Contagious Fire? An Empirical Assessment of the Problem of Multi-shooter,Multi-shot Deadly Force Incidents in Police Work

Abstract

Recent police shootings in which multiple officers fired numerous rounds at suspects have led some observers to assert that such situations involve “contagious fire,” where an initial officer’s shots launch a cascade of gunfire from other officers present. Although there is anecdotal recognition of the contagious fire phenomenon among police and the media, there is not a single empirical study documenting its existence in more than 50 years of deadly force research. This article uses Philadelphia Police Department shooting data to explore the potential for police shootings to become contagious. The article provides a testable definition of contagious shootings and identifies predictors of three outcomes: multiple officer shootings, the average number of shots fired per officer, and multi-officer, multi-shot incidents. Findings show that the requisite preconditions for a contagious shooting rarely occur, and when the preconditions were met, there is no evidence to support the existence of a contagion effect.

Keywords

contagious fire deadly force police shootings

Introduction

The use of deadly force by police officers has long been the subject of notable attention in American society. For example, concerns among members of the general public about the appropriateness of particular police shootings have repeatedly led to civil disturbances and political upheaval in major urban areas (e.g., Geller & Scott, 1992; Klinger, 2004). In the academic community, researchers have conducted numerous studies on topics such as the frequency of police shootings (e.g., Fyfe, 2002; Sherman & Langworthy, 1979), differences in shooting rates across jurisdictions (e.g., Jacobs & O’Brien, 1998; Milton, Halleck, Lardner, & Albrecht, 1977), and the role that race may play in deadly force decision making (e.g., Meyer, 1980; Plant & Peruche, 2005). Both independently and in concert with interested citizens and academics, law enforcement professionals have attempted to develop a better understanding of police shootings and reduce their frequency by, among other things, providing sound training for officers about how to handle potentially violent situations (e.g., Ederheimer & Fridell, 2005; Fyfe, 1996).

A specific aspect of the deadly force picture that has recently captured much attention and caused considerable concern among both the police and the public are incidents in which multiple officers fire numerous rounds at suspects. Two recent multi-officer, multi-shot shootings in major East Coast cities have focused attention on such incidents and raised a number of questions about how and why officers decide to fire their weapons. The first occurred on November 26, 2006, when five New York City police officers working an undercover assignment fired a total of 50 rounds during an encounter with a group of men whom the officers believed might be in possession of an illegal gun. When the officers attempted to stop the men as they tried to drive away, one officer opened fire and the other four followed suit. When the smoke cleared, police gunfire had killed the vehicle’s driver and severely wounded the other two men inside the car (“50 Shots Fired,” 2006). The second notable multi-officer, multi-shot shooting occurred on July 8, 2007, when seven uniformed members of the Philadelphia Police Department (PPD) fired more than 80 shots at a man who was brandishing a semiautomatic pistol on a south Philadelphia street corner. Some 20 rounds struck the armed man, killing him (“Police Fire 80 Rounds,” 2007).

The number of officers who discharged their guns in each of these incidents and the number of rounds they collectively and individually fired (one of the New York officers fired 31 shots) have led some observers to refer to the process in which these events unfold as “contagious fire” or “contagious shootings,” in which an initial officer’s gunshot(s) prompt his or her fellow officers to shoot in a form of collective action. In the wake of the New York City Police Department (NYPD) shooting, for example, there were several stories in major media outlets about this alleged phenomenon, including the New York Times, the Daily News, and ABC News.¹ Among the more stark claims about contagious fire in police work came in an ABC News story in November 2006, which stated that contagious shootings occur when multiple officers “start shooting once an officer opens fire—even if no real threat is present.” Dr. Carol Lieberman, a forensic psychiatrist at UCLA’s Neuropsychiatric Institute, asserted that the phenomenon is more common than people realize and that

the officer who starts the shooting is responsible for where he aims and how many times he shoots at the alleged “criminal.” . . . The other officers who become affected by the “contagion” are responsible for panicking, becoming blinded by the outbursts of gunfire, and not restraining themselves enough to assess the situation. (“How Common,” 2006)

