I argued with someone on r*ddit about this recently and from what I remember the only consistent statistically significant correlations between policing and reduction in crime are auto thefts and theft from autos (private property, who wouldathunk!) Yglesias wrote a shitty article trying to dispute this, and this Twitter thread is a really good account of why Yglesias' points were full of shit and bad misreadings of the data he was citing (even one of the authors said his findings were interpreted incorrectly): https://twitter.com/humanprovince/status/1274007772464644097