Do White Law Enforcement Officers Target Minority Suspects?
Charles Menifield, Geiguen Shin, and Logan Strother
Abstract: The debate over possible bias in the use of deadly force has recently been exacerbated
by highly publicized killings of African American males around the country. While much
research has been conducted examining police behavior, little has been done examining the
impact of race on police behavior. The purpose of this paper is to answer the question: Are white
police officers more likely to use lethal force on minority suspects or persons of a specific race?
To answer this question, we construct a dataset of all confirmed uses of lethal force by police
officers in the United States in 2014 and 2015. Although we find that while minority suspects are
disproportionately killed by police, white officers appear to be no more likely to use lethal force
against minorities than non-white officers.
The vast majority of people killed by police are armed at the time of their fatal encounter,
and more than two-thirds possessed a gun.
African Americans are disproportionately killed by police officers nation-wide.
The disproportionate killing of African Americans by police officers does not appear to
be driven by micro-level racism.
The disproportionate killing of African Americans is likely driven by a combination of
macro-level public policies that target minority populations, and meso-level policies and
practices of police forces.
Fundamental macro-level policy changes, as well as changes to meso-level organizational
practices, will be necessary to address the root causes of racial disparities in police
On August 9th, 2014, Michael Brown, an eighteen-year-old black man was fatally shot by
a white Ferguson, Missouri police officer. The shooting sparked protests and heightened racial
tensions around the country. The events in Ferguson quickly gave rise to the Black Lives Matter
movement, which sought to draw attention to persistent racial disparities in policing, especially
the disproportionate use of force against African Americans. Subsequent killings of black men
by white police officers further inflamed tensions and increased the pitch of the national
conversation. On October 14th, 2014, seventeen-year-old African American Laquan McDonald
was shot and killed by a white Chicago police officer. On April 9th, 2015, Walter Scott, an
African American man in South Carolina man was shot in the back eight times while running
away from a white police officer. On February 25th, 2016, Greg Gunn, another African American
man, was killed outside his neighbor’s home by a white police officer in Montgomery, Alabama.
Charleena Lyles was three-months pregnant when in June of 2017 she was killed by police
officers who had responded to an attempted burglary that Lyles herself had reported.
These and other high-profile police killings have led many to speculate that white police
officers may target nonwhite suspects when exercising lethal force. However, rigorous study of
the use of lethal force by police is extremely difficult. There is very little data on police-citizen
interactions, and even police uses of force are not well accounted for—indeed, the federal
government only recently passed a law (Death in Custody Reporting Act of 2013) requiring
police departments to report uses of lethal force (see Hehman, Flake, and Calanchini 2017). Even
so, uses of deadly force have garnered considerable attention from scholars. Much of this work
has found significant differences in the application of lethal force by police. Numerous studies
suggest that black suspects are killed by law enforcement officers at disproportionate rates
relative to their representation in the population (see Ross 2015). The widespread belief of
racialized policing in general—and racialized police killing in particular—has had a deleterious
effect on levels of trust between minority citizens and law enforcement officers (Brunson and
Even so, the causes of the disproportionate killing of African Americans by police are not
well understood. The dominant explanation for disproportionate police violence and police bias
in general is what has been called the “rotten apple” theory of police misconduct (Mummolo
2018; Haider-Markel and Joslyn 2017). Criminologists, psychologists, and sociologists alike
have routinely argued that the police profession attracts people with certain personality traits,
include “machismo, bravery, authoritarianism, cynicism, and aggression” (Twersky-Glasner
2005, 58). White police officers in particular have been found to hold more racially biased
attitudes than non-police white citizens (LeCount 2017). Moreover, some studies suggest that
police work itself fosters or reinforces aggressiveness, authoritarianism, and racial bias on the
job (Hargrave, et al. 1988; Griffin and Ruiz 1999; Sellbom, et al. 2007; Legewie 2016). These
studies tend to treat all problems of policing as individual, which is to say officer-level,
phenomena. Many police departments and organizations (such as police unions) defend this
perspective as well, arguing that high-profile incidents of egregious police misconduct are
attributable to a few “bad apples” rather than larger systemic problems. Scholars are beginning to
push back against this narrative, as studies show that at least some forms of police misconduct
are demonstrably institutional (Mummolo 2018; Epp, Maynard-Moody, and Haider-Markel
2017; Legewie 2016).
The narrative around police killings of black suspects is not exceptional, then: people on
different sides of the issue suggest that “bad apples”—in this case, racist white cops—are to
blame for an undeniable problem. In this paper, we seek to answer the question: Are white law
enforcement officers more likely to use lethal force on minority suspects or persons of a specific
race? This is an important question with major policy implications for policing in a diverse
society. That is, the disproportionate killing of minority suspects is a well-established problem,
but the causes of that problem are not well understood, and many scholars, pundits, and activists
defend the “bad apple” theory, even if only implicitly. If the racial disparity in police killings is
in fact driven by white officers, then the appropriate remedies may be to reduce “interracial
policing,” retrain white police officers to improve/change their relationships with citizens of
different racial and ethnic backgrounds, reform police recruitment procedures, and the like. If
micro-level officer racism is not the root cause of the discrepancy in police killings, however,
then policymakers and activists should look elsewhere to address the problem. In other words, if
the root problem is more systemic—that is, institutional—in nature, then reforms aimed at
remedying the problem of “bad apples” will be insufficient, at best. Rather, appropriate reforms
for a fundamentally institutional problem would include addressing institutionalized racism in
criminal policy and in the practices of police departments nationwide that result in over-policing
of minority populations, increasing oversight of police uses of force, and the like. For these
reasons, the question of the basic causes of racial disparities in police killings has not only
important theoretical implications for scholars studying police-citizen interaction, but also has
profound real-world implications for policing a diverse society.
Past Studies on Police Uses of Force
While police killings of suspects have become less common over the last four decades
(Zimring and Arsiniega 2015), the persistence of substantial racial disparities in killings has
generated significant recent controversy. Numerous scholars have sought to uncover the basic
causes of this troublesome pattern. Extant research has relied primarily on one of two approaches
to this question: first, analysis of observational data, usually incident reports of deadly police
encounters collected in the FBI’s supplement homicide reports, and second, laboratory-based
“shoot/don’t shoot” experiments in which participants—sometimes including police officers—
respond to images of suspects and objects that may or may not be weapons presented to them in
quick succession. Taken together, this robust body of scholarship suggests that police killing of
suspects are largely independent of the factors that many would expect to drive such phenomena,
including local violent crime rates, neighborhood demographics, and the like. Current research is
conflicted as to whether implicit racial bias influences officers’ decisions to use lethal force.
However, to our knowledge, no study has directly assessed the racial composition of officer
killings of suspects.