Although press accounts (and comments of some police observers²) suggest that contagious fire is both a well-established phenomenon and a notable aspect of deadly force in American police work, not a single empirical study exists that documents the contagion effect or examines its prevalence, characteristics, or causes. As a consequence, we know virtually nothing about the characteristics of multi-officer, multi-shot shootings, most notably, how they differ from other sorts of deadly force encounters and what portion, if any, of such shootings involves so-called contagious fire. Given the potentially severe consequences of police use of deadly force, and the especially sensitive nature of shootings involving large numbers of rounds fired by multiple officers, this lack of research is disconcerting. Moreover, it raises concerns about whether contagious fire is a real phenomenon or if the term is simply a catchphrase that inaccurately describes certain types of officer-involved shootings. If contagious fire is indeed a real phenomenon, there are clear implications for police departments in terms of training, policy, and accountability. Quite simply, in matters of life and death, police officers should never discharge their firearms out of an unconscious, uncontrolled reaction to their fellow officer’s decision to shoot. The far-reaching consequences of police officer decisions to shoot—even when justified—necessitate an investigation into the veracity of the contagious fire thesis.

The lack of systematic evidence on the contagious fire thesis is a function of at least two factors. The first is that the term contagious fire has entered the lexicon of law enforcement with little concern about what, precisely, it means. This lack of conceptual clarity must be overcome before research can proceed. Consideration of the issue, fortunately, indicates that the elements of the thesis are implicit in popular treatments of it. Making these elements explicit, we argue that the contagious fire thesis has three parts: two requisite preconditions and the actual contagion effect. The preconditions for contagious fire are that (a) multiple officers shoot during an incident and (b) a large number of rounds are fired. The last element is the contagion effect itself: Secondary officers fire as a panicked or unconscious response to an initial officer’s firing.

The second reason that there has been no empirical research on the thesis is that data that can directly speak to the issue are not readily available. Because the thesis posits specific mental states on the part of officers who fire their weapons, information about what officers were thinking when they came to pull the trigger is required to directly assess whether any officer who discharged his or her weapon did so solely in response to shots being fired by other officers (i.e., whether any given officer’s actions involved contagious firing). Information about why officers decided to fire their weapons is not routinely collected in any sort of national database.³ Moreover, research on police use of deadly force (see literature review below) has yet to produce any sort of data set that includes detailed descriptions of even a small sample of how officers react when other officers in their presence fire their guns at suspects. As a consequence, no data exist that would permit researchers to conduct the type of research needed to undertake any sort of direct test of the contagious fire thesis.

It is fortunate that data do exist (from one of the two major East Coast agencies mentioned in the introduction that has recently had a highly visible multi-officer, multi-shot shooting) that afford the opportunity to examine several aspects of the general idea of contagious fire in police work. In 1996, the lead author reviewed the official case files of every reported incident in which Philadelphia police officers struck citizens with gunfire for the years 1970-1978 and 1987-1992, creating a data set that included several dozen variables about more than 800 officer-involved shootings (see below for more details). Using these shooting data, the authors explore determinants of two defining features of the contagious fire thesis—the number of officers who fire and the number of shots fired. Specifically, logistic and linear regression are employed to identify predictors of multiofficer, multi-shot incidents and to test whether the number of officers who fire in an incident predicts the average number of shots fired per officer, when controlling for other factors.

Following a review of the research literature on deadly force in American police work, the remainder of this article provides a more detailed description of the Philadelphia data, specifies the analytical plan used to examine it, presents the results of the analyses undertaken, and discusses the implications of these findings for understanding the use of deadly force by police officers.

Prior Research

The legal authority of police officers to use deadly force against citizens represents the most extreme facet of governmental power in our society. Despite the fact that the exercise of this power has been a topic of substantial controversy since officers began carrying firearms in the mid–19th century (e.g., Geller & Scott, 1992; Klinger, 2007), research on police shootings was virtually nonexistent until the 1970s. Since that time, researchers have devoted considerable attention to developing knowledge about the prevalence and determinants of officer-involved shootings. Among the first and most consistent findings of this research is that police officers rarely fire their guns in the line of duty. Fyfe (1988), for example, estimated that in the 1970s and 1980s, the average annual number of justifiable homicides by police officers in the entire United States was no more than 1,000.

The Bureau of Justice Statistics (2001) examined national data on police shootings of felons from 1976-1998 and concluded that the number of fatal shootings has not changed significantly (approximately 400 per year), despite large increases in the U.S. population older than age 13 (47 million) and the number of police officers (200,000). The source of the data used in this report—the FBI’s Supplemental Homicide Reports (SHR)—is notoriously unreliable, however (see Fyfe, 2002, for a discussion of SHR liabilities), so it likely does not provide a sound picture of deadly force patterns over recent decades.⁴

More recent data from the two largest cities in America suggest that the use of deadly force by police officers has been trending downward since scholars first started studying the issue. Unpublished data from the NYPD and Los Angeles Police Department (LAPD) show that in the last half of the 1970s, NYPD and LAPD officers shot an average of 115 and 68 suspects per year, respectively, and in the first 5 years of the new millennium, New York officers shot an average of 34 people, whereas their peers in Los Angeles shot an average of 32.

