by Frank R. Baumgartner –
Imagine a police department separates a city into two zones, one majority white, one majority black, but those zones have no difference in the likelihood that individuals of the two races break the law.
With one more heavily patrolled than the other, but no officer-level racial disparities at all, we can still see dramatic differences in arrests or other outcomes by race.
Imagine these hypothetical patterns. Ten percent of whites and blacks engage in illegal activities, and if they are in an area that the police patrol, they are arrested with certainty. If, however, the police are not there, they do not get arrested.
In this hypothetical example, the police patrol the black area but not the white area. There is no difference in illegal behavior by race. However, the arrests occur only in the area which is patrolled. The result is that the arrests are 90/10 black, even though the behaviors are equal across racial categories.
One logic is to suggest that there is no problem, since by “controlling for geography” one takes out of the equation the decision to patrol certain areas more intensely than others.
Evaluation of racial disparities in policing must incorporate that first step, the decision to devote police resources to certain areas more than others, not eliminate it from the analysis.