Redesigning the Arrest-Related Deaths Program

A rigorous method to track the number of people who die during the process of arrest

The summer of 2020 saw the United States once again grappling with its painful history of racism. Civil unrest broke out after the release of a bystander’s harrowing video that showed the death of George Floyd at the hands of Minneapolis Police Department officers. But George Floyd was only the latest in a long line of individuals who die during an encounter with the police. Each death is a tragedy in itself; but as black individuals are disproportionately killed during encounters with police compared to whites, the losses are a testament to the harm that institutional racism has caused for generations.

RTI is at the forefront of the effort to use scientific methods to the effort to understand the impact of police violence across the country. Beginning in 2011, we worked with the Bureau of Justice Statistics on the Arrest-Related Deaths program, devising a method to create reliable, national data on deaths that occur during the process of arrest. This includes homicides by law enforcement personnel along with deaths attributed to suicide, accidental injury, and natural causes.

Our work advanced the rigor and accuracy of data collection on this critical issue. Previously, the Federal Bureau of Investigation and the Bureau of Justice Statistics maintained separate datasets on arrest-related deaths, relying on voluntary submission of information. We found that these methods did not capture the full scope of the problem—in fact, they missed as many as half of all homicides by law enforcement officers. We set out to fill in the gaps.

Addressing Flaws in Past Efforts to Collect Data on Arrest-Related Deaths

In accordance with the Death in Custody Reporting Act of 2000, BJS began collecting data on deaths that occurred in police custody during transfer and during the process of arrest. The ARD program, launched in 2003, is part of the Death in Custody Reporting Program, which also collects information on deaths that occur in jails and in prisons.

At launch the ARD program relied heavily on voluntary participation from law enforcement agencies to report the information, and many failed to do so. As a result, the ARD program did not capture all reportable deaths in the process of arrest. In fact, a study conducted by RTI International found that those efforts by BJS and the FBI missed as many as half of the homicides by law enforcement officers.

Due to the limitations of the ARD program, BJS suspended the collection and reporting of arrest-related deaths in 2014.

A Hybrid Process for Accurately Coding and Classifying Deaths in Custody

Our latest work demonstrates that combining approaches to identify arrest-related deaths—particularly open-source information review and survey of law enforcement agencies and coroner’s offices—can yield a better estimate of the prevalence and characteristics of arrest-related deaths in the United States.

To gain a comprehensive picture of arrest-related deaths, RTI and BJS devised a hybrid, two-step process to identify, confirm, and collect information about arrest-related deaths. Our data scientists developed an automated text analysis method to identify reports of potential arrest-related deaths from more than 100,000 relevant media outlets. Using sophisticated machine learning algorithms, we flagged and manually reviewed news articles about potential arrest-related deaths. We then surveyed law enforcement agencies and coroner’s offices to confirm deaths identified through the open data review and to provide more complete and accurate incident detail. We surveyed these agencies as well as a sample of additional agencies to identify other deaths not found in media sources.

For this initial effort, reviews of media reports in the United States from June 1, 2015, through March 31, 2016, produced 1,348 potential arrest-related deaths. We found that arrest-related deaths occurred in all 50 states during that period, with the largest number (224) occurring in California. Washington, DC, Wyoming, and New Mexico had the highest rate of potential arrest-related deaths, whereas New Jersey, New York, and Rhode Island had the lowest rates.

We then narrowed the focus to June, July, and August 2015, and conducted a survey of law enforcement agencies and coroner’s offices for official reports and information about the 376 arrest-related deaths identified through open-source review in those three months. This second review identified 424 arrest-related deaths during that time period—13 percent more than reported in the media. Of these 424 deaths, 63 percent were classified as homicide, 18 percent as suicide, 12 percent as accidents, 1 percent as natural causes, 2 percent as undetermined, and 4 percent as pending further investigation.  Law enforcement homicides were largely reported in the media, but other manners of death were likely to be identified from law enforcement agencies directly.

Based on those two surveys, we were able to estimate that approximately 1,900 people die annually while in police custody.

Eliminating Gaps in Data to Inform Public Discourse, Research, and Policy Decisions

Our efforts to develop a reliable and accurate process for tracking arrest-related deaths provide a useful tool for understanding the prevalence and characteristics of these incidents.

Currently, there is no federally sponsored data collection program that adequately measures all manners of arrest-related deaths in the United States, but our new methodology can eliminate that gap.  Using information readily available online, we can standardize data collection, streamline the collection process, and improve the reliance on the voluntary reporting by the more than 18,000 local law enforcement agencies in the U.S. As reported in The Guardian, this system is the most comprehensive official effort so far to accurately record the number of deaths at the hands of American law enforcement and to provide the “national, consistent data” described in 2015 by the U.S. Attorney General.

Such information is invaluable for criminal justice researchers and law enforcement agencies to understand the scope of the issue, as well as to determine whether arrest-related deaths are a growing problem or whether heightened media awareness is simply bringing more attention to the issue.

Equally important, our accurate method can provide policy makers and law enforcement agencies information that can be used to develop new policies and practices to reduce or eliminate deaths during the arrest process.

An Ongoing Need for Reliable Data

The nation’s quest to understand the circumstances that lead to deaths during arrests has grown more intense in the years since we worked on this project. Many law enforcement agencies have stepped up their response.

RTI is currently working with BJS to implement the Federal Law Enforcement Agency Deaths in Custody Reporting Program. We are collecting data on the deaths of individuals in the custody of the FBI, U.S. Marshals Service, Drug Enforcement Administration, and other federal agencies.

In early 2019, the FBI began collecting data on the use of force. This program expands on previous efforts by including incidents that result in serious injury, as well as homicides. However, it still relies on voluntary reporting from law enforcement agencies and does not capture deaths due to accidents and natural causes. Many state and local jurisdictions have begun to require agencies to report these incidents, and media-based and crowdsourced tools have emerged to help fill in the gaps.

Completing the picture of arrest-related deaths through more complete, consistent data is important at this moment in history. As we learn more about the short- and long-term context of these incidents, we improve the ability of law enforcement to serve society.