Recent observational studies have netted a range of important insights into the factors
that contribute to uses of lethal force by law enforcement. For example, Lee, Vaughn, and Lim
(2014) used spatial analysis to examine the relationships between neighborhood factors,
characteristics of the involved parties, and police use of force. In particular, they used
neighborhood crime rates within police command areas to estimate the impact of neighborhood
criminal activities on the police force while controlling for demographic characteristics of the
officers and citizens involved, and the behavior of both parties to the incidents. First, Lee,
Vaughn, and Lim noted that the average citizen in a use of force incident was 29 years old, male
(88.1%), non-white (70.4%); while the officers’ average age was about 34 years old, male
(94.7%) and white (70.4%). Second, they found that officers relied on weaponless tactics in
almost half of all encounters with citizens involving force (so-called soft empty hand control and
hard empty hand control techniques). When police employed weapons in these encounters, they
were more likely to rely on nonimpact weapons like tasers (18.8%) and OC spray (26.6%) than
impact weapons (3.3%). Most important for our purposes, Lee, Vaughn, and Lim (2014) found
no significant relationships between citizen race and officer race on the level of force used in
police-citizen encounters. In another recent study, Legewie (2016) found that murders of police
officers by black men in New York City resulted in significant short-run increases in police uses
of force against black citizens, but when police were murdered by white or Hispanic people, no
evidence of a corresponding increase in uses of force against white or Hispanic populations was
observed. Legewie theorized that this increase in uses of force against African Americans after
the murder of a police officer was attributable to racial bias in policing generally, but his data
does not allow for leverage as to the source(s) of that racial bias—i.e. whether the bias was
institutional or individual (or both).
Other studies have focused more specifically on officer-involved killings of citizens. In
one such study, Smith (2003) examined rates of police killings of felons in cities with
populations greater than 100,000 people in order to see whether police department makeup—that
racial and gender diversity of a city’s police force—effected rates of police violence. Smith
found that police force makeup had no effect on rates of police killings. In other words,
departments with more minority (or female) officers had no fewer police-caused homicides.
Smith did find one factor that was consistently associated with higher levels of police killing: the
proportion of a city’s population that was African American. That is, net of controls, the larger
the black population share, the higher the rate of police homicides.
More recently, Ross (2015) examined the extent of racial bias in the shooting of
American civilians by police officers in recent years by using a geographically-resolved, multi-
level Bayesian model in analysing the US Police-Shooting Database (USPSD) for estimating the
county-level risk ratios of being shot by a police officer as a function of the race/ethnicity of a
suspect as well as the individual’s status as armed or unarmed. Ross found that the median
probability across counties of being black, unarmed, and shot by police was 3.49 times the
probability of being white, unarmed, and shot by a police officer. The median probability across
counties of being Hispanic, unarmed, and shot by an officer was 1.67 times the probability of
being white, unarmed, and shot by an officer. Further, Ross found that, nationwide, the average
risk of being shot by a police officer was the same for unarmed black suspects as it was for
armed white suspects. Thus, Ross’ study suggests significant bias in the killing of unarmed
African Americans relative to unarmed white Americans. Additionally, Ross contends that a
number of county-level factors indicate that police shootings were most likely to emerge in
police departments located in larger metropolitan counties with low median incomes and a
sizable portion of black residents, especially in cases of extreme economic inequality in the
county. Ross found no evidence, however, that the observed racial discrepancy in police
shootings was a function of county-level crime rates.
Following this vein of research, Klinger et al. (2016) examined detailed case attributes
using a micro-spatial analysis of 230 police shootings in St Louis, MO, between 2003 and 2012
in order to examine social determinants of the use of deadly force by the police, racial disparities
in police shootings, or the degree to which racial disparities may reflect biased or discriminatory
police behaviour. Here, the authors examined the influence of neighborhood characteristics such
as violent crime rates and socioeconomic factors on the racial composition of police shootings.
Suspects involved in this study were typically young black males, while two-thirds of the
shooters were non-Hispanic whites, and one-third were non-Hispanic blacks. The study suggests
that the racial and economic composition of the neighborhood were largely unrelated to the
frequency of police shootings, and violent crime rates have only a small effect on the rate of
In another study of the relationship between neighborhood factors and police killings,
Hehman, Flake, and Calanchini (2017) also found no evidence that population factors, economic
conditions, or local violent crime rates influence police killings. They do find some evidence,
however that implicit racial bias (measured using the IAT) of white residents in the region was
significantly associated with the disproportionate use of lethal force on black suspects. Hehman,
Flake, and Calanchini interpret this finding as evidence that prevailing racial attitudes in a
community may shape the decisions officers make in these life-or-death encounters.
In summary, these observational studies have found scant evidence that local violent
crime rates drive the racial disparity in police shootings. The evidence from these studies
regarding the influence of other neighborhood level factors, such as racial composition,
economic inequality, and unemployment on police uses of lethal force is decidedly mixed. One
study found some indirect evidence that implicit racial bias may be a factor contributing to the
disproportionate killing of black suspects. A number of scholars have sought to more directly test
the theory that officer bias may influence the killing of black suspects in experimental settings.
In one such study, James, Vila, and Daratha (2013) examined the influence of race and
ethnicity on police use of deadly force in the line of duty using a robust experimental study
design involving controlled randomized trials across three experiments. Participants – police,
military and civilian samples – were placed in high-fidelity computerized training simulators that
resembled real-life deadly force encounters. The samples were analyzed on decisions to shoot
white, black and Hispanic suspects in potentially deadly situations. James and colleagues found
that participants displayed significant bias favoring black suspects in their decisions to shoot,
rather than discriminating against them. Specifically, the participants took longer to shoot black
suspects than white or Hispanic suspects. In addition, participants were significantly more likely
to shoot unarmed white suspects than black or Hispanic suspects. Moreover, participants were
significantly more likely to fail to shoot armed black suspects than white or Hispanic suspects.
Cox et al. (2014) investigated the influence of suspect race, officer race, and
neighborhood characteristics on officer’s shooting mistakes. Using a realistic plastic gun
apparatus and both dynamic video stimuli and still pictures (to observe differences or similarities
between stimulus types), the researchers examined 54 police officer samples for both reaction
times and error rates for still picture blocks, then for the video blocks. For picture blocks, they
found the tendency to err by failing to shoot armed suspects, rather than mistakenly shooting
unarmed suspects. Further study indicated that officers made fewer errors on unarmed black
suspects relative to armed suspects or unarmed white suspects. However, the study did find that
“interracial patrolling”—white officers in minority neighborhoods and vice versa—increased the
rate of officer errors.
In another experimental study of actual police officers, James, James, and Vila (2016)
examined possible racial biases in police decisions to shoot. In this study of 80 active patrol
officers1 (civilians were also tested for comparison), subjects were faced with highly realistic,
custom-made, high-definition video scenarios in simulators. The officers also took the Implicit
Association Test (IAT) to investigate any association between race and threat. The study found
that officers took 200ms longer to shoot armed black suspects than armed white suspects, while
holding other variables constant, such as suspect demeanour, language, dress, distance from
participant, movement, location, sounds and light levels, constant. IAT scores revealed most
officers showed moderate (40%) or strong (38%) levels of implicit bias. However, the authors
argue that implicit bias was unrelated to decisions to shoot in a deadly force judgement and
decision-making simulator. Findings reveal, despite clear evidence of implicit bias against black
suspects, that officers were slower to shoot armed black suspects than armed white suspects, and
less likely to shoot unarmed black suspects than unarmed white suspects.
