The Justice Department Must Improve Death in Custody Data
Senate Homeland Security and Government Affairs Committee,
Permanent Subcommittee on Investigations on "Uncounted Deaths in America’s Prisons and Jails: How the Department of Justice Failed to Implement the Death in Custody Reporting Act"
Chairman Ossoff, Ranking Member Johnson, and Members of the Subcommittee: Thank you for your attention to the Death in Custody Reporting Act (DCRA) and the Department of Justice’s (DOJ) ongoing failure to implement this important law. There is no doubt that there is a crisis of deaths in custody in this country, and that Black and Brown people are disproportionately the victims of violence, abuse, and neglect at the hands of the criminal-legal system. Due to DOJ’s failure to implement DCRA, there is no single, comprehensive source of data on the number and circumstances of deaths in custody in the United States. This information vacuum deprives policymakers of crucial data to develop policies aimed at reducing these deaths.
Over the last four years, The Leadership Conference and the Project On Government Oversight (POGO), as well as many other organizations, have sent letters, written articles, and met with the DOJ and congressional offices to highlight the urgent need for action on DCRA.1 As the Subcommittee’s investigation and hearing have made clear, however, DOJ continues to fail to collect meaningful data. This statement draws from past advocacy and a forthcoming report by The Leadership Conference and POGO on DCRA to highlight four of the most glaring shortcomings in the DOJ’s approach to the law, as well as outlining a meaningful path forward.
A Flawed Approach
DOJ has shown conflicting approaches to interpreting the scope of DCRA’s mandate since the law’s passage. DOJ’s current plan reflects a narrow reading of DCRA.2 It requires covered jurisdictions to report data on deaths in custody, but it does not commit to validating that data, requiring iterative improvements to collection efforts, or making data available to the public.
We believe this narrow approach undermines the spirit of DCRA, and may in fact prevent the department from fully complying with the statute’s research and policy analysis requirements. It is likely that this approach will lead to incomplete and inaccurate data, making it impossible to meaningfully study deaths in custody or propose ways to reduce them.
A more robust approach is found in DOJ’s since-abandoned 2016 implementation plan.3 That plan included a number of provisions that are not explicitly required by the statute but significantly increase the likelihood that the program will be successful, including data validation, procedures to review and improve collection plans, and a proactive release of anonymized data so that the public can engage in the effort to understand and reduce deaths in custody.
In this statement, we recommend an implementation plan based on that 2016 proposal. We also recommend several technical measures to improve the comprehensiveness and utility of the data DOJ collects.
Challenges to Address
Most of our recommendations address four key areas where DCRA implementation can and should be improved: compliance, data collection methods, research scope, and transparency.
DCRA cannot successfully serve its purpose if the agencies and jurisdictions that are required to report data fail to do so without consequence. To date, it appears that there have been issues gathering data from all required sources at both the federal and state levels. And DOJ’s decision to not impose the statutory penalty of withholding up to 10 percent of a state’s federal justice- related funding from non-compliant states, and its failure to set forth clear guidance for when and how it will be imposed, are missed opportunities to use the full scope of statutory compliance tools.
At the federal level, the Office of the Inspector General (OIG) reported that, as of 2018, DOJ was still unsure of the number of federal agencies even covered by DCRA.4
We do not know the status of internal efforts to determine that number, but defining the universe of covered agencies is an essential prerequisite to full compliance.
At the state level, we recognize that the statute places the burden of collecting data from local departments on each state, creating a patchwork of collection plans and several layers of reporting relationships. Because no data from the state collection has been released yet, we cannot know whether compliance is adequate. However, we are concerned that DOJ’s current effort does not adequately support or oversee state collections. For instance, it does not require states to submit collection plans or improve those plans over time.
Again, DCRA gives the U.S. Attorney General the discretion to withhold up to 10 percent of a state’s funding under the Byrne Justice Assistance Grant program and related funding sources. We recognize concerns that the compliance mechanism of a 10 percent grant reduction to a state’s grant allocation is a blunt tool, as that particular money may or may not affect all local agencies. However, as the penalty has not been imposed yet, we have been unable to gauge its effectiveness. While it is true that the state-level penalty may not give every local agency a financial incentive to comply with the reporting requirement, it may be a sufficient incentive for states to leverage their own power over local jurisdictions to ensure compliance.
In addition to compliance, the methodological weaknesses in how data are currently collected would lead to such poor data quality as to jeopardize its usefulness.
