About 1,200 people were killed by police officers in the U.S. in the 12 months that ended in May, according to a federal report released Thursday. That number is much larger than government counts of police killings for earlier years — and is much more in line with private estimates.
Criminal justice researchers have long argued that official counts of police killings, which rely on voluntary reports from local police departments, are woefully incomplete. Over the past decade, about half a dozen efforts by activists, volunteers and media organizations have sprung up in response to widespread outrage about high-profile killings by police officers to try to fill the breach using information from media reports and other sources. Their annual death toll estimates since 2013 have generally ranged from 1,100 to 1,400, more than twice as high as the counts from official government sources.
Thursday’s report from the Bureau of Justice Statistics is effectively an acknowledgment that the amateurs were right. It used media reports to fill in the gaps in data provided by law-enforcement agencies and reached a figure similar to those from the private groups.1
The corroboration by BJS of other outfits’ numbers “speaks to the power of some of what we’ve been collecting and what so many others have been collecting — that it’s really been able to approximate those numbers,” said Samuel Sinyangwe, who leads one of the volunteer efforts, Mapping Police Violence, and co-founded Campaign Zero, a group that promotes policy ideas that it says can reduce the number of people killed by police officers. Sinyangwe said the new estimate from BJS “sounds correct.” Mapping Police Violence lists 1,198 people killed by police during the period covered by the BJS report; the Guardian, which has been counting police killings since last year, lists 1,127.
Sinyangwe and other activists worry, however, that the government’s new efforts to collect better data on police killings won’t continue under Donald Trump and his nominee for attorney general, Jeff Sessions. If the Senate confirms Sessions, “I am skeptical about whether they will move forward with this,” Sinyangwe said. Spokespeople for Trump and Sessions didn’t return emails seeking comment.
The inadequacy of official statistics on police violence has drawn national attention since Ferguson, Missouri, police officer Darren Wilson, who is white, killed Michael Brown, a black man, in August 2014. Brown’s death and other high-profile police killings raised the profile of the Black Lives Matter movement and sparked a nationwide debate over police violence. But without reliable data, researchers couldn’t answer even basic questions about who was killed by police or whether the number of such deaths was rising or falling. FBI Director James Comey — whose agency, like BJS, is part of the Justice Department — last year called the lack of good data “ridiculous.”
Now the government is trying to improve its data. Last year, BJS researchers found that the agency’s existing methods of counting arrest-related deaths — basically, asking police departments to report totals — were probably missing about half of all cases. (Some researchers think even more were being missed.) Thursday’s report builds on that research by trying to find the deaths that were missing from official counts, using several methods.
First, researchers used automated searches to identify media articles and webpages that might contain information on arrest-related deaths — including shootings and other intentional killings (which BJS classifies as homicides), as well as suicides, accidents and deaths by natural cause. They then sorted those reports manually to find the ones to investigate further. Researchers also took advantage of the BJS’s existing contacts with the 18,000 local law-enforcement agencies around the country to go further than the nongovernmental efforts can. They contacted local law-enforcement agencies and medical examiners or coroners involved with deaths over a three-month period to confirm that the deaths happened and determine whether they should be counted; agencies and the researchers sometimes disagreed. (For example, some agencies resisted counting suicides or accidents that killed someone being pursued by police officers.)
The researchers also asked local agencies whether there had been any deaths that weren’t reflected in media accounts. That allowed them to estimate how many deaths their media-based counts were missing — information that BJS then applied to a full year of data to arrive at its annual death-toll estimate of 1,900, including 1,200 homicides. Deaths by suicide, accident and natural causes were more likely than homicides to get no media coverage.
“This hybrid approach was shown to, we think, really do a great job improving our coverage,” said Michael Planty, who co-wrote the report, is a deputy director at BJS and has been working for five years on measurement of arrest-related deaths. (The BJS counts deaths after arrests separately. In other reports released Thursday that were based on analyses that didn’t use new techniques, BJS reported that in 2014, 1,053 people died in local jails and 3,927 died in state and federal prisons — both figures higher than a year earlier.)
Using media accounts to supplement official reports has become a common research tactic in the grim accounting of terrorism, school shootings and police misconduct. Since 2009, the BJS has used what Planty called an “ad hoc” process of analyzing media accounts to improve its collection of police killings by identifying cases that agencies weren’t proactively reporting.
Speed-reading death reports and turning them into data is difficult and psychologically taxing work. The BJS contracted with RTI International, a nonprofit research organization based in North Carolina, to conduct much of the research. Between June and August 2015, a dozen RTI researchers read through and logged roughly 150,000 media reports that potentially contained information about a death. To try to reduce strain, “we limited folks to no more than 30 percent of their time a week,” said Duren Banks, a senior research criminologist at RTI and co-author of the new report. “It takes a little bit of a thick skin to read about this for hours on end.”
It also isn’t cheap. Even after researchers found ways to dramatically reduce the number of media articles that their algorithm flags, the work costs $40,000 a month, Planty said.

It’s not clear how the Justice Department will collect police-killings data in the future. The department hasn’t committed to adopting the methods described in the new report for its official statistics. (On Thursday afternoon, the department said it would release more information about its data-collection plans on Friday morning.) BJS has more work ahead of it, including reports planned for next year delving into the characteristics of people killed by police — other efforts have found that black Americans are killed at a rate much higher than their share of the population — and comparisons of its data with nongovernmental counts. The FBI is also mounting a parallel effort to get better data.
Activists worry that improving data on police killings, which has proceeded at a pace that many have found disappointing under President Obama, won’t be a priority in a Trump administration. This past summer, Trump blamed Black Lives Matter, without any evidence, for instigating killings of police officers, an issue he emphasized much more strongly in his campaign than killings by police officers. Planty, when asked whether he sensed that the change of administrations would affect BJS’s work in the area, said, “No.”
Even if BJS does proceed with more complete data collection, it won’t necessarily replace private estimates. Sinyangwe said he plans to continue his work regardless, because unlike BJS, which only aggregates deaths for statistical purposes, he and other independent counters provide details on individual deaths, allowing for more granular analysis.
D. Brian Burghart, who founded and runs another police killings database, Fatal Encounters, said the BJS effort was long overdue. He said he started his site because he “realized that the internet meant this information could no longer be hidden.” If the government improves the way it collects data, Burghart said, he would be happy to step aside after nearly five years.
“I want to be obsolete,” Burghart said. “This is the most boring, repetitive work that you can imagine. It’s horrible. I research violent deaths eight to 14 hours a day. So, yeah, it sucks. I’d be totally happy for [BJS] to do that.” He added, “I don’t know what’s taken them this long to get there, to be honest.”