How the DOJ Can Federally Document Every Fatal Case of Police Misconduct

Police Accountability Series: Part 2

Part 1 of the Police Accountability series discussed the critical need for federal, systematic data collection on police misconduct to build the infrastructure for successful legal investigations against officers and police departments abusing their power. It also examined the DOJ’s largely untapped, pre-existing authority to track, document, and investigate police misconduct while calling on President Biden & Attorney General Garland to actualize 3 proposals. This piece will unpack the inner-workings of the first proposal: reinstating the Arrest-Related Deaths program.

The Arrest-Related Deaths (ARD) program — a federal tracking system to document cases of police homicide amid other manners of death emanating from the arrest process & the police interaction at large —  held massive potential to further police oversight goals. It would have been an unparalleled resource for evidence in legal investigations against fatal law enforcement misconduct had it not been discontinued in 2014 over efficacy concerns. Relaunching the ARD program could not be a more pertinent federal tool to curb police violence amid outpouring police assassinations of People of Color.

Not only does re-establishing the ARD program require no legislation, but a template for a revised ARD program has been underway for over 6 years. The pilot study reconstructing the ARD’s methodology yielded a blueprint mapping out how the relaunched ARD program could operate by a hybrid system of open-source media mining and law enforcement agency surveying. Reactivating the ARD program in its contemporary version is one of the easiest & speediest proposals that the Biden administration & Garland’s DOJ can actualize on the matter of criminal justice reform — so, why aren’t they?

What was the Arrest-Related Deaths (ARD) Program?

ARD was an annual data collection project (2003 – 2012) stemming from the legal stipulations of the Custody Reporting Act of 2000. The objective was to record cases of civilian deaths occurring before, during, or shortly after “an arrest event or noncriminal incident.” While the ARD program stored data on police homicide, justified and unjustified, its scope further extended to deaths by means the BJS did not consider to be directly linked to law enforcements’ actions. 

To elaborate, the Bureau of Justice Statistics (BJS) defined “arrest-related” as:

1. “any use of force by state or local law enforcement 
2. injuries sustained while attempting to elude law enforcement or injuries incurred while in custody
3. self-imposed events, such as suicides, accidents caused by the decedent, and intoxication 
4. medical conditions or illness.”

However, these so-called ‘indirect’ immediate causes of death — suicides, accidents, natural causes, etc. — do not exist outside of the totality of the situation, which first and foremost is characterized by the police presence. The cause of death inherently is an implication of the police interaction. Thus both ‘direct’ and ‘indirect’ arrest-related deaths must be weighed equally for research, statistical, and legal purposes alike.

Although the mission of the ARD program wasn’t to comprehensively monitor the police, points 1 and 2 can be useful elements of the program for the DOJ to wield as a metric for police oversight. The second point counters pernicious narratives that Black and Brown people would not be killed if they simply cooperated with law enforcement and complied with their demands. While the “non-compliance” in question tends to actually be People of Color exercising their rights, like questioning on what grounds they’re being arrested or exercising the right to film interactions with the police, non-compliance is frequently weaponized to justify law enforcement homicides. However, neither resistance nor criminality, alleged or confirmed, give authorities a free pass to kill. 

To gather the census data, the BJS allocated State Reporting Coordinators (SRCs) in all 50 states and the District of Columbia. The SRCs were tasked with identifying, confirming, and reporting all cases of arrest-related deaths deemed eligible for the ARD program utilizing any available resources in their respective localities.

While the conceptual principles of the ARD program were well-placed, they were poorly executed. The program was suspended in 2014 over concerns pertaining to the credibility & accuracy of the data, as well as the significant variation in SRC’s strategies to procure the data. The BJS’ assessment of the ARD program’s methodology deduced that the limitations to ascertaining the real statistics on reportable arrest-related deaths impeded data quality. The latter resulted in a substantial underestimation of arrest related deaths.

A year later, in 2015, BJS launched a pilot study to gauge the effectiveness & feasibility of a revamped ARD program with a revised methodology and released the findings in summer of 2019.

The Redesigned ARD Program and its Pilot Study Findings:

First, the pilot study’s terms and definitions for arrest-related deaths did not explicitly include injuries (death being one) endured while eluding law enforcement. It’s unclear at this point whether said provision has been omitted from the redesigned ARD program, whether it’s implied under the umbrella phrase of “while the freedom to leave of [the decedent] was restricted by law enforcement personnel,” or whether the new terminology & definitions continue to operate in accordance with the previous ARD program. Whichever the case may be at the moment, the new ARD program needs to explicitly include the provision for eluding law enforcement otherwise it will buttress harmful rhetoric that cooperation is a precondition for protection from unconstitutional and fatal police misconduct. 

The ARD program pilot study’s methodology consisted of two phases:

1. Using open-source intelligence (OSINT) in a standardized fashion to mine public media (social networks, news outlets, etc) for potential eligible arrest-related deaths 
2. Surveying law enforcement agencies (LEAs), medical examiners’, and coroners’ (ME/C) offices in an effort to confirm arrest-related deaths discovered via the media, identify arrest-related deaths not covered by the media sweep, and inquire about the details of each decedent and the specifics of the police encounter.

The data collection process of phase 1 was accelerated by consulting pre-existing open-source lists (see Table 1) housing data with overlap in scope and criteria with the ARD program. The data retrieved from the media was then reviewed to discard duplicate cases and to inquire further about cases which lacked information.

