> However, I'd say from these that it's not clear that observed effects are the result of "changes in policing"; it's a hypothesis, sure, but not with close to

> However, I'd say from these that it's not clear that observed effects are the result of "changes in policing"; it's a hypothesis, sure, but not with close to the level of confidence you suggest. Yes, because a lot of these studies were from the early days of the BLM riots. Things have solidified in the intervening half decade (although there's some uncertainty about how much of the 2020 surge is attributable to policing versus the pandemic, but I don't think anyone seriously doubts changes in policing played no role at all especially considering the homicide uptick begins promptly and dramatically after May 25, 2000). I don't think there are any remaining hypotheses (there is at least one niche hypothesis that is still a variant of the Ferguson effect--that the protests have caused communities to lose trust in police--but most studies refute it or find no evidence of it). Here's the nber.org link again (I had to do some manual formatting edits when I copy/pasted, and I flubbed this link apparently): https://www.nber.org/system/files/working_papers/w27324/w27324.pdf https://www.nber.org/system/files/working_papers/w27324/w273... > f the hypothesis is that it was the protests that somehow resulted in... changes in policing... so actually, if the hypothesis is changes in policing, why not actually try to measure _changes in policing_ and correlate to homicide rate, instead of correlating protests and murders and assuming "changing in policing" resulted from the protests as an unmeasured middle step? This is precisely what many of these studies do--they find reductions in policing and increases in murder rates. If you read the paragraph above the nber.org link, that's exactly what I describe.