Content uploaded by S. E. Costanza
Author content
All content in this area was uploaded by S. E. Costanza on Jun 08, 2020
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=wecj20
Journal of Ethnicity in Criminal Justice
ISSN: 1537-7938 (Print) 1537-7946 (Online) Journal homepage: https://www.tandfonline.com/loi/wecj20
Contextualizing race: a conceptual and empirical
study of fatal interactions with police across US
counties
Ronald Helms & S.E. Costanza
To cite this article: Ronald Helms & S.E. Costanza (2019): Contextualizing race: a conceptual and
empirical study of fatal interactions with police across US counties, Journal of Ethnicity in Criminal
Justice, DOI: 10.1080/15377938.2019.1692748
To link to this article: https://doi.org/10.1080/15377938.2019.1692748
Published online: 25 Nov 2019.
Submit your article to this journal
Article views: 11
View related articles
View Crossmark data
Contextualizing race: a conceptual and empirical study
of fatal interactions with police across US counties
Ronald Helms
a
and S.E. Costanza
b
a
Department of Sociology, Western Washington University, Bellingham, Washington, USA;
b
Department of Political Science and Criminal Justice, The University of South Alabama, Mobile,
Alabama, USA
ABSTRACT
Recent high profile killings of civilians at the hands of law
enforcement have drawn attention to questions about the
determinants of these violent encounters. The literature is
replete with studies focused on individual characteristics and
situational exigencies. This paper takes a structural approach
to assess alternative explanations. Results show that race,
criminal violence, and general conditions of economic inequal-
ity are strong predictors of police killings of civilians across
3,081 US counties. The empirical findings from this research
provide a broad foundation for conceptualizing a structural
model of police lethal violence against citizens.
ARTICLE HISTORY
Received 8 February 2019
Revised 10 November 2019
Accepted 11 November 2019
KEYWORDS
Police brutality; lethal use
of force; police violence;
minority threat; police and
community; police killings
and race; police and the
law; law enforcement
Introduction
Much research has focused on understanding what happens in police
encounters with citizens triggering lethal violence. While estimates of police
contacts with citizens range from approximately 45-60 million annually
(Stamper, 2016; Hyland, Langton & Davis, 2015), recent evidence suggests
that about 1200 citizens are killed by law enforcement officers in the US
each year (Lowery, 2016)
1
. Research on police-citizen violence is lengthy
and offers a great deal of insight into situational and structural sources of
police violence against citizens.
Politically, when any case of deadly police violence garners national
attention, it generates considerable controversy (Chagnon, Chesney-Lind &
Johnson, 2018) and the potential for a wide range of repercussions for
police and society. This is true even in cases where the killing is ultimately
found to have been justified. Fatal interactions between police and citizens
(FIP’s) attract public attention because each case underscores a tragic story
while also dramatizing irrevocable outcomes of citizen challenges to state
police power. Academic researchers have taken an interest in police-citizen
CONTACT Ronald Helms Ronald.Helms@wwu.edu Department of Sociology, Western Washington
University, Bellingham, WA 98225-9081.
ß2019 Taylor & Francis Group, LLC
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE
https://doi.org/10.1080/15377938.2019.1692748
encounters and have theorized broad foundations of social conflict rooted
in racial threat and income inequality.
The racial threat hypothesis maintains that police seek to regulate and sup-
press threats to the existing social structure ostensibly arising from the pres-
ence of minorities (Holmes & Smith, 2018; Edwards, Esposito & Lee, 2018;
McCarty, Ren, & Zhao, 2012; Holmes, Smith, Freng, & Mu~
noz, 2008;Kent&
Jacobs, 2005;Jacobs&O’Brien, 1998;Liska&Yu,1992). Evidence of minority
overrepresentation as victims of police violence provides a strong basis under-
girding this research focus (Vitale, 2017).JohnsonJr,St.Vil,Gilbert,
Goodman, and Johnson (2018) point out that hostility between police and
minority groups emerges from a historical police focus aimed at controlling
minority areas. Underlying this perspective is a belief that the presence of
blacks and other oppressed groups threaten the distribution of resources con-
trolled by whites. The greater the size of the black population across units, the
more prominent are threats to white control over economic and other resour-
ces. According to the racial threat hypothesis, police serve a conservative role
of stabilizing racial inequality. This article observes whether lethal violence is
more frequent in areas with larger racial and ethnic minority populations.
In addition to racial threat, some researchers articulate a link between eco-
nomic inequality and police aggression. Blalock (1967) claims that political
power in modern society is a function of comparative differences in monetary
resources. Jacobs and Helms (1997) emphasize that researchers who focus
attention on economic inequality typically view large differences in economic
resources as a pervasive characteristic of capitalist societies that requires coer-
cion for maintenance. Chambliss and Seidman (1982,p.31)offerasuccinct
statement on this matter, saying that “themoreeconomicallystratifiedasoci-
ety becomes, the more it becomes necessary for dominant groups to enforce
through coercion the norms of conduct that guarantee their supremacy”.It
follows that one function of economic resource differences is the power to
influence social imperatives about law and social control. Variation in the
level of economic stratification across units has been theorized as playing a
key role in determining criminal justice outcomes (Johnson Jr. et al., 2018;
Stucky, 2005; Jacobs & Helms, 1997;Jacobs,1979). In keeping with this per-
spective, this study observes whether variation in economic inequality is pre-
dictive of fatal interactions with police (FIP’s).
Most people learn about police killings primarily through media reports
that highlight the seemingly split-second nature of decision making in these
encounters (Weitzer, 2015;2002). Unfortunately, such depictions tend to
promote the impression that these events are both stochastic and pervasive
(Dunham & Petersen, 2017; Hirschfield & Simon, 2010).
FIP’s damage institutional legitimacy and citizen trust while generating
renewed questions concerning the police role. Local government risks
2 R. HELMS AND S. E. COSTANZA
financial liability as victims’family members file lawsuits in response to
police actions (Moore, 2016; Cheh, 1996). Perhaps the greatest immediate
challenge for police is maintaining public peace in the aftermath of contro-
versial FIP’s (i.e., the Michael Brown incident in Ferguson, Missouri).
Incidents of widely perceived unjustified killing by police generate unruly
crowds whose actions sometimes turn from civil protest to rioting and
looting (Lowery, 2016; Embrick, 2015; Loftin, Wiersema, McDowall &
Dobrin, 2003). Loftin et al. (2003, p. 1117) write that “These incidents fre-
quently cause large numbers of injuries and deaths, and they disrupt the
social and economic relationships through which essential economic,
health, public safety, and social services are provided to communities.”
In the aftermath of a police killing of a citizen, initial information gener-
ally comes in the form of a nearly ritualistic narrative; the well-worn clich
e
of the officer, in fear for his life, acting in self-defense (Culhane, Boman &
Schweitzer, 2016; Fridell & Lim, 2016). To the extent the media reproduces
police narratives of these violent encounters, they often serve a role in legit-
imizing police violence while muting public discourse that might otherwise
challenge basic assumptions of the existing order.
When a citizen is killed in police custody, people commonly seek to under-
stand the incident by mentally locating themselves in place of the victim or
officer (Lowery, 2016). The media plays to this dominant pathology, convey-
ing relatable information about the individual that was killed and the officer
(or officers) involved. Public anxieties center on common themes: “Was the
victim on drugs?”;“Did the victim have a criminal record?”;and“Was the offi-
cer really in danger?”In focusing on these practical questions, what is lost
through omission are broader questions about social arrangements that draw
police and citizens into these divisive situations. What is missing from the
public narrative is the framework for action, that is to say, an accessible
imagery of the broad structures that cultivate agency, drawing citizens and
police into hostile engagement and deadly violence. The following research
seeks to reduce that gap conceptually by offering an account that links situ-
ational and structural level research.
In most recent cases generating national attention, a key element has
been the discovery of a live recording challenging the police account
(Lowery, 2016). Despite the preliminary nature of typical journalistic
accounts and the shortage of follow up exploration of incidents, the news
media continue to serve as de facto watchdogs over police activities. Their
work is important in registering acts in the public consciousness and stir-
ring collective passions in the face of possible injustice (Bock, 2016).
Social science research has connected patterns of social and economic
structures to police use of force generally, with less attention specifically to
killings (Terrill & Paoline, 2013; Kania & Mackey, 1977). The incident-level
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 3
focus of much of the research is perhaps due to the low frequency of reported
FIP’s,whichmakesastructural-levelapproachimpractical.
The emphasis on incident level determinants surrounding FIPs also reflects
past and present limitations in current data collection and accurate reporting
by local law enforcement agencies (Feldman, Gruskin, Coull & Krieger, 2017;
Jennings & Rubado, 2017). Federal data collection efforts draw on local
records surrounding these incidents, which are sometimes inaccurate or
otherwise distorted, thus undermining well-intended research efforts.
This article utilizes a novel data development approach that overcomes
limitations in data collection and reporting on FIP’s. Focusing on police
contact with criminal populations, minority social demographics, and vari-
ous measures of economic inequality, this article seeks to provide a struc-
tural account of variable risks that any jurisdiction will experience one or
more fatal interactions with police. The following analysis seeks to broaden
insight regarding these highly consequential exchanges between police and
citizens while documenting the social foundations underlying the patterns
of fatal police-citizen encounters.
