Law enforcementBetter tracking of police homicides

Official counts of homicides by police seriously undercount incidents, according to a study from Harvard T. H. Chan School of Public Health, but a relatively new national data system, currently in use in thirty-two states, could be a crucial tool for gathering more comprehensive information, say the researchers.

The study, which was published in the American Journal of Public Health, found that the U.S. Centers for Disease Control and Prevention’s National Violent Death Reporting System (NVDRS) provided more complete and accurate data on police homicides compared with two other national data systems.

“The U.S. is trying to get a handle on police homicides, but how do we learn which policies best prevent these deaths and protect police if we can’t even get an accurate count of them?” said Catherine Barberof the Harvard Injury Control Research Center (HICRC) at Harvard Chan School. “The NVDRS could be the best solution to this vexing problem.”

HSPH reports that counts of homicides by police are currently tracked by two official national data sources: the CDC’s National Vital Statistics System and Supplementary Homicide Reports (SHRs) from the FBI’s Uniform Crime Reports System. But these data sources each rely on a single source — death certificates for the CDC’s system and voluntary police reporting for the FBI’s system — and the study suggests that each undercounts police homicides by as much as half. In addition, the extent of undercounting varies greatly by state, making comparisons among states and municipalities unreliable.

Given the limitations of these data sources, the researchers looked at whether the NVDRS could be a better alternative. A relatively new data system created in 2003, the NVDRS pools information on all violent deaths from participating states. It includes detailed coded data from multiple sources — from death certificates, police reports, coroner or medical examiner records, and crime labs. It also includes short abstracts that describe the circumstances of each incident.

Although thirty-two states currently use the NVDRS, its use hasn’t expanded further because of limited federal funding — which states must apply for in order to implement the system.

For the study, the researchers established standard definitions as to what constitutes a homicide by law enforcement, and compared counts of these homicides derived from NVDRS data against counts from other official sources. For the period 2005-2012, looking at the sixteen states using the NVDRS at the time, the researchers identified 1,552 law enforcement homicides. They found that the NVDRS had identified 1,421 of these homicides – 92 percent of the total — and about twice the number of those identified in SHR data and 71 percent more than those identified in Vital Statistics.

The results showed that the average annual rate of these homicides did not increase over the eight-year period. A key finding was that homicides varied by race and ethnicity and by state. Blacks were nearly three times as likely to be killed in a police homicide as whites. Compared with Massachusetts, which had the lowest rate of such homicides, Alaska’s rate was nearly five times higher and New Mexico’s nearly six times higher.

Extrapolating from the rate of legal intervention deaths found in the 16 NVDRS states, the authors estimated that at least 731 people were killed by law enforcement officers each year in the United States during the study period.

“Expanding the NVDRS to a 50-state system — as President Obama’s budget calls for — and implementing a couple of technical fixes that we identify will supply the nation not only with a more accurate count of these deaths but detailed data on the people, places, weapons, and circumstances involved,” said senior author of the study, David Hemenway, professor of health policy at Harvard Chan School.

— Read more in Catherine Barber et al., “Homicides by Police: Comparing Counts From the National Violent Death Reporting System, Vital Statistics, and Supplementary Homicide Reports,” American Journal of Public Health (17 March 2016) (doi: 10.2105/AJPH.2016.303074)