At Marginal Revolution Alex Tabarrok looks at the first randomized controlled trial of body cameras for police officers:
The results were that police use of force reports halved on shifts when police wore cameras. In addition, the use of force during the entire treatment period (on shifts both using and not using cameras) was about half the rate as during pre-treatment periods. In other words, the camera wearing shifts appear to have caused police to change their behavior on all shifts in a way that reduced the use of force. A treatment that bleeds over to the control group is bad for experimental design but suggests that the effect was powerful in changing the norms of interaction. (By the way, the authors say that they can’t be certain whether the cameras primarily influenced the police or the citizens but the fact that the effect occurred even on non-camera shifts suggests that the effect is primarily driven by police behavior since the citizens would not have been particularly aware of the experiment, especially as there would have been relatively few repeat interactions for citizens.)
It is possible that the police shaded their reports down during the treatment period but complaints by citizens also fell dramatically during the treatment period from about 25-50 per year to just 3 per year.
Here’s a graph of use of force reports before and during the treatment period.