Literature review
The following review of literature highlights the historically incident-based
focus of FIP research literature and also draws attention to structural con-
siderations. This facilitates conceptualizing these respective areas of
research as features of the same phenomenon under study, thereby setting
the stage for comprehending the broad social foundations underlying pat-
terns of fatal interactions between police and citizens.
Most police killings research has sought to dissect incidents, omitting
details that serve as framework and backdrop. A smaller body of research
has focused on organizational factors and structural considerations, often
stressing differences in policing practices in racially and economically
diverse communities (Lichter, Parisi, & Taquino, 2015; Smith & Holmes,
2014; Skogan, 2014; Jacobs & O’Brien, 1998).
Literature on police use of force
Early models of police use of force are oriented in the work of Westley
(1953;1970) who used the personal testimonies of officers to theorize about
their decisions to use violence in the field. Westley’s research highlights the
perceived occupational necessity of the officer to maintain control of situa-
tions as the basis for “…violence in the form of the club or the gun …”
(1953, p. 35). The aforementioned research elevates exigent circumstances,
highlighting officer reactions to rapidly evolving situations.
4 R. HELMS AND S. E. COSTANZA
Research since Westley has shown a great deal of continuity in enforcement
tactics. Less than lethal displays of police coercion are so common in America
that some politicians have openly supported police brutality to appeal to voters
(Bean, 2017). Research by Belvedere, Worrall and Tibbetts (2005) indicates that
use of force is not always misdirected or maleficent. Police sometimes must
resort to abrasive language or aggressive physical posturing to get unruly
crowds or uncooperative suspects under control. They may otherwise assert
authority by touching or “frisking”people or fingering or brandishing weap-
ons. Such actions are tolerated by police agencies (Reiss, 1968) as a necessary
evil in the daily negotiation of social order.
A recent report from the Bureau of Justice Statistics (Hyland, Langton, &
Davis, 2015) documents that police street tactics have remained very con-
sistent. According to the report, among the nearly 45 million citizen con-
tacts recorded in their database: “An estimated 1.6% (715,500 contacts)
experienced the threat or use of nonfatal force by police. Nearly 75%
(535,300 contacts) of those who said police used force during the contact
described it as “excessive”.
Police killings are viewed as the barest exhibit of the state’s coercive cap-
ability exercised against its citizen population (Crawford & Burns, 1998).
Citizen tolerance of coercion quickly evaporates when police act with
deadly force (Weitzer, 2002). Research focusing on individual-level use of
force dynamics turns on assumptions of officer rational decision-making
capabilities and often assumes that escalation of physical force is robustly
determined by individual interactions and “split-second”dynamics (White,
2002; Fridell & Binder, 1992; Binder & Scharf, 1982). With acknowledge-
ment of the occasional rogue officer (rotten apple; see Lee, Weitzer &
Martinez, 2018), force is generally applied under conditions that render it
necessary to gain suspect compliance or situational control (Garner,
Maxwell & Heraux, 2002).
A focus on split-second decisions has provided a useful framework for
understanding the subtle and often ambiguous work of police (Fyfe, 1997).
Using this focus, researchers have documented the importance of suspect
resistance (Terrill, 2003; Tedeschi & Felson, 1994); ascribed characteristics
of the victim (Crawford & Burns, 1998); victim demeanor (Engel, Sobol &
Worden, 2000); the extent of officer education and training (Lipsky, 1980);
and the range of incident level circumstances precipitating violence.
Agency policies and the policing of communities
Although case level studies offer insight into officer decision making in
highly stressful situations, such efforts usually overlook the broader envir-
onmental contexts that may increase the likelihood of a fatal encounter
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 5
(Terrill & Reisig, 2003). Research on police agencies has expanded under-
standing about the effects of bureaucratic practices, organizational subcul-
tures, and rules and regulations (Nowacki, 2015; Eitle, D’Alessio &
Stolzenberg, 2014) on police use of force. Bureaucratic factors such as levels
of organization (Eitle & Monahan, 2009) and number of badged personnel
(Sherman & Weisburd, 1995) have been linked to police enforce-
ment activities.
Wilson (1968) posited that internal organization matched with localized
political environments create a predictable pattern of police behavior.
Raine’s(1967) study of the 1966 Watts uprising indicated that the LAPD
bureaucracy was structured to punish citizens for reporting incidents of
police brutality; a factor that he claimed (among others) led to widespread
use of force by officers. In keeping with this line of inquiry, Ericson (1982)
showed that officers develop a “recipe of rules”that guides their behavior.
At the agency level, a collegial image forms among officers regarding how
to handle everyday aspects of police work (Paoline, Gau, & Terrill, 2016)
2
.
This informal set of rules contributes to the pattern of police-citizen inter-
actions. Lewis and Ramakrishnan (2007) document a link to training prac-
tices as well. Recent work by Jennings and Rubado (2017) indicate that
FIP’s can be curtailed through agency reporting requirements, noting favor-
able effects where officers were required to file a report for every incident
during which they point their weapons at someone.
Structural influences on the policing of communities
Structural research on police violence has focused disproportionately on
the effects of racial conflict (Blalock, 1967) and economic inequality
(Petrocelli, Piquero, & Smith, 2003). Kania and Mackey (1977) studied
police homicide across 32 states and set a precedent in the study of struc-
ture in use of force situations. In their view, police killings of citizens do
not specifically represent an interpersonal decision, but instead is part of a
“greater societal matrix”(Kania & Mackey, 1977, p. 43). Aggregate research
acknowledges the effects of violent crime and economic inequality (Crank,
2003; Petrocelli, et al., 2003). Ecologically based models (Nix, Campbell,
Byers & Alpert, 2017; Ha & Andresen, 2017; Donner, Maskaly, Piquero, &
Jennings, 2017; Klinger, Rosenfeld, Isom & Deckard, 2016; Ross, 2015)of
police use of force examine characteristics of the structured environment in
which citizens and police interact.
Racial composition (Holmes, Smith, Freng & Mu~
noz, 2008) and jobless-
ness have been shown to affect police outcomes. Holmes, Painter, and
Smith (2019) show that neighborhoods with higher concentrations of blacks
or Hispanics are more likely to experience high rates of police killings of
6 R. HELMS AND S. E. COSTANZA
citizens. Given the well-documented connections between race and social
class some researchers have looked at how inequality may augment police
killings. A key idea is that poverty, heightened criminal violence and the
prevalence of inequality across areas sets the stage for conflict (Jones-
Eversley, Adedoyin, Robinson & Moore, 2017; Banks, Eberhardt &
Ross, 2006).
Research by Jacobs and O’Brien (1998) confirms police officers are likely
to use violence in areas where disenfranchized social groups are disruptive
or are more numerous. In a study of police killings of citizens across 170
cities, Jacobs and O’Brien found that the percent of minorities was signifi-
cantly related to police use of lethal force. Law enforcement agencies, in
part, represent the government’s material ability to deal with prospective
challengers to social order.
Other structural research focused on urban areas has shown that politic-
ally disenfranchized groups face disproportionate aggressive action by
police agencies (Holmes et al, 2008; Kent & Jacobs, 2005). Using a struc-
tural framework, researchers have developed empirical links between
diverse indicators of inequality and police agency patterns of discretionary
enforcement and policy decisions (Cline, 2017; Tonry, 2011; Helms &
Costanza, 2010;2009). This area of research presents evidence that citizens
from unequal and racially polarized environments are at heightened risk of
experiencing police aggression. For those whose social behaviors have been
criminalized or whose ascribed characteristics have been successfully stereo-
typed (e.g.,: drug users, gang members, racial minorities, etc.) the risks of
experiencing aggressive policing are heightened (Costanza & Helms 2012).
Limitations in previous research
Among the most serious limitations in reporting on FIPs is a concern that
has only recently come to light. Previous aggregate research has relied on
either the Federal Bureau of Investigation’s Supplemental Homicide
Reports (SHR) or the Center for Disease Control and Prevention’s National
Vital Statistics (NVS). Recently both sources have been challenged by
researchers (Jennings & Rubado, 2017; Campbell, Nix, & Maguire, 2018;
Lowery, 2016; Klinger, 2012) who document far more police homicides
against citizens than have been reported by official sources. This has
resulted in attempts to develop alternative data on the phenomenon using
a crowd sourcing approach.
Some researchers have expressed concern about crowd source data claim-
ing that it may be politically biased (Holmes et al., 2019). These authors
are deeply indebted through their research to the notably flawed govern-
ment produced SHR data. Such a defense of the SHR against other sources
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 7
of police killings data is unfortunate. Other researchers have noted that
official reports are not free from “political bias”while documenting that
the SHR underreports police killings. Loftin et al. (2003) make a convincing
case that official data such as the SHR and NVS have glaring
shortcomings.
3
A structurally informed conceptualization can take steps toward demysti-
fying the dynamics at work in everyday police-citizen encounters while
providing a more transparent account of the broad social dynamics that
put citizens at heightened risk in interactions with police.
4
It also may pro-
vide a basis for comprehending disproportionate minority representation in
these violent encounters (Vitale, 2017). Finally, a study that assesses data
from non-government sources may successfully open the research doors a
bit wider for consideration of social justice, a key factor at the heart of
much police-citizen conflict.
