It is now a well-known fact that black people get killed by police in the United States at a rate disproportionate to their population. Despite making up just 13% of the population, black people constituted 26% of people shot and killed by police in 2015, according to the Washington Post. The Guardian found a similar percentage, 27% for 2015, and Sam Singyangwe's Mapping Police Violence puts this total at 30%. I've also found a similar number through 20 months of data from January 2014 through August 2015 as part of the Lethal Force Database. I've found black people are 27% of the victims of police force acts, and 26% of victims of police gunfire.
Black people get killed at a rate roughly three times that of white people. I've found that while white people were killed by police at a rate of 2.5 deaths per million, black people were killed at a rate of 7.1 deaths per million. The Guardian found these rates to be 2.94 and 7.25 per million, respectively in 2015.
What isn't yet agreed upon is why black people get killed at a much higher rate by police than white people. Some, like University of South Florida criminologist Lorie Fridell argue that police officers may have implicit biases against black people and are therefore more likely to shoot black people due to a perceived threat that may not be real. There are also those who argue that the biases are not so implicit, and that "white supremacy" is the cause of many shootings; in this camp the debate has become whether the explicit biases are systemic or limited to a few "bad apples". For example, Ryan Cooper, writing for The Week, made a case for the number of police shootings being directly tied to poverty, in a piece entitled "To End Police Violence, We Have to End Poverty".
But there are many who say that the root cause of black people getting killed at a higher rate than white people is the fact that they have more interactions with officers. Officers, whose job it is to enforce the law, will naturally be more readily found in high-crime neighborhoods, and black people are more likely to live in high-crime neighborhoods than white people. Thus, if there is an equal risk of a police shooting per encounter, as Harvard economics professor Sendhil Mullainathan supposes in this New York Times piece from October 18, 2015, black people should be overrepresented in police shooting incidents.
David French, conservative writer and #nevertrump presidential candidate for a week, represented this viewpoint in a piece for the National Review on December 29, 2015 entitled "Black Lives Matter is wrong about police". French asserted that "racial disparities in the use of force are largely explained by racial disparities in criminality." He looked at the numbers published by the Washington Post and the Guardian and found an underrepresentation, not an overrepresentation. French pointed out that the black percentage of those killed by police was low compared to the black percentage of homicide and robbery arrests, categories where black people are in the majority (though he neglected to point out that in every other arrest category, from aggravated assault and rape to curfew and loitering law violations, white people are the majority). French argued:
Given these disturbing disparities, no rational person would expect police shootings to precisely track with demographics. Police follow crime, and they tend to operate in high-crime areas. It would be alarming if there were statistically significant racial variations in the use of force even after adjusting for crime rate....
Well, I may have some alarming news for David French.
I combined the zip code location of each death caused by police action in the Lethal Force Database with U.S. Census data on that zip code's population, demographics, and income (technically Zip Code Tabulation Areas). Then I sorted the zip codes by median household income and calculated a rate of death for each income quintile.
It has been shown by the Bureau of Justice Statistics and others that there is a correlation between violent crime and poverty. I have found that there is also a striking correlation between lethal force incidents and poverty. While the rate of death for the entire U.S. is about 3.3 deaths per million people, the rate in the poorest quintile of zip codes (defined by median household income) is nearly double that, at 6.1 deaths per million people. If David French would agree that there is a link between criminality and poverty, then the fact that poor zip codes are the setting for a higher rate of lethal force incidents would not surprise him.
It is also true that black people are more likely to live in the poorest quintile of zip codes than in any other quintile of zip codes. While black people make up 13% of the U.S. population, they constitute 28% of the population in the poorest quintile of zip codes in the U.S.
But here's where David French's argument parts ways with my analysis. If the risk of death from a police encounter were equal for each person in each quintile of zip codes, then black people would constitute only about 15% of the deaths by police in this country overall, despite making up 13% of the population, a slight overrepresentation. (Due to rounding, this difference is closer to 3% than 2%.) But the actual percentage of lethal force incidents in the Lethal Force Database involving black people is much higher: 27%.
In the poorest quintile of zip codes, the rate of death for black people is 9.3 deaths per million. For white people, this rate is only ("only") 4.7 deaths per million. In fact, for each quintile of zip codes sorted by median household income, black people get killed at a much higher rate than white people.
