Improving Police Use of Force: A Policy Essay on National Data Collection

Abstract

Documenting police use of force has been an issue in the United States since at least 1931. As of July 2016, there is still no standardized national data collection effort, despite a call from several presidential and civil rights commissions to do so. Without accurate and timely national data, a moral panic of sorts unfolds that replaces rational thought and debate necessary to enact public policy. Moreover, without such data, it is virtually impossible to estimate the incidence and prevalence of police use of force, which leaves U.S. law enforcement agencies at a tremendous disadvantage for improving practices. This essay briefly examines the history of calls to improve police practices through collecting national use of force data and then offers a practical solution based on rational-technical theory of organizations with a brief analysis of a new promising, but limited, data set. The essay concludes with a proposed research agenda should national data become available through pending legislation H.R. 306, National Statistics on Deadly Force Transparency Act of 2015.

Keywords

police use of force moral panic rational-technical theory evidence-based practice

Introduction

It has been said that

Measurement is the first step that leads to control and eventually to improvement. If you can’t measure something, you can’t understand it. If you can’t understand it, you can’t control it. If you can’t control it, you can’t improve it. (Attributed to U.S. businessman H. James Harrington)

If we wish to improve police–citizen interaction as it relates to use of force, then we must have accurate and timely data about its circumstances. As of July 2016, we still do not have a good national understanding of police use of force because we do not have good measures (Schmidt, 2015). U.S. policing is a fragmented and decentralized system with approximately 18,000 law enforcement agencies (Reaves, 2008) operating and reporting independently. Most of what is known about use of force is the result of sporadic research funded by the National Institute of Justice, the Bureau of Justice Statistics, blue-ribbon investigative commissions, and the research agendas of a few scholars involving single cities or a few cities, which is not necessarily informative for exercising control. The lack of standardized data has led to a fundamental public misunderstanding about police use of force, which is the defining feature of the police role (Bittner, 1970). Virtually every public discussion on police use of force is devoid of context, except for race and sex. Race in particular is socially inflammatory and, as a stand-alone variable in the equation, evokes intolerable perceptions of “racism,” “discrimination,” and “disparate treatment,” yet it is unqualified and never tells the full story. Consequently, emotional arguments, filled with red herrings, ad hominem attacks, and non sequiturs, replace logic and reasoning that is grounded in fact. A handful of anecdotes and isolated use of force incidents do not amount to a pattern or accepted past practice that necessitates wholesale change to public policy; although anecdotes may suggest an emerging problem and each individual incident must be dealt with on its own merits to ensure government accountability, they do not signal a “crisis.”

Before we educate others on police use of force, we must educate ourselves (International Association of Chiefs of Police, 2012). Educating ourselves means basing our positions on sound empirical data instead of vacuous arguments filled with anecdotes and speculation. The intent behind national use of force data collection is to understand what it looks like and how it occurs, which can lead to its control. Control and management may reduce the incidence of use of force as well as prevent organizational accidents (Doyle, 2010) by capturing, analyzing, and reporting on streams of data to ensure a complete picture emerges about trends and patterns across the nation that can inform local policies and practices.

To that end, this article approaches the issue from a knowledge perspective, one where use of force data and information should become an embedded part of police management and be used to routinely examine policies and practices relative to legal principles and community sentiment as a matter of performance (Shane, 2010). This essay briefly reviews previous calls to collect national use of force data and the limitations of existing data, and analyzes a new data set as an extension of microcriminology and rational-technical theory of organizations. The essay concludes with a proposed research agenda should national data become available through H.R. 306, National Statistics on Deadly Force Transparency Act of 2015 (pending as of July 1, 2016).

Statement of the Problem

Background

Since at least 1931, there has been extensive political, social, and academic interest in collecting and analyzing police use of force data in the United States. The National Commission on Law Observance and Enforcement (Wickersham Commission, 1931) concluded that there was considerable evidence of police brutality, particularly during the prohibition era. The Commission did not clearly define “brutality,” except to say “[t]he third degree, that is, the use of physical brutality, or other forms of cruelty, to obtain involuntary confessions or admissions is widespread” (Wickersham Commission, 1931, p. 4). Police leaders vociferously denounced the report and denied the conclusions, and the Commission only briefly and vaguely discussed possible corrective actions, which may have been a political compromise. In 1947, the President’s Committee on Civil Rights noted similar findings but did acknowledge that “. . . many law enforcement agencies have gone far in recent years toward stamping out these evils” (p. 25).

Again, in 1961, the U.S. Commission on Civil Rights found that “police brutality is still a serious problem throughout the United States” (The 50 States Report, 1961, p. 687) but did not have statistical data to support the nature and extent of the allegations, or identify patterns. Following the tumultuous civil rights movement of the 1950s and 1960s, the President’s Commission on Law Enforcement and Administration of Justice (1967), Task Force Report: The Police, concluded that physical abuse by the police was not as serious a problem as it was in the past but also noted their findings were based on a “. . . relatively small number of complaints [that] cannot be considered an accurate measure of the total problem” (p. 181). In 1981, the U.S. Commission on Civil Rights said the Federal Bureau of Investigation (FBI) should collect and make publicly available the data on police assaults and shootings. This recommendation was not acted upon to any meaningful degree, but the value of empirical data was apparent in their declaration. Finally, in 2015, the President’s Task Force on 21st Century Policing recommended, “Policies on use of force should also require agencies to collect, maintain, and report data to the Federal Government on all officer-involved shootings, whether fatal or nonfatal, as well as any in-custody death” (p. 21). The results of this recommendation have not materialized as of July 2016.

Public opinion based on isolated episodes or small samples tends to be personal, idiosyncratic, or unique, which may represent the best (or worse) case of its type. Individuals tend to recall dramatic, media-worthy cases over the mundane, and humans are inclined to place a higher value on remarkable, personal stories to create causal connections. Repeat these stories enough times, present them through mass media, and a high-stakes game of “telephone” is under way, where errors in the details accumulate in each retelling, so the original details vary drastically from the facts. Explaining the wide divergence between fact and fiction may result from anxiousness or impatience, erroneous corrections, deliberate omissions, and deliberate inclusions to ensure a specific message is received. This is one reason why the scientific method emerged as a better process of observation than personal stories, an appeal to tradition, or Church dogma. But, public policy demands a more complete picture that necessitates going beyond a handful of unique or exceptional cases to get the greater context that a large representative sample provides. That complete picture is found in incident-level, also known as unit-record, national data.

