Association for Public Policy Analysis & Management

Chicago has been facing an unprecedented surge in gun violence, with the number of homicides reaching levels not seen since the 1990s. A relatively small share of residents concentrated within a handful of neighborhoods account for most shootings in Chicago. This concentration of risk raises the possibility that narrowly targeting the individuals who face the very highest risk of shooting involvement could have a dramatic impact on serious violence. Historically, targeted interventions for this population have largely involved changes in law enforcement strategies. This project takes a different approach by testing whether a social service intervention offering a supported job and cognitive behavioral therapy (CBT) can reduce violence and criminal justice involvement among a population at extremely elevated risk of shooting involvement.

We are conducting a randomized controlled trial of the Rapid Employment and Development Initiative or READI Chicago, a program created as a citywide initiative to reduce gun violence involvement. It is a partnership between a coordinating non-profit agency, seven community-based organizations with experience working in each study neighborhood, and the research team. We randomly assign which study participants will be offered an 18-month supported job structured around CBT sessions targeting specific behaviors to reduce gun violence. The program also offers support services and referrals as needed based on how often housing instability, mental health struggles, substance abuse, and legal challenges interfere with participants’ ability to attend the program. Unlike many transitional jobs programs, the core goal of READI is to reduce violence and criminal justice involvement rather than improve long-term employment outcomes. Nonetheless, the program works to leverage best practices from the transitional jobs literature to keep participants engaged (and maximize incapacitation effects), including a tiered system of employment and pay rates.

A crucial challenge is identifying the right study population: men in Chicago at the very highest risk of shooting or being shot. Prior research has had some difficulty in predicting gun violence involvement, so we have developed a novel combination of data analysis and agency know-how. Using administrative data on arrests, victimizations, and social networks, the research team used machine learning tools to build a risk assessment. We select the first subset of study participants based on that assessment. Since administrative data only capture part of criminal behavior and risk, we also select a subset of study participants based on referrals from the community agencies that are on the ground in the target neighborhoods. The last subset of participants is a set of high-risk individuals who are exiting jails and prisons, since re-entry is a key time of elevated risk. As program providers refer, recruit, and serve study participants, the research team provides regular data analysis on who is being referred to, participating in, and dropping out of the program. The agencies use that analysis to shape their implementation strategies.

Because it is too early for the outcome evaluation, we will focus on the substantive lessons about prediction and targeting, as well as practical lessons about integrating data analysis and the implementation process.