Study
Hedberg, Katz, and Choate (2017) used a quasi-experimental design to assess the effectiveness of body-worn cameras on arrests, citizen complaints, and citizen resistance in two patrol districts in the Maryvale precinct of Phoenix, Ariz. The Maryvale precinct served as the study site. The precinct was operationally and geographically divided into two similarly sized squad areas that provided first-response coverage to calls for service on a 24-hour basis, 7 days a week. There generally were between 100 and 110 patrol officers equally divided between these two squad areas in the precinct. Area 81 served as the comparison group, and Area 82 served as the “target group” in which officers wore body-worn cameras. Fifty-six camera systems were purchased and deployed in Area 82 on April 15, 2013. Departmental policy involving the use of the cameras was formulated before implementation and was also an integral part of the training by the Phoenix Police Department.
The two areas were geographically similar (Area 81 was 7.4 square miles, and Area 82 was 7.9 square miles) but differed in population size (71,676 individuals in Area 81, compared with 56,630 individuals in Area 82) and portion of population under age 18 (39.5 percent in Area 81, and 43.1 percent in Area 82). Area 82, the target area, had a larger proportion of Hispanic or Latino residents (82.5 percent, compared with 71.1 percent in Area 81, the comparison area) but fewer Black residents (3.9 percent in the target area, compared with 6.4 percent in the comparison area). Further, comparison Area 81 was more affluent (as measured by mean household income: $53,646 in the comparison area, $44,895 in the treatment area). At the time of the study, the two areas had similar levels of total crime; however, the comparison area had a slightly greater share of property crime (20.3 percent of the crime in treatment Area 82, versus 17.6 percent in comparison Area 81).
A total of 44,380 incidents (
n
= 22,720 in the comparison Area 81;
n
= 21,660 in the treatment Area 82) from April 1, 2013, through March 31, 2014, were recorded in the Phoenix Police Department’s Call-Aided Dispatch/Records Management System. Data from the cameras were also coded by incident, and dichotomous indicators were created that coded whether an officer involved in the incident used a camera (“video”), and whether the officers involved in the incident were assigned cameras (“assign”). A local average treatment effect (or “LATE,” also known as the effect of “treatment on the treated”) was employed to account for camera activation. The authors found camera activation was relatively limited, with cameras being activated in only about 32 percent of incidents; specifically, a body-worn camera was activated in about 39.0 percent of incidents involving violent offenses, 26.5 percent of incidents involving property crime offenses, and 6.5 percent of traffic offenses.
Then, for each incident, outcome data were coded, including whether an arrest was made (“arrest”), whether the officers faced resistance (“resist,” consisting of flight, passive resistance, or forceful resistance), and whether the officers received complaints based on the actions related to that incident (“complaint”). Analysis occurred at the incident level using bootstrap resampling of the data. Both linear and generalized linear (relative risk) regressions were employed to estimate the effect of body-worn camera assignment and activation on likelihood of arrest, citizen complaints, and citizen resistance in the case of arrest. No subgroup analysis was conducted.
Study
Huff, Katz, and Hedberg (2020) used a randomized controlled trial to examine the impact of body-worn camera assignment and activation on officer-initiated activity (officer-initiated contacts are proactively initiated by an officer who observed an event and chose to contact an individual, as opposed to responding to a citizen request for service), arrests, officer use of force, and citizen complaints against officers in the Phoenix (Ariz.) Police Department. Officers assigned to patrol units in six of the seven precincts in the city were included. Patrol officers assigned to the Maryvale precinct were excluded from this study because it served as the location of the 2013 body-worn camera pilot test (Study 1, described above). Of the 841 officers eligible for inclusion, 668 were approached and asked to participate in a voluntary survey about body-worn cameras. Of the 668 approached officers, 467 participated in the survey.
The body-worn cameras were deployed to officers in two phases: 1) a volunteer phase and 2) a mandated phase. In the volunteer phase, 144 officers from the pool of 467 officers who participated in the survey were randomly selected and asked to voluntarily wear a body-worn camera. Forty-seven of the randomly selected officers agreed to wear a camera (“body-worn camera volunteers”). There were 97 officers who declined to wear a camera; they were referred to as “body-worn camera resistors.” In the mandated phase, all officers who had not agreed to volunteer to wear a body-worn camera were randomized into either a treatment or a nontreatment condition. Those who were randomly selected during the mandated phase were required to wear a body-worn camera without the option to decline (
n
= 35 “body worn camera mandated”). The 281 officers who were not asked or assigned to wear a body-worn camera during either the volunteer or the mandated phase served as the comparison group. The final sample included incidents involving 47 body-worn camera volunteers, 94 body-worn camera resistors, 34 body-worn camera mandated officers, and 277 comparison officers. All body-worn cameras were deployed on May 24, 2017.
Computer-aided dispatch data (all Phoenix Police Department incident reports for crime and disorder events, including records for dispatched incidents and officer-initiated contacts), arrest data, use-of-force reports, complaints, and body-worn camera metadata were used in the analysis. The time period was the 18 months after body-worn cameras were deployed (May 24, 2017, through Nov. 23, 2018). All use-of-force reports were completed by the involved officers’ supervisor. Complaint data were gathered from the Phoenix Police Department’s Professional Standards Bureau and included all complaints related to an officer’s job performance, regardless of the source of the complaint (including citizen, supervisor, or the Phoenix Police Department website). Metadata were collected from the camera vendor to obtain a record of every body-worn camera activation and were used to examine whether the camera was activated in each individual police–citizen contact.
To account for differences between officers who agreed to volunteer to wear a body-worn camera and those who were mandated to wear the device, a body-worn camera volunteer variable was included in the analyses. Two separate multivariate models were used to assess the effects of body-worn cameras on the outcomes. First, the impact of body-worn cameras was assessed using an intent-to-treat approach to examine the impact of assigning the devices to officers. Then, the impact of the treatment on the treated was estimated using an instrumental variable analysis. A probit path model was used to predict each outcome using body-worn camera activation, the incident type, and the body-worn camera volunteer variables, offset by the number of responding officers. Overall, the second portion of the analysis identified the effect of body-worn cameras on incidents involving officers who were assigned to wear the device and who activated it, compared with two other groups of incidents: 1) those involving officers who were assigned to wear a body-worn camera but did not activate it, and 2) incidents in which the responding officer was not assigned to wear a camera. Subgroup analyses were conducted with the 47 body-worn camera volunteer officers.