This is not to say anecdotal evidence or unique stories have no place in legal or policy debate; rather, empirical data are the counterweight to individual experiences, and such experiences should be used sparingly and with great caution to help society make value judgments while the science drives estimates on incidence and prevalence (Moore, 2002). Unfortunately, individual experiences on police use of force have supplanted scientific evidence because the United States has failed to properly collect empirical data. That we already have the data from the FBI, the Centers for Disease Control and Prevention (CDC), or a few isolated incidents from across the country represent the national picture of police use of force is absolutely incorrect and does not justify wholesale policy change. As we reflect critically on policing history between the Wickersham Commission (1931) and the President’s Task Force (2015), one thing is clear: For all of the improvements in U.S. policing, there still exists a tremendous knowledge gap that fuels the rancor and discord surrounding police use of force.

What You Do Not Know May Harm You: What Happens When There Is Limited, Incomplete, or No Data Available

There is limited research, incomplete data sets, and no national data, which limits our knowledge on the risks and effectiveness of use of force decisions. There is no research to predict the type of incidents that pose the greatest risk of use of force, or what the odds are of being shot compared with less-lethal force options. The lack of data makes it difficult to definitively establish how prevalent use of force may be, or to what extent the problem exists. Incomplete or no data limit the ability of police leaders to speak cogently about use of force patterns and trends within their own agency or region. Without data, police leaders do not have a mechanism to train their officers on emerging trends in assaults and risky confrontations, or strengthen training to avoid confrontations and accidents in a unified manner like the U.S. military’s War Colleges approach leadership. In short, policing absorbs a great deal of uncertainty by failing to collect use of force data.

The problem for the public is that they are denied the benefit of knowing how their police agency compares with similar police agencies on use of force incidents and how the specific circumstances and environment may have contributed to the event. The public will close that knowledge gap with emotional arguments that mirror a moral panic of sorts resulting from exaggerated claims and fictitious information. Cohen (1972) noted how this occurs (with my annotations):

A condition, episode, person or group of persons emerges to become defined as a threat to societal values and interests [a use of force incident by police occurs]; its nature is presented in a stylized and stereotypical fashion by the mass media [unaccountable, militarized, racist police officers]; the moral barricades are manned by editors, bishops, politicians and other right-thinking people [“moral entrepreneurs” emerge in community activists, politicians, and religious leaders]; socially accredited experts pronounce their diagnoses and solutions [police are an occupying army; the chief must resign] ways of coping are evolved or (more often) resorted to [minority parents have the “police talk” with their kids, Whitaker & Snell, 2016]; the condition then disappears, submerges or deteriorates and becomes visible [social movements arise]. (p. 1)

The features of a moral panic are present in several notable police use of force cases, Oscar Grant (Oakland, California, 2009), Eric Garner (Staten Island, New York, 2014), Michael Brown (Ferguson, Missouri, 2014), and Freddie Gray (Baltimore, Maryland, 2015), which gave rise to the Black Lives Matter and the “Hands UP . . . Don’t Shoot” movements.

Limitations of Existing Use of Force Data

The use of force data collected by the FBI Uniform Crime Report titled “Justifiable Homicide” and the CDC National Violent Death Reporting System are incomplete and rely on aggregate data. The U.S. Bureau of Justice Statistics readily acknowledges,

Justifiable homicides by police for an entire State are sometimes missing from the [FBI] SHR [supplementary homicide reports] database . . . The opposite problem—too many rather than too few records of justifiable homicide by police in the database—also exists. (J. M. Brown & Langan, 2001, p. 29; see also Maxfield, 1989)

Furthermore, data measured at the aggregate level and the statistics that flow from them summarize a set of observations that communicate the largest amount of information as simply as possible, but they do not measure the fine details; for example, the data collected as part of 42 U.S.C. § 14142, Use of Excessive Force, provides a succinct overview but does not contain any contextual details (Hickman, 2006). The aggregation problem is defined as the “information loss which occurs in the substitution of aggregate, or macro-level, data for individual, or micro-level, data” (Clark & Avery, 1976, p. 428), which limits their utility, for example:

Aggregate data can summarize average characteristics for a group, but they cannot assume those characteristics apply to every member of the group. It is inappropriate to assume that relationships at the aggregate level will also hold at the individual level. It is entirely possible to find a relationship at the national level that does not hold at the local level. The reverse is also true, which is reaching a group conclusion (e.g., national-level picture) based on exceptional cases (e.g., individual local cases). Both of these errors point to traps in everyday reasoning, which leads to stereotyping and hasty judgments. Society must determine empirically what individual incidents look like, not just rely on group averages.

Aggregate data typically do not allow easy manipulation of variables, so there may be a tendency to overlook influences that may affect individual incidents. By relying on aggregate data, we lose the opportunity to use the data to generate additional hypotheses, and it makes testing theory difficult.

Aggregate data will show averages but not individual attributes. Summary data mask incident-level details that are necessary to identify trends and patterns, as well as the factors that may be correlated with use of force across similar cities and various contexts, such as age, sex, race of the officer and the offender, level of offender resistance, officer assignment, years of service, neighborhood composition, offender resistance, type of incident, organizational composition, and geographical location. In short, context is absolutely essential but is concealed by aggregate data—a shortcoming that is, at times, exploited by those who use such data to support an oversimplified or emotional argument.

The limitations of the current data system are well known among police researchers dating to at least 1979 (Sherman & Langworthy, 1979; see also Fyfe, 1988; Police Executive Research Forum [PERF], 2016) and the empirical interest in measuring police use of force dates to at least 1963 (Robin, 1963; see also Harding & Fahey, 1973; Jacobs & Britt, 1979; Kobler, 1975).

