Which policies should police departments adopt? On February 9 and 10, the Policing Project and the Police Foundation convened over twenty experts on policing practices and quantitative methods to explore one possible answer to this question: those policies whose benefits outweigh their costs.

Making their way through the blizzard that stormed New York at the conference start, participants gathered at New York University School of Law to discuss the importance and challenges of applying cost-benefit analysis to policing. Cost-Benefit Analysis (CBA) enables decision-makers to compare various policies using a common denominator and determine what is worthwhile. CBA is commonplace across government, but notably absent from policing. With generous support from the Laura and John Arnold Foundation, we have set out to change that.

Before the conference, participants read materials that the Policing Project prepared in advance, and submitted short thought pieces in response to specific prompts. (The short pieces will be available here in a few weeks, and some expanded versions will be published in the Journal of Benefit Cost Analysis.) The conference then featured four sessions – each more concrete than the previous – ultimately identifying a set of police practices susceptible to CBA.

Session 1 asked what should get measured and included in a policing CBA. Specifically, when we do a CBA of, say, stop and frisk practices, do we include the social costs and benefits of incarceration, transferred effects across parties, distributional effects among groups, and intermediate versus ultimate effects? Lisa Robinson set the stage by recommending counting everything, because the analyst’s job is to present all options to the decision-maker who will then weigh them. Michael Hanemann advocated avoiding inter-personal comparisons of utility as impractical, while Jeffrey Fagan emphasized accounting for the preferences and political ecology of police departments. Police experts like Hassan Aiden, Tracie Keesee, and Stuart Greer offered their practitioner’s perspective: with 18,000 police departments across America, only the highest functioning ones can do CBA – still, all departments must increasingly consider the downstream effects of their actions, such as on community trust.

While Session 1 focused on abstract concepts like spillover effects and unanticipated consequences, Session 2 asked what research methods can be used to test these effects. Randomized-control trials are the gold standard, but are expensive and rarely practical in this area. Thus, expert methodologists like Richard Carson, Jon Krosnick, Amanda Geller, Jillian Carr, and Rebecca Goldstein discussed alternative research methodologies like quasi-experiments and observation. Both rollouts and (in the case of stop-and-frisk) rollbacks of policy can afford a good opportunity for study. The group recommended adding questions about law enforcement to city-level surveys that are administered regularly.

Session 3 asked how we quantify what is hard to quantify. For example, what is the privacy cost of using license-plate readers; how do we measure community trust; and what are the social costs of proactive policing? Effects are difficult to measure because social costs and social benefits are intangible, encounters with police are heterogeneous, and perceived harms may be unaligned with actual harms. Mark Cohen raised the difficulty of quantifying avoidance behavior, or the costs borne by those who fear targeting and take added steps to prevent police contact. Richard Brooks offered one avenue into an objective measure of how communities assess reputational harms: jury verdicts. Body camera footage could be another data source, but Paul Heaton warned that the data would be massive, mostly uneventful, and must be handled securely. Collectively, conference participants also suggested conducting a contingent valuation study that asks respondents to watch videos and gauges their willingness-to-pay to avoid varying sorts of police actions. Cynthia Lum raised the possibility of leveraging TSA security to study police consent searches, while Glenn Blomquist reminded us that revealed preferences are still an option.

In Session 4, we finalized a list of pilot projects that members of the group will pursue, balancing their appropriateness for CBA, scholarly interest, and real-world considerations. The Policing Project’s ultimate goal in introducing CBA to policing is to offer police departments one tool for determining what works, and therefore what course of action they should take. At the start of the conference, Jack Glaser said he prefers calling it “Benefit-Cost Analysis,” because it focuses on the benefits – if there are no benefits, why do the analysis? From this shared foundation of enhancing policing’s benefits, the Policing Project hopes that CBA (or BCA) will advance policing accountability and minimize the inevitable costs of keeping society safe. Stay tuned for announcements on our forthcoming CBA pilot projects.