Data analysis and use often occur only within an executive team and may lack a clear connection to office goals. Prosecutors may not see data as a decision-making tool but instead as a means of oversight, said Stemen.

Measures for Justice is a nonprofit organization with the goal of bringing data transparency to the criminal justice system, said Gipsy Escobar, Measures for Justice. A national data portal contains core data on the performance of the criminal justice system from 20 states and 1,200 counties, said Escobar. The organization has worked with communities, to allow them to engage with the data, and with prosecutors, to create and track policy goals. Further, Measures for Justice developed tools to evaluate the data culture, data infrastructure, and data quality in prosecutors’ offices. Through a partnership with the Association for Prosecuting Attorneys, aiming to collect data from 10 prosecutors’ offices, Measures for Justice identified several common data challenges. One challenge involves missing information across fields; for example, at one site, 90 percent of cases were missing judge information. Escobar noted that there are artificial intelligence (AI) tools in development to analyze judges’ decision-making patterns; however, these tools are ineffective if 90 percent of the necessary data are missing. Invalid values are another common challenge; this happens when the information entered in a field does not fall into one of the assigned value categories. Systems also exhibited inconsistent case outcomes, said Escobar, with cases being marked as “open” but with all charges dismissed. In addition, event dates were sometimes improperly recorded—11 percent of cases had an arrest date that preceded the incident date. Finally, many cases, such as expunged or declined cases, are removed from datasets. All these missing or incorrect data hinder the ability to use data to inform and guide practice, said Escobar.

Oren Gur, director of the District Attorney’s Transparency Analytics (DATA) Lab, Philadelphia’s District Attorney’s (DA) Office, said using data teams, developing data-informed policy, researching policy and practice, and influencing public discourse through data are all emerging areas of focus for prosecutors. Multiple agencies within the criminal justice system generate and use data. Police generate data on incidents and arrests, prosecutors on charges and cases, and courts on outcomes, he said. Prosecutors may have access to both police and court data and can combine these data to systematically examine current trends and practices. However, prosecutors’ offices have been historically underresourced and underleveraged from a data and research perspective, limiting this ability. In Philadelphia, the DATA Lab works within

the DA’s office to “use data, research, and advocacy to inform policies and practices; increase equity, transparency, and accountability; and reduce harms through prosecutorial and systems reform in Philadelphia and beyond.” Gur called attention to the phrase “and beyond,” noting that learnings from Philadelphia can be applied more broadly. One objective of the DATA Lab is to become an integral component of the DA’s office; Gur said that such institutionalization is essential for sustainability. The DATA Lab received initial funding from foundations but is working to procure additional funding. The daily work of the DATA Lab falls into five areas, said Gur: operations, research, policy/practice, transparency, and community. This multipronged approach allows shifts in capacity, to focus on areas in need of extra attention, he said.

Aurélie Ouss, University of Pennsylvania, described how collaborations between criminal justice researchers and prosecutors’ offices can improve programs, policies, and practices. The research partnership between University of Pennsylvania and the Philadelphia DA’s office, established in 2020, has several goals, said Ouss. The first goal involves understanding the role prosecutors play in shaping criminal justice outcomes, and the effects of policies such as cash bail reform or diversion programs. Second, the partnership aims to develop data opportunities, both for graduate students and for researchers outside of the University of Pennsylvania. For example, several researchers are embedded within the DA’s office and can access all data available to prosecutors, said Ouss; they can develop their own research agendas around prosecution and criminal justice. Third, beyond prosecution, the partnership is leveraging this collaboration to improve understanding of how penal decisions are shaped. Prosecutors are central to many processes, she explained, because they receive data from both police departments and courts, and they generate their own data through their case management systems.

