Missouri looks like it's in the middle of the pack. But we desperately need better data.
The death of 18-year-old Michael Brown at the hands of a Ferguson, Missouri, police officer has reintroduced police-related killings as a topic of major national debate. Brown is just the latest in a long line of young, unarmed black men killed by law enforcement agents.
It's been widely reported that roughly 400 Americans die at the hands of police per year. And yet, that figure is likely a significant underestimate, as Reuben Fischer-Baum details at FiveThirtyEight.
We ask a slightly different question: Where are Americans more likely to die at the hands of police or while under arrest?
With the help of my colleagues Charlotta Mellander and Nick Lombardo of the Martin Prosperity Institute (MPI), we mapped data from two sources: “arrest related deaths” from the Department of Justice’s Bureau of Justice Statistics, and from the FBI’s annual Supplementary Homicide Report (SHR) on “felons killed by police.” We also got input from three leading American criminologists: Alfred Blumstein and Daniel Nagin, my former colleagues at Carnegie Mellon, and John Roman of the Urban Institute.
It’s important to reiterate that both data sources suffer from serious deficiencies, not the least of which is under-reporting. Roman worries about "reporting bias," particularly the possibility that "more responsible agencies"—those least likely to use force in the first place—"are more likely to report, and less responsible agencies are less likely to report." But he also adds that what looks like missing data may not be. "It might be that few policing agencies have an officer-involved shooting and the agencies that don't simply don't report any data," he writes in an email.
But, taken together and in light of their limits, the maps are broadly suggestive of the geography of U.S. police killings as well as the states where arrests are likely to result in more deaths. As Roman puts it: "It is important to shine a light on the subject. Because there is such limited data, our ability to define the scope of the problem greatly limits our ability to form an appropriate response."
The Geography of Arrest-Related Death
We start with the BJS data on arrest-related deaths. Nearly 5,000 (4,813) arrest-related deaths were reported between 2003 and 2009. These contain everyone who died in the custody of law enforcement officials, including suicides, deaths from intoxication and accidents. Homicides by police accounted for roughly six in ten of all arrest-related deaths between 2003 and 2009, as the chart to the left shows, and ranged from a low of around 55 percent in 2005 to a high of 68 percent in 2009. BJS also cautions “Data are more representative of the nature of arrest-related deaths than the volume at which they occur.”
The first map charts all arrest-related deaths throughout the country. The BJS data cover 2003 through 2009, from which we calculated the average annual amount for those periods.
As the map shows, California, Texas and Florida – all big states – had the largest average numbers of arrest-related deaths per year, with 111, 99, and 75 respectively. New York (38), Arizona (34), Pennsylvania (31), and Illinois (30) also had significant numbers of arrest-related deaths. Missouri, where Michael Brown was killed, falls in the second lowest quintile, with an average of 4 arrest-related deaths annually.
The next map charts average annual arrest-related deaths per million state residents, to control for population size.
These range from a high of 6.5 deaths per million people to a low of 0.27. Some states, in other words, have nearly twenty times the number of deaths than others. At the top end of the scale is the District of Columbia, which we note as usual is an outlier because it is 100 percent urban, as opposed to states.
Leaving the District aside, Arizona leads with 5.2 deaths per million, followed by New Mexico (4.1 deaths per 1 million), Florida (3.9 deaths) and Texas (3.9 deaths). Note the broad swath of dark red running across the Southwest. The West Coast follows closely behind, with Utah (3.5 deaths), California (2.9 deaths) and Oregon (2.8 deaths).
Northeastern states have relatively low levels of arrest-related deaths. New York has just 1.9 deaths per million residents, and New Jersey is even lower, with 1.3 deaths. Massachusetts is among the states with the lowest arrest-related deaths, at 0.9.
A few Southern states have low rates as well. Arkansas had 0.4 deaths per one million residents; Georgia had 0.27. And Missouri posted one of the lowest arrest-related rates in the nation, at 0.6 per million.
The Geography of Felons Killed by Police
The second set of maps cover felons killed by police. These data are from the FBI’s Uniform Crime Reports for 2012, the most recent year available. These figures, too, are problematic. The FBI’s data is self-reported and not all police departments participate in the reporting process. The data only capture the deaths of those who were convicted of or were in the process of carrying out a felony. Furthermore, as Fischer-Baum points out, "unjustifiable homicide by police" is not a classification, so deaths like that of Michael Brown might not be counted. Also, a number of states – New York, for instance – have zeroes entered, suggesting there are missing values and serious under-reporting.
The third map, above, charts the total number of felons killed by police by state in 2012. Again, the Southwest and West Coast have the highest levels of police killings. California tops the list with 114, followed by Texas (54), Pennsylvania (29), Arizona (27) and Georgia (20). Missouri falls into the upper band of states, with 11 police killings of felons, below New Jersey (13) and above Michigan (10).
The fourth map, below, shows the number of felons killed by police per one million people, again controlling for population. Once again, Southwestern states have the largest concentrations. Arizona leads with 4.2 deaths per million, followed by Maryland (3.3), California (3.0) and Nevada (2.9).
On the opposite end of the spectrum, there are low concentrations throughout the Upper Great Lakes. The states with the least killings per one million residents were Ohio (0.3), Connecticut (0.3), Mississippi (0.3) and Utah (0.4). Missouri, with 1.8 felons killed by police per million people, again falls in the middle of the pack.
• • • • •
There are similarities and differences between the two sets of maps. The Southwest and West Coast have relatively high values according to both measures. Conversely, Mississippi, Minnesota, and Kentucky have relatively low levels of both.
Still, differences persist. The Sunbelt, which fares far worse on arrest-related deaths, does not have nearly the same level of police killings of felons. The Northeast does better on arrest-related deaths than police killings, with Pennsylvania, Maryland, and Vermont all scoring higher on the rate of felons killed by police. There is a particularly drastic change that occurs between the two maps in the Deep South and lower Midwest. Missouri, Tennessee, and Georgia have low levels of arrest related deaths per million, but rise to the top on felons killed by police.
• • • • •
The divergences on these maps highlight the need for better data on people killed by police. Knowing how many Americans are killed by police officers is important far beyond the outcry over the case of Michael Brown. A 2003 American Journal of Public Health study notes that the perception that someone was killed by police officers or by government agents has been the spark for “almost every major civil insurrection that has occurred in the United States in the past century.” Such incidents can and often do precipitate even more injuries and deaths, and cost their communities dearly, both economically and socially. When people no longer trust the police to handle things nonviolently, the study further points out, they are much less likely to cooperate with them or even to report crime at all. "The ability to accurately assess the incidence and characteristics of justifiable homicides committed by police officers is central to the development and evaluation of policies that promote public health and safety," the report concludes.
Meanwhile, some journalists are trying to fill in the gaps in data on their own. At Deadspin, writers are crowdsourcing data collection by asking readers to use Google’s search tools to find and submit shooting deaths via a special public submission form.
The bottom line: We need complete and transparent data on Americans who die in the custody or at the hands of police. Crime data is notoriously problematic, with under-reporting a common occurrence. But we are talking about people dying while under arrest or at the hands of police – very different from a random, simple, nonviolent crime, like somebody stealing a bicycle. The prospect of the state killing its own people is a very serious one. And the U.S. must mandate reporting of all such incidents, collecting and publishing detailed accounts of how, why and when these killings have occurred.
Only when we have that information can we even begin to pinpoint the problems within our justice system – and to begin the process of fixing them.