Defender’s Office of Lake County, Illinois.
GPL Project Leader Hena Rafiq presented to the Safety and Justice Challenge Network Meeting on using data-driven performance management to overhaul decision-making around placing conditions on pretrial release.
Rafiq focused on what pretrial systems could learn from the use of data-driven performance management (DDPM), a core GPL tool which powered the pilot project in Harris County. The approach involves high-frequency reviews of key performance metrics with critical stakeholders to collaboratively improve target outcomes. “Data-driven performance management means using data in real-time to push forward important decisions,” Rafiq explained. The GPL applies data-driven performance management across all of its policy areas – including children and families, criminal justice, economic mobility, housing and homelessness, and procurement systems – to support state and local governments to develop effective and efficient public management systems.
The data-driven performance management approach, audience members shared, is far from common practice. For instance, one participant shared how they currently only report key metrics annually but want to review them more regularly. Another said that they collect and review monthly indicators within their agency but do not yet share these metrics with judges.
Data-driven performance management requires moving from data analysis to informed action.
In Harris County, the GPL’s guidance on implementing data-driven performance management led to three process improvements that helped judges make more informed decisions on the use of pretrial release conditions:
1. Making the feedback loop for judges more accurate
“Before the pilot,” Rafiq said, “judges would only be notified when someone violated their supervision conditions. This gave them the mistaken impression that most people were violating their conditions.” In reality, she pointed out, 70% of clients were fully compliant with all conditions of their bond.
Judges also lacked aggregated data on how they applied conditions across their caseloads. Rafiq added, “We were finding that judges were making decisions about bond conditions at an individual case level, and they had no clue that most of their caseloads had conditions.”
Through the pilot, the GPL created weekly reports on key metrics, including the percentage of defendants compliant with the conditions of their release, judges’ rates of applying conditions, and re-arrest rates among defendants. The GPL shared the reports with judges and convened them quarterly to give them a more accurate understanding of the interaction between defendants’ behavior and their own judgements.
The data-driven performance management model applied in Harris County included a range of touchpoints with judges to share and discuss key indicators of the program's success.
2. Proactively identifying defendants eligible to have their pretrial release conditions lowered, or "stepped down"
Before the pilot, a Harris County judge or defense attorney had to initiate the process of lowering a defendant’s bond conditions. Then, Harris County Pretrial Services would review the compliance history for each client. Under this system, reductions to conditions were ad hoc, as judicial officers and defense attorneys lacked access to comprehensive data on who was following the conditions of their release. Judges would typically set conditions at an initial hearing and would not revisit them throughout the course of the trial, even though most cases took over a year to be resolved. Step-downs only occurred if capacity-strained judges and attorneys were willing and able to take extra time on an individual’s case to pursue it. This system made it unrealistic for judges, who had over 1,000 cases on their docket at a time, to consistently apply step-downs to all defendants following the conditions of their bond.
To reduce this operational friction, the GPL convened judges to develop criteria that could make someone eligible for a step-down (for example, having two consecutive negative drug tests or complying with all bond conditions for two to three months). With these standards, along with updates to Pretrial Services’ record-keeping process, the GPL built an automated system that identified defendants eligible for step-downs. Judges then reviewed their lists, approving 99% of proposed step-downs. This transition streamlined the step-down process for judges, Pretrial Services staff, and attorneys, all while preserving judicial discrepancy.
3. Giving judges a point of reference to evaluate their decision-making
Prior to the introduction of data-driven performance management in Harris County, judges did not have a frame of reference for their use of bond conditions; they lacked a mechanism to assess their own decision-making relative to other judges in the county. To address this, the GPL brought judges together every four months to review data on the use of pretrial release conditions across courtrooms. As a result, judges who assigned conditions at higher rates than their peers began to reassess their approach and scale back their use of conditions. It also provided a shared space for peer-to-peer problem-solving among judges to brainstorm how to resolve challenging cases without remanding a defendant or placing additional supervision mandates on them.
As part of the GPL’s work in Harris County, judges came together to review and compare their use of pretrial release conditions. Providing this frame of reference caused some judges to reduce the intensity of their condition use. Note: A small amount of random noise has been added to the data to preserve courtroom anonymity.
In jurisdictions across the country, the GPL has worked with governments on using data-driven performance management to improve government operations and outcomes for residents. The Harris County pilot strengthened information flows to judges, allowing them to better calibrate their use of pretrial release conditions. Doing so reduced caseloads for Harris County Pretrial Services, allowing staff to focus on their most serious cases. At the same time, it reduced financial and time costs for defendants since they had to comply with fewer conditions.
“Initially, the GPL created a feedback loop for County Criminal Court judges. The pilot project that the GPL ran shared data with judges with the intent to adjust the use of pretrial release conditions based on defendant compliance, court appearance rates, and rates of new arrest among defendants awaiting trial,” said Natalie Michailides, Director of Harris County Pretrial Services. “As the judges saw the successes of removing conditions while maintaining important measures of safety, they became very invested and were eager to participate actively in the initiative. From my perspective, it laid the groundwork for better communication between judges and Pretrial Services.”
A recording of the presentation at the Safety and Justice Challenge Network Meeting is available on their website.