One critical aspect of the partnership is the dissemination of the work. Ouss gave an example of research conducted and disseminated to address an issue within the criminal justice system. Ouss and her colleague researched bail reform (Ouss & Stevenson, 2023) and noticed that the term “Failure to Appear” (FTA) was associated not just with defendants but often with other court actors such as police officers, victims, civilian witnesses, or defense attorneys. The team dug deeper into court records to understand the scale of this issue, said Ouss, and found that nondefendants were far more likely to miss court than defendants (Graef et al., 2024). For example, defendants missed at least one court hearing in 19 percent of cases; by contrast, police

officers missed court in 30 percent of cases they were ordered to attend. In addition, Ouss and colleague found that cases in which at least one witness missed court were twice as likely to be dismissed as cases in which no one missed court; in fact, witness FTA is the most significant predictor of a case being dropped—more so than charge, criminal history, and demographics combined. Ouss and colleagues presented their findings internally to the prosecutor’s office, working with unit supervisors to present results in the most understandable format, including simple visuals. Alongside the DATA Lab, they conducted research-to-action meetings to help disseminate results internally and externally. The results did not surprise prosecutors, she said, but providing data helped elevate the issue. The work has continued, said Ouss, through monthly meetings with stakeholder groups who are working to address the issue. This example illustrates how doing research with the DA’s office can help elevate issues that reach beyond the office and can help bring stakeholders together to work on improving the criminal justice system, said Ouss.

Researchers can engage in mutually beneficial collaborations with DA offices in multiple ways, said Ouss. When DA offices give researchers access to data, researchers can both answer important academic and policy questions and find ways to make data and data analysis helpful to prosecutors’ daily work. DA offices may be unsure how to use much of the data they collect, and researchers can help organize and interpret these data to help prosecutors identify challenges and develop solutions. Data analysis may, in turn, reveal areas in which more data collection would be helpful to provide a clearer understanding of criminal justice system dynamics and operations. The first step in this collaboration, said Ouss, is for researchers to take time to listen to and understand the office’s needs and questions. Once researchers understand the office’s needs, they can use available data or collect new data, translating these data into formats helpful to all stakeholders, from defendants to victims to prosecutors.

Ann Davison, City Attorney for Seattle, Washington. is the city’s independently elected prosecutor, overseeing the criminal division with 80 attorneys and staff and the civil division with 110 attorneys and staff. The criminal division receives approximately 10,000–12,000 gross and simple misdemeanor referrals for prosecution each year from the Seattle Police Department. Davison presented two case studies of challenges in Seattle and described the approach she and her office took to address those challenges.

tracking definitions document, and a timeline of policy and process changes to help provide context to the data. The team has also focused on data literacy, said Wong, by providing trainings and community presentations, and has taken steps to grow the office’s data culture by increasing data accessibility and incorporating feedback from community members and practitioners. The goal, she said, is for prosecutor’s office attorneys and staff to have the data they need to do their jobs well.

Wong explained that she serves as a bridge between the practices of prosecution and data analytics. She was assigned to work with nontraditional diversion programs and found herself collaborating with outside researchers to ensure that alternatives to traditional prosecution were effective. Wong said she learned that collaboration and communication with other prosecutors and the community are essential. Rather than just asking a DA to collect data for a project, it is better to sit with the DA to explain the process and why the information is needed, she said. Similarly, Wong spent time listening to the community to find out which types of information they would find meaningful, so the office could collect and share these data. Meaningful, two-way communication about data needs is important, she said. For example, the data team can ask prosecutors which kinds of data could make their jobs easier, either in the process of prosecution or in talking with victims and the community. Obtaining feedback from practitioners and gathering and sharing relevant data, said Wong, have been powerful ways to build data culture and to secure support from practitioners.

PROSECTUOR’S OFFICE: COOK COUNTY

In an earlier workshop session, DA Kimberly Foxx shared that her approach to prosecution involves using data to show results. When she was first elected in 2016, Foxx said she knew that constituents would want to see results of the changes made during her term. Her office built open data portals to collect and share county-wide crime and justice information. While data and research are essential for making evidence-based changes, Foxx said, data are not always valued—instead, people tend to rely on whether they feel safe.