Concerning the determinants of deadly force, research has focused on three basic classes of variables: features of the situations in which officers and suspects come together in time and space, conditions of the social environment in which officers work, and aspects of the police agencies that employ officers. Research on the organizational forces shaping police shootings is dominated by inquiry into the role that administrative policies and practices play in officers’ decisions to employ deadly force. Research has consistently demonstrated that administrative policies—when enforced—can substantially curtail the rate of police shootings (Fyfe, 1988; Gain, 1971; Geller & Scott, 1992). Alternatively, administrative permissiveness can also lead to higher rates of police shootings (Fyfe, 1979). White (2001), for example, found that the number of police shootings in Philadelphia increased significantly after a restrictive administrative policy was abolished in 1974. Research has also explored the influence of other organizational forces on police shooting rates, such as racial composition of the force (Fyfe, 1978; Geller & Karales, 1981a, 1981b).

The environmental forces most commonly considered by researchers are the laws that govern police behavior and features of the communities in which officers work. With regard to the legal climate of policing, research has focused on the roles that state statutes and federal court rulings concerning deadly force usage by police officers play in officers’ decisions to fire their weapons. This research suggests that statutory and case law exert little direct influence on officers’ use of their firearms (e.g., Skolnick & Fyfe, 1993; Tennenbaum, 1994; White, 2003). Research on the link between community characteristics and police shootings has examined the roles of several factors, such as racial composition, arrest rates, economic inequality, and various measures of public violence. The most consistent finding from this line of research is that levels of deadly force are positively related to the prevalence of violence between citizens (e.g., Kania & Mackey, 1977; Liska & Yu, 1992; Matulia, 1985).

As for situational features, scholars have focused on factors such as the degree of threat that suspects who were shot posed to officers or others, status characteristics of officers and suspects, and the role that third parties might play in police shootings. Research has found that the majority of suspects taken under fire by the police presented an imminent danger to officer(s) who fired (Binder & Fridell, 1984; Fyfe, 1980, 1981a; Klinger, 2004). Shooting victims are disproportionately minorities, especially Blacks, although the causes of this overrepresentation are unclear and may vary by location (Fyfe, 1981b; Geller & Karales, 1981b; Milton et al., 1977). Research has also examined a host of officer characteristics including officer race, gender, age, duty and uniform status, assignment type, and experience (Fyfe, 1978; Geller & Karales, 1981a, 1981b). Although research has established patterns between certain officer characteristics and elevated rates of deadly force (e.g., Black officers are disproportionately likely to shoot suspects), Fyfe (1989) states that “there is virtually no empirical support for assertions that individual officer characteristics are measurably related to any type of performance in office” (p. 478).

Several scholars have stressed the importance of viewing police–citizen encounters as fluid events, almost like a chess match where each participant makes “moves” that influence the moves of the other participant (e.g., Bayley, 1986; Binder & Scharf, 1980). In the realm of deadly force, researchers have noted that how officers conduct themselves at earlier stages of encounters can influence the way in which encounters unfold and, hence, the likelihood that they will resort to gunfire (Fyfe, 1986, 1989; Geller & Karales, 1981b). Binder and Scharf (1980) assert, for example, that

a police “decision” to use, or not to use, deadly force in a given context might be better described as a contingent sequence of decisions and resulting behaviors—each increasing or decreasing the probability of an eventual use of deadly force. (p. 116)

The notion that police shootings are social events that unfold in a series of stages in which what transpires during previous stages affects subsequent ones has substantial implications for the notion of contagious fire, as the core assertion of the contagious fire thesis is that the actions of an initial officer who pulls the trigger lead others to subsequently fire.⁵ Scholars have yet to assess this aspect of the temporal sequencing of police–citizen interactions in which multiple officers discharge their weapons. There has, however, been a bit of research on other components of police shootings relevant to the contagious fire thesis: the number of officers who discharge their weapons and the number of rounds they fire (which has received the lion’s share of scholarly attention).