These experimental studies, then, cast some doubt on the common assertion that racial
disparities in police killings are a function of racial bias on the part of police officers. The
existing literature is hardly dispositive though. If anything, these studies indicate a long list of
factors that appear not to be significantly related to police killings: local violent crime rates,
neighborhood demographic factors, and explicit racial bias, to take some prominent, if
unexpected, examples. One central question from the contemporary debate over police violence
remains wholly unaddressed, likely due to the dearth of available data, by this literature: is the
racial disparity in killings driven, in whole or in part, by the disproportionate killing of minority
suspects by white police officers?
This question is crucially important for two reasons. First, if white police officers are
disproportionately killing minority suspects, that finding would lend credence to the racist-cop
narrative advanced by many activists. On the other hand, if white police officers do not
disproportionately kill minority suspects, or black suspects in particular, this would corroborate
the experimental findings of the research discussed above, and cast deep doubt on the racist-cop
narrative. However, the existing laboratory experiments, despite their rigor and admirable
attempts to mirror real-world conditions, are necessarily lacking the high degree of external
validity necessary to draw firm conclusions on this important question. Second, and perhaps
more importantly, the answer to this question has important policy implications. If the racial
disparity in minority killings by police is driven by white police officers, this may indicate the
need to reduce “interracial policing”—in particular, the policing of minority neighborhoods by
white police officers. If white officers are not the driving force behind the racial disparity in
killings, however, then the policy remedies are rather different. In this case, the appropriate
remedies may include officer training techniques that emphasize de-escalation, as well as larger
changes in public policy that result in the over-policing of minority neighborhoods more
Theoretical Perspective and Expectations
We theorize that racial disparities in police killings are driven by a combination of
macro-level criminal policies and meso-level organizational factors, and not by micro-level
racism. In a recent study, Epp, Maynard-Moony, and Haider-Markel (2017) found that African
American motorists were disproportionately stopped for “investigatory stops” by police officers
of all races (cf. Antonovics and Knight 2009). In a similar vein, Nicholson-Crotty, Nicholson-
Crotty, and Fernandez (2017) found that an increase in the number of black police officers on a
police force does not necessarily reduce the incidence of fatal encounters between police and
African American citizens, though they do find that such encounters do decline after forces reach
a “critical mass” of black officers. Earlier work suggested that police force diversification does
not reduce police-caused homicides (Smith 2003). In summary, these studies strongly suggest
that there is more at work here than intergroup bias among officers on the ground. For these
reasons, we contend that the same processes that drive disparities in investigatory stops and the
like are at work in police uses of force, including killings.
Macro-level factors include well-established racial disparities in criminal policy and
policing across-the-board (e.g. Alexander 2012; Miller 2010; Weaver 2007). To take just one
example, the “War on Drugs” has been disproportionately waged on minority populations, such
as by targeting drugs that are more often used by poorer and minority groups compared to White
residents or the more affluent (e.g. Provine 2007). The discriminatory effects of these policies
have been inflamed by American’s taste for punitive justice, especially for black offenders (Bobo
and Johnson 2004; Enns 2016).
Meso-level organizational factors include: police force level policies, both formal and
informal, as well as utilizing various methods to engage the general population as well as
suspects, and the amount of training. In many places, police force-level policies such as stop-
and-frisk (or more recently, the “investigatory stop”) institutionalize differential treatment of
minority suspects (e.g. Legewie 2016; Epp, Maynard-Moody, and Haider-Markel 2017). For
example, the Department of Justice’s investigation of the behavior of police officers in Ferguson,
Missouri, in the wake of the killing of Michael Brown, found “a pattern or practice of unlawful
conduct”: the department was targeting black residents for fees and fines, and treating them as a
source of “alternative revenue” for the city (DOJ 2015). A study of traffic stops in San Francisco
found significant “racial disparities regarding S.F.P.D. stops, searches, and arrests, particularly
for Black people” (San Francisco Blue Ribbon Panel 2016, 28; see also Eterno, Barrow, and
Silverman 2017). In sum, these policies have the cumulative effect of leading to disproportionate
policing of minority communities in general, and of minority citizens in particular.
It is important here to note that it is well-known that organizational rules and culture
powerfully shape individual (bureaucratic) behavior on the job. In other words, the bureaucratic
organization itself is an interactive place where the behavior and belief among individuals are
assimilated and disseminated (Wilkins and Keiser 2004, Adams and Balfour 2015). In this sense,
all manner of public-police interactions, ultimately including even the use of deadly force, are
heavily influenced by the policies and norms of police departments (see generally Gerrish 2016;
O’Toole and Meier 2015; Jennings and Hall 2012; Wilson 1978).2 Additionally, differences in
training regarding how to respond to perceived threats and how to interact with citizens and
suspects, as well as police force culture, can have profound effects on the use of police violence
(see, for example, Balko 2014; Eterno, Barrow, and Silverman 2017; Prenzler, Porter, and Alpert
2013; Lersch and Mieczkowski 2005). With respect to police department-level operational policy
in particular, Mummolo (2018) found that department-level policy changes can profoundly
impact police officer behavior on the ground. In particular, Mummolo (2018) showed that a
memorandum from New York City Chief of Patrol James P. Hall requiring that officers submit a
narrative justifying stops of people suspected of criminal action (the infamous stop-question-
frisk encounters) dramatically reduced the number of police stops and increased the “hit rate” of
such stops (the fraction of stops which produce a violation).3
In light of the evidence presented in the studies discussed in the preceding section, we
expect these macro- and meso-level factors, and not micro-level police officer racism, are
responsible for the disproportionate killing of black and other minority suspects (Griffin and
Ruiz 1999). Put differently, we contend that racial discrepancies in police killings are an artifact
of racial disparities in policing more generally: criminal policy and even police departments may
target black and minority citizens, but we do not believe that individual officers of any race are
intentionally targeting African Americans for lethal force. To be clear, unintentional—which is it
so say institutional—disproportionate killings of African Americans is no less racist than micro-
level targeting. The distinction is crucial, however, because institutional (macro- or meso-level)
causes require very different remedies than would micro-level targeting (see generally Legewie
2016; Mummolo 2018).
Recent studies indicate that roughly seventy-five percent of American police officers are
white (Ashkenas and Park 2015; Reaves 2015). The large predominance of white police officers,
then, means that, ceteris paribus, white officers will likely be responsible for most police
killings—specifically, about seventy-five percent of them. Furthermore, if black residents are
disproportionately killed by police, they will be disproportionately killed by white police
officers, precisely because police departments are predominately white. Thus, we expect that
African Americans and other minority suspects are disproportionately killed by white police
officers, but not significantly more than by officers of other races. Put differently, we expect that
police killing of suspects will be spread more or less evenly spread across officer race, indicating
that factors other than the officer’s race drives the observed discrepancies in police killings.