Working with researchers to collect data is essential for making policy changes, said Foxx, because “you cannot fix what you cannot measure.” However, such work requires lawyers to exhibit a level of humility, said Foxx. Tension can exist between the lawyer’s real-life experiences and the researcher’s academic expertise. Foxx gave an example from her early

career illustrating the importance of working with researchers. Researchers were working to analyze the composition of the county jail population and the underlying reasons for their incarceration. Data showed that many people who were in jail were there because they could not afford bail. It was Foxx’s job to explain the research to other actors in the criminal legal system and the community.

Chicago has an array of academic institutions that are eager to work with prosecutors and others to conduct research on the criminal legal system, said Foxx. Over the past eight years, the field of prosecution has embraced data and formed research-practice partnerships. This demonstrates an emerging acknowledgement among prosecutors, said Foxx, that “we cannot practice the way that we’ve practiced before”—instead, it is important to engage with research and be thoughtful in the work. Prosecution is a risk-averse field, she said, and prosecutors might worry that research could show that their policies are not having the intended effect, or that findings of unfairness or racism might suggest that individuals working in the criminal justice system are themselves unfair or racist. Relying on research rather than anecdotal evidence is not the traditional culture of prosecution, said Foxx. However, over the last few years, this culture has started to shift as prosecutors begin to engage in research. “Institutional bravery” is needed to continue the evidence-based prosecution approach, said Foxx.

DATA ON RACE AND ETHNICITY

In a workshop session focused on data use and data culture in prosecutors’ offices (see Appendix B), Kutateladze asked panelists to comment on how race and ethnicity data are captured. Stemen replied that, in most jurisdictions he works with, race and ethnicity data are captured by law enforcement and transferred to prosecutors’ offices when the case is referred. Some offices have tried to confirm these data by cross-checking it against other available data. For example, he said, last names can be checked against census records and classified as Latino if an individual’s last name is among those for which 85 percent or more people self-identify as Latino. Driver’s license records can also be used to cross-check; linking with the Department of Motor Vehicles’ records allows access to race and ethnicity data self-identified by individuals. Escobar agreed that, in her experience, most race and ethnicity data come from the arresting officer. Prosecutors generally have little contact with the defendant outside of court, she noted, resulting in a lack of opportunity to collect data firsthand. Gur suggested that, in addition to

linking records, AI could accelerate the process of classifying complex race and ethnicity data. Escobar added that challenges exist when using technology to identify race based on images of people.

Race and ethnicity data can be challenging to capture, Escobar said, because of the difference between actual and perceived race and ethnicity. A new law in California, the Racial and Identity Profiling Act (RIPA), requires law enforcement officers to report the perceived race of people who have certain types of contact with law enforcement.¹⁵ Perceived race and ethnicity are important, said Escobar, because any bias or advantage associated with a defendant’s perceived race follows them throughout the criminal justice system. Ideally, data on both actual and perceived race and ethnicity would be collected to study issues like bias and disparities. Steven Raphael added that the RIPA law requires officers to check boxes for any racial identity they perceive for the defendant, and there is concern that officers might undermine the system by checking every box. Another challenge, he said, is that perception is malleable—a defendant’s officer-perceived race may change depending on many factors. Gur said that he has seen cases in which a person’s recorded race and ethnicity change each time they interact with the criminal justice system; this is due to data obtained from a variety of sources that collect data differently (e.g., self-report vs. perceived). Gur agreed with Stemen that collaborating with other agencies that collect self-reported race and ethnicity data could increase data reliability. Wong concurred with the need for more reliable data and said that an effort in Multnomah County, Oregon to reconcile race records among the jail, court, and prosecutor systems found that 19 percent of all cases had mismatched data on race. Another challenge with self-reported race and ethnicity data, said Kutateladze, is that people may not see themselves as fitting into the prescribed categories. For example, in his work on hate crimes in Miami, he found that people of Cuban descent systematically identified themselves as “not Hispanic,” and wrote “Cuban” in the notes section. These people would be classified merely as “White” unless someone checked all the notes, said Kutateladze. To make sense of race and ethnicity information, it is important for those collecting and analyzing those data to understand the cultural issues in a given jurisdiction, he said.

