A new study released Thursday suggests that two major data sources used by federal officials to tally police homicides — instances in which a police officer fatally shoots someone while on the job — dramatically undercount the number of deaths.
Researchers from Harvard and Northeastern University found that the number of police homicides, also known as “legal intervention deaths,” recorded by the FBI’s Supplementary Homicide Reports (SHRs) undercounted the number of police-caused deaths by more than half in 16 states from 2005 to 2012, including New Jersey, Kentucky, Oregon, and South Carolina. The Centers for Disease Control and Prevention’s (CDC) National Vital Statistics System also missed a significant percentage of such incidents. The researchers conducted their analysis by comparing police homicides as recorded by those two data systems to the National Violent Death Reporting System, which is considered the most comprehensive surveillance program for violent deaths, but does not collect data from every state.
Across the three databases, in the 16 states examined, researchers found a combined 1,552 police homicides during that time period. The NVDRS identified 1,421 of those deaths (or 92 percent of the total), Vital Statistics only reported 906 deaths, and SHRs reported only 742.
“The main message is that NVDRS reports over twice the number of homicides by law enforcement as the FBI’s Supplementary Homicide Reports and 71 percent more than Vital Statistics,” Cathy Barber, the study’s lead author, tells The Trace. She says the findings are important because federal officials still use SHRs or Vital Statistics as their primary source of data for deaths caused by law enforcement.
On a national level, data from Vital Statistics and SHRs indicate that about 400 civilians are killed by law enforcement officers every year. The results from the NVDRS, if extrapolated nationwide, suggest that number is closer to 730.
NVDRS, which is also maintained by the CDC, works by gathering data from multiple sources — including police reports, coroner reports, and death certificates — to provide as much information as possible about each violent death, including the victim, suspect, weapon used, victim-suspect relationship, location, and circumstances. After that information is collected, an “abstractor” — the person responsible for analyzing that data — compiles it into one report and assigns codes for every detail of the death. One of those codes is “type of death,” which includes an option for “legal intervention.” That code is only used, according to the NVDRS manual, when “decedent was killed by a police officer or other peace officer (persons with specified legal authority to use deadly force), including military police, acting in the line of duty.”
In contrast, Vital Statistics only compiles information from death certificates. This poses a problem when it comes to identifying law enforcement homicides, because police intervention isn’t always explicitly mentioned on the documents. There are also serious limitations to using SHRs as a way to tally police homicides. The supplementary reports are provided to the FBI by local law enforcement offices as part of the agency’s voluntary Uniform Crime Reports system. While the majority of local departments use that system, many do not file SHRs, which are the only way the FBI can track legal intervention deaths.
Though the NVDRS is significantly more accurate than Vital Statistics or SHRs, it’s important to note that abstractors still missed at least 158 cases of police homicide over the course of the study (or 10 percent of the legal intervention deaths identified by Barber and her colleagues). Most of the under-reporting stemmed from human error, and Barber notes that it could be fixed by a simple technological solution within the system.
“It’s possible that if state NVDRS offices aren’t aggressive about circling back and checking on open cases to see if they are closed and ready to be abstracted, then some will be missed,” Barber says. “But this would have an easy fix, because offices would just need to periodically check back on open cases.”
Ultimately, Barber and her colleagues conclude that because the NVDRS provides a more comprehensive count of police homicides than either Vital Statistics or SHRs, the program should be expanded to all 50 states and adopted as the main method of measuring deaths at the hands of law enforcement officers. It’s still unclear whether this is an achievable goal — President Barack Obama requested $23.5 million for the NVDRS in his 2017 budget request, but many lawmakers have already described his plan as “dead on arrival.”