Fatal police shootings and race

The claim that police lethal force is disproportionately used against blacks has gained a lot of attention in recent years. Some data show the strength of this perception, though there is also data questioning this assertion. Unfortunately, the government does not mandate the reporting of lethal force by police departments, so it has been difficult to learn more about the frequency and context of these incidents. Nonetheless, the narrative that blacks are disproportionately killed by the police has inspired rapid policy changes.

Some databases have attempted to collate information on police use of lethal force, but their numbers are not always consistent and are subject to their own limitations. Other researchers have examined the relationship between race and police lethal force at a more detailed level, where they are able to see whether other case characteristics influence the relationship. A recent report by Robert VerBruggen of the Manhattan Institute explains several of the approaches used to examine racial bias in fatal police shootings, which are discussed below.

Official Data

Currently, there is no official data showing the number of people killed by the police per year. There are, however, some ongoing endeavors that hold promise for the future.

National Use of Force Data Collection (Federal Bureau of Investigations) 

The FBI’s National Use of Force Data Collection was the first attempt by the federal government to collect data on police use of force at a national level. Data collection began on January 1, 2019, with the intention to collect information on uses of force, including basic information on the circumstances, subjects, and officers involved in these incidents. Thus far, the publicly available numbers only reflect program participation, and no data has been released about actual use-of-force incidents. Unfortunately, law enforcement participation in the program has been so low that it could cause the database to shut down, and the data may never even be published. As of September 2021, 7,559 out of 18,514 (about 41%) of federal, state, local, and tribal law enforcement agencies throughout the nation had participated and provided use-of-force data.

Arrest-Related Deaths Program (Bureau of Justice Statistics)

The Arrest-Related Deaths Program is a national census of arrest-related deaths that is maintained by the Bureau of Justice Statistics (BJS). It includes all civilian deaths that occurred during, or shortly after an arrest or restraint process. The data are collected from news reports and interviews with agencies. Preliminary findings were provided in Arrest-Related Deaths Program Redesign Study, 2015-16: Preliminary Findings. In this report, researchers identified 1,348 potential arrest-related deaths occurring in the United States from June 2015 through April 2016. There was an average of 135 deaths per month, ranging from a low of 87 in June to a high of 156 in December. Shortly after that report was released, BJS began making improvements to the program,  but the improved iteration has not produced publicly available data as of this time. However, in 2019, BJS reported more detailed information about the changes that were made to the program.

National Violent Death Reporting System (Centers for Disease Control)

The National Violent Death Reporting System (NVDRS) is a database maintained by the Centers for Disease Control (CDC) that collects facts from death certificates, coroner/medical examiner reports, law enforcement reports, and toxicology reports. Data elements also collect information about the context surrounding violent deaths, such as relationship problems, mental health conditions, toxicology results, life stressors, and financial problems. The program was created in 2002 and initially collected data from six states. This expanded to 17 states by 2006. The most recent expansion of NVDRS includes all 50 states, the District of Columbia, and Puerto Rico. The recently added states do not have publicly available data yet, so the data are not currently nationally representative. Hopefully over time though, this database will improve our ability to understand the national scope of the problem.

National Vital Statistics System (Centers for Disease Control)

Another source that tracks officer-involved fatalities is the CDC’s National Vital Statistics System (NVSS), which tracks death certificates and causes of death in the country. Data from coroners/medical examiner reports is collected regarding type of death, which sometimes specifies whether a police shooting was related to the death. The most recent estimates are from 2019, at which time the database had tabulated 142,440 deaths, 14,730 of which were firearm deaths. The data showed only 5 deaths by “official authorities” (the rate is not reported as the number is too small to be reliable). Additionally, there were 662 handgun-related deaths with “unspecified intent,” so it is unclear whether this could have been related to authorities as well. Unfortunately, the results regarding police shootings may be undercounted by more than half, per a recent study in The Lancet. One potential reason for the undercounting is that a medical examiner or coroner must fill out the cause of death on the death certificate, however, their training and expert opinions can still vary, and they may not mention whether police violence contributed to the death.

Open-Source Data

Due to the lack of current official data on police use of force, independent researchers have taken it upon themselves to collate information on police-related fatalities using information from news reports, department websites, public records requests, anonymous tips, and other sources. Four major non-governmental, open-source databases on police killings include the Washington Post’s “Fatal Force,” the Massachusetts Institute of Technology’s “The Counted,” Campaign Zero’s “Mapping Police Violence,” and the University of Southern California’s “Fatal Encounters.” Each project has a somewhat different focus, different methods, and different inclusion criteria, which means the overall numbers vary from database to database. For example, “Fatal Force” includes only fatal shootings (not all fatalities) by police in the line of duty; thus, it excludes some of the most protested police-involved deaths, such as George Floyd. In contrast, “Fatal Encounters” collects information on all deaths that happen when police are present — even if the person died from suicide, car crash, or the like. The latter would obviously have more cases than the former.

