Report: increases in police use of force in the presence of body-worn cameras are driven by officer discretion: a protocol-based subgroup analysis of ten randomized experiments

Objectives: Our multisite randomized controlled trial reported that police body-worn cameras (BWCs) had, on average, no effect on recorded incidents of police use of force. In some sites, rates of use of force decreased and in others increased. We wanted to understand these counter-intuitive findings and report pre-specified subgroup analyses related to officers’ discretion on activating the BWCs. Methods: Using pre-established criteria for experimental protocol breakdown in terms of treatment integrity, ten experimental sites were subgrouped into “high-compliance” (no officer discretion applied to when and where BWCs should be used; n = 3), “no-compliance” (treatment integrity failure in both treatment and control conditions; n = 4), and tests where officers applied discretion during treatment group but followed protocol in control conditions only (n = 4). Results: When officers complied with the experimental protocol and did not use discretion, use of force rates were 37 % lower [SMD = (−.346); SE = .137; 95 % CI (−.614) – (−.077)]; when officers did not comply with treatment protocol (i.e., officers chose when to turn cameras on/off), use of force rates were 71 % higher [SMD = .392; SE = .130; 95 % CI (.136) – (.647)], compared to control conditions. When full discretion (i.e., overall breakdown of protocol) was applied to both treatment and control conditions, null effects were registered [SMD = .009; SE=.070; 95 % CI (−.127) – (.146)], compared to control conditions. Conclusions: BWCs can reduce police use of force when then officers’ discretion to turn cameras on or off is minimized—in terms of both case types as well as individual incidents. BWCs ought to be switched on and the recording announced to suspects at early stages of police–public interactions. Future BWCs tests should pay close attention to adherence to experimental protocols.