Controlling for crime rates can reverse or erase racial use-of-force disparities

Evidence

For example, officers use force approximately 1 percent to 2 percent of the time across all contact with community members, but about 15 percent to 20 percent of the time when making an arrest (cited in Smith, Kaminski, Alpert, Fridell, MacDonald, & Kubu, 2010). … Crime rate is an important benchmark. Measures of specific kinds of crime, such as arrests for violent crime as opposed to arrests for property crime, can serve as a proxy for the types of provocative or criminal behavior officers are trained to perceive as a threat and respond to with force (Cesario, Johnson, & Terrill, 2018; Goff et al., 2016). … After accounting for crime, the raw and population-indexed racial disparities in use of force tend to narrow to the point of no longer being statistically significant, or the gap is even reversed, such that Whites become statistically more likely than Blacks to have force used against them (Cesario et al., 2018; Goff et al., 2016). … After controlling for crime, statistically significant racial disparities have been found for less lethal force, but not lethal force (Fryer, 2019; Goff et al., 2016). — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 375

Categorization

• XREF: Connects to the general debate on statistical controls in disparate-impact analysis; the Cesario/Fryer findings are widely cited in both popular and academic writing on policing.
• QUESTION: The chapter itself flags that arrest and crime rates are likely racially skewed, so the control may itself be biased — a methodological question worth chasing.