Recruit Screening, Representation, and the Moral Hazard Problem in Policing

Abstract

Calls for more representative police forces, made regularly over the past four decades, rest in part on the assumption that hiring minority officers will help departments overcome adverse selection and moral hazard problems that lead to overly aggressive and discriminatory policing. However, the empirical evidence regarding the relationship between force representativeness and improved policing outcomes has been remarkably mixed. This article explores the effectiveness of another method that departments can use to overcome those same problems, as well as the degree to which those methods may interact with the demographic composition of police departments. Results from a difference-in-differences analysis of more than 500 police departments in 2002 and 2008 suggest that screening recruits for conflict management skills reduces racial disproportion in discretionary arrests, particularly in departments that are more representative of the communities they serve. The analyses also suggest that conflict management screening, when combined with recruit screening for sensitivity to diverse cultures, reduces the lethal use of force by police.

Keywords:

• Agent problems
• candidate screening
• policing
• principal
• representation

Notes

1 For similar examples of mixed findings regarding representation in the criminal justice literature, see Bradbury and Kellough (2011).

2 Departments appear in our sample if they had 80 or more officers in at least one of the years. We restrict the analysis to these years because, while the LEMAS was fielded in 2013, the instrument did not include questions about screening techniques.

3 See https://www.bjs.gov/index.cfm?ty=datool&surl=/arrests/index.cfm#

4 See crowd-sourced data such as fatalencounters.com; public interest group compiled data from mappingpoliceviolence.org; and data compiled through investigative journalism such as the “Fallen” dataset maintained by the Guardian or the “Fatal Force” dataset compiled by the Washington Post.

5 We also estimated models using the proportion of African American officers (rather than the proportion relative to the proportion in the larger population). The findings are very similar except that the interaction between this measure of representation and both screenings (see Figure 2) is no longer statistically significant. It is important to note that the race of officers does not have a significant impact on our dependent variables regardless of whether we use measure representation relative to population or simply as a proportion of total officers.

6 The results are unchanged if we use the total violent crime rate.

7 This assumption in difference-in-differences analyses is that treatment effects are properly estimated because the trends in the dependent variable are parallel in treated and untreated units until after the treatment.

8 We believe that our measure of African American officers relative to African American population is the best measure of representation, as opposed to a measure that simply captures the total proportion of the force that is African American. Nonetheless, it is important to note that if we use the latter measure, the direct impact of screenings reported herein remains unchanged, but the effect is no longer significantly moderated by force make-up.