Counted among the few studies that have addressed the number of officers who fire is White’s (1999) study of officer-involved shootings in Philadelphia, which found that one quarter of the more than 800 Philadelphia police shootings during the years 1970 through 1992 involved multiple shooters. Klinger’s study of 113 shootings involving officers from 19 police departments in four states found a substantially higher proportion of multiple shooters; at least two officers discharged their weapons in approximately 45% of the cases examined (Klinger, 2004).⁶

The number of shots fired has been examined in three general contexts: the widespread switch in American policing from revolvers to semiautomatic handguns that has taken place in the past several years, the portion of suspects hit by police gunfire who die from their wounds (i.e., fatality rates), and shooting accuracy (i.e., hit rates). There is some research to suggest that the adoption of semiautomatic handguns has resulted in more shots fired per incident (Verhovek, 1992). In an analysis of police shootings in Portland, Oregon, over a 4-year period, for example, Snell (1992) found that the average number of shots fired by officers using revolvers was 2.6, compared with 4.6 for officers using semiautomatics. Other research, however, has shown little difference in rounds fired across handgun types (Brown, 1992; Cerar, 1990; Kindrick, 1992). Matulia (1982), whose research was undertaken when most agencies still equipped their officers with revolvers, reported that agencies using semiautomatic pistols experience a “significantly higher justifiable homicide rate” (see also Matulia, 1985). In Snell’s (1992) study of Portland shootings, he reported that

there is a strong relationship between the volume of shots fired by police and the probability of killing the suspect. In 17 incidents in which police fired three times or less, only two persons died. In 12 incidents in which four or more shots were fired, nine persons died. (p. A22)

A fair amount of research has examined the relationship between the number of shots fired and shooting accuracy, as well as the factors that influence accuracy (Geller & Scott, 1992; Morrison & Vila, 1998; Vila & Morrison, 1994). Research has consistently shown that hit rates (i.e., shooting accuracy) typically dip well below 50% (with significant variation by department; Copay & Charles, 2001; Geller & Scott, 1992). Research in New York City, for example, showed consistently low rates from year to year for the NYPD: 1987 (26%), 1988 (31%), and 1990 (23%) (Cerar, 1990; NYPD, 1988; Commanding officer, NYPD Firearms and Tactics Section, personal communication, 1989). Factors that have been linked to shooting accuracy include the following:

work environment and availability of back-up (Geller & Scott, 1992)

severity of the threat to the officer (Donahue & Horvath, 1991; Schade & Bruns, 1989; White, 2006)⁷

distance between participants (White, 2006)

officer approach and preparedness (White, 2006)

Although research on the number of officers shooting, the number of rounds they fire, and the number of rounds that find their mark has clear relevancy for the notion of contagious shootings, scholars have yet to tie such issues directly to the contagious fire thesis. In this article, we seek to do so by first examining the micro structure of officer-involved shootings to see how frequently the requisite preconditions for contagious fire exist during police–citizen interactions in which at least one officer discharges his or her weapon. We then conduct a series of statistical analyses to examine the role of potential predictors of officers’ actions in multi-officer, multi-shot shootings to shed light on conditions that might give rise to such shootings. We also conduct analyses to assess the effects of several potential predictors on the number of shots officers fire during shootings, including the role that other officers’ actions play in an individual officer’s firing behavior. Attention now turns to the data and methods we used to examine these issues.

Method and Data

This article uses a sample of shootings that includes all incidents in which members of the PPD struck citizens with gunfire during the years 1970-1978 and 1987-1992.⁸ The data come from PPD Internal Affairs shooting investigation files, which were obtained as discovery material during civil litigation against the department. The lead author examined each case file and coded the material therein to yield a data set containing some 75 distinct variables for each case (N = 808). Among the variables measured were the number of officers who discharged their firearms in each incident and the number of rounds each shooter fired.

A Brief Note on the Data

There are two aspects of the data that warrant discussion. First, this study is based on 30-year-old shooting data. Although the age of the data may raise concerns over the study’s applicability to 21st-century policing, the timing and locale allow for consideration of the contagious fire thesis in a historical context. Central to this historical context are the downward trends in police shootings over time and the increasing use of semiautomatic pistols with increased firepower. In effect, this study lays an important historical foundation for examination of multi-shooter, multi-shot incidents and the extent to which they might involve contagious fire by exploring data from a time period when shootings were more common (especially in the study site), but firearms were not as high powered as they are today.

