Abstract: Whether police discriminate on the basis of race and ethnicity when making stops is a topic of frequent debate among academics, in courts, and beyond. However, due to implausible assumptions about police behavior, the most commonly used tests are quite susceptible to indicating discrimination when it is not present or to indicating a lack of discrimination when it is present. This is true of research on the New York Police Department’s (NYPD) practice of Stop, Question, and Frisk (SQF), a particularly contentious case where findings have been mixed. Using data from 2008 to 2012 on over 700,000 NYPD weapon stops, I develop a three-step technique with which to test for discrimination that uses machine learning algorithms to estimate stop-level hit rates which are then compared through matching and semi-parametric models. This technique addresses several of the most worrisome inferential challenges faced when testing for discrimination in police stops. I find that the NYPD was discriminatory against blacks, and to a lesser extent against Hispanics, in its SQF stops for suspected criminal possession of a weapon. However, differences in the contexts in which stops happened—especially where they happened—account for most of the raw disparities in hit rates.
Roland Neil is a PhD candidate in the Department of Sociology at Harvard University.