Theory and hypotheses
Addressing threats: Police contact with criminal populations
The first set of hypotheses emphasizes police contact with criminal popula-
tions as a source of heightened risk that police will be drawn into violence
against citizens. We emphasize the key role of police coercion inherent in
every encounter with citizens, regardless of the nature or severity of the
activities being addressed. To assess police capacity for contact with citi-
zens, a hypothesis is developed concerning the total number of badged
police personnel available for duty across police agencies
5
. A measure of
violent crime rates is introduced to isolate the population of citizens
engaged in serious criminal activities locally, which is a unique component
of risk.
A measure of arrests for drug offenses provides an indicator of police
activation against an identifiable target group that has been vilified in both
national and local political discourse and media. Increased rates of crime
and heightened attention to the drug offenses with its pragmatic targeting
of local users presents risks for officers and citizens (Tyler, Fagan & Geller,
2014; Costanza & Helms 2012; Helms & Costanza 2010). We note that the
discourse on legalization has shifted in recent years and so the historical
antagonism in police-citizen encounters may have diminished substantially.
The outcomes of these types of encounters may be opposite to conven-
tional claims. An indicator of disorder arrests is also introduced, highlight-
ing risks associated with police engagement with low end criminal
populations. Every police-citizen encounter poses the risk of escalation into a
deadly exchange and so we include these contacts in the empirical mod-
els below.
8 R. HELMS AND S. E. COSTANZA
Addressing racial threats: Social demographics and police social control
Researchers have focused intensively on hypotheses linking threats emanat-
ing from the presence of racial minority groups and heightened levels of
police violence (Smith & Holmes, 2014). Racial threat theorists emphasize
that aggressive policing measures are a response to social uncertainties
associated with the presence of ethnic and racial groups (primarily African
Americans and Hispanic Americans) who are often socially disenfran-
chized. Police are theorized as playing a key role in the suppression of
blacks and others who present a threat to existing stable inequality favoring
Whites. Allegations of mistreatment of minorities are often at the center of
anti-Police protests. Across units, the size of minority population groups,
with implied threats to existing social resource allocations, is hypothesized
as a source of enhanced social control response and the potential for
lethal violence.
Two indicators were developed based upon common themes in the lit-
erature regarding social demographics and police violence. Historical evi-
dence documents tense relationships between African Americans and local
police (Schlesinger & Tonry, 2013; Tonry, 2011). The history surrounding
the policing of racial minorities implies that any study of police violence
against citizens would be missing a key dimension without modeling this
critical social dynamic. Recent literature has been critical of studies focus-
ing only on African Americans to the exclusion of other social demograph-
ics (Segura & Rodrigues, 2006). This research also introduces a measure of
Hispanic population size. Hispanics have experienced unique historical ten-
sions with local police and are a growing population group throughout the
US today. As with African Americans, Hispanics have been grossly stereo-
typed both in terms of drug use and interpersonal violence. Therefore,
indicators for both African American and Hispanic population percentages
are included. A positive association between the size of these groups and a
heightened risk of police-citizen lethal violence is expected.
Addressing Economic-Based threats: Economic inequality and police coercion
Economic inequality has been theorized and empirically assessed for its
effects on police violence (Cline, 2017; Rogers, 2015). Central to this argu-
ment is the idea that as private economic resources become increasingly
skewed, those with the greatest stakes in existing arrangements increasingly
exhibit both the need for expanding social control measures and a height-
ened ability to call upon the state to provide enhanced protection (Jacobs,
1979). This extends to budgeting and personnel hiring (Helms, 2008;2007)
and includes actions in the form of permissive policies governing police
use of force and muted reactions to police violence generally. While
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 9
research has documented a link between heightened comparative differen-
ces in resources and outcomes favorable to the affluent such as strength-
ened police departments, the argument is extended here to incorporate
insights about the corrosive effects of inequality on police-citizen relation-
ships and the risk of lethal violent encounters between members of these
antagonistic groups in society.
Adding to these claims, the “American Dream”argument (Messner &
Rosenfeld, 2012) highlights the centrality of economic values and competition
for social prestige and economic rewards. Under conditions of gross economic
inequality, economic tensions are conceptualized as contributing to a general
heightened risk of conflict between police and citizens. Economic inequality
places pressure on groups striving to achieve material success goals. Across
communities, conditions of inequality put everyone at heightened risk of failing
to meet basic needs and fulfill economic expectations (Cloward, 1959;Merton,
1938). The effects of economic strain would likely be magnified for the most
socially marginalized (all too often Hispanic and Black citizens) as they are the
least well situated to successfully compete for reduced economic rewards. It is
under just such conditions that police and citizens in any encounter might
approach a critical threshold for angry aggression (Bernard, 1990) and the
attendant risk that aggression will spill over into deadly violence.
To capture the social dynamics of disruptive economic conditions, this
article focuses on a range of economic indicators, including median
incomes, household economic inequality (Gini), and area unemployment
rates. Adult population ages 20-34 is also included since this group faces a
unique challenge in assuming their position in the economic system
6
. Each
of the economic indicators provide context for economic-based tensions,
particularly in the presence of young adults striving to establish a footing
on the economic success ladder. We expect that each of these indicators
will exhibit a positive association with the dependent variable. Regional
indicators are dummied to isolate the effects of possible spatial variation
that otherwise would not be accounted for in the models.
Data and methods
Outcome data and sample units
The outcome data utilized in this research are counts of police killings of
citizens taken from the Killed by Police database. Operating since 2013,
Killed by Police is a project that compiles web-based crowdsourcing data
(Campbell et al., 2018) on deadly police-citizen encounters. Campbell et al.
(2018, p 10) describe Killed by Police (KBP) as follows:
“Each time a civilian death occurred involving a police officer over this period, KBP
documented the following: date; name of the deceased (along with a picture in most
10 R. HELMS AND S. E. COSTANZA
instances); the state in which the death occurred; the deceased’s age, gender, and
race/ethnicity; the manner of death (gunshot, TASER, vehicle, chemical, and/or
restraint); and a link to at least one local news source.”
Data on police killing of citizens were collected for the period of May
2013 through April 2014. At the time of collection, neither earlier nor
more recent data were available to researchers through the Killed by Police
archive
7
. For this time period this represents the most complete data avail-
able outside of the SHR data commonly used in structural research on
police killings. This period of collected data predates the highly controver-
sial Michael Brown killing, which attracted considerable national attention
to the phenomenon of police killings of citizens. The police killings count
indicator ranges from 0 to 64 (for additional insight, see the methodology
section below), with a majority of counties experiencing 0 or 1 police kill-
ing during the year. This indicator was merged with US Census data and
other sources to form the final database for this research. The unit of ana-
lysis is U.S. Counties (N¼3,081).
Measurement of indicators
In addition to outcome data, data were taken from The U.S. Census,
Uniform Crime Report and other sources to develop operational indicators
for this research. The law enforcement rate indicator is based on geocoded
data for all jurisdictions within each county. The police personnel indicator
consists of all law enforcement officers, both full and part time, from all
jurisdictions (2500þ) within each respective county. Underlying counts are
drawn from 2 sources, the FBI’s Uniform Crime Report, which provides
counts of county deputies, and the American Community Survey. After
geocoding the data, these were merged to form law enforcement counts.
The data were then constructed as per capita rates and logged to address
skew. For all logged variables used in this analysis a constant of 1 was
added prior to logging as a necessary step in order to avoid attempting to
log any 0 values.
The violent crime rate indicator includes interpersonal crimes that are
part of the Uniform Crime Series from the FBI. These are constructed as
per capita rates and logged to address skew. Drug arrests statistics are taken
from the FBI data and are measured as all arrests for sales and manufactur-
ing, distribution, and possession of any drug. The variable is standardized
as a per capita rate. Disorder arrests are taken from the UCR (Part 2) and
include approximately 20 categories of arrests for low level offenses
(excluding drug arrests)
8
. These are constructed as per capita rates,
summed, and logged. Social demographic indicators for African American
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 11
and Hispanic population are percentages of total unit population and are
logged to address skew.
Three economic variables drawn from the US Census (2010) are intro-
duced into the analysis. Along with median incomes, economic inequality
is operationalized with the Gini index of household incomes. Finally, the
unemployment percentage is the percent of the working age citizen popula-
tion that is currently unemployed. Percent working age population (20–34)
is from the US Census (2010) and is logged. Following convention, we
operationalize four regional dummy indicators (West, South, Midwest,
Northeast) based on Census regions to capture any unmeasured sources of
variation, with the omitted West region serving as the reference category in
the analyses.
Methodology
This study presents negative binomial regression estimates in the analyses
below. The dependent variable for this study is a count of police killings of
citizens. The non-normal distribution of count data, including an excess of
0’s, means that a least-squares approach is not appropriate. The count indi-
cator for police killings of citizens ranges from 0 to 64. Among those with
any killings, a majority of jurisdictions experienced only 1 killing. Initial
tests showed that the indicator is over-dispersed and therefore a Poisson
regression model was rejected in favor of the negative binomial model. The
UCLA statistics division offers the following supporting logic: “[Negative
binomial regression] can be considered as a generalization of Poisson
regression since it has the same mean structure as Poisson regression and
it has an extra parameter to model the over-dispersion.”(UCLA, 2017).