While the open-source databases are not perfect, they do provide some basic details about people who were shot and in what situations. While the numbers vary slightly across databases, it appears that the police shoot and kill about 1,000 people throughout the country each year and that changes in the number and racial composition of the shootings have been minor in recent years.

Fatal Force (Washington Post)

Fatal Force, led by researchers at the Washington Post, collects data on fatal police shootings occurring since 2015. The data are collected from news accounts, social media posts, and police reports, and includes a variety of important case characteristics (e.g., race, gender) that can be broken down by city and state. As stated previously, the Post only documents fatal shootings committed by an on-duty police officer, and does not track non-shooting deaths or instances where people died in custody.

The project has tallied more than 7,100 events since its infancy, or about 1,000 per year — from a low of 958 in 2016 to a high of 1,055 in 2021. People were armed in more than 90% of cases, 58% of whom were armed with a gun. The racial breakdown also remained steady since 2015. Per the data, approximately 25% of people shot and killed by the police are black. Blacks comprise an even higher percentage (34%) of unarmed people shot and killed by the police. Because these percentages are more than the black share of the overall population, on its face, it could be a sign of potential bias. However, not all shootings of unarmed civilians are unjustified, and justified officer-involved shootings tend to cluster in specific high-crime neighborhoods. These neighborhoods often have predominantly minority populations, which may explain some of the racial discrepancies. Based on the data we have now, it is hard to know whether discrepancies are due to racial bias specifically or whether they could be due to other factors. Overall though, it does seem that fatal police shootings have increased slightly over time.

The Counted (Massachusetts Institute of Technology & the Guardian)

The Counted was collated by researchers at the Massachusetts Institute of Technology (MIT) and the Guardian. This project tracked all instances of on-duty fatal police violence, including in-custody deaths, using data from police reports, local governments, news reports, research organizations, open-source reporting projects, and crowdsourced accounts from witnesses. The data offers snapshot descriptions of the killings, detailing a wide range of circumstances that can be broken down state-by-state. Unfortunately, the data only span 2015 and 2016.

The project identified annual numbers that were slightly higher than those seen in Fatal Force — roughly 1,100 per year instead of 1,000 (1,146 in 2015 and 1,093 in 2016). Obviously these numbers are inflated in comparison to Fatal Force because the inclusion criteria was a little broader (i.e., they included all fatalities and not just shooting-related fatalities). Though, data only included information on case context (e.g., race, weapon type) for 2016. Nonetheless, this data showed that blacks accounted for 13% of police killings. 84.5% of cases involved armed subjects, 46% of whom were armed with a gun. These data actually show a small decrease in the number of police killings over time, but with only two years of data, this is barely a finding.

Mapping Police Violence (Campaign Zero)

Mapping Police Violence was collated by researchers from Campaign Zero, which is a non-profit research organization. This data has gathered information from police and sheriff’s departments across the country from 2013 forward, which can be broken down by state and police department. The data included information on all fatalities caused by on-duty police officers. In some cases, researchers were able to obtain data on individual-level police officers, including things such as uses of force and disciplinary records.

From 2016-2021, the researchers tabulated information on 9,901 police-related fatalities in the U.S from 2016-2021. Their annual numbers showed around 1,100 killings per year, ranging from a low of 1,070 in 2016 to a high of 1,144 in 2018.  Of the 9,901 cases, about 16% of these cases involved black people. 71% of cases involved armed subjects, 53.7% of whom were armed with a gun. 16.4% of cases involved people who appeared to be under the influence of drugs or having a mental health crisis. This data shows that fatalities have increased since 2016, but it also suggests that they have decreased from 2018-2021.

Fatal Encounters (University of Southern California)

Fatal Encounters is collated by researchers at the University of Southern California. It tracks all killings that occurred in the presence of police, including suicides, car crashes, and killings committed by off-duty officers. The data span from the year 2000 to present day, and can be broken down by city, state, and county. This data is collected from news reports and public records requests from police departments. However, one limitation is that police departments don’t always respond, reports contain inconsistent amounts of information, and many departments redact information on important variables, leading to a fair amount of missing data. It does appear to have higher numbers of deaths compared to other the other publicly-available datasets, likely because of the wide inclusion criteria, and possibly due to additional information gleaned from public records requests.

