Home / Articles / Measuring What Matters: Addressing Police Reform Must Start with Accurate Data

USAFacts has received several requests for facts on excessive police force, but the reality is that robust data simply isn’t out there. We addressed the need for recent, standardized data and what governments can do about it in a recent post on Medium and are re-sharing it here.


One of the most striking things about data on excessive police force is what is lacking. Federal data on excessive force and police shootings is often too dated and inadequate to inform current discussions. Old, incomplete information is an ongoing trend across government data that needs addressing. Some, but not all, local governments provide policing statistics; however, what’s reported is incomplete and nonuniform, making national comparisons all but impossible.

This is not to undermine the painful human discussion that is taking place across the US. Data is just one part of the conversation and it does not capture an individual’s or a community’s lived experience. And that data must be shared in a way that’s easy to understand, free of distancing government terminology. Timely, accurate statistics are part of the puzzle to have honest conversations about excessive police force and communities of color.

National level datasets are dated and inadequate

No federal agency publicly provides timely, comprehensive data on excessive use of force and officer-involved shootings annually, a minimum bar, though monthly or weekly data could bring even more transparency. The Bureau of Justice Statistics (BJS) does provide two excessive police force datasets, but the most recent report is from 2016.

One of these datasets, the Arrest-Related Deaths Program (ARD) identifies and confirms arrest-related deaths and checks them against information from law enforcement, reviews of news articles, and surveys of medical examiners/coroners. The ARD’s preliminary results identified 1,348 potential arrest-related deaths between June 1, 2015 and March 31, 2016. Considering everything that can happen in five years—elections, new policies, economic shifts, demographic changes—lags in data like this make it difficult to hold ourselves accountable as a nation.

The BJS also surveys people who have had contact with police in the last year. Data is updated every three years beginning in 1996—survey data from 2015 was released in October 2018. And in 2015, roughly 996,000 people experienced force during their most recent police-initiated contact. While 60% of Black people and 53% of Hispanic people called the force in their police contact excessive, 43% of white people did.

The Centers for Disease Control and Prevention has data on the underlying cause of death documented on death certificates. Data for 2018 was released in early 2020. According to the Underlying Cause of Death dataset, 614 people were killed in encounters with the police in 2018. It also maintains the National Violent Death Reporting System, which catalogs violent deaths documented in 34 participating states and Washington, DC. In 2016, it captured 515 deaths due to police officers.

In 2015, an advisory board to the FBI recommended that the FBI develop a new data collection on fatal and nonfatal officer-involved shootings. The data collection effort began in 2019. No metrics have been shared publicly, yet.

Local data is timelier, more detailed, far from standardized

Some city police departments report on excessive use of force and/or deaths due to police intervention. However, the shape and definitions of this data varies from city to city. Examples include:

  • Minneapolis: Use of force and officer-involved shootings data is available. The most recent use of force data is from early June 2020, and the most recent officer-involved shooting data is from December 14, 2019.
  • Seattle: Use of force and officer-involved shootings data is available. Both datasets detail the race of the subject. The most recent use of force data is from May 28, 2020, but the most recent officer-involved shootings data is from June 19, 2016. There are three records of disciplinary action for officer-involved shootings in the dataset for 2005-2016.
  • Atlanta: The Atlanta Police Department produces an annual report and publishes weekly crime data, but these reports don’t mention use of force. The most recent data on officer-involved shootings is from 2015. Six Atlanta police officers were recently charged after allegedly using excessive force.
  • Dallas: The City of Dallas provides officer-involved shooting data, though it doesn’t include demographic information on the subject. The Dallas Open Data portal also provides some data on use of force by police officers, but it is unclear if anything more recent than 2016 is available.
  • New York City: The New York Police Department provides use of force incidents by quarter and excessive use of force by year; however, the data only describes the type of force used, the command group/precinct, whether the officer was on-duty or off-duty, and in a separate unlinked dataset, the severity of the subject’s injury. The data does not include subject demographics.

What do non-government datasets do?

Considering the dearth of timely, standardized numbers on officer-involved deaths, non-government datasets have sprung up to fill the holes. These non-government aggregators generally use a similar methodology to what surfaced the greatest proportion of cases in the ARD report: comprehensive review of cases reported in the media.

Our government can and must do better

USAFacts believes facts deserve to be heard; shared, comprehensive and transparent data can ground public debates in facts. The nation is engaging in a discussion about police brutality in Black communities nationwide. We need clear, comprehensive, contextual numbers to help advise decisions on policing. Reporting standards must be applied evenly across the US and the data needs standardization and clear methodology.

Data is just one aspect of this discussion, but we can’t use it to inform opinions if it’s out of date. The BJS has demonstrated that the ARD methodology is the most holistic of any national dataset on the topic, plus the BJS has an existing data pipeline and machine-learning classifiers to do some of this work. Americans can demand Congress consider shifting funds towards better data to shed light on excessive use of force and measure any implemented changes. Government data alone will not solve these issues, but it should be available and accessible to everyone to measure change and progress.