A note on the data we used
One of the first questions we asked when we embarked on this reporting was, “What data exists on U.S. gun deaths?” The answer is that there is lots, but all of the sources are incomplete and flawed in different ways. As a result, we’ve had to stitch together different data sets to try to get the most complete picture.
The main source of data for these articles, as well as for the interactive graphic, is the Multiple Cause of Death database from the Centers for Disease Control and Prevention. The MCD, part of the CDC’s National Vital Statistics System, is based on death certificates filled out by funeral directors, doctors, medical examiners and coroners across the country. Because nearly everyone who dies gets a death certificate, and because gun deaths are generally pretty easy to identify, the MCD is widely considered the most comprehensive count of U.S. firearm deaths.
But the MCD isn’t perfect. Its counts of accidents and police shootings are considered unreliable. And the database doesn’t include any information on the circumstances behind a shooting — how many other people were killed, or who did the shooting. So the MCD can’t identify mass shootings or incidents of domestic violence.
Another CDC data set, the National Violent Death Reporting System, tries to fill some of those gaps by linking data from coroners and medical examiners with information from law enforcement sources. But only 32 states participate in the system.
The Federal Bureau of Investigation collects data on shootings through its Uniform Crime Reporting system, which is based on widespread (but voluntary) reporting by local law enforcement agencies. For homicides, the FBI collects additional details of an incident through its Supplementary Homicide Reports. Researchers say, however, that these reports are often incomplete and out of date; based on the number of homicides listed in the FBI data compared to the number listed by the CDC, the FBI’s count misses roughly 20 percent of all gun homicides. And the FBI doesn’t collect any information on suicides or accidental deaths.
Throughout these stories, we’ve used the CDC’s Multiple Cause of Death data as our source for the overall number of gun deaths and for most of the information on the demographics of the victims. But where the MCD fell short, we’ve supplemented it with data from the violent-death system and the FBI, as well as with information from state and local agencies. Where government data was lacking, we also turned to private sources, including the Global Terrorism Database, a database of mass shootings from Mother Jones magazine and a variety of counts of police shootings.
Data and code for this project are available on our GitHub page.
— Ben Casselman
Credits
This project was reported and written by Anna Maria Barry-Jester, Carl Bialik, Ben Casselman, Maggie Koerth-Baker, Leah Libresco and Hayley Munguia.
The interactive graphic was by Ben Casselman, Matthew Conlen and Reuben Fischer-Baum. Additional contributions by Kshitij Aranke. David Nield contributed research.
Charts by Ella Koeze with additional contributions by Ritchie King.
Photography by Daniella Zalcman. Illustrations by Hannah Drossman.
Simone Landon was the editor and Meghan Ashford-Grooms, John Forsyth and Blythe Terrell were the copy editors. Additional editing by Chadwick Matlin.
Creative direction by Kate Elazegui.
Design and art direction by Kate LaRue. Photo production by Amy Hoppy.
Development by Ritchie King, Justin McCraw, Paul Schreiber and Gus Wezerek.