Theoretical Framework for National Data Collection

Capturing police use of force data at the incident level is an extension of microcriminology, which is concerned with small units of analysis and processes. The incident and all of its contextual attributes can illuminate where opportunities for use of force arise as three important factors converge: police officers, offenders, and the environment. In this sense, incident-level use of force data parallel micro-level crime data that provide a framework for understanding how humans and their environment interact to create opportunities for using force (Clarke, 1997; Groff, Weisburd & Yang, 2010). Although a number of theories exist to explain why crime occurs (e.g., Cullen & Agnew, 2006), those that focus on proximate causes instead of distant, tenuous causes best inform crime prevention—or, in this case, preventing use of force. Collecting and analyzing incident-level data examines proximate causes, which focuses on the immediate situational context by analyzing the opportunities that give rise to the incident, then seeking measures to block those opportunities. The environmental approach views the logistics of an incident (how) as a more important practical consideration than motivation (why). When viewed from this perspective, it is easier to control the convergence setting through strategies and tactics that increase the effort to attack an officer, increase the risk of apprehension to an offender who launches an attack, or reduce the provocations for an attack through de-escalation techniques.

To accomplish this, U.S. law enforcement needs a more flexible, in-depth data system similar to the National Incident-Based Reporting System (NIBRS). Where FBI Uniform Crime Report (UCR) is an aggregate count of crime, NIBRS presents incident-level information about crime and its circumstances. Micro-level data are inherently flexible and can unmask relationships between victims, offenders, locations, and other incident-level details. NIBRS data are superior to UCR data for inferential purposes and have been widely used to uncover relationships and patterns involving sexual assault of young children (Snyder, 2010), prostitution of juveniles (Finkelhor & Ormrod, 2004b), predictors of homicide clearance (Roberts, 2007), kidnapping of juveniles (Finkelhor & Ormrod, 2000), child pornography (Finkelhor & Ormrod, 2004a), and intimate partner violence (Vazquez, Stohr, & Purkiss, 2005), among the many. A similar system could link detailed incident-level information, so we can understand and control the sequence of events that lead to police use of force—that is, the who, what, where, when, how, and possibly why of use of force transactions.

The impetus for law enforcement agencies to track and report use of force transactions is the community mandate that these agencies redefine their business methods in more rational ways. Rational in this sense is aimed at measuring aspects of the agency’s business process such as input, output, intermediate objectives, outcome, strategies, work processes, and resources, all of which combine to potentially improve performance. The rational-technical theory of organizations implies the organization behaves and is structured in a manner designed to optimize efficiency and effectiveness by adopting rational, evidence-based practices that are best suited for achieving technical goals (e.g., controlling use of force; Blau & Schoenherr, 1971; Thompson, 1967). This is contrasted against institutional theory, where the organization responds to the cultural features about how the organization should be structured and is rooted in politics, perceptions, symbolism, tradition, and cultural beliefs. Police agencies operating in the institutionalized environment are not likely to invest the necessary resources in evaluating its performance, making police use of force—and the record-keeping systems therein—appear to be a poster child for institutional theory. The institutional environment operates independently of what the rational-technical environment suggests, and the two perspectives are not necessarily congruent (Crank, 2003). If use of force is a technical aspect of police performance (and one could hardly argue against it), then incident-level data are the counterweight to institutionalism to help stand up a rational management framework.

A Brief Analysis of Promising (but Limited) Data

In an attempt to close the gap in deficient use of force data at the national level, the Washington Post (“Fatal Force Database,” 2016) has undertaken a project to collect publicly available reports on use of force incidents that result in death from news reports, law enforcement websites, social media, and independent databases such as Killed by Police and Fatal Encounters. Although the data are collected at the incident level, which is an improvement over aggregate data, it suffers from issues raised by Jacob (1984) concerning published data, which may lead to problems drawing accurate conclusions. Other limitations are as follows: (a) All of the cases involve fatalities,¹ so comparing nonfatal shootings is not possible; (b) only deadly force is captured, so comparing less-lethal force options is not possible; (c) no data on the officer’s characteristics, which limits demographic comparisons; (d) no data on environmental characteristics, which limits controlling for the tactics, approach, crime type, offender resistance, and the immediate situation; (e) no data on organizational composition, so it is not possible to identify agency correlates; (f) the data on offenders and the situation are very limited; and (g) the data do not differentiate intentional and unintentional shootings. Nevertheless, the data do provide some insight into this important social and political issue and represent the type of data that could be used routinely by police management.

Table 1 shows the descriptive statistics, and Table 2 shows a chi-square test of independence between select variables and the manner of death. The unit of analysis is the shooting incident. All data are binary coded for the presence (1) or absence (0) of the condition except weapon type, which is coded 1 to 4, and age (in the regression model), which is continuous. The database is rolling, which means that it is updated periodically as police shootings occur; the data for this analysis are from January 2, 2015 to April 4, 2016 (n = 1,254). Every state and the District of Columbia are represented, where 54.5% of the fatal encounters (n = 683) are from 20% of the states; 827 cities are represented, of which 166 cities (20%) represent 46.6% of the incidents (n = 584). The top 11 cities account for 10.1% of the incidents (n = 127).

Table 1.

Descriptive Statistics for the Washington Post Use of Force Data.

Variables and metrics	n	%	M	SD
Dependent
Manner of death	1,254	100.0	—	—
Shot (1)	1,176	93.8	—	—
Shot and tasered (0)	78	6.2	—	—
Predictor
Age (continuous)	1,233	—	36.5	12.8
Age (grouped)	1,233	100.0	—	—
<35 (1)	620	50.3	—	—
>35 (0)	613	49.7	—	—
Race	1,197	100.0	—	—
White (1)	620	51.8	—	—
Non-White (0)	577	48.2	—	—
Race	937	100.0	—	—
White (1)	620	66.2	—	—
Black (0)	317	33.8	—	—
Sex	1,254	100.0	—	—
Male (1)	1,197	95.5	—	—
Female (0)	57	4.5	—	—
Signs of mental illness	1,254	100.0	—	—
Yes (1)	323	25.8	—	—
No (0)	931	74.2	—	—
Control
Weekend only	1,254	100.0
Weekend (1)	353	28.1	—	—
Weekday (0)	901	71.9	—	—
Officer body camera	1,254	100.0	—	—
Yes (1)	100	8.0	—	—
No (1)	1,154	92.0	—	—
Offender was fleeing	1,211	100.0	—	—
Yes (1)	322	26.6	—	—
No (0)	889	73.4	—	—
Offender was armed	1,215	100.0	—	—
Yes (1)	1,105	90.9	—	—
No (0)	110	9.1	—	—
Offender was attacking	1,254	100.0	—	—
Yes (1)	870	69.4	—	—
No (0)	384	30.6	—	—
Weapon type	1,090	100.0	—	—
Guns (1)	713	70.6	—	—
Cutting instruments (2)	173	19.5	—	—
Blunt instruments (3)	71	2.1	—	—
Other (4)	136	7.8	—	—

Table 2.