With respect to phase 2, the BJS implemented a dual-selection strategy in deciding which LEAs & ME/C offices to survey: 

1. All LEAs & ME/C offices with jurisdiction over the media-identified deaths were surveyed and
2. A sample of 498 LEAs and 232 ME/C offices with no media-identified deaths were surveyed under the assumption that a media sweep could not capture all incidents of arrest-related deaths that occured in the respective time period under study.

Using OSINT, the BJS unearthed the state & local LEAs & ME/C offices of all media-identified deaths with the intent to have law enforcement confirm the death and provide characteristics of the incident to ensure they met the ARD’s metrics for arrest-related deaths. 

This remains a problematic feature of the redesigned ARD program as the role to claim media-identified deaths appoints law enforcement to be the gatekeepers of which cases will be included in the federal ARD census, all the while dominating the narrative regarding the circumstances of the death — which are also documented in the ARD census. This taints the authenticity of the ARD’s data as LEAs are called on to fill-in the blanks, so to speak, regarding the decedent’s demographic information, “the reason the LEA was involved with the decedent, actions the decedent took during the incident, and actions the LEA took during the incident” via the CJ-11A incident form — which, judging by the history of law enforcement’s cooperation, or lack thereof, with oversight mechanisms, can be easily falsified or manipulated at the LEAs’ discretion. 

The BJS’ own analysis of CJ-11A forms disclose the following trends:

• More than 1 in 10 cases were missing the following information:
  • “Number of shots fired (17%)
  • Number of officers responding (14%)
  • Decedent’s weapon type (17%)
  • Whether the decedent attempted to injure others (12%)”
• Details outlining law enforcement’s actions, information acquired to measure the extent of the use of force, was missing for approximately 7% of cases
• In all cases, LEAs reported that:
  • 93% of decedents had allegedly committed a crime
  • 18% of decedents seemed to be battling mental-health problems
• With respect to law enforcement homicides, LEAs reported that:
  • In 97% of cases, decedents had allegedly committed a crime
  • In 89% of cases, decedents were either carrying a weapon or had appeared to be in possession of a weapon, however only 28% of these decedents had actually fired a firearm
  • In 83% of cases, the decedent reportedly tried to harm police or others
  • In approx. 50% of cases, decedents reportedly either attempted or were successful in eluding law enforcement 
• In 20% to 30% of law enforcement homicides, measures to describe the circumstances up to the event of death were missing, largely due to agency non-response.

Does it not strike one as a biased account of the ‘truth’? How can responses voluntarily provided by law enforcement be considered reliable when specific information is willfully omitted in adverse cases? Is it not scheming when LEAs foreground the punitive & disproportionately villainized rhetoric of a ‘criminal’ and perpetuate the fear-mongering around ‘criminal activity’ to foster a sense of deference for law enforcement? What about when LEAs try to garner sympathy by the perceived presence of a weapon threatening their life — is a wallet, bible, hairbrush, iPod, skittles, slice of pizza, sandwich, pill bottle, underwear (a non-exhaustive list) a weapon? 

One secondary reason for the adopted hybrid data collection process was to judge whether OSINT in and of itself would be sufficient to procure all eligible arrest-related deaths or if it would need to be buttressed by LEA reports. Considering that 99.6% of all law enforcement homicides were tracked via media sources, coupled with the fact that 63% of the arrest-related deaths were law enforcement homicides, media-identified deaths must stand independent of LEAs’ confirmation to ensure that the revamped ARD program’s data retains its integrity & accuracy. 

Seeing as the conflict of interest predominantly lies in cases of law enforcement homicides and that OSINT does not carry the same promising results in cases of arrest-related deaths by other causes — suicide (83%), accidents (54%), natural causes (17%) — LEAs and ME/C offices can continue to be surveyed to assist in collecting data on cases of arrest-related deaths fitting within the aforementioned classifications until methods proving more successful are discovered & tested. The redesigned ARD program can look to existing non-governmental police monitors who’ve sought this data independent of law enforcement by inquiring about the information from public defenders and non-law enforcement first responders (i.e. firefighters, paramedics, EMTs) and whatnot. The search for alternative methodology to gather arrest-related deaths statistics not classified as law enforcement homicides should not postpone the relaunching of the revised ARD program. 

In essence, a final refinement to eliminate a distorted vision of a verification procedure by LEAs must be made and the BJS’ original definition of “arrest-related” which explicitly includes injuries (inclusive of death) acquired while trying to elude law enforcement must be preserved. Otherwise, the pilot study of the redesigned hybrid methodology for the ARD program proves optimistic in assembling federal data for police oversight objectives. It addressed the former ARD program methodology’s limitations & rightful criticisms of the data’s accuracy/fullness and reliability. Fundamentally, the pilot study’s methodology standardized SRCs’ data acquisition mechanisms, “resulting in significant improvements in data completeness and quality compared to the previous ARD program methodology,” all the while empowering the public to contribute in checking the police as OSINT-identified deaths tend to take root in civilian footage published on (social) media outlets.

Please stay tuned in the next article of the Police Accountability series that unpacks the following proposal focused on the C.O.P.S. Office. The Police Accountability series explores executive branch criminal justice reform promoting law enforcement oversight through the DOJ’s latent Office of Justice Programs — housing both the Bureau of Justice Statistics & the office of Community Oriented Policing Services.

Header image: “Police The Police Sign – Dog wearing cardboard Black Lives Matter Protest Sign – Justice For Regis – Not Another Black Life rally and March – May 30, 2020 – Creative Commons Photos Here Later Today – Toronto Christie Pitts Park down Bloor Street to Queens” by Jason Hargrove is licensed under CC BY 2.0