Data and Variables
To address this question, we constructed an original database of all confirmed uses of
deadly force by police officers in the United States in 2014 and 2015. We chose 2014 and 2015
because these are the first years for which there are complete data during the contemporary
moment of heightened salience of police killings of citizens, especially young black men. The
database was constructed in a multi-step process. We began by drawing on data gathered by
KilledByPolice.net, a nongovernmental entity that tracks police killings reported in the news and
updates its data set each day. We chose this source as a base because the site links each killing
with a news story that we could locate using the internet. In order to ensure that the data was
accurate, we cross checked it with two other online websites that collects data on police killings
(lethaldb.silk.co and fatalencounters.org).4 All three data sets have been used by other scholars
studying police killings (Lott and Moody 2016; Nicholson-Crotty, Nicholson-Crotty, and
The KilledByPolice data contained the victim’s name, race, age, date of birth, gender,
date and time of killing, city, state, and a news account of the killing. We supplemented this data
with other variables available in news accounts and other police killing datasets, including local
population demographics, cause of death, geographic location of the killing, type of offense,
presence of a non-police witness, and whether or not there was a warrant for the suspect. One
important variable for our analyses was the presence and type of weapon, that is whether a
suspect had a gun, some other type of weapon, or was unarmed. In some cases, though, it was
impossible to ascertain from existing sources whether a victim was armed at the time of his
death; even so, we were able to collect this data for more than eighty-three percent of the killings
in our data.
Most importantly for our purposes, we also coded for a range of variables about the
officers themselves whenever possible, such as officer race and gender, years of police service,
and type of officer. Due to missing data, we had to thoroughly analyze every news story that we
could locate on each killing. In some cases, we determined the race of the officer based on the
name and physical appearance in photographs in the news stories. For example, in the case of
Dino Smith Jr. in Nashville, TN, the news story revealed that Michael Gooch and Josh Reece
were the officers who shot the suspect. There were pictures of both officers on the website
(Metro Nashville Police Department 2014). In other cases, the news stories simply stated the race
of the officer. For example, Christopher Maurice Jones was shot by Officer Nicholas Stone in
Berkeley, MO; in this case the race of the officer was indicated in the local media account of the
incident (Kohler 2014). Unfortunately, in many cases (32.02%), we were unable to determine the
race of the officer involved.
After constructing the dataset with the total number of persons killed by police (n=1,108
in 2014, n=1,200 in 2015), we coded for whether the killing was accidental or non-accidental.5
The killing was an accident or unusable when: a person was hit by a police officer’s car while the
officer was in pursuit of another person; a person died when the evidence was clear that the
officer was targeting a known suspect (for example babies in cars or a hostage); or if the person
was killed by an officer who was off duty and committing a crime. Thus, our data considers only
intentional “line-of-duty” killings (even if the officer was off-duty at the time of the incident).
This procedure netted a final count of 1,952 (860 in 2014, 1,092 in 2015) non-accidental police
killings which were used for analyses in this study.
Analysis and Findings
We begin by presenting basic trends in officer killings in our data. Figure 1 depicts the
trend graphically. The panel at left in Figure 1 shows that a bare majority of victims of police
killings are white (51%) followed by black (28.1%), and Hispanic/Latino victims (19.3%), while
only a small fraction of victims are Asian American (1.7%). The panel at right in Figure 1
presents a ratio of the percentage of victims of police killings made up by each race to the
proportion that race makes up of the population at large. Here we see that African Americans are
killed by police more than twice as often as we would expect, ceteris paribus, given their share of
the general population. That is, while only about thirteen percent of the American population is
black, twenty-eight percent of people killed by police are black. Latinos are killed slightly more
often that we would expect, and white citizens considerably less often, while Asian Americans
are killed by police far less often than we would expect if killings were randomly distributed
throughout the population.
[Figure 1 about here]
Another component of the national debate is that police are wantonly killing unarmed
suspects, especially if they are black. We find no support for this claim in our data. Figure 2
depicts whether people killed by police possessed a gun, possessed some other type of weapon,
or were unarmed at the time of the incident that resulted in their death. Overall, the majority of
people killed by police were armed with a firearm at the time of their death. This is also true for
every racial subgroup of victims, except Asian Americans. Most importantly, less than one
percent of the victims of police killing in our data were unarmed. In other words, police killings
of unarmed suspects—especially unarmed black men—garner massive media coverage (and not
without reason) but they are far less common than the prevailing narrative suggests.
[Figure 2 about here]
Next, we turn to the age of persons of different racial groups killed by police, because
age—particularly youth—is a major component of the national narrative regarding police
violence against minority populations. Figure 3 presents histograms which depict the age
distribution of people killed by police for Whites, African Americans, and Hispanic/Latinos,
respectively. The histograms indicate that the average age of minority persons killed by police is
considerably younger than the average age of White persons killed by police. The mean age
among Black persons killed by police is 32.5 years (median is 30 years), the mean age among
Hispanic/Latino persons killed is 32.5 years (median is 31), while the mean age among White
person killed is 40.3 years (median is 39). Furthermore, Figure 3 indicates that the age
distribution for black persons killed is much more skewed (and skews younger) than the
distributions for the other groups. Distribution skewness for Black victims is 1.2 (highly
skewed), and 0.7 (moderately skewed) for Hispanic/Latino victims, while it is only 0.4 (slightly
skewed) for White victims.
[Figure 3 about here]
As a last word on the basic demographics of those killed by police, we note that while
police violence against women is more common than often thought and underreported (e.g.
Ritchie 2017), 95.5% of people killed by police are men. In other words, while women are
undoubtedly often victims of police violence, it is relatively rare for women to be killed by
[Figure 4 about here]
This brings us to our central question: is the disproportionate killing of African
Americans by police driven by white police officers? To answer this question, we present a
simple contingency table that depicts the percentage of those killed by police officers within each
racial group who were killed by white and non-white officers, respectively. As noted above,
police departments in the United States are disproportionately white: while white citizens make
up only 61 percent of the total U.S. population, roughly 75 percent of police officers are white.
For this reason, we expect that most police killings of people of all races will be done by white
officers; however, we do not expect that white officers will use lethal force on black or other
minority suspects disproportionate to their share of the police force. Table 1 depicts the core
finding in two contingency tables; Figure 5 depicts the relationship graphically.
[Table 1 about here]
These analyses presented in Table 1 and Figure 5 clearly indicate that the
disproportionate killing of black suspects is not driven by white police officers. In our data, just
under eighty-four percent of the officers are white6, thus white police officers actually kill
slightly fewer black and Latino suspects, and slightly more white suspects, than we would expect
if the killings were random. Non-white officers correspondingly kill slightly more black and
Latino suspects, and slightly fewer white suspects, than we would expect if the killings were
randomly distributed among officers. Moreover, while African Americans are disproportionately
killed by police, they are killed at much higher rates by non-white officers than by white officers.
This is likely due to the fact that minority police officers tend to be assigned to minority
neighborhoods, and as such, minority officers have more contact with minority suspects.