While point estimates tend to be nearly identical across Poisson and
Negative binomial models, the second model is more conservative, since “if
the conditional distribution of the outcome variable is over-dispersed, the
confidence intervals for the Negative binomial regression are likely to be
narrower as compared to those from a Poisson regression model.”
(UCLA, 2017).
Results
Table 1 shows the number of valid cases, minimum and maximum values,
means, and standard deviations for all indicators presented in the analysis.
Table 2 shows zero-order correlations for all indicators used in this ana-
lysis. Correlations with the outcome indicator are generally consistent with
expectations and, as is evident from the table, preliminary evidence does
not indicate problems with collinearity. Substantial shifts in variables
12 R. HELMS AND S. E. COSTANZA
included in respective models show results with generally stable coefficients,
which further strengthens our belief that collinearity is not disproportion-
ately affecting estimates.
Negative binomial regression results
Table 3 shows results of a series of increasingly saturated models, with the
dependent variable being a count of police killings of citizens. A negative
binomial model is run after running a Poisson regression model and
assessing possible over-dispersion. For each successive model the resulting
likelihood ratio values suggested that lnalpha was above zero, indicting over-
dispersion, and that a Poisson model is inappropriate. A negative binomial
model adjusts for over-dispersion, providing more conservative standard
error estimates and may modify parameter estimates (Hilbe, 2011).
Model 1 focuses on police engagement with criminal populations. Model
2 retains indicators from Model 1 while introducing several additional
social demographic indicators to isolate the effects of racial and ethnic
minority population on police killings of citizens. Finally, in model 3 add-
itional indicators are introduced that together express the effects of eco-
nomic stress and affiliated social and labor force tensions. Each of the
models includes the series of dummy indicators as well.
The base model (Model 1, Table 3) introduces regional dummies and a
police rate indicator to address police saturation across the respective com-
munities. A measure of violent offenses is used to isolate the level of local-
ized threat. Finally, 2 arrest indicators are introduced to address police
activation around drugs and a wide range of disorder crimes. The basic
insight of this model is that police activation against criminal populations
increases the risk of violent encounters.
Table 1. Descriptive statistics.
NMinimum Maximum Mean Std. deviation
KILLED BY POLICE 3081 0.00 64.00 0.36 1.83
LN OFFICER RT 3081 0.00 6.61 2.82 0.61
LN VIOLENT CRIME RT 3081 0.00 4.81 2.14 0.99
LN DRUG ARREST RT 3081 0.00 8.34 3.27 1.18
LN DISORDER ARREST RT 3081 0.00 5.30 2.24 1.22
LN BLACK POP 3081 0.00 4.47 1.55 1.19
LN HISPANIC POP 3081 0.10 4.59 1.65 0.94
MEDIAN INCOME 3081 20577 119075 43075 10641
GINI INDEX 3081 33.22 59.94 43.54 3.50
% UNEMPLOYMENT 3081 1.50 29.00 9.16 3.10
LN POP 20-34 3081 2.25 3.79 2.90 0.20
D. SOUTH 3081 0.00 1.00 0.46 0.50
D. MIDWEST 3081 0.00 1.00 0.34 0.47
D. NORTHEAST 3081 0.00 1.00 0.07 0.25
D. WEST 3081 0.00 1.00 0.14 0.35
Valid N (listwise) 3081
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 13
Table 2. Zero-order correlations.
1 2 345678 9 101112 13 1415
1. KBP COUNT 1.00
2. LN T. POL RT. .047 1.00
3. LN V CRIME RT. .098 .165 1.00
4. LN TDRUG RT. .048 .128 .666 1.00
5. LN DISORDER .441 .072 .491 .165 1.00
6. % BLACK POP .127 .192 .319 .178 .181 1.00
7. % HISP POP .197 .272 .195 .161 .146 .084 1.00
8. MED INCOME .122 .103 -.076 .006 .167 -.089 .153 1.00
9. GINI INDEX .127 .081 .213 .132 .143 .383 .089 -.353 1.00
10. % UNEMPL .887 .064 .222 .061 .531 .157 .225 .211 .141 1.00
11. % POP 20-34 .166 .106 .240 .205 .226 .416 .253 .072 .237 .195 1.00
12. D. SOUTH -.021 .178 .193 .186 -.082 .605 .088 -.248 .343 -.059 .157 1.00
13. D. MIDWEST -.078 -.112 .311 -.272 -.297 -.425 -.308 .080 -.293 -.070 -.179 -.656 1.00
14. D. NORTHEAST .029 -.281 .073 .085 .218 -.018 -.050.202 .017 .109 .019 -.243 -.191 1.00
15. D. WEST .116 .098 .097 .045.002 -.277 .331 .102 -.080.102 .005 -.367-.288 -.107 1.00
Significant at the .05 level.
Significant at the .01 level.
14 R. HELMS AND S. E. COSTANZA
Model 1 produces a reasonable overall prediction result with a log likeli-
hood of -1920.9466 and a highly significant model Chi
2
(169.68, p<.001).
The Pseudo R
2
for this model is .0423. Stata calculates McFadden’s pseudo
R
2
as 1 –ll(model)/ll(null) ¼0.0423. The pseudo R
2
measure is not strictly
comparable to the Pearson’sR
2
found in OLS regression analyses and so
should be interpreted with caution. Its value here is to provide a baseline for
assessing the performance of increasingly saturated models (UCLA, 2017).
The logged officer rate indicator (.418, p <.001) in Model 1 exhibits a posi-
tive and significant relationship with the citizens killed count indicator.
Substantively, after holding constant other predictors in the model, a one unit
change in the officer rate is associated with a .418 difference in the logs of
expected counts of citizens killed. Also, violent crime is significant in Model 1
(.719, p <.001). One potential interpretation of this result is that under condi-
tions of growing violent crime police may be quicker to escalate violence as a
defensive measure. Model 1 does not include social demographic and eco-
nomic sources of police killings and so noise in the data is likely substantial.
An indicator of community violence may be reduced after social demographics
that heighten volatility in police–citizen relationships is statistically controlled.
Social disorder arrests (.017, not significant) appear to have no discern-
able influence on police killings. By contrast drug arrests (-.393, p <.001),
Table 3. Negative binomial regression analyses of police killings of civilians, May
2013–April 2014.
Dependent Variable ¼Count of police killings of civilians
Independent variables Model 1 Model 2 Model 3
Constant 1.853 2.978 16.697
LN OFFICER RT .418 .069 .128
LN VIOLENT CRM RT .719 .194.251
LN DRUG ARREST RT .393 .127.238
DISORDER ARREST RT .017 .053 .012
LN BLACK POP –.868 .434
LN HISPANIC POP –.695 .468
MEDIAN INCOME ––.056
1
GINI INDEX HH INCOME ––.119
% UNEMPLOYMENT ––.092
LN POP 20–34 ––2.050
D. SOUTH .867 1 .576 .862
D. MIDWEST 1.381 .953 .644
D. NORTHEAST .208 .214 .282
Log likelihood 1920.9466 1750.0119 1668.3702
Model chi-square 169.68 511.55 674.83
Number of cases 3081 3081 3081
Pseudo R
2
.0423 .1275 .1682
L-R Test
Model 1 vs. Model 2 341.87
Model 2 vs. Model 3 163.28
Significant at the .05 level with one-tail test.
Significant at the .01 level with one-tail test.
Significant at the .001 level with one-tail test.
1
coefficient multiplied by 1,000
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 15
exhibit an inverse association with the study outcome. Finally, the regional
dummies are negative and excepting the Northeast are significant in this
model. We note that estimated coefficients on the region dummies are rela-
tive means. When compared with the omitted category, which is the West
region, the regional dummies all exhibited negative coefficients, implying
that the West region, on average, experienced more police killings than
other regions.
In sum, Model 1 provides initial evidence regarding effects of police con-
tact with criminalized populations on the citizen killings count indicator.
Model 2 adds two social demographic indicators into the regression equa-
tion, while retaining the indicators introduced in Model 1, to assess
whether theorized social tensions associated with the presence of African
American and Hispanic population groups are useful in predicting the pat-
tern of citizen killings across communities.
Model 2 documents improved performance associated with the introduc-
tion of the social demographics indicators. The log likelihood is reduced
(Model 1 ¼-1920.9466 v. Model 2 ¼-1750.0119) and the model Chi
2
is
large and significant (511.55, p <.001). The Pseudo R
2
also reflects
improvements in the model’s predictive accuracy (Model 1 Pseudo R
2
¼
.0423 v. Model 2 Pseudo R
2
¼.1275). The Likelihood Ratio test comparing
the two models is highly significant (L-R Test Result ¼341.87, p <.001),
indicating the block of social demographic indicators together produce a
significant enhancement when compared with the base model.
Turning to results for Model 2, after including the African American
and Hispanic population indicators the base model estimates are modestly
altered. The police rate indicator is no longer statistically significant as a
predictor of citizen killings. One interpretation is that the presence of
minorities rather than the number of police is a key to predicting fatal
police–citizen encounters. Police are a scarce resource with generally no
more than 2-4 officers per 1,000 population. And so, social dynamics loom
large as a proximal explanation for the variable risk of fatal encounters
between citizens and police.
