How many people are killed by police?
It’s a simple and important question, but, until this year, the shocking answer was “we have no idea.” There is no central authority systematically keeping track of killings by police, and police in most states are not required to report the use of lethal force to anyone.
The FBI’s annual Crime in the United States report does keep track of how many homicides were reported by police that were deemed to be justified killings by police. As defined by the FBI, “justifiable homicide by law enforcement” means “the killing of a felon by a peace officer in the line of duty.” From 1995 to 2014, we averaged about 380 justifiable homicides by police each year, with a very slight (but erratic) upward trend.
The obvious question is, how many unjustifiable killings were committed by law enforcement? There is no report for that: unjustified homicides by police would, in theory, be recorded as generic murders. Moreover, because police departments themselves decide what to report, and what is justifiable, there’s plenty of room for doubt about whether those homicides really were justified — or whether police were actually reporting all of their killings.
From 2003 to 2009, the Bureau of Justice Statistics (BJS) tried to track police killings more systematically, using data (once again, voluntary reported by police) from its Deaths in Custody Reporting Program (DCRP), using a subset of that data on “arrest-related deaths: homicides by law enforcement.”
This data doesn’t differentiate between justified and unjustified killings, so you might think that you could simply subtract the FBI’s “justifiable” killings from the BJS total killings, and you’d get the number of “unjustifiable” killings. If you do that math, from 2003 to 2009, the BJS recorded 2,931 “arrest-related homicides” by police, and the FBI recorded 2,660 “justifiable homicides” by police. That’s a difference of 271 (presumably unjustifiable) homicides, about 9 percent of the total, over seven years.
But not so fast. The BJS and FBI numbers conflict in weird and impossible ways. As the charmingly titled blog “numbskulduggery” pointed out, from 2003 to 2005, California police reported 160 homicides to the BJS, while reporting 354 justifiable homicides to the FBI — “implying that in California there were 194 more justified police shootings than there were total police shootings.” In addition, some states participated in the BJS program but not the FBI’s, or vice versa, but they are lumped into the total comparison anyway.
So despite two federal databases, there was no clear consensus on the number of police killings, and the data was hopelessly muddled by states and agencies not participating or reporting conflicting numbers. That was the situation in 2014, the last time I wrote about this question, but nonetheless, I concluded: “Still, this record is the best data there is and probably tracks the true figure pretty well.”
A lot has changed since then. And, boy, was I wrong. Cops do not report anywhere close to the true number of people they kill, either because of neglect, confusion, or perhaps not wanting to draw attention to their department.
In 2015, a number of private groups started systematically tracking police killings from news reports.
The Washington Post started tracking people “shot dead” by police (weirdly, specifically only counting shooting deaths). At the same time, the Guardian began counting all people killed by police, regardless of the means. Brian Burghart, former editor of the Reno News & Review, founded a website called “Fatal Encounters,” which ambitiously attempts to count all police killings across the United States going back to the year 2000. Fatal Encounters’ database uses numbers from other public databases, public records requests, paid researchers, and crowdsourced submissions from verified media reports.
In 2015, the Washington Post counted 990 people shot dead by police; the Guardian counted 1,146 people killed; Fatal Encounters counted 1,357 killed.
These are two, three, or four times higher than the numbers typically reported by the FBI or the BJS. (This, of course, does not say anything about how many are justified or not.)
Fatal Encounters’ data from earlier years is admittedly incomplete, but they have so far collected more than 14,000 records of people killed by police from January 2000 through June 2016. From 2000-2014, FE records at least 12,137 people killed by police, compared to just 5,830 reported to the FBI by police over the same period.
So are police killings becoming rarer or more common?
It’s not yet possible compare national trends over time. While FE believes they have complete records from all 50 states since 2013, we don’t know how complete their records are from prior years. There are many obstacles: searching and finding old media and government reports is difficult, and the likelihood of a police killing being reported on at all probably decreases the further back in time you go. Nonetheless, FE states:
We believe we include complete records for these 28 states back to 2000: Alabama, Alaska, Arizona, California, Colorado, Connecticut, Delaware, District of Columbia, Florida, Idaho, Louisiana, Maine, Massachusetts, Minnesota, Mississippi, Montana, Nevada, New Hampshire, New York, North Carolina, North Dakota, Oregon, Rhode Island, South Dakota, Utah, Vermont, Washington and Wyoming.
So a comparison of just these 28 “complete” states should give us a true indication of the frequency of police killings over time, at least in half the country. What it shows is alarming and incredible: a steep and steady increase from 359 killings in 2000 to 739 killings in 2015.
My rule of thumb with numbers that look unbelievable is generally “don’t believe them,” at least without corroboration. There are lots of reasons to think that reporting might be a lot better in recent years, and that reports in prior years (if they were made at all) may be increasingly difficult to find the further back you go. In addition, FE’s totals for 2013-2015 — the years they consider most complete — are pretty flat.
So I can’t testify to the quality or completeness of the historical data FE has assembled, but if the trend is significantly upwards, that would be disturbing, since violent crime has been steadily declining over the same period, as have been attacks on police officers.
So here’s what we know now about how often cops have been killing people in the United States:
The federal data is crap, the private data is better but still incomplete, and there’s a bunch of recent data points and a couple suggestive (but not definitive) possible trends. The numbers are interesting, but not illuminating.