[Figure 5 about here]
Next, we turn to the important question of whether there are racial disparities in officer
killings of unarmed or less-threatening suspects. However, as noted above, the extremely low
number of killings of unarmed suspects undercuts this claim from the start. Indeed, there are so
few killings of unarmed suspects that those killings (n=4) cannot be statistically scrutinized.
Instead, we compare rates of officer killings of suspects of different races conditional on the type
of weapon a suspect had at the time of his/her death: a gun, or some other weapon. Table 2
presents the contingency tabulations; Figure 6 depicts the findings graphically. The findings
presented in Table 2 and Figure 6 show that white officers are not disproportionately killing
lower-threat (non-gun wielding) minority suspects. Overall, the pattern of findings here
reinforces those presented in Table 1 and Figure 5: patterns of police killings appear to be driven
by who polices what communities.
[Table 2 about here]
[Figure 6 about here]
With these basic findings in mind, we turn now to multivariate analysis, in order to
account for the possibility that contextual factors, such as characteristics of the officer, the
suspect/victim, their interaction, or the community in which the incident took place influence the
likelihood that white officers kill black or other minority suspects. In other words, we conduct
multivariate analyses to confirm the robustness of the relationship identified in the non-
parametric analyses above. We rely on existing research to identify potentially important control
variables. Prior studies have focused primarily on two types of contributing factors to police
killings: individual level characteristics such as officer race and victim race, class, and the like
(e.g. Nicholson-Crotty, Nicholson-Crotty, and Fernandez 2017, Paoline, Gau, and Terril 2018),
and ecological characteristics including local population characteristics (racial demographics,
poverty, etc.), local crime rate, and so on (e.g. Smith 2003; Lee, Vaughn, and Lim 2014; Ross
2015; Nix, et al. 2017).
Thus to test the robustness of our findings above, we estimate a series of multinomial
logistic regression models: in the first model, we regress victim race against a number of
variables capturing the individual-level characteristics of the victim and the police officer,
including the victim’s age, whether the victim had warrants, and whether the commission of a
violent crime led to the interaction between the officer and the victim; in the second model, we
regress victim race against a series of measures of community-level (i.e. ecological) factors, such
as the city size, population demographics and economics, minority population shares, and crime
rate. In the third model, we consider all of the above covariates together. We utilize multinomial
logistic regression to estimate the models because the outcome of interest is a categorical
placement variable (i.e. the race of the victim). In all models, the omitted comparison group is
white victim. Additionally, we omit Asian Americans from these analyses because there are not
enough Asian Americans in our data (n=31) to validly estimate the models for that racial group.
The findings are presented in Table 3.
Data for the community-level covariates were gathered from a variety of sources. We
gathered zip-code-level median income data as part of our original data gathering process; for
analysis we group these data into income quintiles using data from the Brookings Institution’s
Tax Policy Center. We take this approach to estimating the effect of socio-economic class
because it is highly unlikely that police would often know exactly what socioeconomic class any
individual suspect is part of. Rather, we think officers likely assume the class characteristics of
suspects on the basis of neighborhoods and communities. Population data from the U.S. Census
Bureau was gathered for every city in which a police killing took place, including the racial
composition of the population at the zip-code level. In order to address potential differences in
effects across cities of different sizes, we group cities into five categories based on size: small
towns (population under 50,000), large towns (50,001-150,000), small cities (150,001-250,000),
medium cities (250,001-500,000) and large cities (over 500,000). Finally, we utilized the FBI’s
Uniform Crime Report to measure city-level violent crime rate.
[Table 3 about here]
The models presented in Table 3 strongly reinforce the bivariate analyses discussed
above. In the first model (two columns at left), that considering individual-level factors, we see
that officer race is negatively signed and statistically significant; this indicates that a person
killed by a white police officer is significantly less likely to be either black or Latino, compared
to white, net of controls. In these models, victim age is also significant and negatively signed,
indicating that black and Latino victims of police killings are on average significantly younger
than white victims. The influence of the other individual-level factors, including whether the
person killed had a gun, and whether a violent crime led to the fatal interaction, were not
statistically distinguishable for black, white, and Latino victims.
In the second model, which considers community-level factors (presented in center two
columns), officer race does not appear to be significantly related to victim race. More precisely,
white officers are no more (or less) likely to kill black or Latino persons than white persons.
Furthermore, this model indicates, unsurprisingly, that as the share of the black population rises,
a person killed by a police officer is increasingly likely to be black (relative to being white); the
same is true for Latinos given increasingly Latino population share. Very interestingly, though,
the model also indicates that a victim of lethal force is also increasingly likely to be black
(relative to white) as the Hispanic/Latino population share rise, but Latinos are not significantly
more likely to be killed as black population share increases. Importantly, though, we find no
evidence that neighborhood violent crime rates, poverty, or city size meaningfully influence the
likelihood that people killed by a police officer will be members of minority populations (relative
to whites), after controlling for minority population share.
Finally, the third model (presented in the columns at right) includes both individual- and
community-level factors. In this combined model, we see again that black and Latino victims are
significantly younger than white victims, net of controls. And we see again that the race of
victim is significantly related to the share of the local population that belongs to his racial group,
but not to local violent crime or poverty. Finally, this model indicates that officer race is not
significantly related to victim race. Most importantly for our purposes, these models, taken
together, strongly indicate that white police officers are no more likely than non-white officers to
kill black suspects, net of controls. Additionally, the models indicate, consistent with prior
research, that community factors and crime factors do not much influence the incidence of police
killings (e.g. Ross 2015; Cox, et al. 2014). The fact that the relationships presented in the model
are almost uniformly null is consistent with our expectations: the models indicate that police
killings in general, and killings of members of various racial groups in particular, does not hinge
on police officer, suspect, or community level factors.
In summary, our analyses clearly indicate two things: African Americans are
disproportionately killed by police, and this disproportionate killing is not driven by white
officers targeting black suspects. Consistent with prior research, our study finds no support for
the popular narrative of the “racist white cop” as being the key driver of the killing of black
suspects. The present research provides real-world corroboration of recent laboratory-
experimental research finding no evidence of race-based targeting of suspects for killing. Indeed,
we find that non-white officers kill both black and Latino suspects at significantly higher rates
than do white officers, likely due to the high rate of deployment of minority officers in minority
neighborhoods. Further, we find that the overwhelming majority of suspects killed by police are
armed, and fully two-thirds are armed with a gun, when they are killed.
This study began with the observation that many perceive, as a result of recent shootings
of young African American males, that white law enforcement officers are more likely to
exercise lethal force when the suspect is a young black man, even when suspect is unarmed. This
perspective has been fueled by the tendency of media to fixate on such cases, even though our
data indicate that these cases are highly unusual. It is perhaps unsurprising, though, that these
particularly egregious cases of lethal police misconduct receive massive media attention: the
controversial, the unpopular, the unusual, the bizarre, are all well-known indicators of
“newsworthiness” (e.g. Straubhaar, LaRose, and Davenport 2009; McCombs 2014; Strother
2017; Strother 2018).
