Study 1
Lawrence and Peterson (2020) conducted a randomized controlled trial to examine the effectiveness of body-worn cameras on Milwaukee Police Department (MPD) proactive, officer-initiated community interactions between July 18, 2015, and Nov. 22, 2016. This examination sought to determine how body-worn cameras affect different types of officer-initiated community interactions, in response to general concern in the field that officers may pull back from these policing functions if forced to wear cameras. The four most common categories of proactive activities by MPD officers during this time were 1) traffic stops (when an officer pulls a vehicle over because of suspected traffic or vehicle violations), 2) business checks (when officers enter a local business to check for proper licenses/permits or to speak with the owner and staff about neighborhood concerns or activities), 3) subject stops (all officer-initiated stops of community members, such as when a person is behaving suspiciously or matches the description of a person suspected of criminal activities), and 4) park and walks (when officers walk around assigned neighborhoods to have a visible presence and to meet community members).
The MPD deployed body-worn cameras to all officers by the end of 2016. The deployment took place over four phases, the first being a pilot program. The randomized controlled trial was initiated in March 2016 during phase 2 of the deployment. Officers who received a body-worn camera as part of the pilot program, who were not patrol officers, and who were on any form of limited duty were considered ineligible for the study. Officers from District 5 and the Neighborhood Task Force (who operated outside district boundaries and received body-worn cameras during the pilot deployment) were also considered ineligible for inclusion, since their working conditions differed considerably from the other districts where cameras were not yet deployed.
A total of 666 patrol officers and sergeants from Districts 1, 2, 3, 4, 6, and 7 remained eligible for inclusion after filtering the roster on these criteria. A stratified random sampling procedure was conducted in which the strata included the officer’s race (non-Hispanic white or nonwhite) and shift (power, late, day, or early). The sample of officers from each district was proportional to the size of that district relative to the overall department. With this, the appropriate number of officers from each district were randomly selected and then randomly assigned by strata into the treatment and control groups. The treatment group comprised 252 officers who were randomly assigned a body-worn camera. The business-as-usual control group comprised 252 officers who did not wear a camera. The study period occurred across 247 days (approximately 8 months) from March 21, 2016, to Nov. 22, 2016. An equal period of time was used for the pre-intervention period, which was July 18, 2015, to March 20, 2016. Panel data were used to measure change across time, where the data were aggregated into sixteen 31-day periods, creating eight pre- and post-intervention periods across the two experimental groups. All of the panels were 31 days long except the first and last panels, which were 30 days long.
There were no statistically significant differences between the treatment and control group officers on officer characteristics or any pre-intervention proactive activities. Officers in the treatment group were predominately male (85.7 percent) and white (62.3 percent). The remaining were 17.5 percent Black, 13.5 percent Hispanic, and 6.7 percent Asian, and they had on average 12.23 years of tenure. The control group was mostly male (89.7 percent) and white (62.3 percent). The remaining were 21.0 percent Black, 13.5 percent Hispanic, and 3.2 percent Asian and had on average 12.7 years of tenure. During the pre-intervention period, officers in the control group conducted an average of 312.4 proactive activities, including 80.8 business checks, 136.6 traffic stops, 43.4 subject stops, and 15.8 park and walks. Treatment group officers conducted an average of 308.2 total proactive activities, including 80.8 business checks, 138 traffic stops, 41 subject stops, and 13.8 park and walks.
The data came from administrative records and the MPD’s proactivity database, where officers were required to record the start and end times of all proactive activities using the dispatch system on their in-car computers. The proactivity data were at the officer/event unit of analysis to calculate the total count of proactive activities for each officer and to identify which officers were present at the event. Random-effects negative binominal panel regression analyses were used to assess the effect of body-worn cameras on the total count of proactive activities and traffic stops, business checks, subject stops, and park and walks. Difference-in-differences analyses were used to estimate the difference between the treatment officers’ pre-intervention and post-intervention outcomes, relative to the same difference for the control officers. Three models were tested for each outcome. The first was an unconditional regression that included only the variables for the difference-in-differences, to assess change in the outcomes from random assignment with no other controls. The second model built on the first by adding body-worn camera contamination levels. Contamination was considered present in an event when a control group officer was with a body-worn camera-wearing officer at the scene, and where a treatment group officer was at the scene with any officer not wearing a camera. The third model included all other covariates (race, officer tenure, and whether the officers were sergeants). No subgroup analysis was conducted.
Study 2
Peterson and colleagues (2018) used the same sample from the randomized controlled trial detailed in Study 1 (Lawrence and Peterson, 2020) to examine the effectiveness of Milwaukee Police Department (MPD) body-worn cameras on arrests, citizen complaints, and use-of-force incidents from the pre-intervention period (the 9 months before March 21, 2016) to the post-intervention period (the 9 months after March 21, 2016).
The methodology was the same as described in Study 1. The body-worn camera treatment group officers received their cameras around March 21, 2016, and the control group officers received their cameras around December 20, 2016, allowing for a 9-month study period. During the pre-intervention period, officers in the treatment group made an average of 13.79 arrests, had on average 0.11 citizen complaints, and averaged 0.34 use-of-force incidents, while officers in the control group made an average of 13.49 arrests, had on average 0.08 complaints, and averaged 0.40 use-of-force incidents. There were no statistically significant differences between the treatment and control group officers on officer characteristics or any pre-intervention activities.
Difference-in-differences analysis with both Poisson and logistic regression models was used to estimate the difference between treatment officers’ pre-intervention and post-intervention outcomes, relative to the same difference for the control officers. The models included a binary variable identifying whether an individual officer was in the treatment group or control group. A second binary variable was included to distinguish between the intervention and pre-intervention periods. A third binary variable represented the product of the first two. The coefficient associated with this third binary variable represented the difference-in-differences estimate of the effect of the body-worn camera treatment on the outcome. No subgroup analysis was conducted.