The violent crime rate was a significant predictor in Model 1 and
remains so after statistically holding constant social demographic indicators
(.194, p <.05). The drug arrest rate (-.127, p <.05) and disorder arrests
(.053, not significant) are shown to be reasonably consistent coefficients
with the drug rate indicator exhibiting a significant negative association
with the police killings indicator.
Essential to the conceptualization of this paper are findings regarding
race and ethnicity. It was theorized that the presence of large racial and
ethnic minority populations would be associated with heightened social
uncertainty and potential for social unrest. A positive effect for both the
16 R. HELMS AND S. E. COSTANZA
African American and Hispanic population indicators was expected. The
tabled results provide supporting evidence since both coefficients are posi-
tive and significant in the model (African American ¼.868, p <.001;
Hispanic ¼.695, p <.001).
These empirical results along with the previously discussed finding on the
insignificant police rate indicator are in line with claims that variable social
threats are a primary feature in the heightened risk that citizens will experi-
ence violence at the hands of local police. Recent empirical evidence suggests
police actions are not informed only by implicit bias but that overt attitudes
very likely affect police actions in everyday encounters (Johnson, 2016).
Police are likely subject to a range of social and psychological filters that
condition their sensitivities. When police disproportionately engage with
minorities there may be an all too real heightened sense of insecurity and
fear. Police perceptions of disproportionate minority involvement in violence,
whether justified or not, may be a factor resulting in police being quicker to
resort to aggressive violence to counter such perceived threats. Results for
the regional indicators are substantively unchanged across Models 1 and 2.
Model 3 advances in a parallel manner with previous models. Base model
predictors remain, along with demographic indicators, while 4 indicators
linking theorization about expectations of labor market competition and
related economic tension are introduced. The model shows a strong
improvement in performance when compared with Model 2. Comparisons
between Models 2 and 3 reveal that the log likelihood is further reduced
(Model 2 ¼-1750.0119 v. Model 3 ¼-1668.3702), while Model 3’s Chi
2
value is large and significant (674.83, p <.001). The Pseudo R
2
again
reflects improvements in the model’s predictive accuracy (Model 2 Pseudo
R
2
¼.1275 v. Model 3 Pseudo R
2
¼.1682). The Likelihood Ratio test com-
paring Model 2 and Model 3 is highly significant as well (L-R Test Result
¼163.28, p <.001), indicating that the block of economic indicators
together produce a significant enhancement over the less-saturated model.
The added indicators in this model are significant with the correct signs
on the coefficients. After holding constant respective indicators for police
contact with criminalized populations and several social demographic indi-
cators and regional dummies, increased median incomes (.056, p <.001),
expanded income inequality (.119, p <.001), heightened unemployment
(.092, p <.001), and the presence of a large relatively young working age
population with rising economic expectations (2.050, p <.001) are each
statistically associated with change in the logs of counts of citizen killings.
These results present an opportunity to develop a more complete conceptu-
alization of economic-based risks.
The various economic indicators each exhibit a close link with height-
ened risks of police-citizen violence. Challenging economic conditions are
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 17
for many people a basic reality that must be navigated perpetually in an
individualized pursuit of middle-class living arrangements. In the leading
capitalist society, materialism and open-ended competition are core cultural
values compelling the general pursuit of material success goals (Messner &
Rosenfeld, 2012). Under conditions of expanded economic stress, police
undoubtedly face a heightened risk that economically and socially margi-
nalized citizens will be drawn into increasingly focused exchanges with
police and that police in turn will be challenged to aggressively confront
those who present direct challenges to existing patterns of inequality. It is
in just these situations that deadly conflict between citizens and police is all
too often actualized.
Model 3 appears consistent with this type of conceptualization of the
structured settings that frame police-citizen encounters. In Model 3 both
social demographic indicators (African American and Hispanic population
indicators) are strong, significant, and positive predictors of differences in
the log of expected counts of citizens killed, a result that is consistent with
previously stated hypotheses.
Strengths and limitations of this study
This study offers a number of benefits. The outcome data in this article
were developed using an open source investigation protocol (Campbell
et al., 2018) with individual cases drawn from published media reports of
police killings (Bernstein & Isackson, 2014). The data collection protocol
seeks to address a documented downward bias in official reporting while
presenting an opportunity to assess the phenomenon of police killings in a
new and independent way. Moreover, this effort reflects a shift in research
reliance on local and state police agencies, who may have incentives to be
less than fully transparent in their reporting. Research based on such data
sources has potential to contribute to an enhanced level of police account-
ability and also police legitimacy.
The outcome data do not differentiate between diverse sources of police-
induced deaths such as gunshot, asphyxiation, traffic accident, or other
causes while in police custody. We note that the overwhelming number of
cases of police-induced deaths arise from a police officer shooting
their firearm.
This study uses a unit of analysis that is less common for the study of
police killings since most research emphasizes cities with large urban police
departments. Osgood and Chambers (2000) note when talking about crime,
that the rural-urban dimension is a key feature of communities, and
research would benefit by extending theoretical insights across the full
range of this dimension. This article extends this idea to the study of police
18 R. HELMS AND S. E. COSTANZA
killings of citizens. Many FIPs have not been the focus of research due to
their geographic location falling outside central cities. This study uses a
broader unit of analysis that allows inclusion of these FIP cases.
Among the most noteworthy limitations in this study is that the study
uses one year of data on FIP’s. Ideally, a multiyear count would potentially
increase confidence in the stability of the indicator. Furthermore, we use a
comprehensive indicator of deadly violence incidents without distinguishing
whether police actions were justified or unjustified. A refinement in future
research would be to model these separately.
Conclusions
The results of this research provide empirical support for previously devel-
oped insights regarding lethal police aggression. After controlling for mul-
tiple indicators of police contact with criminal populations, this study finds
support for the racial threat argument. The size of Hispanic and African
American populations are both significant predictors of police killings. The
evidence also supports an economic-based hypothesis of police aggression.
These findings provide critical validation of much previous research that
uses official police data (Ozkan, Worrall & Zettler, 2018).
Police size and contact with criminal populations appears initially to
have important effects on FIP’s. While personnel size is not always a robust
predictor, several criminal population metrics are important predictors of
citizen deaths. Contacts generated around criminalized behaviors are often
controversial. Violent crime offenses and drug arrest rates are both predict-
ive of police killings of citizens. While violent offenses appear to predict
heightened risks for citizens, drug offenses show a very stable negative
result, which is noteworthy given the historical discourse demonizing drugs
and drug users, along with more recent liberalization of marijuana laws.
Racial minority indicators are predictive of police killings of citizens.
Historically, police have been criticized for both over-enforcement of laws
againstminoritiesandafailuretointerveneonbehalfofminoritygroups
(under-protection). Recent high-profile shootings of minorities and the imme-
diate social reactions of minority groups testify to the dangerous volatility per-
colating just beneath the surface of everyday encounters. Finally, economic
indicators were significant predictors of FIPs. Of note, the inclusion of eco-
nomic indicators in the models does not alter the result regarding race effects.
Discussion
Much of the literature criticizing police contact with minority populations
emphasizes police training as a key to reform. The foregoing analyses do
not so much dispute this focus but rather encourage a logical proposition
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 19
that runs somewhat counter to that prescription. Police are a scarce
resource. Police administrators allocate limited personnel strategically to
maximize efficient response to citizen calls for assistance. This places police
disproportionately in the most racially divided and economically distressed
areas of most communities (Weisburd, Davis & Gill, 2015).
Under conditions of heightened racial and economic tensions, challenges
to police authority become increasingly likely. Meanwhile, police defensive
efforts must expand in the attempt to contain intensifying risks. These
countervailing imperatives place police and citizens in a mutual environ-
ment of escalating risk. Police uncertainty and reactive aggression may be a
function of broad social and economic environments that contextualize
threats and draw both police and citizens into high risk exchanges.
One area of interest in police training focuses on implicit bias.
Empirically, the evidence is mixed regarding the belief that implicit bias is
a key factor affecting police violence (Oswald, Mitchell, Blanton, Jaccard &
Tetlock, 2013). In contrast to this focus on hidden personal biases, we sug-
gest a plausible alternative view, which is that police who serve a repressive
state function are tasked with preserving an existing social (racial) order
whose conditions are for many citizens untenable. In the most divided
environments, the tensions present in any standoff between police and citi-
zens has the potential to shift suddenly from a state of latent threat to actu-
arial violence and the risk of deadly force.
Economic inequality and frustrated economic success goals very likely
contribute to high risk encounters. The presence of younger working age
adults who are thwarted in their efforts to discover viable economic path-
ways may be a key factor underlying a broader social dynamic. This phe-
nomenon is one in which pressures multiply and concentrate in a downward
progression so that those most economically and socially marginalized, who
are all too often members of minority groups, are the least well-situated to
navigate the increasingly constrained opportunity structure. Police mean-
while are drawn into confrontations with members of the community in
their physical defense of a social order increasingly dominated by unequal
economic exchange processes (Jacobs & Helms, 1997).
An unwritten rule of policing is that police demand more respect than
they give. Since police officers are tasked with the actions of street level
enforcement, their perceptions of social antagonism and direct hostility
readily translate into an aggressive sensibility that tips the balance in favor
of force when tensions are rising. And while officers do not always pursue
the full logic of this sensibility, the potential to do so is always latent in
their encounters with citizens.
