Using Value-Added Models to Identify Outlier Officers in Traffic Stop Data

• In Event: What Works, What's New, and What's Troubling in Investigating Racially Biased Policing

Abstract

While many law enforcement agencies are concerned about racially biased policing, identifying whether it is occurring in traffic stops is difficult to do. A variety of methods, long standing methodological issues like the “denominator problem,” and divergent jurisdictions make establishing a means to investigate racially biased policing difficult. To approach this issue, we borrow from the field of educational statistics value added models (VAMs). VAMs identify the teacher-specific contribution to students’ outcomes. Using hierarchical, slopes-as-outcomes models where stops are nested within officers, we examine officer-specific contributions to stop outcomes, like arrest, citation, and search. Our data – which includes all self-initiated stops for the fiscal years of 2014-2016 – comes from a southwestern law enforcement agency. Officer who impact the race-specific stop outcomes net of situational, driver, and contextual controls may be having an undue influence on the race-specific stop outcomes and warrant further investigation into their policing behavior by their supervisors. Our results show that VAMs are an effective way of identifying behavioral patterns in officers net of controls, as well as organization-wide behavioral patterns.

Author

• Danielle Wallace, Arizona State University