Analyzing police presence

In the wake of George Floyd, Breonna Taylor and Ahmaud Arbery’s tragic deaths, thousands have marched in protest of police brutality across the U.S. and beyond. Their deaths have reignited discussions surrounding the ongoing violence and discrimination Black Americans experience, and have been experiencing, at the hands of authority for over 400 years in this country.

And to be clear — Black people and activists never stopped having these discussions. Black Lives Matter is a sustained movement that, most of the time, receives very little of the public’s attention. So while calls for police accountability are in the national spotlight, I think it’s important that we listen to Black people who have consistently been in this space, and who understand bias in policing far more than those of us who are newly passionate about the issue.

I think we intuitively feel that Black communities face heavier police presences than middle/white America does. Living in a suburb, I can count the number of times I’ve even seen a cop car on one hand. I was surprised to see few accessible data points on this issue, so went and analyzed the data myself — here are my findings:

Mapping police presence

These maps compare the number of police officers in a city’s department per 1,000 residents against three factors — race, crime rate and median income. Interact with the maps below, and click each respective button to change the county-level layer.

Correlation between Police Presence and Race, Crime rates, Income

The following analysis considers data from the 100 most populated cities and metropolitan areas in the U.S.

The percentage of a population that is Black, number of known offenses per 1,000 residents and median household income all, to some degree, have a relationship with the number of police officers per 1,000 residents in a given city. The r² values are low, but not insignificant. And it’s worth noting that according to these simple linear regressions, Black populations have the strongest correlation with police presence — even more so than crime rates.

And here are the residual plots for each of the variables:

After combining these variables to do a multiple regression, it’s clear that a relationship exists between police presence and these three variables. The r² value of the multiple regression is 0.93, meaning that 93% of the variance in police presence can be explained by race, income and crime rates. The Mean Absolute Error is considerably lower as well.

The residual plots below are effectively the same graph repeated three times. I changed the x-axes for easier comparison with the residual plots from the simple linear regression above.

And for good measure, here’s the correlation matrix:

Again, Black populations emerge as the most statistically significant variable in explaining this relationship.

On a closing note, the political, historical and personal experiences of non-Black people are not enough to inform our understanding of racism in this country. The way you understand the world, contemporary issues and the people in it, are shaped by the people you listen to and those you don’t.

Your view of the police and your trust in them may depend on whether or not you live in a heavily-policed community. This analysis is far from comprehensive, but looking at the data, it’s easy to see why Black people may have differing views than the average American, and harbor deep mistrust for institutions like the police.

In moments like these, it’s our job as non-Black people to listen — then act.