The findings from this research are cautionary; they do not align
squarely with a broadsided attack on police since the vast majority of police
20 R. HELMS AND S. E. COSTANZA
contacts do not result in police violence against citizens. The results also
do not easily align with calls for extensive police reforms. Moreover, they
do not imply that police reform efforts will be decisive in reducing these
types of events (for extensive discussion, see Vitale, 2017). To the extent
that any restructuring of police is implemented, this study encourages a
tempered expectation regarding the likelihood of success. With this precau-
tion in mind, we note that several plausible reforms are currently under
policy consideration with some being presently operational. We note that
many of today’s popular reform efforts are rooted in incident-level insights
and a police management focus of channeling police behavior rather than
focusing on gross inequalities evident in today’s complex society.
One hopeful reform involves issuing body cameras as standard gear for
all law enforcement officers in the field (Moore, 2016). When an official
account is challenged, it is often due to the emergence of film contradicting
the officer’s account. While empirical evidence is mixed regarding the
effects of body cameras (Ariel, Farrar & Sutherland, 2015), one benefit to
officers is that camera evidence has the potential to vindicate officer
accounts of actions taken in these violent exchanges.
The aforementioned work by Jennings and Rubado (2017) showed a sig-
nificant negative correlation between citizen deaths and required reporting
by police officers of gun-related behaviors. The written reporting require-
ment seems to have had a discouragement effect on police killings of citi-
zens. This type of research provides a very optimistic turn in regards to
policy remedies at the organizational level.
Skolnick and Fyfe (1993; Fyfe, 1997) noted that in evaluating police per-
formances in split-second situations, investigations should focus less on
outcomes of potentially violent situations and more on questions of
whether officers respond in ways likely to reduce the potential for violence.
This advice remains highly relevant to any inquiry into police violence
against citizens. Where officers violate their professional responsibilities, in
addition to remedial training, a viable model of criminal prosecution needs
to be established. As a corollary, a recommendation that would serve the
interest of strengthening community confidence in local enforcement is
that officer shootings should be investigated by a fully independent body
(i.e., representatives outside of local law enforcement or local prosecutors’
office personnel) since the conflict of interest at the local level is too strong
to guarantee an objective inquiry.
While reforms generally focus on police technology and management
considerations, researchers also have a continuing role to play in the assess-
ment of police violence. Future research should seek to assess neighbor-
hood level segregation of minorities as a feature in police killings. While
the foregoing data, with its focus on county level units, did not allow for a
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 21
full analysis of segregation effects, it is imperative for racial threat research-
ers to seek improved measures of neighborhood segregation in order to
more fully assess the impact of race on police killings of citizens.
A final recommendation is appropriately directed to elected officials and
their political appointees as they confront inequalities and injustices in
American society. It is imperative that political leadership deliver resource
support, not just with words but also with actions, to move in the direction
of a more socially just and economically viable social order. Alongside the
monumental tasks of addressing gross inequalities in the economy and
society generally, this means advancing a political agenda to redirect police
culture and practices so that questions regarding whose law and what order
(Chambliss & Mankoff, 1976) increasingly can be answered in ways that
reflect equal protection, equal enforcement, and justice as cornerstones in
any citizen’s response.
Notes
1. Various structural studies of police killings of citizens have been published (Jacobs &
O’Brien 1998; Holmes et al., 2019; Liska & Yu, 1992; Ross, 2015; Smith, 2003;
Sorensen, Marquart & Brock, 1993; Willits & Nowacki, 2014) using Supplemental
Homicide Report (SHR) data but each of these regardless of other redemptive qualities
is susceptible to the criticism that the SHR data are misleading due to their failure to
accurately report the number of police killings.
2. Under conditions of heightened inequality, agency directives may overlook acts of
police aggression since these are instrumental to the maintenance of order. This
tendency to overlook police aggression extends to internal affairs review processes and
direct supervisory oversight of police street level actions. This organizational
phenomenon contributes to a conceptual model privileging police discretionary
decisions about the use of lethal force in the field.
3. The SHR reflects limitations associated with non-reporting or underreporting of citizen
deaths (Loftin et al., 2003). The NVS series relies on medical certifiers to provide
information about cause and manner of death, whereas other case level information is
typically added by funeral directors. Certifiers are not required to identify whether a
killing was legally justifiable or specifically caused by contact with the police (Black,
2015; Bernstein & Isackson, 2014). According to Ross (2015) data on police homicides
compiled by these sources is incomplete under current collection protocols.
4. We note that police, unlike other public sector personnel (unemployment
specialists, social security bureaucrats, postal letter carriers, among others), engage
in street level activities that are by their very nature threatening. Police coercion is
implicated in every formal police contact with citizens, whether that contact is
focused on a trivial matter such as a broken tail light or a failure to signal before a
turn or is linked to a more serious issue such as questioning a citizen suspected of
a felony. Structural accounts often fail to articulate the basis for police violence that
starts with police coercion potential in everyday encounters. This is why it is
imperative to not lose sight of the link between incident studies and
structural research.
22 R. HELMS AND S. E. COSTANZA
5. This conceptualization is operationalized with a comprehensive measure of full-time
law enforcement officers with arrest authority from small towns (2500þpopulation) to
the largest central cities within each county along with all county level badged officers.
6. We acknowledge that police violence may be associated with this population age
cohort due to its heightened levels of participation in crime. This possibility is not
inconsistent with the conceptualization of young adults experiencing difficulties finding
their place in the stratified US economy.
7. We have no reason to believe that these counts are derived from a process that would
generate systematic bias in parameter estimation processes. Collecting more cases over
additional years is desirable but not crucial to the purposes of this research.
8. Inclusion of arrests for drug and disorder offenses is well-justified as a means of
isolating the effects of police contacts with criminalized populations in their various
dimensions. Much of policing is correctly characterized as being focused on
interactions with these marginal low-end criminal populations and as was noted
previously all police-citizen contacts present risks to both officers and citizens even if
most contacts do not result in violence. Also, Johnston’s(1984) remarks on the issue
of inclusion versus exclusion of additional control variables is highly relevant. He says
“it is more serious to omit relevant variables than to include irrelevant variables since
in the former case the coefficients will be biased, the disturbance variance
overestimated, and conventional inference procedures rendered invalid, while in the
latter case the coefficients will be unbiased, the disturbance variance properly
estimated, and the inference procedures properly estimated. This constitutes a fairly
strong case for including rather than excluding relevant variables in equations. There
is, however, a qualification. Adding extra variables, be they relevant or irrelevant, will
lower the precision of estimation of the relevant coefficients”(p. 262), so inclusive
empirical specifications typically will lead to more conservative significance tests.
References
Ariel, B., Farrar, W. A., & Sutherland, A. (2015). The effect of police body-worn cameras
on use of force and citizens’complaints against the police: A randomized controlled trial.
Journal of Quantitative Criminology,31(3), 509–535. doi:10.1007/s10940-014-9236-3
Banks, R. R., Eberhardt, J. L., & Ross, L. (2006). Discrimination and implicit bias in a
racially unequal society. California Law Review,94(4), 1169–1190. doi:10.2307/20439061
Bean, B. (2017). United States: The reason trump cheered cop violence. Green Left Weekly,
114(1148), 15.
Belvedere, K., Worrall, J. L., & Tibbetts, S. G. (2005). Explaining suspect resistance in
police-citizen encounters. Criminal Justice Review,30(1), 30–44. doi:10.1177/
0734016805275675
Bernard, T. J. (1990). Angry aggression among the “truly disadvantaged. Criminology,
28(1), 73–96. doi:10.1111/j.1745-9125.1990.tb01318.x
Bernstein, D., & Isackson, N. ( (2014). ). The truth about Chicago’s crime rates. Chicago
Magazine,7. Retrieved from https://www.chicagomag.com/Chicago-Magazine/May-2014/
Chicago-crime-rates/(retrieved Nov 2017).
Binder, A., & Scharf, P. (1982). Deadly force in law enforcement. Crime & Delinquency,
28(1), 1–23. doi:10.1177/001112878202800101
Black, C. (2015). How Chicago tried to cover up a police execution. The Chicago Reporter.
Retrieved November 2018 from https://www.chicagoreporter.com/how-chicago-tried-to-
cover-up-a-police-execution/
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 23
Blalock, H. M. Jr, (1967). Status inconsistency, social mobility, status integration and struc-
tural effects. American Sociological Review,32(5), 790–801.
Bock, M. A. (2016). Film the police! Cop-watching and its embodied narratives. Journal of
Communication,66(1), 13–34. doi:10.1111/jcom.12204
Campbell, B. A., Nix, J., & Maguire, E. R. (2018). Is the number of citizens fatally shot
bypolice increasing in the post-Ferguson era? Crime & Delinquency,64(3), 398–420. doi:
10.1177/0011128716686343
Chagnon, N., Chesney-Lind, M., & Johnson, D. T. (2018). Cops, lies, and videotape: Police
reform and the media in Hawaii. Crime, Media, Culture: An International Journal,14(2),
171–190. doi:10.1177/1741659016677328
Chambliss, W. J. & Mankoff, M. (Eds.) (1976). Whose law? What order? A conflict approach
to criminology. New York, NY: Wiley.