Chi-Square Analysis Comparing Manner of Death.

Variable and coding	Question	Finding	n	χ²	p	V
Age grouped (<35, >35)	Are those younger than 35 more likely to be shot than those older than 35?	No	1,254	1.493	.222	.035
Race (White, non-White)	Are non-Whites (includes Hispanic, Asian, Others) more likely to be shot than Whites?	No	1,197	0.008	.931	.003
Race (White, Black)	Are Blacks more likely to be shot than Whites?	No	937	0.032	.859	.006
Sex (male, female)	Are males more likely to be shot than females?	No	1,254	0.094	.521^a	.009
Sex and race (White, Black)	Are Black males more likely to be shot than White males?	No	934	0.012	1.00^a	.006
Unarmed status × Race (White, Black)	Are unarmed Blacks more likely to be shot than unarmed Whites?	No	911	0.001	.973	.004
Signs of mental illness (yes, no)	Are those who show signs of mental illness more likely to be shot than those who do not show signs of mental illness?	No	1,254	2.496	.114	.045
Weekend only (yes, no)	Are offenders more likely to be shot during the weekends than weekdays?	No	1,254	0.074	.786	.008
Body camera (yes, no)	Are officers without a body camera more likely to shoot someone than officers with a body camera?	No	1,254	0.113	.736	.010
Fleeing (yes, no)	Are those who were fleeing from an officer more likely to be shot than those who were not fleeing from the officer?	No	1,211	1.612	.204	.036
Armed (yes, no)	Are those who are armed more likely to be shot than those who are unarmed?	Yes	1,215	24.366	.000	.142
Attacking (yes, no)	Are those who were attacking an officer more likely to be shot than those who were not attacking an officer?	Yes	1,254	14.701	.000	.108
Weapon type (gun, cutting instrument, blunt instrument, other weapon)	Are those who are armed with a gun more likely to be shot than those who are armed with another type of weapon?	Yes	1,090	92.633	.000	.292

Fisher’s exact test due to expected cell counts less than 5.

Most offenders were shot (93.8%) compared with those who were shot and also tasered (6.2%) as the manner of death; the offender’s mean age was 36.5 with a standard deviation of 12.8 and a modal age of 29; when grouped by age, most offenders were under 35 (50.3%); most offenders did not display signs of mental illness (74.2%); most fatal encounters occurred during weekdays (71.9%); most officers did not possess a body camera (92%); the offender was armed during most incidents (90.9%); most offenders were not fleeing (73.4%); most offenders were White (51.8%; Black = 26.5%, Asian = 1.4%, Hispanic = 17.8%, Other = 2.5%); when race is dichotomized into White and Black, most offenders were White (66.2%); offenders were overwhelmingly male (95.5%); and most offenders were attacking when they were shot (69.4%). As for weapons, 44 different types were used in 1,090 incidents; unarmed, undetermined, and unknown weapons accounted for 161 incidents that were removed from the armed-encounter analysis; three cases were missing. Nine different weapons (20.5%; gun, knife, vehicle, toy, machete, box cutter, sword, hammer, and taser) were used in 95.4% of the incidents. The weapons data were recoded into guns (=1), cutting instruments (=2), blunt instruments (=3), and all other weapons (=4) for regression analysis. During six incidents, the offender was armed with multiple weapons, so the data were recoded using this scheme: (a) gun and knife were recoded to gun (n = 2), (b) baseball bat and fireplace poker were recoded to blunt instrument (n = 1), and (c) gun and explosives were recoded to gun (n = 3).

To establish a baseline measure, estimates from the 2008 police–public contacts survey were used (Eith & Durose, 2011) along with U.S. census data. Assuming all contacts had an equal chance to end in a use of force (which is not likely), the rate of fatal encounters is .003135%, or a ratio of 1:31,898 encounters. After standardizing for estimated population based on the U.S. Census Bureau (2016b), the rate of fatal shootings per million people is, by race, (a) White = 249.2, (b) Black = 727.2, (c) Asian = 97.9, (d) Hispanic = 380.9; by age, (a) <35 = 4.6, (b) >35 = 3.6; and by sex, (a) males = 7.9, (b) females = 0.36 (U.S. Census Bureau, 2016a).

All of the tests for relationships were not significant, except for whether unarmed offenders were shot and tasered, whether offenders who were not attacking were shot and tasered, and whether those who were armed with a gun were more likely to be shot (Table 2). Offenders who were unarmed were shot and tasered at counts higher than expected. This may reflect that the offender was first tasered with no effect and subsequently shot by the officer, although the sequence of events is not known. If this speculation is correct, then this suggests proper escalation of force insofar as the offender did not respond to less-lethal options and continued to attack the officer. Similarly, offenders who were not attacking the officer were shot and tasered at counts higher than expected, which may reflect that the offender was first tasered with no effect and was subsequently shot by the officer. This too is speculation because the sequence of events and circumstances are not known. Finally, those who were armed with a gun were more likely to be shot than those who were armed with a cutting instrument, a blunt instrument, or another weapon. Said differently, those who were armed with a cutting instrument, blunt instrument, or other weapon were more likely to be shot and tasered than those armed with guns. This may reflect that the offender was first tasered with no effect and subsequently shot by the officer. Again, this is speculation because the sequence of events and circumstances are not known.

