Association for Public Policy Analysis & Management

On June 16, 2009, Professor Henry Louis Gates Jr. returned home to Cambridge, MA. He found his front door jammed shut, so he attempted to push the door open with the help of his driver. Little did Gates know that Sgt. James Crowley, an 11-year veteran with the Cambridge police force, was on his way to the address in response to a 911 call about a possible break-in. Six minutes later, Sgt. Crowley arrested one of the leading African American scholars in the U.S. for “exhibiting loud and tumultuous behavior in a public place.” The struggle between Gates and Crowley reignited a national conversation about race and policing.

The analysis following this incident by leading academics and public officials focused on how Gates and Crowley behaved during the interaction, mirroring a broader tendency in discussions of police reform to focus on the moment of contact between police and citizens. Although this approach has led to valuable insights, it fails to recognize how actions by 911 operators shape police responses. The 911 audio-recording from this incident reveals that the call-taker played a pivotal role in the Sgt.'s response by taking an ambiguous, cautious call and entering it as a high-priority run. Sgt. Crowley shared with The Cambridge Review Committee that he had legitimate concerns about his safety because he was responding to a high-priority in-progress call and that this contributed to his abrupt demeanor.

The human processing of 911 calls likely shapes much of how police perceive incidents, yet has been largely unexplored by scholars. In this paper, I use 911 call-for-service administrative data from 2015-2017 (N=46,669) and participant observation field-notes to examine the process of call classification and prioritization within a Public Safety Answering Point (PSAP) in Michigan. I exploit a natural experiment for my research design where call-takers are randomly assigned to calls. This allows me to test for the existence of discretion among call-takers in how they classify and prioritize calls, and whether there is variability in how it is exercised.

I use a least-squares-dummy-variable model (LSDV) to test whether significant heterogeneity exists among operators in their likelihood of classifying the same types of calls as high-priority. Findings from this model indicate that an officer responding to a call processed by a call-taker from the top of the call-taker distribution is 30 percentage points more likely to receive a high-priority run than an officer responding to the same type of call processed by a call-taker from the bottom of the distribution. This discrepancy is based solely on the level of aggression of the 911 operator who happened to pick-up the phone. In the world of policing, call-taker heterogeneity can have huge effects in terms of whether an officer draws a weapon, makes an arrest, or uses force. For police reforms to be successful, public policy must also account for the actions of 911 operators.