Chambliss, W. J., & Seidman, R. (1982). Law order and power (2nd ed.) Reading, MA.
Addison-Wesley Publishing.
Cheh, M. M. (1996). Are lawsuits an answer to police brutality? In W. A. Geller and H.
Toch, eds., Police violence: Understanding and controlling police abuse of force.
NewHaven, CT: Yale University Press.
Cline, A. (2017). Poverty, race, and police: A conversation. Poverty & Public Policy,9(1),
96–102. doi:10.1002/pop4.166
Cloward, R. A. (1959). Illegitimate means, anomie, and deviant behavior. American
Sociological Review,24(2), 164–176. doi:10.2307/2089427
Costanza, S. E., & Helms, R. (2012). Street gangs and aggregate homicides: An analysis
ofeffects during the 1990s violent crime peak. Homicide Studies,16(3), 280–307. doi:10.
1177/1088767912449623
Crank, J. P. (2003). Institutional theory of police: A review of the state of the art. Policing:
An International Journal of Police Strategies & Management,26(2), 186–207. doi:10.1108/
13639510310475723
Crawford, C., & Burns, R. (1998). Predictors of the police use of force: The application of.
Police Quarterly,1(4), 41–63. doi:10.1177/109861119800100403
Culhane, S. E., Boman, J. H., IV,., & Schweitzer, K. (2016). Public perceptions of the justifi-
ability of police shootings: The role of body cameras in a pre-and post-Ferguson experi-
ment. Police Quarterly,19(3), 251–274. doi:10.1177/1098611116651403
Donner, C. M., Maskaly, J., Piquero, A. R., & Jennings, W. G. (2017). Quick on the draw:
Assessing the relationship between low self-control and officer-involved police shootings.
Police Quarterly,20(2), 213–234. doi:10.1177/1098611116688066
Dunham, R. G., & Petersen, N. (2017). Making black lives matter. Criminology & Public
Policy,16(1), 341–348. doi:10.1111/1745-9133.12284
Edwards, F., Esposito, M. H., & Lee, H. (2018). Risk of police-involved death by race/ethni-
city and place, United States, 2012-2018. American journal of public health,108(9),
1241–1248.
Eitle, D., & Monahan, S. (2009). Revisiting the racial threat thesis: The role of policeorgani-
zational characteristics in predicting race-specific drug arrest rates. Justice Quarterly,
26(3), 528–561. doi:10.1080/07418820802427817
Eitle, D., D’Alessio, S. J., & Stolzenberg, L. (2014). The effect of organizational andenviron-
mental factors on police misconduct. Police Quarterly,17(2), 103–126. doi:10.1177/
1098611114522042
Embrick, D. G. (2015). Two Nations, Revisited: The Lynching of Black and Brown Bodies,
Police Brutality, and Racial Control in ‘Post-Racial’Amerikkka. Critical Sociology,41(6),
835–843. doi:10.1177/0896920515591950
24 R. HELMS AND S. E. COSTANZA
Engel, R. S., Sobol, J. J., & Worden, R. E. (2000). Further exploration of the demeanorhypo-
thesis: The interaction effects of suspects’characteristics and demeanor on policebehav-
ior. Justice Quarterly,17(2), 235–258. doi:10.1080/07418820000096311
Ericson, R. V. (1982). Reproducing order: A study of police patrol work (Vol. 5). Toronto:
University of Toronto Press.
Feldman, J. M., Gruskin, S., Coull, B. A., & Krieger, N. (2017). Killed by police: Validity of
media-based data and misclassification of death certificates in Massachusetts, 2004 2016.
American Journal of Public Health,107(10), 1624–1626. doi:10.2105/AJPH.2017.303940
Fridell, L. A., & Binder, A. (1992). Police officer decision making in potentially violent con-
frontations. Journal of Criminal Justice,20(5), 385–399. doi:10.1016/0047-2352(92)90075-
K
Fridell, L., & Lim, H. (2016). Assessing the racial aspects of police force using the implicit-
and counter-bias perspectives. Journal of Criminal Justice,44,36–48. doi:10.1016/j.jcrim-
jus.2015.12.001
Fyfe, J. J. (1997). Good policing. In: Dunham R, Alpert GP (Eds.), Critical issues in
policing: contemporary readings, 5th edn. Waveland, Prospect Heights, pp. 194–213.
Garner, J. H., Maxwell, C. D., & Heraux, C. G. (2002). Characteristics associated with the-
prevalence and severity of force used by the police. Justice Quarterly,19(4), 705–746.
doi:10.1080/07418820200095401
Ha, O. K., & Andresen, M. A. (2017). Unemployment and the specialization of criminal
activity: A neighborhood analysis. Journal of Criminal Justice,48,1–8. doi:10.1016/j.
jcrimjus.2016.11.001
Helms, R. (2007). The impact of political context on local law enforcement resourcing:
Ananalysis of deputy employment rates in US counties. Policing and Society,17(2),
182–206. doi:10.1080/10439460701302743
Helms, R. (2008). Locally elected sheriffs and money compensation: A quantitative analysis
of organizational and environmental contingency explanations. Criminal Justice Review,
33(1), 5–28. doi:10.1177/0734016808315588
Helms, R., & Costanza, S. E. (2009). Race, politics, and drug law enforcement: An analysis
of civil asset forfeiture patterns across US counties. Policing and Society,19(1), 1–19. doi:
10.1080/10439460802457578
Helms, R., & Costanza, S. E. (2010). Modeling the politics of punishment: A contextual
analysisof racial disparity in drug sentencing. Criminal Justice Review,35(4), 472–491.
doi:10.1177/0734016810373114
Hilbe, J. M. (2011). Negative binomial regression. Cambridge England: Cambridge
University Press.
Hirschfield, P. J., & Simon, D. (2010). Legitimating police violence: Newspaper narratives
of deadly force. Theoretical Criminology,14(2), 155–182. doi:10.1177/1362480609351545
Hirschfield, P. J., & Simon, D. (2010). Legitimating police violence: Newspaper narratives
ofdeadly force. Theoretical Criminology,14(2), 155–182. doi:10.1177/1362480609351545
Holmes, M. D., & Smith, B. W. (2018). Social-psychological dynamics of police-minority
relations: An evolutionary interpretation. Journal of Criminal Justice,59,58–68. doi:10.
1016/j.jcrimjus.2017.05.003
Holmes, M. D., Painter, M. A., & Smith, B. W. (2019). Race, place, and police-caused hom-
icidein US municipalities. Justice Quarterly,36(5), 751–786. doi:10.1080/07418825.2018.
1427782
Holmes, M. D., Smith, B. W., Freng, A. B., & Mu~
noz, E. A. (2008). Minority threat, crime
control, and police resource allocation in the Southwestern United States. Crime &
Delinquency,54(1), 128–152. doi:10.1177/0011128707309718
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 25
Holmes, M. D., Smith, B. W., Freng, A. B., & Mu~
noz, E. A. (2008). Minority threat, crime-
control, and police resource allocation in the Southwestern United States. Crime &
Delinquency,54(1), 128–152. doi:10.1177/0011128707309718
Hyland, S., Langton, L., & Davis, E. (2015). Police use of nonfatal force, 2002–2011.US
Department of Justice.
Jacobs, D. (1979). Inequality and police strength: Conflict theory and coercive control in
metropolitan areas. American Sociological Review,44(6), 913–925. doi:10.2307/2094716
Jacobs, D., & Helms, R. (1997). Testing coercive explanations for order:The determinants
of law enforcement strength over time. Social Forces,75(4), 1361–1392. Vol. doi:10.1093/
sf/75.4.1361
Jacobs, D., & O’Brien, R. M. (1998). The determinants of deadly force: A structural analysi-
sof police violence. American Journal of Sociology,103(4), 837–862. doi:10.1086/231291
Jennings, J. T., & Rubado, M. E. (2017). Preventing the use of deadly force: The relation-
shipbetween police agency policies and rates of officer-involved gun deaths. Public
Administration Review,77(2), 217–226. doi:10.1111/puar.12738
Johnson, O., Jr, St Vil, C., Gilbert, K. L., Goodman, M., & Johnson, C. A. (2018). How
neighborhoods matter in fatal interactions between police and men of color. Social
Science & Medicine, doi:10.1016/j.socscimed.2018.11.024
Johnson, R. R. (2016). Does the implicit bias test predict discriminatory behavior?
DolanConsulting Group.
Johnston, J. (1984). Econometric methods (3rd ed.). New York, NY: McGraw-Hill.
Jones-Eversley, S., Adedoyin, A. C., Robinson, M. A., & Moore, S. E. (2017). Protesting
Blackinequality: A commentary on the civil rights movement and Black lives matter.