Additional analysis was conducted at the multivariate level. Because the outcome variable is dichotomous (shooting or shooting and tasered), binary logistic regression was used to model which independent variables predicted the type of outcome (Field, 2009). The odds ratio, exp(B), was calculated to evaluate the effect size of the predictor variables, where an exp(B) value above 1 implies a greater likelihood of a fatal shooting than a fatal shooting also involving a taser (Table 3). A test of the full model shows a statistically significant improvement over the baseline constant-only model, indicating the predictors as a set reliably distinguish between fatal shootings and fatal shootings that also involve a taser, χ²(11) = 75.816, p < .000, −2LL = 332.803. The model correctly classified 94.8% of the cases. The 95% exp(B) confidence intervals were positive, indicating that the model was stable for interpreting the predictive effect of each variable and the range of effects was generally narrow, which suggests generalizing from the results is viable (Field, 2009). The Nagelkerke pseudo R² was moderate and indicates that the model accounts for 21.8% of the total variance in the outcome variable. The Wald test shows that all of the predictor variables, except for being armed with a cutting instrument (β = 36.359), exp(B) = .077, and being armed with a blunt instrument (β = 21.367), exp(B) = .053, did not significantly predict being shot as the manner of death. Controlling for other factors, those who were armed with a cutting instrument were .077 times less likely to be fatally shot than those who were armed with a gun, and those who were armed with a blunt instrument were .053 times less likely to be fatally shot than those who were armed with a gun.

Table 3.

Logistic Regression Predicting Being Shot as the Manner of Death (n = 1,251).

Variables	B	SE	Wald	df	p	exp(B)	95% CI for exp(B)
Variables	B	SE	Wald	df	p	exp(B)	Lower	Upper
Predictors
Age	−0.014	.012	1.317	1	.251	0.986	0.963	1.010
Sex	0.111	.651	0.029	1	.864	1.118	0.312	4.005
Race (White, non-White)	−0.179	.314	0.326	1	.568	0.836	0.452	1.546
Sign of mental illness	−0.212	.330	0.412	1	.521	0.809	0.423	1.545
Control
Body camera	0.172	.521	0.109	1	.741	1.188	0.428	3.295
Fleeing	−0.018	.415	0.002	1	.965	0.982	0.436	2.213
Attacking	0.139	.352	0.155	1	.694	1.149	0.576	2.292
Weekend	0.280	.362	0.598	1	.439	1.323	0.651	2.689
Armed with weapon (gun, referent)			41.316	3	.000
Cutting instrument	−2.558	.424	36.359	1	.000	0.077	0.034	.178
Blunt instrument	−2.939	.636	21.367	1	.000	0.053	0.015	.184
Other weapon	−1.088	.713	2.327	1	.127	0.337	0.083	1.363
Model
Constant	4.624	.909	25.876	1	.000	101.880
−2 log likelihood	332.803
Model χ²	(11) = 75.816, p < .000
Nagelkerke R²	.218
Hosmer–Lemeshow	p = .466

Note. CI = confidence interval.

How Would The National System Work?

Participation should be mandatory for all U.S. law enforcement agencies at all levels; voluntary participation equals limited participation, and limited participation equals limited coverage. Full participation is required for a national outlook but also because police use of force is a very low-frequency event that any single agency, even very large agencies, will not produce enough data in 1 year to make statistically reliable predictions, or to identify patterns and trends. Use of force data would be captured at the point of origin (the agency where it occurred). The data would be entered directly into a national database following uniform specifications using the existing web-based infrastructure of the National Crime Information Center (NCIC). The FBI would set policy and protocols, act as custodian of records, and grant access to federal law enforcement agencies. The states would operate their own systems, providing access to all local law enforcement agencies, and could be funded though federal grants authorized by the Police Reporting Information, Data and Evidence Act of 2015 (S.1476, pending as of July 1, 2016). The data would be categorized into police officers, offenders, and the environmental conditions to ensure all of the contextual details are provided. All records would be free of personal identifiers; the data would become public record as soon as legally feasible, and legacy data would eventually be migrated to the Inter-University Consortium for Political and Social Research (ICPSR) or the Bureau of Justice Statistics for public archival and access (see similarities at Klinger, Rosenfeld, Isom, & Deckard, 2016).

Such an open system would be consistent with the core mission of (a) the National Police Research Platform, funded by the National Institute of Justice; (b) the National Data Collection Committee of the Division of Policing at the American Society of Criminology; (c) the findings from a joint report issued by the National Sheriff’s Association and the Treatment Advocacy Center (2013) on justifiable homicides by law enforcement officers involving the mentally ill; (d) the findings from the Police Executive Research Forum (2012) on being proactive about preventing use of force situations; (e) the Police Foundation’s report titled “5 Things You Need to Know About Open Data in Policing”; and (f) the President’s “Police Data Initiative.”

What Is the Intended Outcome?

An indispensable tool for managing a police agency is a steady flow of information that indicates performance (Shane, 2010, Sparrow, 2015). Data analysis is the primary mechanism through which law enforcement agencies can address internal and external concerns that arise from their policies and practices. As Ramirez, Farrell, and McDevitt (2000) noted, “. . . By collecting information on the nature, character, and demographics of police enforcement practices, we enhance our ability to assess the appropriate application of the authority and broad discretion entrusted to law enforcement” (p. iii). In this case, it is the ability to identify the circumstances when use of force is likely to take place, what form it is likely to take, and the characteristics of the officer, the offender, the location, and the situation.

The advantages of national, incident-level use of force data include the following: (a) identifying the opportunity structure for police use of force; (b) identifying relationships between officers, offenders, and situational conditions; (c) identifying the nature and scope of emerging use of force issues; (d) identifying patterns and trends, and evaluating changes over time; (e) understanding causal connections through path analysis; (f) predicting relative risk for a use of force encounter as a matter of officer safety; (g) reducing or preventing organizational accidents; (h) becoming more self-reflective (a “learning organization,” M. M. Brown & Brudney, 2003); (i) promoting government transparency and accountability; (j) driving decision making; (k) providing respect to contextual differences of individual incidents; (l) building upon what is already known; (m) comparing police agencies and the type of incidents that resulted in use of force; (n) testing hypotheses; (o) elevating the standards of professional policing; (p) estimating progress toward the agency’s goals (achievements and shortcomings); (q) making a stronger case to acquire the resources the agency needs to be more efficient and effective at delivering service; (r) proactively managing community expectations; and (s) developing common terminology useful for discussions with the community.

Police and the community can use the data to confirm or dispel doubts they may have and answer mutually important policy questions such as the following:

Where does the perception of “widespread” or “epidemic” of police killings of citizens originate? How can we control those perceptions?

