Enhancing Accountability and Transparency: The Impact of Body-Worn Cameras in the Presence of Elevated Crime and Police Lethality

We study the implementation of body-worn cameras (BWCs) by the Military Police in the state of Bahia, Brazil. The state has one of the highest criminality rates and the second highest police lethality in the country. The process is starting with a pilot run in a few precincts of the capital city, Salvador, making use of 85 cameras. We randomized the roster of police officers with public facing duties across the precincts to either a treatment or a control group in a 1:1 ratio. Those selected to use the cameras will do so in every shift; control officers will not wear the camera in any shift. We stratify by police precinct and type of policing; when a shift had two or more units with differential spatial coverage, we further stratified by shift. Our analysis is done at the police dispatch/event level in which police-citizen interactions unfold.

The Military Police (Polícia Militar, or PM) is the institution responsible for law enforcement and maintaining public order within states and municipalities in Brazil. Our partner is the PM of the state of Bahia (PMBA), the second state with highest index of police lethality in the country. Upon taking office in 2023, the Lula government singled out the adoption of body-worn cameras as a key initiative. Bahia is procuring the devices for implementation across the entire workforce (of over 3k officers) across the state, in a push that was widely publicized (see, for example, https://www.cnnbrasil.com.br/nacional/bahia-contrata-cameras-corporais-para-uniforme-de-policiais-militares/).

This process is starting with a pilot run with staggered adoption. First, three precincts of Salvador, the capital city, are making use of 85 cameras. As per PBMA request, we implemented a randomization of the cameras. We randomized the roster of 167 police officers with public facing duties across the precincts to either a treatment or a control group in a 1:1 ratio. Those selected to use the cameras will do so in every shift; control officers will not wear the camera in any shift. We stratify by police precinct and type of policing; when a shift had two or more units with differential spatial coverage, we further stratified by shift. In a second moment, the BWCs are planned to be rolled out to additional 3 to 5 precincts. Our analysis is done at the police dispatch/event level in which police-citizen interactions unfold.

The presence of police body-worn cameras may impact the dynamics of the dispatch in ways that will be measured and have not been fully studied in the literature. On the one hand, one could hypothesize that the presence of the cameras, while inducing pacifism in the police-society relations, reduces the chances of use of force which are associated with longer and more complex operations. On the other hand, preliminary evidence from other studies may suggest that the presence of the cameras increase the reporting of misdemeanours which may suggest that they spend more time in pursuing investigations against those cases vis-a-vis the absence of those cameras. Overall, the potential consequences of adopting BWCs have been considered by PMBA in their implementation program and are not a direct consequence of this study, given that BWCs will be implemented regardless of the PIs involvement.

The Randomised Control Trial (RCT) will be used in conjunction with observational methods such as Differences-in-differences and synthetic control. We will measure compliance of police officers with the use of BWCs as mandated by their operational protocols and experience with the equipment. The study seeks to evaluate whether the cameras reduce the use of force by police officers, the incidence of disobedience or charges of contempt, and civilian complaints against officers. We will assess how BWCs affect the efficiency of judicial proceedings and the quality of reporting, in particular we verify whether it introduces police passivity or underpolicing, and/or increase the reporting of cases of misdemeanours.

2024-06-09

2024-09-09

(i) Indicator if the event generated a police report; (ii) Indicator if there was any victim reported; (iii) "Citizen Behavior"; (iv) Use of Force; (v) Indicator for Arrests and Handcuffs; (vi) Number of complaints against police officers; (vii) Duration of judicial proceeding; At the time of registration it is unclear whether we will have access to data that will allow us to construct all measures.

Citizen Behavior is an indicator if there was any charges of contempt, disobedience or resistance towards officers; Use of Force is an indicator if there was any deployment of physical, non-lethal or lethal force by the police;

(a) indicator if case of misdemeanour; (b) indicator if police officer activate the equipment

BWC Assignment: From the roster of 167 sworn police officers with public facing duties we stratified them into blocks based on (i) precinct and (ii) officer main activity type (tactical, preventive or ostensive operations); we further stratify by shift when a shift had two or more units with differential spatial coverage (5 out of the total 22 stratification blocks). We then randomize units within each strats in a 1:1 ratio between treatment and control group. Units/officers selected to use the cameras will do so in every shift; control officers will not wear the camera in any shift.

BWC Assignment: From the roster of 167 sworn police officers with public facing duties we stratified them into blocks based on (i) precinct and (ii) officer main activity type (tactical, preventive or ostensive operations); we further stratify by shift when a shift had two or more units with differential spatial coverage (5 out of the total 22 stratification blocks). We then randomize units within each strats in a 1:1 ratio between treatment and control group. Units/officers selected to use the cameras will do so in every shift; control officers will not wear the camera in any shift.

Computer Randomization

BWC randomization unit is the spatial coverage (within shift)

Yes

22 stratification blocks

180 events per day on average before the randomization x 104 experimental days = 18.720 events

31 treated units (82 officers), 34 control units (85 officers)