This is where it gets interesting. We need to talk about "crime data" and selection bias. We also need to recognize that more police is not the same as less crime.
The prevalence of crime in a given area is in and of itself, socially constructed: how do we define "crime", who decides when one has been committed, and under which circumstances someone is charged. This sets us up for data collection problems before the first obs is logged.
Areas of "high crime" tend to already have substantially more policing. In the current data collection process, police are all three: arbiters of criminality, those who enter preliminary encounter data, and those with the most to gain/lose based on the picture thus painted.
The FIO dataset offers a unique opportunity to look at this phenomenon at the city level. These plots, generated from the Boston PD's Field Interrogation (FIO) dataset, show the stark reality of stop & frisk in Boston. Racially motivated S&F is very much a reality here. #mapoli