Journal of Community Practice,25(3/4), 309–324. doi:10.1080/10705422.2017.1367343
Kania, R. R., & Mackey, W. C. (1977). Police violence as a function of community charac-
teristics. Criminology,15(1), 27–48. doi:10.1111/j.1745-9125.1977.tb00047.x
Kent, S. L., & Jacobs, D. (2005). Minority threat and police strength from 1980 to 2000: A
fixed-effects analysis of nonlinear and interactive effects in large US cities. Criminology,
43(3), 731–760. doi:10.1111/j.0011-1348.2005.00022.x
Klinger, D. A. (2012). On the problems and promise of research on lethal police violence:
A research note. Homicide Studies,16(1), 78–96. doi:10.1177/1088767911430861
Klinger, D., Rosenfeld, R., Isom, D., & Deckard, M. (2016). Race, crime, and the micro –
ecology of deadly force. Criminology & Public Policy,15(1), 193–222. doi:10.1111/1745-
9133.12174
Lee, A. S., Weitzer, R., & Martinez, D. E. (2018). Recent police killings in the United States: A
three-city comparison. Police Quarterly,21(2), 196–222. doi:10.1177/1098611117744508
Lewis, P. G., & Ramakrishnan, S. K. (2007). Police practices in immigrant-destination cities
political control or bureaucratic professionalism? Urban Affairs Review,42(6), 874–900.
doi:10.1177/1078087407300752
Lichter, D. T., Parisi, D., & Taquino, M. C. (2015). Toward a new macro-segregation?
Decomposing segregation within and between metropolitan cities and suburbs. American
Sociological Review,80(4), 843–873.
Lipsky, M. (1980). Street-level bureaucracy: Dilemmas of the individual in public services.
New York: Russell Sage.
Liska, A. E., & Yu, J. (1992). Specifying and testing the threat hypothesis: Police use of
deadly force. Social Threat and Social Control,53–68.
Loftin, C., Wiersema, B., McDowall, D., & Dobrin, A. (2003). Underreporting of justifiable
homicides committed by police officers in the United States, 1976–1998. American
Journal of Public Health,93(7), 1117–1121. doi:10.2105/AJPH.93.7.1117
26 R. HELMS AND S. E. COSTANZA
Lowery, W. (2016). They can’t kill us all: Ferguson, Baltimore, and a new era in America’s
Racial Justice Movement. UK. Hachette Publishing.
McCarty, W. P., Ren, L., & Zhao, J. S. (2012). Determinants of police strength in large US
cities during the 1990s: A fixed-effects panel analysis. Crime & Delinquency,58(3),
397–424. doi:10.1177/0011128709336942
Merton, R. K. (1938). Social structure and anomie. American Sociological Review,3(5),
672–682. doi:10.2307/2084686
Messner, S. F., & Rosenfeld, R. (2012). Crime and the American dream. Belmont, CA.
Cengage Learning.
Moore, M. A. (2016). The Next Stage of Police Accountability: Launching a Police Body-
Worn Camera Program in Washington DC. Seattle Journal for Social Justice,14(1),
11–16.
Nix, J., Campbell, B. A., Byers, E. H., & Alpert, G. P. (2017). A bird’s eye view of civilian-
skilled by police in 2015: Further evidence of implicit bias. Criminology & Public Policy,
16(1), 309–340. doi:10.1111/1745-9133.12269
Nowacki, J. S. (2015). Organizational-level police discretion: An application for police use
oflethal force. Crime & Delinquency,61(5), 643–668. doi:10.1177/0011128711421857
Osgood, D. W., & Chambers, J. M. (2000). Social disorganization outside the metropolis:
Ananalysis of rural youth violence. Criminology,38(1), 81–116. doi:10.1111/j.1745-9125.
2000.tb00884.x
Oswald, F., Mitchell, G., Blanton, H., Jaccard, J., & Tetlock, P. (2013). Predicting ethnic
andracial discrimination: A meta-analysis of IAT criterion studies. Journal of Personality
andSocial Psychology Studies,105(2), 171–192.
Ozkan, T., Worrall, J. L., & Zettler, H. (2018). Validating media-driven and crowdsourced
policeshooting data: A research note. Journal of Crime and Justice,41(3), 334–345. doi:
10.1080/0735648X.2017.1326831
Paoline, E. A., III, Gau, J. M., & Terrill, W. (2016). Race and the police use of force
encounter inthe United States. The British Journal of Criminology,58(1), 54–74.
Petrocelli, M., Piquero, A. R., & Smith, M. R. (2003). Conflict theory and racial profiling:
Anempirical analysis of police traffic stop data. Journal of Criminal Justice,31(1), 1–11.
doi:10.1016/S0047-2352(02)00195-2
Raine, W. J. (1967). Los Angeles riot study: The perception of police brutality in South
Central Los Angeles (Vol. 109). Institute of Government and Public Affairs, UCLA
Reiss, A. J. (1968). Police brutality-answers to key questions. Society,5(8), 10–19. doi:10.
1007/BF02804717
Rogers, A. (2015). How police brutality harms mothers: Linking police violence to there-
productive justice movement. Hastings Race & Poverty Law Journal,12(2), 205–235.
Ross, C. 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. doi:10.1371/
journal.pone.0141854
Schlesinger, T., & Tonry, M. (2013). Punishing race: A continuing American dilemma.
Oxford: OxfordUniversity Press.
Segura, G. M., & Rodrigues, H. A. (2006). Comparative ethnic politics in the United States:
Beyond black and white. Annual Review of Political Science,9(1), 375–395. doi:10.1146/
annurev.polisci.9.072204.175806
Sherman, L. W., & Weisburd, D. (1995). General deterrent effects of police patrol in crime
“hotspots”: A randomized, controlled trial. Justice Quarterly,12(4), 625–648. doi:10.1080/
07418829500096221
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 27
Skogan, W. G. (2014). Using community surveys to study policing. In The Oxford hand-
book of police and policing (pp. 449–470). Oxford, England: Oxford University Press.
Skolnick, J. H., & Fyfe, J. J. (1993). Above the law: Police and the excessive use of
forceNewYork: Free Press.
Smith, B. W. (2003). The impact of police officer diversity on police-caused homicides.
Policy Studies Journal,31(2), 147–162. doi:10.1111/1541-0072.t01-1-00009
Smith, B. W., & Holmes, M. D. (2014). Police use of excessive force in minority commun-
ities:A test of the minority threat, place, and community accountability hypotheses.
Social Problems,61(1), 83–104.
Sorensen, J. R., Marquart, J. W., & Brock, D. E. (1993). Factors related to killings of felons
bypolice officers: A test of the community violence and conflict hypotheses. Justice
Quarterly,10(3), 417–440. doi:10.1080/07418829300091911
Stamper, N. (2016). To protect and serve: How to fix America’s police. New York, NY:
Nation Books.
Stucky, T. D. (2005). Local politics and police strength. Justice Quarterly,22(2), 139–169.
doi:10.1080/07418820500088739
Tedeschi, J. T., & Felson, R. B. (1994). Violence, aggression, and coercive actions.
Washington,DC: American Psychological Association.
Terrill, W. (2003). Police use of force and suspect resistance: The micro process of the
police suspect encounter. Police Quarterly,6(1), 51–83. doi:10.1177/1098611102250584
Terrill, W., & Paoline, E. A. III, (2013). Examining less lethal force policy and the force
continuum: Results from a national use-of-force study. Police Quarterly,16(1), 38–65.
doi:10.1177/1098611112451262
Terrill, W., & Reisig, M. D. (2003). Neighborhood context and police use of force. Journal
of Research in Crime and Delinquency,40(3), 291–321. doi:10.1177/0022427803253800
Tonry, M. H (2011). Punishing race: A continuing American dilemma.Oxford,United
Kingdom, Oxford University Press.
Tyler, T. R., Fagan, J., & Geller, A. (2014). Street stops and police legitimacy:
Teachablemoments in young urban men’s legal socialization. Journal of Empirical Legal
Studies,11(4), 751–785. doi:10.1111/jels.12055
UCLA: Statistical Consulting Group (2017). Negative binomial regression jstatadataanalysis
examples.Retrievdfromhttp://stats.idre.ucla.edu/stata/dae/negative-binomial-regression/
Vitale, A. S. (2017). The end of policing. London: Verso.
Weisburd, D., Davis, M., & Gill, C. (2015). Increasing collective efficacy and social capital
atcrime hot spots: New crime control tools for police. Policing,9(3), 265–274. doi:10.
1093/police/pav019
Weitzer, R. (2002). Incidents of police misconduct and public opinion. Journal of Criminal
Justice,30(5), 397–408. doi:10.1016/S0047-2352(02)00150-2
Weitzer, R. (2015). American policing under fire: Misconduct and reform. Society,52(5),
475–480. doi:10.1007/s12115-015-9931-1
Westley, W. A. (1953). Violence and the police. American Journal of Sociology,59(1),
34–41. doi:10.1086/221264
Westley, W. A. (1970). Violence and the police: A sociological study of law, custom, and
morality (Vol. 28). Cambridge, MA: MIT Press.
White, M. D. (2002). Controlling police decisions to use deadly force: Reexamining theim-
portance of administrative policy. Crime & Delinquency,47(1), 131–151. doi:10.1177/
0011128701047001006
28 R. HELMS AND S. E. COSTANZA
Willits, D. W., & Nowacki, J. S. (2014). Police organisation and deadly force: An examin-
ation ofvariation across large and small cities. Policing and Society,24(1), 63–80. doi:10.
1080/10439463.2013.784314
Wilson, J. Q. (1968). Dilemmas of police administration. Public Administration Review,
28(5), 407–417. doi:10.2307/973756
JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 29