But it's not just the poorest zip codes that show this disparity. A black person in the richest quintile of zip codes is more likely to be killed by police (5.2 deaths per million) than a white person in the poorest quintile of zip codes (4.7 deaths per million). While a black person in the poorest quintile is only ("only") twice as likely to die at the hands of a police officer than a white person in the poorest quintile, a black person in the richest quintile of zip codes is 3.6 times more likely to be killed by police than a white person in the richest quintile of zip codes.
An almost identical relationship is present when the poverty rate (percent of individuals within the zip code with a household income below the poverty level) is used instead of median income. The more poverty exists within a zip code, the higher the rate of lethal force incidents for each racial group, but the rates for black people are always higher than the rates for white people.
There are at least three problems with this analysis, however. The first is that these death rates are calculated with a denominator of residents of the zip code, but victims of lethal force incidents are not always killed in the zip code in which they reside. I just did a quick analysis of the first thirty incidents (sorted alphabetically by first name, so all the Aarons, Adams and Adrians basically), and I found that thirteen incidents took place in the zip code in which the decedent probably resided, five were likely out of the zip code (the results of car chases mostly), six were unknown, and six were not counted in the database. If this was a representative sample for the entire database, then the fraction of lethal force incidents that occur in the zip code in which the decedent resided is greater than half but less than four fifths. The fewer incidents that take place inside the zip codes of the decedents' households, the less valid this analysis is.
The second problem is that zip codes may not be a small enough geographical area to meaningfully portray crime and poverty. High-crime areas and high-poverty neighborhoods often share a zip code with safer and more affluent neighborhoods. If all the lethal force incidents come from the high-crime neighborhoods of the zip codes, and the high-crime neighborhoods have a large relative share of black people, then David French's assertion that black people are only targeted by police gunfire because black people commit more crimes would not be incompatible with this data. One solution to this issue would be to focus on census tracts, which have only a couple thousand people per tract. This was actually done in a project started by Reuben Fischer-Baum and Ben Casselman of Fivethirtyeight.com in June 2015, shortly after the Guardian launched their own database of police killings. Using data from the first five months of 2015, they found that 42% of lethal force incidents involving black people occurred in the poorest quintile of census blocks while 44% of black people nationwide live in such census blocks. (I found that while 41% of lethal force incidents involving black people took place in the poorest quintile of zip codes, only 33% of black people nationwide live in such zip codes.) This finding echoes my thesis, that black people get killed by police at higher rates than white people no matter how economically prosperous their part of the city is. If poverty and crime were the reason black people were killed at a higher rate than white people, then one would expect to see a much greater percentage of lethal force incidents in poor census tracts than the percentage of black people living in such tracts.
The third problem with my analysis is the possibility that individual actions that lead to lethal force incidents might not be clarified by statistics and generalizations. Lethal incidents most often take place because an individual points or fires a gun at or near police officers, no matter the race. If, for whatever reason, black people do this more often than white people, black people are going to be overrepresented in the police shooting data. And those reasons may be diverse. Black people, 13% of the population, are overrepresented in the numbers of mentally ill people shot by police (24%),
in the number of people shot by police while known to be intoxicated by drugs or alcohol (24%),
and in the number of people tased by police who later die (often aided by a combination of intoxication and obesity) (48%).
It is possible that black people are shot by police at high rates because a variety of social problems can affect people's choices, including "bad apple" cops, poverty, criminality, obesity, drug abuse, more contact with police officers, or lack of access to mental health facilities. In some sense, this can all be lumped in with racism, either on an individual level or a systemic level. But in most cases, police officers are found (justly or not) to be justified in their actions because they face a legitimate threat from an individual, not from a society, a race or class of people.
Interpretation of this zip code analysis should be done with caution for these reasons. But I think it's important to recognize that the anger that drives the Black Lives Matter movement is based in part on this high rate of black homicides by police officers that spans neighborhoods of all income levels and crime rates, and this is a true thing, regardless of one's partisan affiliation.
(Cross posted at the Lethal Force Database — Lethaldb.com)
(Explore the data here)