Professor, Criminology and Sociology Professor, Law and Epidemiology Using Shifts in Deployment and Operations to Test for Racial Bias in Police Stops John M. MacDonald Professor, Criminology and Sociology Jeffrey Fagan Professor, Law and Epidemiology
Infra-marginality Challenge to Using Outcome Tests
A Model to Address Infra-marginality
Increase incentive to stop, question, and search suspects Increase incentive to stop, question, and search suspects Locations and context is all the same aside for the surge in police force If police are not targeting race then relative % change of stops that result in an arrest, frisk, search, and yield of contraband should in principle similar to other areas of the city that are unaffected Extensive measures to control for similar stops characteristics
We Exploit Change in Impact Zone Locations Examine stops in Impact 9 areas that don’t overlap with Impact 8
Estimate Outcomes Before and After Impact Zone Forms Compared to Non-Impact Areas Outcomes: Arrest, Summons, Frisk, Search, Hands, Wall/Car, Finding of Contraband or Weapon Impact Zones (pre) Jan-June (post) July-Dec Blacks= 26,330 (71%) Hispanics=7,451 (20%) Whites & Asians=3,229 (8.7%) Non-Impact (pre) Jan-June (post) July-Dec Blacks= 174,878 (66%) Hispanics=11,633 (4.4%) Whites & Asians=77,284 (29.3%)
Context of Stops Measured Suspected crime violent, weapons, property, drug, or other offense reason (yes=1, no=0) Crime suspicions (cs) carrying an illegal object in plain view, fit a crime description, casing a place or victim, serving as a lookout for a crime, engaging in a drug transaction, exhibiting a furtive movement, observed committing a violent crime, had a suspicious bulge, or any other non-specified criminal suspicion (yes=1, no=0) Stop context Radio call, day of the week, the patrol shift, general age of individual stopped, gender (male=1)
Entropy Balancing Weights Cases To Be Similar on All Stop Contexts Before and After Impact Zone
Blacks Stopped in Other Areas Blacks Stopped in Impact Zones Balanced on Crime Suspected, Radio, Day of Week, Shift, Age of Suspect, Gender (Blacks, Hispanics, Whites-Asians Separately) Blacks Stopped in Other Areas Blacks Stopped in Impact Zones
Stops Weighted to Be Similar on All Stop Factors
Blacks More Likely to be Arrested, Summons, and Frisked After Impact Zone Forms Relative to Unaffected Areas
Hispanics More Likely Arrested, Frisked, and Hands on After Impact Zone Forms Relative to Unaffected Areas
Whites-Asians Are Not More Likely to be Subject to Any Outcome After Impact Zone Forms Relative to Unaffected Areas
Conclusions Evidence pointing to racial disparities in frisks associated with declaring an area an impact zone No evidence of difference in recovery rates from frisks and searches We tested for average differences but cannot assign a true value of searches. Future research should explore the productivity of searches from policy experiments, by estimating whether policies produce differences in search thresholds and hit rates by race (Simoiu, Corbett-Davies, & Goel, 2017).