In terms of data on race and ethnicity of victims, Escobar said that prosecutors who have contact with victims may be reluctant to ask questions about race and ethnicity. Victim advocates

___________________

¹⁵ For more information, see https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=201520160AB953

might be a potential solution, Davison said, in that race and ethnicity information could come from self-identification through conversations between victims and advocates. As victims are at risk of becoming revictimized or committing an offence themselves, she argued that data on victims is important to collect, track, and report publicly.

BUILDING ANALYTICAL CAPACITY

Analytical capacity in prosecutors’ offices can be built in a variety of ways, said Kutateladze, from partnering with external researchers to building in-house expertise. He asked panelists about the advantages and disadvantages of each model, and whether any one model is preferred. Using multiple approaches is ideal, Gur responded. “There is more work than there is time in the day and people to do it,” he said, so utilizing both in-house and external workers can help to distribute the workload. Regardless of the model, he noted, it is critical to spend time and resources to improve processes and create efficiencies, to learn as much as possible from each data analysis and maximize every project’s impact. In the Philadelphia prosecutor’s office, researchers are jointly employed by both the office and University of Pennsylvania. This is “incredibly beneficial,” said Gur, because these researchers not only have more research-specific expertise than others in the office, but they also gain perspective and insight from their day-today experiences in the office.

Wong agreed that a hybrid model, with both in-house and external research expertise, is the best approach. In Multnomah County’s DA office, some internal staff work on daily data requests, while third-party researchers help with bigger projects. Having a neutral third party is important when evaluating the office’s programs and policies, said Wong. In addition, third-party researchers have unique perspectives and expertise. Several years ago, said Wong, a local university helped conduct qualitative research on the workings of the prosecutor’s office. Attorneys were more willing to hold honest conversations with the third-party researchers than they would have been if those conversations were conducted in-house, she said.

Stemen shared his experiences working with prosecutors’ offices in Colorado. It was the first time his institution had done work with an entire state, he said, and both Loyola University of Chicago and the local prosecutors’ offices wanted to have a research partner embedded locally. Stemen’s organization partnered with Lauren Gase at the University of Denver/University of Colorado; she could travel to offices frequently and convene cross-site

workgroups. For two years, coordinating work across the state offices was her full-time job. This put a “face on the work,” said Stemen, and created sustainable relationships. University of Colorado continues to work with prosecutors’ offices across Colorado and has applied for funding and grants to support the partnership, Stemen said.

A workshop participant, Adam Gamoran, William T. Grant Foundation, shared another benefit of partnerships between in-house and external researchers—the practitioner office will often build its internal capacity in response to the opportunity to work with the outside partner. This internal capacity growth, in turn, helps to strengthen and institutionalize the partnership. Stemen agreed that this is an important benefit and observed that, when his organization started work in Colorado, no prosecutors’ offices had analysts. Once offices began partnering with external researchers, several offices hired their own analysts.

DATA ON PLEA BARGAINING

One aspect of data-driven work that has been especially challenging, said Kutateladze, is capturing information around plea bargaining and evidence. Assessing how and why charges change without accounting for the quality of evidence is difficult, he said, and researchers are often required to make assumptions. He asked panelists whether they have seen improvements in this area of data over time. “No,” said Escobar—most offices continue to capture only the final plea accepted by the defendant, and previous offers are generally not captured in the case management system. Often, she said, the entire process is conducted via text messages between the defense attorney and the prosecutor, which makes capturing data extremely difficult. Similarly, prosecutor recommendations are frequently not captured in the case management system, or if they are captured, the office often does not analyze or report the data. Several potential reasons exist for the failure to collect data in these areas, Escobar said. First, the information may be material to the case, so attorneys do not want it shared or entered into the case management system. Second, the culture of prosecutors and defense attorneys is characterized by quick, transactional interactions, often using text message. Finally, said Escobar, attorneys might not understand the purpose of studying the “life of a plea.” While a researcher may see value in studying how a plea changes over time, attorneys may largely focus on the final result.

view of what the team, as a collective, is doing. Sharing data can connect prosecutors to their work and educate the public on why such work is important, said Davison.