What can the police and community members do together to reduce hostility and misunderstandings between the police and the community arising from use of force?

What can the police and community members do together to reduce violent crime and anti-social behavior, which often occasions negative police–citizen interaction?

What can the police and community do to maintain open communication about police use of force, particularly involving officers and citizens of different racial or ethnic groups, which is a perennial source of tension? (Johnson, n.d.)

Unarmed does not necessarily mean not dangerous, so what are the implications for the community when an officer resorts to deadly force after perceiving a dangerous threat from an unarmed person? What are the implications for the officer?

An incident-driven data system would also support comparing neighboring cities on emerging use of force patterns, trends, and triggering conditions that extend beyond local boundaries. Comparison has a significant effect on human beings’ ability to make sound decisions through the concept of relativity: the process by which people assign value to something else. Humans typically have two minds when making decisions: the heart and the head. Most decisions occur at the heart level (emotional) based on incomplete information, which affects attitudes, perceptions, trust, and confidence. However, the head can step in and overrule heart’s decisions, provided sufficient information exists. When people make decisions in isolation—without sufficient information—the heart fills in the blanks with attributes that are easy to evaluate (e.g., age, race, sex, vicarious experiences, neighborhood characteristics, social class, one’s own definitions; R. A. Brown, Novak, & Frank, 2009; Hurst, McDermott, & Thomas, 2005; Rosenbaum, Schuck, Costello, Hawkins, & Ring, 2005; Schuck, Rosenbaum, & Hawkins, 2008) rather than those that are relevant (e.g., specific articulable facts). When a person evaluates something in isolation, it is difficult to determine whether that thing is good (e.g., the use of force rate, use of force against specific groups). However, when people can make comparisons, the head will readily have a more complete picture to judge the value of things (Hsee, 1996). Because people do not possess an intrinsic ability to judge the value of something in isolation, value is determined by comparing one thing to another—in this case, use of force incidents.

People make judgments and decisions against a backdrop of available information, the quality of that information, and how that information relates to something specific. Data on police use of force is that backdrop and can easily become part of an agency’s management framework, where the focus is on improving individual and collective performance (Shane, 2010).

A Proposed Research Agenda

Should national data become available, here is a proposed research agenda that will provide a systematic and comprehensive understanding of police use of force and its circumstances relevant to policy and practice:

Baseline Description

Who is using force against whom, under what circumstances, and armed with what type of weapon? Answering this requires determining an abundance of contextual details including the antecedents (pre-incident conditions and awareness), the approach, actions, reactions (incident conditions and information), and training and policies involving use of force (post-incident conditions). Because officers, offenders, and the environment differ across time and location, comparative research is necessary to identify how these elements interact using a common format to describe what is occurring.

How frequently is force used by local, state, county, federal, and special jurisdiction law enforcement officers? What is the citizen complaint rate for excessive force?

What organizational characteristics of the law enforcement agency are correlated with use of force?

Does the nature and scope of use of force vary across cities, states, and regions? Is violence against the police increasing, decreasing, or about the same over time? What type of encounters most and least often result in injury and death? What type of police tactics produce greater survivability and fewer injuries to the officer and the offender?

If more than one agency is involved in a use of force incident, then how many? Do they have overlapping jurisdiction?

Public Perception and Social Impact

What is the sociodemographic profile of the population served and its relationship to more or less force?

How is use of force changing over time?

Are the data being used to discuss law and policing policy with neighboring jurisdictions, the community, and national police advocacy groups?

To what extent does community-based justice (e.g., dispute resolution, mediation, restorative justice) contribute to fewer use of force incidents?

What is the relationship between the crime rate and police use of force? To what extent can police shift and share responsibility among other government, private, and nonprofit groups to reduce crime and disorder to help reduce use of force?

Police Use of Force Policies and Substantive Law

To what extent does use of force policies vary across local, state, county, federal, and special jurisdiction law enforcement agencies? How do those policies contribute to more or less force? Do different law enforcement agencies have different operational approaches to use of force? How do those differences contribute to more or less force?

What factors explain the differences in use of force policy? What role do labor agreements, collective bargaining, or arbitrators’ decisions play in such policies?

What are the variations in substantive and procedural state laws governing use of force? What police actions are legally permissible and impermissible?

What aspects of policy and substantive law do police leaders, elected officials, and other experts believe may require more regulatory attention? Are any aspects of law and policy too restrictive that place officers at unnecessary risk?

How do U.S. law enforcement agencies compare with their counterparts across the world in use of force frequency, circumstances, and policy prescriptions?

As a matter of policy, to what extent does a law enforcement agency use data to improve performance?

If officers’ interpretations of threat, risk, and permissible use of force differ, then what accounts for those differences? Several possibilities exist: the nature of the policing jurisdiction; the conditions under which the officers work including staffing levels, geographical dispersion, and workload volume; extent of training and supervision; law and agency policy.

Some Final Thoughts

The media frequently seizes on isolated use of force episodes, which distorts the public’s perception about the justifiability and rate of use of force to the point where a moral panic of sorts erupts. National consensus about the issue is often reached based on a select few incidents. Only with sufficient and reliable data can we estimate the incidence and prevalence of use of force, and only after we have a factual understanding of the nature and extent of use of force can we apply terms such as crisis, epidemic, widespread, discrimination and disparate treatment in the public discourse. To do so beforehand—as has been done in the recent past—is irresponsible and counterproductive to promoting desirable police–community relationships.

Overall, the results of the Washington Post data do not support some commonly held notions about police shootings involving age, race, sex, and mental illness that are prevalent in popular culture and mass media. The wonder lies not in the data that are present but in the data that are absent. Answers to why the rates are higher for some groups may rest in the contributing circumstances of each incident that context reveals, which is why more robust data are needed. As a nation, we are best served by experts who act and opine from the head based on data that are directly relevant, by those who have devoted themselves to understanding the complexities of police work and who can guide rational discussion toward a productive outcome. We are not well served by dilettantes who act and opine from the heart based on data that are easy to evaluate, or with incomplete information.

