A few weeks ago I wrote about a government report that got us a little closer to knowing how many people the police kill in the U.S. each year. But it still left a lot of doubt as to the real number. This week, a new study makes a bit more progress. It suggests the government estimate was 25 percent too low.
The study, by human-rights statisticians who specialize in estimating unreported violent deaths in other countries, essentially puts a number on a qualitative critique they made last month, just after the Bureau of Justice Statistics issued its report. The authors, Kristian Lum and Patrick Ball of the Human Rights Data Analysis Group, said in the earlier critique that the BJS was likely undercounting the number of omitted killings. Now they’ve extrapolated from their work in five other countries (Colombia, Guatemala, Kosovo, Sierra Leone and Syria) to estimate that the BJS study, which said the police killed 7,427 people in 2003 through 2009 and 2011, missed approximately 2,500 to 3,000 killings by police — about one quarter of the total number.
Michael Planty, chief of victimization statistics for the BJS, declined to comment on the report by Lum and Ball.
Lum emphasized how uncertain she and Ball were about their estimate, which suggests that about 1,200 to 1,300 Americans were killed by the police annually between 2003 and 2009 and in 2011. They don’t know if that number is right, but they do know that the true number “could be much bigger” than the latest BJS estimate of about 930 annual deaths, Lum said in a telephone interview.1
Lum emphasized that this finding shouldn’t be interpreted as a condemnation of the BJS report, which, she said, was the best analysts could do with the data they were working with. She isn’t confident enough in the new method outlined in her paper to use it in other countries. The estimates have lots of uncertainty; it’s possible, though not likely, that the BJS overestimated killings by police in the U.S. instead of undercounting them.
These analyses would be far more reliable if the government produced a single, complete, updated national count of killings by police. “No amount of modeling can compensate for good data,” Lum said. “We try, but at the end of the day, complete enumeration of all the people who have been killed by police would be much better.”