Weekend Long Reads: The Problem With Police Violence Stats

by Kevin Schofield

This weekend's long read is a research paper from the leading medical research journal in the U.K., The Lancet. The paper, however, has local roots: It was authored by researchers at the University of Washington's Institute for Health Metrics and Evaluation. The paper attempts to ascertain the accuracy of statistics on U.S. deaths caused by police in the National Vital Statistics System (NVSS), the Centers for Disease Control and Prevention's (CDC) official repository on births and deaths.

The NVSS collects data from death certificates, including the cause of death. Usually a physician fills out the death certificate, but it could be completed by a coroner or medical examiner instead if there is suspicion of crime, foul play, or police violence. However, that creates a conflict of interest, as the paper describes, if the same government responsible for police violence is also responsible for reporting it.

The researchers relied on open-source databases of police violence and killings, compiled and maintained by non-governmental organizations, as a second source to compare and verify the NVSS data for the years from 1980 to 2017. And it found serious problems: The NVSS significantly underreports police killings. The researchers found that nationwide in any given year the NVSS database misclassifies between 53% and 68% of police killings as attributed to another cause. Out of the 30,800 deaths caused by police between 1980 and 2017, it did not correctly report about 17,000 of them. There were also racial disparities: Black deaths at the hands of police were misclassified 59.5% of the time; white deaths 56.1%; Hispanic deaths 50%; and other races only 32.6%.

Their state-by-state analysis shows that misclassification isn't a red-state/blue-state thing; it's widespread, and the exceptions are rare. Here's the map for misclassification of deaths of Black people at the hands of the police:

After completing their estimate of the underreporting, the researchers then got busy reconstructing a more accurate count of police killings in the United States since 1980. They found some interesting trends over time: Between 1980 and 1990 the number of police killings (measured as mortality rate per 100,000 persons) dropped dramatically, especially for Black and Hispanic victims. But since 2010, police killings have been increasing across all demographics.

The researchers also dove into state-by-state comparisons of the mortality rates. It's probably no surprise that the mortality rate for white people is uniformly lower across the country:

The Black mortality rate, however, is far from uniform — and again it doesn't fit the picture you might expect. It's highest in the southern Rockies, the Midwest, and the deep-blue West Coast; lowest in the northeast, the northern Rockies, and the deep South.

There is a wealth of other insights in the paper, including state-by-state details on how the mortality rate has changed over the past four decades.

The researchers speculate a bit about but don't explore why the NVSS data on police killings is so inaccurate: Is it poor training for medical examiners, bias, a conflict of interest for the government trying to cover up its sins, or some combination of the three? It does, however, suggest three things that might help: better training on how to report deaths, whistleblower-style protections for medical examiners who accurately report police killings, and sustained funding for independent watchdog organizations to collect and verify the data free from government interference.

While it's important to have accurate data, and we can applaud efforts that improve data accuracy, we also need to remind ourselves that the data is simply a means to an end. The real goal is to get the number of deaths at the hands of the police to zero, across the entire country. And with a clearer focus on the real numbers, we can see that not only do we still have a long way to go to reach that goal, but in recent years we have been moving in the wrong direction.

Find the full long read here: Fatal police violence by race and state in the USA, 1980-2019: a network meta-regression

Featured Image: Thurman Blevins "Please don't shoot me!" March in Minneapolis, Minnesota. Photo is attributed to Fibonacci Blue (under a Creative Commons, CC BY 2.0 license).