Critique of the 2022 Research Report from the Center for Applied Science in Public Security (FGV)
Rafael Erthal Corrêa de Sá
Juliano de Andrade Gomes
Introduction
The Olho-Vivo Program was an initiative by the State of São Paulo, launched during the João Doria administration in 2020, aimed, among other objectives, at reducing the number of deaths in police confrontations through the implementation of continuous-recording body cameras in certain PMESP (São Paulo State Military Police) battalions.
The use of body cameras by police officers has sparked intense public debate in Brazil. Recently, the Supreme Federal Court (STF), specifically through its president, Justice Luís Roberto Barroso, ruled that the São Paulo State Military Police (PMESP) must install body cameras on its officers, citing cases of police abuse.
In his decision, Justice Barroso explicitly references the study analyzed in this article, conducted by researchers from the Center for Applied Science in Public Security (CCAS) at Fundação Getulio Vargas (FGV):
Considering the relevance and timeliness of the discussion, the goal of this article is to analyze and provide potential critiques of the study produced by CCAS. This study was selected because it is one of the few in Brazil that has actually evaluated the impact of body cameras on police officers and is frequently cited, particularly by advocates of this technology.
For any public policy implementation, it is essential first to establish its fundamental premises regarding effectiveness—that is, whether the tool is suitable for achieving its intended purpose. Although it is challenging to reach absolute certainty on social
phenomena, which are influenced by multiple variables, public policies should, at the very least, demonstrate a reasonable level of effectiveness beyond a reasonable doubt. Scientific uncertainty is inherent to any serious research, but when interventions involve millions in public spending, the majority of studies should point in one clear direction—effectiveness—which is not observed in this case.
So far, most academic literature presents inconclusive results regarding whether body cameras effectively reduce violent police behavior. This underscores the importance of this debate, and this paper aims to contribute to the discussion by critically analyzing the study in question.
a. Summary of the Study
The study examined the impact of portable body cameras (COP) in São Paulo Metropolitan Region battalions on police use of force, exploring possible mechanisms behind the observed results.
The treatment group consisted of PMESP units equipped with body cameras between June 2021 and July 2022.
The control group included police units that did not use the technology.
According to the study's results:
In areas where officers used cameras, deaths resulting from police intervention (MDIP) decreased by 57% compared to areas without cameras.
Injuries caused by police intervention (LCDIP) also dropped by 63%.
Arrests and crime reports showed no reductions, suggesting no major change in police efforts.
b. Methodological Critiques
The study's treatment group included police units in specific areas of the metropolitan region that allegedly had higher levels of police violence. The control group consisted of units without cameras.
However, the study does not specify which battalions were involved or which metropolitan areas were analyzed. Were all regions included? Or only those with higher or lower crime rates?
It is practically impossible to design a double-blind study on police body cameras, where neither the observers nor the police officers know who is in the control or treatment group. However, the lack of a true control group should have been explored more thoroughly, given the many variables that could influence officers' behavior, such as temporary compliance or spillover effects.
The study assumes there was no spillover effect, meaning that police officers did not patrol areas outside their designated zones. However, this does not reflect reality. In complex cases, police from different battalions often collaborate, which could affect the results.
Additionally, the study does not discuss the possibility of behavioral contamination—a phenomenon where officers without cameras modify their behavior because they work alongside those who are being recorded. This effect has been documented in several other studies on police body cameras.
c. Questioning the Results
The study claims a 57% reduction in MDIP (police intervention deaths), but this percentage was obtained by dividing an average effect (-0.220) by the global mean (0.385) instead of the actual mean for the treated units (0.415). Using the correct method, the reduction would be 53%, assuming the result is statistically significant, which is debatable.
Similarly, LCDIP (police intervention injuries) was reported as decreasing 63%, but a more precise calculation suggests a 57.6% reduction. The study also estimates that 104 deaths were "prevented" due to body cameras, but the methodology for reaching this number is unclear.
Furthermore, Figure 3 in the study presents a confidence interval overlap in the reported MDIP values, suggesting that the differences are not statistically significant. This weakens the claim that body cameras significantly reduce police intervention deaths.
Other graphs (Figures 4) display similar statistical issues, raising further doubts about the robustness of the findings.
The study suggests that police effort remained unchanged, meaning officers did not reduce their activity due to camera use (no depolicing effect). However, this assumption is questionable.
For example, incidents of domestic violence (Maria da Penha cases) and drug possession increased significantly, while more complex crimes (homicides, robberies, etc.) did not. This suggests that police shifted their focus to less risky situations—a qualitative rather than quantitative reduction in policing efforts.
Discussion
Body cameras have been promoted as a solution to reduce police abuse in Brazil, but other, less expensive interventions could be equally or more effective without raising constitutional concerns.
The policy of continuous recording—even during bathroom breaks—violates the dignity of police officers, as protected by Article 1, III of the Brazilian Constitution.
Alternative policies, such as intermittent recording, would be more reasonable and cost-effective.
Financial Costs
The annual cost per 100,000 inhabitants ranges from $297,900 to $785,400.
The cost per camera varies between $1,221 and $3,219 per year.
With 10,000 cameras, the estimated annual cost is between $12.2 million and $32.1 million USD (R$73 million to R$193 million).
If 20,000 cameras were implemented, costs could reach R$266 million annually, exceeding the budget for police training and administration in São Paulo.
Would it not be more effective to invest in police training rather than spending hundreds of millions on a policy with uncertain results?
Conclusion
This article aimed to highlight overlooked aspects of the body camera debate in Brazil. The CCAS-FGV study has serious methodological issues, and the scientific literature remains inconclusive on the effectiveness of body cameras in reducing police violence.
Public policies must be based on solid evidence, not just political choices. Given the uncertainties, body camera policies in Brazil remain a speculative and costly initiative, with unproven effectiveness.