Excluding cases of suicide, the project has tallied 28,153 deaths from 2000 to 2021, approximately 1,300 per year. This ranged from a low of 795 in 2000 to a high of 1,804 in 2020. The data showed that 23.6% of cases involved people who were black. Data on weapon type was limited to approximately 12,311 cases; of these cases where data was available, 57.4% of people were armed and 41.7% were armed with a gun. While this data source has the highest counts out of the open-source databases, this is likely because it casts a wide net of cases that are eligible for inclusion. Even after excluding suicides, the numbers are inflated compared to others. It does suggest that fatal encounters has increased over time though, with 2020 having the highest number so far.

Research has in some cases, tried to take advantage of the various data sources mentioned previously. Other times, researchers are able to work with police departments directly, and this might give them exclusive access to more detailed data, allowing for more rigorous research. As reported by VerBruggen, there are a few common methodological approaches used to examine fatal police shootings; these are highlighted below.

Methodological Approaches

When it comes to studying whether racial bias exists in policing, there are many things that cannot be easily measured that may conflate this relationship. There are several methodological approaches that researchers use to examine this issue, but there tends to be a lot of debate regarding the “best” methodological approach.

Benchmarking Studies

The simplest approach to examining racial bias in policing is through “benchmarking” studies. These studies compare the racial breakdown of those shot by police with the racial breakdown of another similar population, such as the arrested population or a city’s overall population. Unfortunately though, there is no ideal population that should be used for comparison. For example, in the former, the composition of the arrested population is affected by police discretion and possible racial bias, limiting the utility of this variable in testing for bias. In the latter, the racial breakdown of the overall population isn’t necessarily the same as the breakdown in the offending population. Additionally, this approach ignores the context associated with shootings, so it is unclear whether there were other factors (e.g., weapon type) affecting the relationship between race and lethal force.

Some benchmarking studies found no racial differences when looking at fatal injuries caused by the police. For example, a benchmarking study published in 2018 compared racial breakdowns in police shootings with the racial breakdowns of 16 different crime rates, finding no difference between blacks and non-blacks in terms of fatal shootings.

Others have found that blacks are more likely to be fatally shot, but that the relationship changes when controlling for violent crime arrests or weapons-related arrests. One study published in 2018 compared racial breakdowns of fatal police shootings with the racial breakdowns in three different populations: total population, population of police-citizen interactions, and total arrests. They found that overall, blacks were more likely than whites to be fatally shot by police officers. However, when controlling for violent crime arrests or weapons offense arrests, this relationship reversed, such that black citizens were less likely to be fatally shot by police.

A benchmarking study published in 2020 found even more nuanced conclusions upon expanding the sample to include fatal and nonfatal shootings. The study used three different data sets of police shootings: the Washington Post’s “Fatal Force” counts of fatal officer-involved shootings, fatal and injurious officer-involved shootings in Texas, and all firearm discharges by officers in California. These data were compared with demographic data on people who has assaulted cops in the relevant jurisdictions. The authors found that, both nationally and in Texas, blacks were not more likely than whites to be fatally shot based on the benchmark used. However, this was not the case in California, where blacks were still more likely than whites to be shot by police. These more nuanced conclusions begin to show how complicated this relationship can be.

Based on benchmarking studies, it seems that other factors (such as weapon type, violent crime) might explain some of the relationship between race and fatal police shootings. But these studies still are not strong enough to prove nor disprove racial bias in policing.

Examining Level of Threat Posed by Subjects

The level of threat posed by subjects is an important factor to control for when studying the relationship between race and police use of force. Researchers sometimes limit their analysis to only unjustified shootings as a way to isolate people who were, in theory, less threatening. However, whether or not a shooting is justified will always be based on interpretation and one’s subjective perception of whether an individual was an imminent threat.

For example, one common incident that falls in the grey area of a threat is when a suspect produces a knife and refuses to drop it. They don’t actually attack the officer — but they are possibly only moments away from doing so. Another common occurrence is when an officer perceives a “furtive movement,” such as a quick reach toward one’s pocket or vehicle when instructed not to do so. In these situations, an officer might presume that the person is about to pull a gun, feeling the need to react before the suspect does. Furtive movements do not always indicate a potential attack, though. For example, a person could be reaching for a cell phone, or they might have a disability that causes sporadic limb movements, such as Tourette’s syndrome. Nonetheless, these so-called furtive movements are a common precursor to use of force that can be difficult to measure — they may indicate a potential threat or they may not.

Some studies have tried to isolate “less threatening” people by focusing on police killings of unarmed persons. A 2020 study using the Washington Post data found that black people who were fatally shot by police were more likely to be unarmed than their white counterparts. A 2021 study similarly found that among people shot by the police, blacks were more likely than whites to be unarmed, but this was only seen among people who were 55 or older, mentally impaired, or living in the south.