The DCRA data collection instruments are forms designed for agencies and respondents to use to capture and report the DCRA data. The Bureau of Justice Assistance (BJA) uses one form for state deaths; the Bureau of Justice Statistics uses two forms for federal deaths (for deaths during arrests and in corrections custody).5 Inconsistencies between these forms and changes to the forms over time cause data integrity issues. For example, if questions change from year to year, it is difficult to compare data over time. Additionally, if the forms ask substantively similar questions in different ways, it becomes more difficult to compare data collected from different forms.
More fundamentally, the current forms are insufficient for capturing all the information necessary to address the research questions mandated in DCRA. For example, on BJS Form CJ-13A, the 2016 and 2017 instruments required federal law enforcement agencies to report the number of officers who responded to the original service call, the number of officers discharging weapons, and the number of shots fired.6 By 2018, these questions were deleted and as of 2021 had yet to be added back to the instrument.7
Between the 2018-2019 iterations and the 2021 iterations of BJS Form CJ-13B, multiple questions were dropped.8 The dropped questions are critical for research, transparency, and accountability. They included questions about the existence and source of a death certificate; additional questions about people who died by suicide; additional injuries to personnel, other confined people, or the victim; law enforcement actions (e.g., use of a weapon); and, if the death was due to a preexisting medical condition, information on any medical treatment the decedent received.
The current form for the BJA state collection is also inadequate. The form requires very little data and reporting, especially when compared to the federal collection forms (CJ-13A and CJ- 13B). Most notably, the BJA form omits questions regarding weapons used by the decedent and the officers/personnel; the reason for law enforcement use of force, injuries to law enforcement personnel and civilians; decedent behavior including types of resistance; or the decedent's perceived state of mind. Instead, the BJA form attempts to capture this in the qualitative text field with the following instructions: “Please provide a brief description of the circumstances leading to the death (e.g., details surrounding an event that may have led to the death, the number and affiliation of any parties involved in the incident, the location and characteristics of an incident, other context related to the death, etc.).”
This meets the statutory requirement to collect “a brief description of the circumstances” surrounding deaths. However, this open-ended approach runs the risk that not all relevant information will be captured and complicates subsequent data analysis by using text rather than checkboxes.
All these inconsistencies in data collection harm data integrity and usability. The inconsistent wording of questions across years diminishes our ability to compare trends or agency types. This also applies to comparisons between federal, state, and local agencies. The result of these challenges is a likelihood that the department will be unable to use the data to answer the statutorily mandated research questions.
Another challenge to compliance and collecting quality data is that collection methods likely vary by state. With these different methods of collection, it may be methodologically unsound to compare data from different states.
Finally, the track record of the BJS Arrest Related Deaths program suggests that relying solely on self-reported data will likely result in undercounts.9 While we recognize that the plain language of DCRA only requires the department to collect data from state reporting agencies and federal agencies, some sort of audit using open source data is necessary to ensure the collection program is working and will be an essential aspect of data collection for DOJ to satisfy the research requirements in the statute.
The DCRA of 2013 mandates that DOJ answer two broad research questions.10 The first is to analyze how to use the data collected to reduce the number of deaths in custody. The second is to find relationships, if any, between deaths in custody and administrative policies. But there are no data collection plans that would produce the information necessary to deliver on this requirement. Most notably, answering these questions requires an analysis of agency and facility policies, such as those pertaining to use of force. These policies change over time and vary across agencies and departments. To obtain the data necessary to conduct an accurate and relevant analysis of the impact of policies on deaths, data collection on agency policies would have to be essentially ongoing. Based on the data currently being collected at this time, DOJ will be unable to fulfill the research requirement outlined in the statute.
The DCRA statute does not explicitly require the public release of data beyond the report to Congress. However, the spirit of the law is plainly to increase public understanding of deaths in custody. This congressional intent is reflected most recently in the Senate Appropriations Committee’s preliminary explanatory report accompanying the 2023 Commerce, Justice, Science, and Related Agencies appropriations bill, which calls on the department to provide an explanation of how to “improve the quality and transparency of future data collected” under DCRA.11
It is true that other federal laws, like the Privacy Act, limit how the government handles individually identifying information, and we understand that OJP is concerned about privacy when it comes to releasing DCRA data to the public.12 While we appreciate the constraints imposed by other laws, we do not agree that they are a major stumbling block to the release of ideally incident-level data, but at least agency- and facility-level data. As the department acknowledged in its 2016 plan, removing individually identifying information should resolve Privacy Act concerns, making it possible to release at least agency- and facility-level data. The identities of agencies that report deaths in custody are not protected by privacy considerations. And BJA, as a non-statistical agency, does not face the same restrictions on the use and aggregation of data as BJS does.