Despite the best intentions of those leading the effort to collect national data, they seem to underestimate the power of tradition within the police profession. The tradition of local government control, home rule, and the power of the Tenth Amendment to generally devolve criminal justice authority to the states instead of the federal government reigns supreme when discussing U.S. policing policies and practices. The police are traditionally insular, suspicious, and provincial. They “circle the wagons” and adopt a “bunker mentality” when policy changes loom as to how they should conduct their business and who is best suited to judge an officer’s actions—especially when use of force arises. The appeal to tradition is intoxicating, albeit fallacious; said differently, “this is the way we have always done it” does not mean because the method is old that it is better than one that is new. This sort of “reasoning” is appealing because people tend to value what is older or traditional. It is a common psychological characteristic, which may emanate from the fact that people take comfort in that which is familiar and has been generally accepted over time. From a purely functional perspective, preserving tradition is typically easier than testing and implementing new ideas; therefore, people tend to adhere to traditions out of laziness or fear of the unknown. Obviously, age does have a bearing in some contexts such as when it is relevant to quality, but we have already established that this is not the case with existing use of force data. A national database on use of force can help overcome outmoded thinking and parochial views rooted in tradition.

Compelling participation from law enforcement agencies will be a significant challenge given the constitutional changes that would be required. Self-governance through the Tenth Amendment ought not to be considered sacrosanct given the social and political importance of the issue; however, it is no surprise that the data collection required by the Violent Crime Control and Law Enforcement Act of 1994 never resulted in mandatory reporting on officers’ use of (excessive) force by police agencies (Adams, 1996). Perhaps a compromise to national-level data is to standardize use of force data collection at the state level. State legislatures could mandate such data be collected by police agencies without constitutional challenge. There is some precedent with similar state-level initiatives (Missouri, Connecticut, and Rhode Island) that mandate collecting data on police stops to better evaluate how the police interact with motorists; these initiatives are often known as racial profiling data collection. State-wide data collection and comparisons drawn from agencies that operate in the same legal and other contexts may be equally informative and persuasive to elected leaders and police executives. Even if the focus is restricted to deadly force incidents, state-level data will be useful.

As for H.R. 306, National Statistics on Deadly Force Transparency Act of 2015, it has potential, but the scope of the data is limited. The legislation is only weakly tied to sanctions for failing to comply; a state or unit of local government must first receive a specific grant award under Subpart 1 of Part E of Title I of the Omnibus Crime Control and Safe Streets Act of 1968 (42 U.S.C. 3750 et seq.) and then fail “substantially to comply with the requirement under section 2 for a fiscal year,” before the U.S. Attorney General will reduce the grant award by just 10%. “Substantially comply” is vague and undefined, and if the governing body is not a grant recipient, then the law does not apply.

Any effort to improve police policy and practice in this area will require some degree of standardization across the United States, home rule, and self-regulation notwithstanding. But complying with uniform national standards may compete with other priorities among local law enforcement agencies. The majority of law enforcement agencies are accountable to a local constituent base, and satisfying higher units of government has little or no influence on local politics, which means meager (if any) political support for the authorizing legislation. This brings us back to similar calls to improve policing in 1931: The Wickersham Commission concluded that the law cannot really solve the problems plaguing law enforcement and that solutions ultimately depend on the will of the community. Not a good prospect for H.R. 306.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

Jon M. Shane is associate professor of criminal justice in the Department of Law, Police Science and Criminal Justice Administration at John Jay College. His research focuses on police policy and practice issues, envronmental criminology and situational crime prevention.

References

Adams

(1996). Measuring the prevalence of abuse of force. In Geller

W. A.

Toch

(Eds.), Police violence: Understanding and controlling police abuse of force (pp. 61-97). New Haven, CT: Yale University Press.

Bittner

(1970). The functions of police in modern society. Washington, DC: U.S. Government Printing Office.

Blau

Schoenherr

(1971). The structure of organizations. New York, NY: Basic Books.

Brown

J. M.

Langan

P. A.

(2001). Policing and homicide, 1976-98: Justifiable homicide by police, police officers murdered by felons. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

Brown

M. M.

Brudney

J. L.

(2003). Learning organizations in the public sector? A study of police agencies employing information and technology to advance knowledge. Public Administration Review, 63, 30-43.

Brown

R. A.

Novak

K. J.

Frank

(2009). Identifying variation in police officer behavior between juveniles and adults. Journal of Criminal Justice, 37, 200-208.

Clark

W. A.

Avery

K. L.

(1976). The effects of data aggregation in statistical analysis. Geographical Analysis, 8, 428-438.

Clarke

R. V.

(1997). Situational crime prevention: Successful case studies. Albany, NY: Harrow and Heston.

Cohen

(1972). Folk devils and moral panics: The creation of the mods and the rockers. London, England: MacGibbon & Kee.

10.

Crank

(2003). Institutional theory of police: A review of state of the art. Policing: An International Journal of Police Strategies & Management, 26, 186-207.

11.

Cullen

F. T.

Agnew

(2006). Criminological theory past to present: Essential readings (3rd ed.). Los Angeles, CA: Roxbury.

12.

Doyle

J. M.

(2010). Learning from error in American criminal justice. The Journal of Criminal Law & Criminology, 100, 109-148.

13.

Eith

Durose

(2011). Contacts between police and the public, 2008. Washington, DC: Bureau of Justice Statistics.

14.

Fatal force database. (2016). Washington Post. Retrieved from https://www.washingtonpost.com/graphics/national/police-shootings/

15.

Field

A. P.

(2009). Discovering statistics using SPSS (3rd ed.). London, England: Sage.

16.

The 50 States Report, Submitted to the Commission on Civil Rights by the State Advisory Committee. (1961). Washington, DC: U.S. Government Printing Office.

17.

Finkelhor

Ormrod

(2000). Kidnaping of juveniles: Patterns from NIBRS. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

18.

Finkelhor

Ormrod

(2004a). Child pornography: Patterns from NIBRS. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

19.

Finkelhor

Ormrod

(2004b). Prostitution of juveniles, patterns from NIBRS. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

20.

Fyfe

J. J.

(1988). Police use of deadly force: Research and reform. Justice Quarterly, 5, 165-205.

21.

Groff

E. R.

Weisburd

Yang

S. M.

(2010). Is it important to examine crime trends at a local “micro” level? A longitudinal analysis of street to street variability in crime trajectories. Journal of Quantitative Criminology, 26, 7-32.

22.

