In the 1990s, District of Columbia police shot and killed more people per resident than any metro police department in the country – including New York, Los Angeles, Chicago, Detroit and New Orleans.
The Washington Post story, "Deadly Force," resulted from a close examination of the FBI's Supplementary Homicide Report, one of the computer databases included in the annual release of the Uniform Crime Reports.
That examination revealed that the 1995 SHR data for the country (released in 1996) did not have a single "81" – the code in the circumstance field for justifiable homicide by a police officer. Neither did any of the other years of data that I checked.
This suggested that whole records were missing from the dataset. The FBI's own documentation revealed another corroborating clue: The data is provided on nine-track tapes, and an "analysis of records written" printout accompanies each tape. This printout describes how many records were "read" in the government's computer versus how many were "written" on the tape provided to the purchaser. These should match; in 1995, however, the number of records written was short by 287.
They shall be released
Several conversations with the FBI revealed that the bureau did collect justifiable homicides but that the numbers were not part of the "standard data release." We ordered a nonstandard release, specifically asking for justifiable homicides.
Six weeks passed before the nine-track tapes arrived. When they did, we eagerly downloaded the data, opened the records, scrolled to the circumstance field and looked for 81s. There were none.
Despite a very specific written request, which had been reinforced with several telephone conversations, the FBI had sent the "standard release," minus justifiable homicides. We reordered the data and reiterated the request.
Several more weeks passed, new tapes arrived, we processed the data, and there they were: hundreds of 81s.
A hot tip
The raw numbers of justifiable homicides by police officers alone were remarkable. Only officers in a handful of cities – all much larger than Washington, D.C. – had shot and killed more people. Ultimately, these records served only as a starting point.
As the project progressed, The Post decided against relying on the FBI's justifiable homicide reports for several reasons. Some criminologists have questioned the data's accuracy. By definition, FBI reports are jurisdictional, not departmental: If an officer kills someone in another jurisdiction, according to Uniform Crime Report guidelines, the death should be reported by the agency in the jurisdiction where it occurred rather than by the officer's own department.
Additionally, reports to the FBI may include justifiable homicides by armed security guards, transit police, housing police or other armed police. There is no flag in the data to indicate whether an officer is a member of the reporting agency's force.
Because of these and other concerns, The Post chose to survey directly police departments to collect fatal shootings and woundings by each department's officers for 1990 to 1997. We focused on the 27 cities that had at least 500,000 residents at any point from 1990 to 1996, the latest year that U.S. Census population estimates were available.
For verification, we compared the numbers we collected to those in the SHR. We searched news clippings for previously published reports. Often, there were small differences; sometimes, bigger ones. To address these variations, we contacted the departments for further clarification or sought other sources, such as state UCR reporting programs or state attorneys general charged with reviewing fatal shootings by police officers.
Five-gauge analysis
After consulting criminologists and reading literature on the subject, we determined that no single measure was adequate to gauge a department's number of officer-involved shootings. We decided instead to measure fatal shootings against five other categories: population, violent crime, arrests for violent crime, homicide and the number of uniformed police officers.
Our data sources, all electronic, included:
• Return A for violent crime and homicide totals. (This is the base table in the FBI's UCR.) Because it is the FBI standard, we used the Return A for homicides instead of the SHR. We also found its numbers to be more accurate than the SHR's. For example, in 1996, the SHR recorded four homicides for Austin, while the Return A recorded 40.
• Age, sex and race of persons arrested for the number of violent crime arrests
• Police employees for each department's number of uniformed officers
• Instead of the FBI's population estimates for cities, which are extrapolated from state estimates, we used the more sophisticated Census Bureau city estimates for per capita calculations. (These figures are not ready early enough for inclusion in the FBI's annual reports.)
Arriving at violent crime, homicide and arrest totals required adding up numerous fields of coded data. To verify our data work, we checked each number that we planned to use against the appropriate volume in "Crime in the United States," the FBI's annual publication. To further vet the FBI data, we checked the District's crime numbers with the police department and called a sample of other departments as well.
More shootings by all measures
To help offset possible discrepancies and to compensate for years with accurate but unusual spikes in fatal shootings, we averaged the number of fatal police shootings over the period for each department. Likewise, we averaged the numbers of violent crimes, arrests for violent crimes, homicides, uniformed officers and population.
We used these averages to calculate how many fatal shootings there were per 100 homicides, per 1,000 arrests for violent crimes, per 10,000 violent crimes and per 100,000 residents in the District.
We aggregated the other 26 cities' numbers and compared the District to the group. We also aggregated the data for nine other particularly violent cities and compared the District to that subset as well. Because of variations in how departments define and report violent crime, we used homicides to identify violent cities, choosing those with the highest homicide rates for the period.
By each of these measures, the number of fatal shootings by District officers was higher than average – often significantly higher. Only when measured against homicide did D.C. approach the norm. Some criminologists say a high rate of homicide helps explain officer-involved shootings.
To confirm the statistical significance of our findings, The Post consulted with Sarah Cohen, then NICAR's training director (The Post has since hired her), and later submitted the findings to James Fyfe, a nationally recognized criminologist, for review.
Local digs
While we mined these federal databases, we also dug in locally, obtaining at least one electronic database, building others and scouring the courts for civil and criminal cases.
With a FOIA request, we obtained an electronic database from the District of Columbia Use of Service Weapon Review Board that purported to detail every firearm discharge by a District police officer. The records outlined the circumstances leading to each shooting, but the department redacted officer names.
We collected and built a database of nearly seven years of press releases from the D.C. police department on officer-involved shootings. We also found seven press releases on fatal shootings that had never made it to the department's Use of Service Weapon Review Board, the body charged with determining whether a shooting was justified or not.
Another FOIA request yielded a list of lawsuits against the District, the Metropolitan Police Department or its officers that involved a shooting. We examined each of these lawsuits, gleaning more details about the shootings, the officers and sometimes the department's own investigations.
Eventually, we expanded our search to include police brutality and built a database of more than 300 lawsuits against the District, the police department or police officers that had been settled or adjudicated, recording the settlement amount and the nature of the complaint.
Finally, we mapped the local data, which provided a stark and startling picture of the volume and location of fatal shootings and woundings in the District.