The federal data tables BJS currently releases are insufficient for the purpose of enabling true public engagement with the information. To determine how the data may help reduce deaths in custody, stakeholders and outside researchers will need more granular data than what has been released to date. The basic data tables prevent a fuller understanding of the context of each incident and the discovery of overall patterns in the data. Because the data are aggregated and presented at a national level, we cannot use them to analyze the relationships of variables within the data. As an example, the BJS data tables cannot shed light on the racial breakdown of people placed in prone positions by law enforcement before their death, because the tables only sort by one variable at a time.
Sources for Reform
Many of our recommendations are derived from prior DOJ proposals and examples from other jurisdictions with death in custody reporting programs. We provide a brief overview to emphasize that roadmaps for DCRA improvement already largely exist.
2016 DCRA Collection Plan
The first proposal from BJA for state data collection was announced in the Federal Register on December 19, 2016. It remains the most detailed and rigorous plan for implementing DCRA and still provides a model for how the department could realize both the spirit and letter of the law.13 The most notable features of the 2016 plan include:
- Precise definitions of deaths that must be reported.
- A requirement for states to develop and submit plans for collecting the data they are required to report to the department. Those plans were to be reassessed annually to ensure their effectiveness.
- A plan for the department to use open-source data to evaluate the accuracy of data reported by states, modeled on the BJS redesign of the Arrest Related Deaths program.14
- A commitment to release anonymized data at the agency and facility level.
Crucially, the plan shows that the department itself has previously endorsed actions that would broadly address many of the concerns we raise in this statement. Readopting them would not require breaking new ground.
State Efforts: Texas Death in Custody Act of 1983
DOJ has an important role to play in sharing best practices with states. However, several states have practices that may help inform federal efforts as well.
The Texas DCA of 1983 requires law enforcement agencies in the state to report deaths in custody which includes deaths occurring during physical detainment, arrest, and incarceration.15 These reports are submitted to the office of the state attorney general and are due within 30 days of the death.
The Texas office of the attorney general currently posts information from each individual death in custody to its website, amounting to more than 15,000 entries dating back to the early 1980s.16 This, in turn, has enabled non-governmental organizations to create interactive data visualizations.17
In recent years, journalists and advocates have pointed out weaknesses in the implementation of the Texas law. In a five-year period since 2015, hundreds of reports were filed after the 30-day limit, and more than 100 reports lacked required medical examiner information.18 The punishment for failing to comply is a class B misdemeanor and could potentially result in up to 180 days in jail. However, like the federal DCRA, this punishment has never been levied.19
Even with these issues, the Texas program provides an example of how to make data on deaths in custody relatively available to the public and how additional transparency can highlight challenges and lead to improvement in data collection processes.
Other State Efforts
California collects data on deaths in custody pursuant to Government Code § 12525, first passed in 1961.20 Again, in-custody deaths are more broadly defined in this state context to include deaths during any type of detainment, including during arrest. California also creates visualizations so the public can better understand the trends in the data.21 Additional states like Illinois passed state laws that require all law enforcement agencies to report deaths in custody.22
The DCRA has also encouraged states to be more transparent about deaths in custody even when the state has not passed its own law. For example, the Indiana Criminal Justice Institute (ICJI) is the state statistical agency responsible for collecting DCRA data. In March 2021, the ICJI released a report covering 2020 data, and subsequently created an online statistical dashboard.23 These efforts make it easier to understand patterns that could be lost when data are aggregated at the national level. Simple visualizations combined with the reporting requirements of the DCRA give advocates and researchers another opportunity to monitor deaths in custody within a state.
As we look at existing statutes and current systems for reporting, our recommendations below are straightforward, within the already existing statutory framework, and will dramatically improve the accessibility and usefulness of DCRA reporting.
- Refine and update coordination, definitions, and guidance for data collectors at the federal and state levels. This consistent attention to updated, clear, standardized guidance and collaborative technical assistance will positively impact compliance, data collection, research, and transparency.
- Go beyond the plain text of the statute in designing a DCRA implementation plan. The 2016 plan illustrates how steps beyond those explicitly described in the law are necessary to fully realize the stated requirements and underlying intent of DCRA.
- If it does not exist already, we recommend a mechanism for state and local personnel to provide feedback on training and guidance so that the BJA may iteratively improve these resources.
First, every local, state, and federal agency with arrest or custodial authority must participate fully in data collection and reporting. The power to arrest and detain people demands a companion responsibility to do so with integrity. Collecting and reporting quality data is the mechanism for ensuring integrity. Compliance cannot be optional. However, to avoid the risk that imposing the penalty will make it harder for states to fund compliance, DOJ should readopt the provision from the 2016 plan allowing states to use the 10 percent that would be penalized to fund DCRA implementation.
