An Experimental Evaluation of the BJA DMI Training and Technical Assistance Initiative

Open-air drug markets impose enormous costs on public safety and public health. Traditional enforcement approaches are often unsuccessful and sometimes exacerbate violence and racial tensions as well as damage citizens' perceptions of procedural justice and police legitimacy. Drug Market Intervention (DMI) programs attempt to do more than shut these markets down, they also attempt to heal communities as well as provide services to those seeking change. The goal of this proposal is to conduct process, outcome, and cost evaluations of several DMIs, with a focus on methods and outcome measures that are typically excluded in DMI evaluations. RAND is extremely well suited to conduct these analyses with experts on drug markets, policing, respondent driven sampling, and advanced statistical modeling. A six-member Scientific Steering Committee will play substantive roles as they will help the team develop instruments, review drafts, co-author papers in their respective research areas, and attend annual team meetings. Since the 12 DMI sites have not been selected, RAND proposes a flexible research design that prioritizes randomization but will also work with quasi-experimental conditions. Our preferred strategy involves 1) Encouraging jurisdictions to submit applications to Michigan State University (MSU) for DMI training and technical assistance, 2) Working with MSU to identify promising applicants, 3) Placing these promising applicants into strata based on key variables (e.g., geography, alleged size of open-air market, crime rates) and then randomly choosing one jurisdiction from each strata to participate in the DMI training (those not chosen will serve as controls), and 4) For jurisdictions receiving the DMI training that have multiple open-air drug markets, randomly assigning the order in which the DMI is applied (compliance with this will be an initial condition of receiving DMI training). Randomization across and within jurisdictions will maximize our potential to draw strong inferences and make sound policy recommendations. If randomization is not possible, RAND will apply advanced propensity score matching methods developed at RAND to create the most appropriate control markets. In addition to collecting traditional criminal justice statistics for the treatment and control markets (and surrounding areas), RAND will obtain information about community experiences and perceptions as well as information about drug market activity from three novel sources: 1) respondent driven sampling (RDS) of heavy drug users, 2) community surveys, and 3) business surveys. The general approach will be to collect outcome data before the call-in and 3, 12, and 24 months after. Given the costs associated with RDS and surveys, RAND will be selective about the markets and time periods in which they are administered. Finally, RAND will perform a process evaluation to understand the importance of program fidelity as well as monetize the costs and benefits of DMI. ca/ncf