One unfortunate feature of the media focus on atypical cases is the persistence of the
“bad apple” theory of police misconduct: disproportionate killing of young African American
men, the theory goes, is driven by racist white cops—the bad apples. However, much research in
organizational theory suggests that the problem of disproportionate killing may be fundamentally
institutional. In this article, we present and analyze a new and comprehensive dataset of all
police killings in 2014 and 2015 in order to address this fundamental question in current policy
debates: is the disproportionate killing of black suspects driven by the targeting of minorities by
The research presented here strongly suggests that the answer to this question is no: we
find that white police officers actually kill black and other minority suspects at rates lower than
we would expect if killings were randomly distributed among officers of all races. The fact that a
majority of police officer killings are committed by white officers is a function of the
predominance of white officers in police departments nationally. In other words, white officers
do not kill black suspects at a higher rate compared to non-white officers. Simply stated: in our
study of actual police killings, as well as in prior laboratory experiments involving officer
shoot/don’t shoot trials, there is no compelling evidence that micro-level racism drives the killing
of black suspects. This finding is strongly indicative that the bad apple theory of police conduct
has limited explanatory value when it comes to police killings of African Americans. In sum, our
findings indicate that an institutional and organizational perspective offers greater leverage for
explanation of disproportionate killings of citizens of different racial groups.
To be very clear, we are not arguing that the disproportionate killing of black suspects is
racially innocuous. Indeed, law enforcement officers of all races disproportionately kill black
suspects. The killing of black suspects is a police problem, not a white-police problem. We
believe that the disproportionate killing of black suspects is a down-stream effect of
institutionalized racism in macro-level criminal policy and meso-level organizational factors
within many police departments. Put differently, our research contributes to the perspective that
persistent racial disparities in police killings are driven primarily by prior disparities in racial
policing generally: disproportionate killing is a function of disproportionate police contact
among members of the African American community. In this light, the finding that minority
police officers are actually more likely to kill minority suspects is not surprising, given that many
police departments make efforts to assign minority police to minority neighborhoods.
Addressing this important problem will not be easy. Our research, along with prior work
on the topic, suggests that the necessary remedies for this problem involve high-level policy
changes in the criminal code along with changes in many organizational features that combine to
produce observed racial disparities in policing in America. Changes in the criminal code might
include eliminating differential legal treatment of drugs used by white citizens compared to
members of minority groups (e.g. Provine 2007). Moreover, other fundamental changes in non-
criminal policy could well dramatically improve differential rates of policing, such as policies to
improve the educational and employment opportunities of minorities, especially in blighted
urban neighborhoods (e.g. Lucas 2001). Meso-level reforms might include eliminating
“investigatory stops,” which are a key source of racial differences in police-citizen interactions
(e.g. Epp, Maynard-Moody, and Haider-Markel 2017; Mummolo 2018), and removing
incentives for police officers to make formal and informal contact with citizens and suspects, and
meeting ticketing and arrest quotas (e.g. Mummolo 2018; Balko 2017; see also DOJ 2015).7
Last, we note that police departments may engage in what Adams and Balfour (2015) refer to as
“administrative evil”. That is, many behaviors such as “investigatory stops,” are not viewed as
inherently wrong despite the implications of engaging in this racially motivated behavior. As a
result, it is important that police departments thoroughly examine their formal and informal
processes so as to discern whether these processes create a culture that has negative implications
for persons of a particular race or status.
This article presents systematic analysis of police killings of civilians drawn from an
original dataset of all confirmed police killings in the US in 2014 and 2015. Still, its limitations
must be acknowledged. First, our data, while comprehensive, cover only two years of police
killings, and these two years may not be strongly representative of larger overall trends. Second
is perhaps the largest problem in doing this sort of research, it is created by the lack of
mandatory reporting of police shootings: it is that there is no authoritative source of data on the
subject. Absence of official (i.e. police department) reporting means we must rely on news media
coverage of police killings. However, as local media is in decline around the country, there is
tremendous variation in the extent and depth of coverage of killings. Ultimately, this leads to
incomplete data on many police killings, and incomplete data makes sound analysis more
difficult. Finally, the issue of police killings—especially of black men—has become extremely
salient since the killing of Michael Brown in August of 2014. As a result of this increased
salience of police killings, it is possible that police forces may be even more secretive about use
of force incidents, and even more protective of their officers, which could potentially lead to
ever-increasing data problems in the future. Even so, future work should seek to expand on our
efforts to build comprehensive datasets of all confirmed killings (going both backward and
forward in time). Additionally, scholars interested in police uses of force would do well to
further probe the organizational sources of discriminatory policies.
Police officers have a difficult and sometimes dangerous job, and the realities of the job
sometimes make it necessary to use force—even lethal force—on citizens. Long running racial
discrepancies in the way that force is applied to suspects, however, has significantly eroded trust
between law enforcement and the public whom they serve. This unfortunate state of affairs is
unlikely to improve until fundamental changes in public policy and policing are undertaken.
1. Sample characteristics were: 71 male, 9 female; 76 White, 1 Asian, 2 Hispanic; average age
was 40.4 years; average police force experience was 14.5 years.
2. Recent survey research has found that white police officers are more racially biased, and
more likely to see African Americans as inherently violent, than white non-police officers
(LeCount 2017). With this in mind, we note that individual-level racism among police is not
wholly independent from institutional racism; rather, institutional racism is perpetuated in
part by organizations’ internal climate, policies, and procedures, which are themselves in part
functions of individual officer attitudes (see Griffith et al. 2007).
3. Unfortunately, Mummolo’s (2018) data does not speak to whether this reduction in overall
stops also reduced the policy’s disproportionate impact on minority populations, especially
the African American community (see Legewie 2016).
4. Additionally, FiveThirtyEight audited KilledByPolice.net’s data and found it to be highly
reliable. See Fischer-Baum and Johri (2014).
5. These numbers are slightly different than those presented at killedbypolice.net, because
Killedbypolice includes, but does not assign a number to persons who were killed while in
police custody. For example, in 2014, the data set created by Killedbypolice indicated that
1,114 persons were killed by police. However, we excluded 6 people (Valerie Morrow,
December 14; Richard Tavera, December 7; Christina Tahhahwah, November 17; Mark E.
Cannon Jr, August 30; Joseph Murphy August 14; and John Patrick Walter, April 20) who
died while in police custody. Similarly, for 2015, Killedbypolice indicates 1,222 people
killed, but a number of these died in police custody. To take an example, on January 14,
2015, 10 people were killed when a prison bus was involved in an accident. In summary, the
numbers we report do not exactly match the Killedbypolice totals for reasons such as these
6. One could reasonably ask if this fact alone—that white officers appear to be over-represented
in a database of police killings—is not prima facie evidence that police killings are driven by
white officers. However, because we only have reliable officer race data for about thirty-six
percent of our data, we believe the data that currently exist are insufficient to draw that
7. Additionally, some police departments around the country have garnered praise for recent
changes in recruitment and training that have drastically reduced violent police officer—
citizen interactions and improved police-public relationships (Lantigua-Williams 2016). For
example, the Dallas (TX) PD recently retrained its officers in emotion-management in high-
stress situations, and in dealing with suspects with mental challenges, and this effort seems to
have improved public relations for the department.