- Use all available data, including from open sources, to assess compliance. Full compliance requires reporting complete and accurate information, which cannot be known without some sort of audit.
- Issue clear standards for when and how the 10 percent penalty will be imposed. Impose the penalty on non-compliant states.
- Give states the option to use the 10 percent that would otherwise be withheld to improve and standardize data collection and reporting processes.
Data Collection Methods
Collection forms must be designed to collect the data we need to answer the critical questions at hand. This requires re-engineering the forms with purpose, intent, and technical expertise. We cannot collect inferior data and expect quality research as a result. The two are mutually exclusive. To this point, we recommend a more robust data collection plan modeled largely after the December 2016 collection plan. Critical components of the 2016 plan include clear and robust definitions and annual state-level data collection plans. Additionally, the BJA and the BJS should collaborate to create uniform forms across programs and departments that can be used for several years.
- Redesign federal and state forms so that:
- Demographic variables for race, ethnicity, and age brackets match the U.S. Census Bureau classifications for easy comparability and use. Going forward, DCRA forms should always be updated to reflect any future adjustments to the U.S. Census classifications. Getting DCRA collection on the census standard for race could have positive implications for other law enforcement collections at the federal, state, and local levels.
- Questions relevant to arrest-related and in-custody deaths should appear on both collection forms.
- Questions shared by both forms should have identical wording and answer sets.
- To the extent possible, questions should be specific with check box or multiple- choice answers.
- Ensure forms capture killings by law enforcement officers as a separate category, rather than capturing these deaths in an “other” category.
- Keep the text box with a clear prompt on the incident report forms.
- All forms should include one question at the end of the survey with an open text box answer where additional details about the death in custody can be reported. However, the forms should not rely on text boxes as the primary or only means of eliciting key information.
- Ensure that the forms also capture deaths of bystanders that are a consequence of police action.
- Structure questions about cause of death that clearly distinguish between accidents and intentional deaths.
- Once the forms are redesigned, keep them consistent over several years. Additions can be made, but standardized questions and answers between forms should be kept the same for research and evaluation.
- When changes to forms are needed, first study how changes would affect the quality of data. This is a typical practice for federal agencies like the U.S. Census Bureau, and we recommend that steadied approach here.
- Restore the 2016 requirement for states to submit data collection plans and review and revise those plans annually.
The U.S. Attorney General must commit to studies that address the DCRA-mandated research questions. In order to facilitate this, it must clearly define the universe of data necessary to answer those questions. We recommend the December 2016 definitions as a model to help define reportable deaths, as it provides specific and granular categories of covered situations. DCRA data collection must also address the fundamental issue of policy data collection to satisfy the second part of the DCRA-mandated research.
- Identify the most appropriate definition of deaths to include in the DCRA reporting. Use this definition consistently across agencies, forms, and years. We recommend the definition used in the December 2016 data collection plan.
- Develop a plan to gather data on relevant agency and facility policies.
- When an investigation is open/pending at the time the data are reported, create a process for respondents to follow up after the investigation is closed and report required data.
DOJ and other federal agencies must commit to timely data reporting. As a good faith effort, we recommend the responsible agencies release more data now. Specifically, publish homicides broken out by use-of-force/police-involved and other homicides — a step taken for the first time in the July 2022 data tables published by BJS.24 Release the existing data at a more granular level by agency, facility, and perhaps even at the level of individual incidences in machine readable formats.
- Create, maintain, publicly publish, and annually update a list of federal agencies subject to the DCRA.
- Where applicable, clarify whether deaths occurred in agency-owned facilities or contract facilities.
- Improve data reporting with more visual reports and public data dashboards.
- Create a unique individual identifier (a number) system like those commonly used by other agencies to protect privacy while disaggregating data for research and evaluation.
Publicly accessible data on deaths in custody is an essential aspect of law enforcement accountability in the United States. We need accurate data to research in custody deaths and learn how to prevent them. Without sound DCRA implementation, decisionmakers do not have the information they need to make policy, advocates do not have a clear picture of the full impact of the criminal-legal system, and DOJ cannot provide the oversight needed to reduce in custody deaths. The recommendations made in this report will strengthen DCRA administration and compliance, improve data collection, expand the ability to conduct DCRA required research and increase transparency.
The Constitution Project seeks to safeguard our constitutional rights when the government exercises power in the name of national security and domestic policing, including ensuring our institutions serve as a check on that power.