But focusing exclusively on unarmed people also isn’t necessarily the best approach, because unarmed does not always mean not dangerous. While it is true that suspects armed with guns do account for the overwhelming majority of police officers killed, there are also instances where officers are killed with their own weapons or killed by unarmed persons. A 2016 book examining 153 cases where unarmed people were killed by police found that most of the people killed were engaging in criminal behavior at the time of the event and typically posed direct threats to the officers or other bystanders.

However, this is not to say that all shootings of unarmed citizens are justified. Some studies try to clear this up by incorporating some proxy for “level of threat” imposed by the subject. One approach to this is to limit the sample to only situations where the person was unarmed and not attacking (rather than just being unarmed). Others have attempted to code variables that measure suspect “level of resistance” or “level of threat” using ordinal categories. For example, a lower level would be insubordination while a higher level would be attacking the officer with hands/feet, and an even higher level would be brandishing a knife or pointing a firearm. However, officers often misperceive levels of resistance just as they misperceive threats. Further, this information is not always included in police reports, so missing data is a typical limitation in these types of studies.

Examining Cases where Lower Levels of Force Were Used

Many studies attempt to compare cases where lethal force was used with similar cases where lethal force could have been used but was not. This allows researchers to compare black and white suspects who were in roughly similar situations and their likelihood of being shot. To do this, researchers strategically choose a comparison group by selecting cases that look similar on a range of factors. One common approach is to compare cases using lethal force with cases using lower levels of force, as these types of cases tend to have similar case characteristics. This type of analysis is much more complicated to conduct, but it provides far more detail than aggregate-level studies are able to do. Of course, there are limitations. In particular, no set of control variables can perfectly account for all of the relevant factors, and there is always a potential variable that is not being examined. Relatedly, critics often claim that racial bias helps determine who is included in the comparison group as well — for example, whoever gets tasered or has a gun pointed at them.

A famous 2017 study compared officer-involved shootings with a comparison group of cases that involved an arrest for at least one of the following: attempted capital murder of an officer, aggravated assault on an officer, resisting arrest, evading arrest, and interfering with an arrest. This comparison group was based on these arrest types, as these types of cases would likely involve physical confrontations and therefore could conceivably result in force. Interestingly, this study found evidence of racial bias in nonfatal shootings, but not in fatal shootings. The analysis relied on data from only one city though (Houston), and may not apply to other areas of the country. Further, there is also the possibility that the selection of the comparison group is also racially biased, which may have impacted the validity of these comparisons.

Other studies have compared police shootings with situations where officers drew and pointed their guns but did not shoot. In a 2018 study in Dallas, researchers found that when an officer shoots someone, it is usually driven by two key situational factors: whether a subject was armed, and whether an officer was injured. When controlling for these factors, blacks were actually less likely than whites to be shot. Another 2018 study in an unnamed department similarly found that black suspects were less likely to be shot, both overall and when controlling for various incident characteristics.

Of course, these studies are limited by the fact that officers may not always document decisions to draw their guns without shooting. Further, there is always a possibility of racial bias in an officer’s decision to draw their gun, which could distort the comparison group. In an attempt to remedy this problem, a 2019 study (also in Dallas) limited analysis to “arrest and active aggression cases,” where they again found that blacks were not more likely than other races to have weapons drawn on them by the police.

Recommendations for Future Research

Recommendations for future research include conducting more rigorous studies that look beyond aggregate rates and examine greater detail on a smaller subset of incidents. This would help us examine where problems lie specifically, rather than hunting for answers in aggregate nationwide numbers.

It would also be useful to study more occurrences where lethal force is applied but no death occurs. Governments should collect this information in a systematic fashion and make it available to the public. States could require this, or the federal government could provide funding for creation of a comprehensive national lethal-force database.

Similarly, we should also continue to study cases where lethal force was not intended but nonetheless occurred. Some of the most controversial police incidents have involved situations where suspects died following lower level uses of force. These cases should be examined in greater detail on their own as they are qualitatively different from incidents were suspects were fatally shot.

Overall, we know far more about the use of lethal force by police than we used to, and existing datasets are improving over time. Still though, it is difficult to fully prove or disprove existence of racial bias based on the current research. At this time, the numbers are very mixed, and it is still hard to draw adequate conclusions. At the same time, there are probably some departments that have racial bias issues, but it is probably not as widespread as it is often portrayed. Relatedly, it is likely wrong to assume that every police department in the U.S. needs a total revamp in order to address racial bias. Instead, we need to be honest about how often racial bias occurs, where it occurs, and in what instances it occurs. This will help us address other factors associated with the “root” of the problem, and it would also be more efficient than demanding extreme policy changes based on inadequate research.