Harding

R. W.

Fahey

R. P.

(1973). Killings by Chicago police, 1969-70: An empirical study. Southern California Law Review, 46, 284-316.

23.

Hickman

M. J.

(2006). Citizen complaints about police use of force (NCJ 210296). Washington, DC: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

24.

Hsee

C. K.

(1996). The evaluability hypothesis: An explanation for preference reversals between joint and separate evaluations of alternatives. Organizational Behavior and Human Decision Processes, 67, 247-257.

25.

Hurst

Y. G.

McDermott

M. J.

Thomas

D. L.

(2005). The attitudes of girls toward the police: Differences by race. Policing: An International Journal of Police Strategies & Management, 28, 578-593.

26.

International Association of Chiefs of Police. (2012). Emerging use of force issues: Balancing public and officer safety. Alexandria, VA: Author.

27.

Jacob

(1984). Using published data: Errors and remedies (Vol. 42). Thousand Oaks, CA: Sage.

28.

Jacobs

Britt

(1979). Inequality and police use of deadly force: An empirical assessment of a conflict hypothesis. Social Problems, 26, 403-412.

29.

Johnson

(n.d.). Examining the prevalence of deaths from police use of force. Unpublished presentation by Dr. Richard Johnson, University of Toledo, OH. Retrieved from www.forcescience.org/forcepresentation.ppt

30.

King

W. R.

Sanders

B. A.

(1997). Nice guys finish last: A critical review of killed in the line of duty. Policing: An International Journal of Police Strategies & Management, 20, 392-407.

31.

Klinger

Rosenfeld

Isom

Deckard

(2016). Race, crime, and the micro-ecology of deadly force. Criminology & Public Policy, 15, 193-222.

32.

Kobler

A. L.

(1975). Police homicide in a democracy. Journal of Social Issues, 31, 163-184.

33.

Maxfield

M. G.

(1989). Circumstances in supplementary homicide reports: Variety and validity. Criminology, 27, 671.

34.

Moore

M. H.

(2002). The limits of social science in guiding policy. Criminology & Public Policy, 2, 33-42.

35.

National Commission on Law Observance and Enforcement. (1931). Report on Lawlessness in Law Enforcement. Washington, DC: U.S. Government Printing Office.

36.

National Sheriffs’ Association & Treatment Advocacy Center. (2013). Justifiable homicides by law enforcement officers: What is the role of mental illness? (NCJ# 248116).

37.

Police Executive Research Forum. (2012). An integrated approach to de-escalation and minimizing use of force (Critical Issues in Policing Series). Washington, DC: Author.

38.

Police Executive Research Forum. (2016). Guiding principles on use of force (Critical Issues in Policing Series). Washington, DC: Author.

39.

The President’s Committee on Civil Rights. (1947). To secure these rights. New York, NY: Simon & Schuster.

40.

The President’s Commission on Law Enforcement and Administration of Justice. (1967). Task force report: The police. Washington, DC: U.S. Government Printing Office.

41.

The President’s Task Force on 21st Century Policing. (2015). Final report of the President’s Task Force on 21st Century Policing. Washington, DC: Office of Community Oriented Policing Services.

42.

Ramirez

D. A.

Farrell

A. S.

McDevitt

(2000). A resource guide on racial profiling data collection systems: Promising practices and lessons learned (NCJ# 184768). Washington, DC: U.S. Department of Justice.

43.

Reaves

B. A.

(2008). Census of state and local law enforcement agencies, 2008 (NCJ# 233982). Washington, DC: Bureau of Justice Statistics.

44.

Roberts

(2007). Predictors of homicide clearance by arrest: An event history analysis of NIBRS incidents. Homicide Studies, 11, 82-93.

45.

Robin

G. D.

(1963). Justifiable homicide by police officers. The Journal of Criminal Law, Criminology & Police Science, 54, 225-231.

46.

Rosenbaum

D. P.

Schuck

A. M.

Costello

S. K.

Hawkins

D. F.

Ring

M. K.

(2005). Attitudes toward the police: The effects of direct and vicarious experience. Police Quarterly, 8, 343-365.

47.

Schmidt

M.S.

(2015, April 8). Scant data frustrates efforts to assess number of shootings by police. New York Times. Retrieved from http://www.nytimes.com/2015/04/09/us/us-has-limited-data-on-shootings-involving-police.html

48.

Schuck

A. M.

Rosenbaum

D. P.

Hawkins

D. F.

(2008). The influence of race/ethnicity, social class, and neighborhood context on residents’ attitudes toward the police. Police Quarterly, 11, 496-519.

49.

Shane

J. M.

(2010). Performance management in police agencies: A conceptual framework. Policing: An International Journal of Police Strategies & Management, 33, 6-29.

50.

Sherman

L. W.

Langworthy

R. H.

(1979). Measuring homicide by police officers. Journal of Criminal Law and Criminology, 70, 546-560.

51.

Snyder

H. N.

(2010). Sexual assault of young children as reported to law enforcement: Victim, incident, and offender characteristics (A NIBRS Statistical Report). Collingsdale, PA: DIANE Publishing.

52.

Sparrow

M. K.

(2015). Measuring performance in a modern police organization. New Perspectives in Policing Bulletin (NCJ 248476). Washington, DC: U.S. Department of Justice, National Institute of Justice.

53.

Thompson

J. D.

(1967). Organizations in action. New York, NY: McGraw-Hill.

54.

U.S. Census Bureau. (2016a). Age and sex composition in the United States, 2012. Retrieved from http://www.census.gov/population/age/data/2012comp.html

55.

U.S. Census Bureau. (2016b). QuickFacts, United States. Retrieved from https://www.census.gov/quickfacts/table/PST045215/00

56.

U.S. Commission on Civil Rights. (1981). Who is guarding the guardians? Washington, DC: Author.

57.

Vazquez

S. P.

Stohr

M. K.

Purkiss

(2005). Intimate partner violence incidence and characteristics: Idaho NIBRS 1995 to 2001 data. Criminal Justice Policy Review, 16, 99-114.

58.

Whitaker

T. R.

Snell

C. L.

(2016). Parenting while powerless: Consequences of “the talk.” Journal of Human Behavior in the Social Environment, 26, 303-309. doi:10.1080/10911359.2015.1127736