Adams, Guy B., and Danny L. Balfour. 2015. Unmasking Administrative Evil. 4th Ed. New York:
Alexander, Michelle. 2012. The New Jim Crow: Mass Incarceration in the Age of
Colorblindness. New York: The New Press.
Antovonics, Kate, and Brian G. Knight. 2009. A New Look at Racial Profiling: Evidence from
the Boston Police Department. Review of Economics and Statistics 91: 163-177.
Ashenkas, Jeremy, and Haeyoun Park. 2015. The Race Gap in America’s Police Departments.
New York Times, April 8. https://www.nytimes.com/interactive/2014/09/03/us/the-race-
gap-in-americas-police-departments.html [accessed April 8, 2018].
Balko, Radley. 2017. Public Choice Theory is Crucial to Understanding the Criminal Justice
System. The Washington Post, July 21. https://www.washingtonpost.com/news/the-
justice-system/ [accessed April 8, 2018].
Balko, Radley. 2014. Rise of the Warrior Cop: The Militarization of America’s Police Forces.
New York: Public Affairs.
Bobo, Lawrence D., and Devon Johnson. 2004. A Taste for Punishment: Black and White
Americans’ Views on the Death Penalty and the War on Drugs. Du Bois Review 1(1):
Brunson, Rod. K., and Jacinta M. Gau. 2015. Officer race versus macro-level context a test of
competing hypotheses about Black citizens’ experiences with and perceptions of Black
police officers. Crime & Delinquency 61(2): 213-242.
Cox, William T.L., Patricia G. Devine, E. Ashby Plant, and Lauri L. Schwartz. 2014. Toward a
Comprehensive Understanding of Officers’ Shooting Decisions: No Simple Answers to
This Complex Problem. Basic and Applied Social Psychology 36(4): 356-364.
Department of Justice (DOJ). 2015. Investigation of the Ferguson Police Department. Report by
United States Department of Justice Civil Rights Division, March 4.
Enns, Peter K. 2016. Incarceration Nation: How the United States Became the Most Punitive
Democracy in the World. New York: Cambridge University Press.
Epp, Charles R., Steven Maynard-Moody, and Donald Haider-Markel. 2017. Beyond Profiling:
The Institutional Sources of Racial Disparities in Policing. Public Administration Review
Eterno, John A., Christine S. Barrow and Eli B. Silverman. 2017. Forcible Stops: Police and
Citizens Speak Out. Public Administration Review 77(2): 181-192.
Fischer-Baum, Reuben, and Al Johri. 2014. Another (Much Higher) Count of Homicides by
Police. FiveThirtyEight, August 25. https://fivethirtyeight.com/datalab/another-much-
higher-count-of-police-homicides/ [accessed April 8, 2018].
Gerrish, Ed. 2016. The Impact of Performance Management on Performance in Public
Organizations: A Meta-Analysis. Public Administration Review 76(1): 48-66.
Griffin, Catherine, and Jim Ruiz. 1999. The Sociopathic Police Personality: Is It a Product of the
‘Rotten Apple’ or the ‘Rotten Barrel?’ Journal of Police and Criminal Psychology 14(1):
Griffith, Derek M., Mondi Mason, Michael Yonas, Eugenia Eng, Vanessa Jeffries, Suzanne
Plihcik, and Barton Parks. 2007. Dismantling Institutional Racism: Theory and Action.
American Journal of Community Psychology 39: 381-392.
Haider-Markel, Donald P., and Mark R. Joslyn. 2017. Bad Apples? Attributions for Police
Treatment of African Americans. Analyses of Social Issues and Public Policy 17(1): 358-
Hargrave, George E., Dierdre Hiatt, and Tim W. Gaffney. 1988. F+4+9+Cn: An MMPI Measure
of Aggression in Law Enforcement Officers and Applicants. Journal of Police Science
and Administration 16(4): 268-273.
Hehman, Eric, Jessica K. Flake, and Jimmy Calanchini. 2017. Forthcoming. Disproportionate
Use of Lethal Force in Policing is Associated with Regional Racial Biases of Residents.
Social Psychological and Personality Science.
James, Lois, Stephen M. James, and Bryan J. Vila. 2016. The Reverse Racism Effect: Are Cops
More Hesitant to Shoot Black than White Suspects? Criminology and Public Policy 15(2)
James, Lois, Bryan Vila, and Kenn Daratha. 2013. Results From Experimental Trials Testing
Participant Responses to White, Hispanic and Black Suspects in High-Fidelity Deadly
Force Judgment and Decision-Making Simulations. Journal of Experimental Criminology
Jennings, Edward T., and Jeremy L. Hall. 2012. Evidence-Based Practice and the Use of
Information in State Agency Decision-Making. Journal of Public Administration
Research and Theory 22(2): 245-266.
Klinger, David, Richard Rosenfeld, Daniel Isom, and Michael Deckard. 2016. Race, Crime, and
the Micro‐Ecology of Deadly Force. Criminology & Public Policy 15(1): 193-222.
Kohler, Jeremy. 2014. Family wants answers in case of fatal shooting by Pine Lawn police
officer. St. Louis Post-Dispatch, November 2. http://www.stltoday.com/news/local/crime-
be00-5195-8639-0d33df03594a.html [accessed April 8, 2018].
Lantigua-Williams, Juleyka. 2016. How Much Can Better Training Do to Improve Policing?”
The Atlantic, July 13. https://www.theatlantic.com/politics/archive/2016/07/police-
training/490556/ [accessed April 8, 2018].
LeCount, Ryan Jerome. 2017. More Black than Blue? Comparing the Racial Attitudes of Police
to Citizens. Sociological Forum 32(S1): 1051-1072.
Lee, Hoon, Michael S. Vaughn, and Hyeyoung Lim. 2014. The Impact of Neighborhood Crime
Levels on Police use of Force: An Examination at Micro and Meso Levels. Journal of
Criminal Justice 42(6): 491-499.
Legewie, Joscha. 2016. Racial Profiling and Use of Force in Police Stops: How Local Events
Trigger Periods of Increased Discrimination. American Journal of Sociology 122(2): 379-
Lersch, Kim Michelle, and Tom Mieczkowski. 2005. Violent Police Behavior: Past, Present, and
Future Research Directions. Aggression and Violent Behavior 10(5): 552-568.
Lott, John R., and Carlisle E. Moody. 2016. Do White Police Officers Unfairly Target Black
Suspects? SSRN, November 15, 2016. https://ssrn.com/abstract=2870189 [accessed April
Lucas, Samuel R. 2001. Effectively Maintained Inequality: Education Transitions, Track
Mobility, and Social Background Effects. American Journal of Sociology 106(6): 1642-
McCombs, Maxwell. 2014. Setting the Agenda: Mass Media and Public Opinion, 2nd ed. Cambridge:
Metro Nashville Police Department. 2014. Armed Carjacker on Parole Fatally Wounded by
MNPD Officers. Press Release, March 22. http://www.nashville.gov/News-Media/News-
[accessed April 8, 2018].
Miller, Lisa L. 2010. The Invisible Black Victim: How American Federalism Perpetuates Racial
Inequality in Criminal Justice. Law & Society Review 44(3): 815-842.
Mummolo, Jonathan. 2018. Modern Police Tactics, Police-Citizen Interactions, and Prospects for
Reform. Journal of Politics 80(1): 1-15.
Nicholson-Crotty, Sean, Jill Nicholson-Crotty, and Sergio Fernandez. 2017. Will More Black
Cops Matter? Officer Race and Police-Involved Homicides of Black Citizens. Public
Administration Review 77(2): 206-216.
Nix, Justin, Bradley A. Campbell, Edward H. Byers, and Geoffrey P. Alpert. 2017. A Bird’s Eye
View of Civilians Killed by Police in 2015. Criminology and Public Policy 16(1): 309-
O’Toole, Laurence J., and Kenneth J. Meier. 2015. Public Management, Context, and
Performance: In Quest of a More General Theory. Journal of Public Administration
Research and Theory 25(1): 237-56.
Paoline, Eugene A., Jacinta M. Gau, and William Terril. 2018. Race and the Police Use of Force
Encounter in the United States. British Journal of Criminology 58(1): 54-74.
Prenzler, Tim, Louise Porter, and Geoffrey P. Alpert. 2013. Reducing Police Use of Force: Case
Studies and Prospects. Aggression and Violent Behavior 18(2): 343-356.
Provine, Doris Marie. 2007. Unequal Under Law: Race in the War on Drugs. Chicago:
University of Chicago Press.
Reaves, Brian A. 2015. Local Police Departments, 2013: Personnel, Policies, and Practices.
Bureau of Justice Statistics Report, May 14.
http://www.bjs.gov/index.cfm?ty=pbdetail&iid=5279 [accessed April 8, 2018].
Ritchie, Andrea J. 2017. Invisible No More: Police Violence against Black Women and Women
of Color. Boston, MA: Beacon Press.
Ross, Cody T. 2015. A Multi-Level Bayesian Analysis of Racial Bias in Police Shootings at the
County-Level in the United States, 2011–2014. PloS one 10(11): e0141854.
San Francisco Blue Ribbon Panel. 2016. Report of the Blue Ribbon Panel on Transparency,
Accountability, and Fairness in Law Enforcement. July 2016.
http://sfdistrictattorney.org/sites/default/files/Document/BRP_report.pdf [accessed April
Sellbom, Martin, Gary L. Fischler, and Yossef S. Ben-Porath. 2007. Identifying MMPI-2
Predictors of Police Officer Integrity and Misconduct. Criminal Justice and Behavior
Smith, Brad W. 2003. The Impact of Police Officer Diversity on Police-Caused Homicides.
Policy Studies Journal 31(2): 147-162.
Straubhaar, Joseph, Robert LaRose, and Lucinda Davenport. 2009. Media Now: Understanding
Media, Culture, and Technology. Belmont, CA: Wadsworth.
Strother, Logan. 2018. The National Flood Insurance Program: A Case Study in Policy Failure,
Reform, and Retrenchment. Policy Studies Journal. DOI: 10.1111/psj.12189
Strother, Logan. 2017. How Expected Political and Legal Impact Drive Media Coverage of
Supreme Court Cases. Political Communication 34(4): 571-589.
Twersky-Glasner, Aviva. 2005. Police Personality: What Is It and Why Are They Like That?
Journal of Police and Criminal Psychology 20(1): 56-67.
Weaver, Vesla M. 2007. Frontlash: Race and Development of Punitive Crime Policy. Studies in
American Political Development 21: 230-265.
Wilkins, Vicky M., and Lael R. Keiser. 2004. Linking Passive and Active Representation by
Gender: The case of Child Support Agencies. Journal of Public Administration Research
and Theory 16(1): 87-102.
Wilson, James Q. 1978. Varieties of Police Behavior: The Management of Law and Order in
Eight Communities. Cambridge, MA: Harvard University Press.
Zimring, Franklin E., and Brittany Arsiniega. 2015. Trends in Killing of and by Police: A
Preliminary Analysis Commentaries. Ohio State Journal of Criminal Law 13(1): 247-
Table 1. Percentage of People Killed by Police, by Officer and Victim Race
Race of Victim
Race of Officer
Frequency reported in parentheses.
Pearson chi-squared: 44.36, p < 0.001
Totals may add to less than 100 because of rounding.
Table 2. Racial Makeup of those Killed by White and Non-White Police Officers, When
Armed with a Gun compared to Other Weapons
Race of Officer
Victim Had Gun
Victim Had Weapon
Other than Gun
With Gun: Pearson chi-squared: 26.48, p < 0.001
With other Weapon: Pearson chi-squared: 5.89, p = 0.117
Totals may add to less than 100 because of rounding.
Table 3. Community, Officer, and Suspect/Victim Characteristics Do Not Significantly
Influence Police Killings
Victim Had Gun
Violent Crime led to Interaction
Victim Had Warrant
Black Population Share (zip)
Hispanic Population Share (zip)
Violent Crime Rate (city)
Median Income (quintile, zip)
City-Size Fixed Effects
Population Over 500K
Robust standard errors in parentheses. *** p<0.001, ** p<0.01, * p<0.05
Figure 1. Racial Breakdown of Victims of Police Killing
Population shares calculated from 2014 Census data. African Americans are killed at more than
twice the expected rate given their share of the population. Hispanics and Latinos are killed
slightly more often than expected as well. Whites are killed by police somewhat less often than
their expected share, while Asian Americans are killed much less often than expected given their
relative size in the population.
Black White Latino Asian
Race of Victims of Police Killings
Black White Latino Asian
Proportion Killed to Population
Figure 2. Percentage of People Killed by Police Who are Armed, by Race
In our data, less than 1 percent of victims of police killings were unarmed. Across all racial
groups, 65.3 percent of those killed possessed a firearm at the time of their death.
Black White Latino Asian
Gun Other Weapon Unarmed
Figure 3. The Age of Persons of Different Races Killed by Police
The histograms above present the age distributions of Black, White, and Hispanic/Latino persons
killed by police, respectively. The distributions are overlaid with a kernel density plot which
smooths the distribution and indicates the mean (peak of curve).
Figure 4. The Gender of Persons of Different Races Killed by Police
The histograms above present the gender distributions of Black, White, and Hispanic/Latino
persons killed by police, respectively.
020 40 60 80 100
Black Person Killed
020 40 60 80 100
Hispanic Person Killed
020 40 60 80 100
White Person Killed
Gender of Persons Killed by Police
Figure 5. Graphical Depiction of Victim Race by Officer Race
Figure 6. Graphical Depiction of Victim Race by Officer Race, Across Suspect Weapon Type