Second, the Internal Affairs reports for the intervening years (1979-1986) were originally obtained through the discovery process but were lost in the early 1990s. PPD destroys all shooting investigations after 5 years, so the authors were unable to reconstruct the missing time period. Although the data gap limits exploration of the contagious fire thesis, this methodological limitation creates an interesting natural experiment because of changes that occurred during the intervening years. During the 1970s, the PPD operated under the formal and informal control of Frank Rizzo, and in 1979, the U.S. Justice Department took the unprecedented action of filing suit against the PPD for tolerating brutality. In 1980, the PPD instituted and enforced rigorous administrative controls on police shooting behavior, resulting in a well-documented 60% drop in officer-involved shootings (Fyfe, 1988; White, 1999). In effect, we are able to explore the prevalence of multi-shooter, multi-shot incidents—and the extent to which they might involve contagious fire—under two very different internal environmental climates: one where decisions to shoot were essentially unfettered (1970-1978) and one where those decisions were strictly controlled (1987-1992).⁹

Contagious Fire Preconditions

An obvious precondition for contagious shootings is that at least two officers are present during an incident in which police bullets are discharged.¹⁰ Therefore, the first step in any attempt to empirically examine issues related to contagious fire is to examine how frequently two or more officers are present in police shootings. In these data, slightly less than half of the shooting incidents (369 of 806, two cases missing; 46%) involved a single police officer. In nearly half of the cases, then, there existed no potential for contagious fire for the simple reason that the officer who fired was the only officer on-scene. Among the 437 cases in which at least two officers were present, just one officer fired any shots in 243 (56%) of them. Overall, then, three quarters of the cases in these data clearly involved no contagious fire, largely because (a) in almost half of the cases, the officer who fired was the only officer present, and (b) in another 25% of cases, there were multiple officers on scene but only one officer fired.¹¹

Another precondition of the contagious fire thesis is the notion that the multiple officers who pull the trigger in such shootings fire a sizable number of rounds. Implicit in this notion is the idea that officers in multi-shooter incidents would tend to fire more rounds than officers who are sole shooters. Across all shootings, the total number of shots fired ranged from 1 to 42 (n = 795; 13 missing cases). As displayed in Table 1, the mean number of shots was 3.83 and a single shot fired was the modal category (which occurred in 36% of the cases). Table 1 also shows that 2 to 4 shots were fired in 35.4% of the cases, 5 to 9 shots were fired in 20.9%, 10 to 19 were fired in 6.6%, and 20 or more shots were fired in just 1.2% of the shootings.

Table 1.

Distribution of Shots Fired by Number of Shooters

Number of Shots Fired	All Shootings (n = 795)	Single-Shooter Incidents (n = 598)	Multiple-Shooter Incidents (n = 197)
1 shot	35.8% (285)	47.7% (285)	0.0% (0)
2-4 shots	35.4% (282)	38.2% (228)	27.5% (54)
5-9 shots	20.9% (166)	13.5% (80)	43.7% (86)
10-19 shots	6.6% (52)	0.9% (5)	23.7% (47)
20-39 shots	1.1% (9)	0.0% (0)	4.5% (9)
40 or more shots	0.1% (1)	0.0% (0)	0.5% (1)
Mean shots fired	3.83	2.39	8.21
Mode shots fired	1	1	5

Among the 598 cases in which just one officer fired, the number of shots fired ranged from 1 to 16, with a mean of 2.39 and a mode of 1 (which occurred in 47.7% of the cases; see Table 1). Among the 197 shootings in which two or more officers fired, the total number of shots fired ranged from 2 to 42, with a mean of 8.21 and a mode of 5 (which occurred in 10.7% of the cases; see Table 1). Finally, the mean number of shots fired per officer in multiple shooter incidents was 3.13, which indicates that officers tend to fire slightly more shots in incidents in which other officers also discharge their weapons.¹² This comparison of the number of shots fired by officers who are sole shooters and those who fire in concert with peers is consistent with the notion implicit in the contagious fire thesis that the average number of shots fired per officer would increase when more than one officer fires. It is not the last word on the topic, however, for it is possible that features of shootings beside the number of officers firing might account for the differences observed. This matter is one of the key topics we address in the following section, in which we report the results of several multivariate analyses of the Philadelphia data.

Multivariate Findings

The Number of Shooters

The first step we took in our multivariate analysis was to identify factors that predict whether two or more officers would discharge their weapons among the 437 shootings in which two or more officers were present (i.e., whether a given incident in which at least two officers could shoot would be a multi-shooter shooting). Table 2 shows the list of potential predictor variables, many of which either seem intuitively related to the number of shooters (or shots fired) or have been linked to these shooting outcomes by prior research (e.g., suspect weapon, lighting conditions, assignment to a specialized unit, officer approach, etc.). To predict the number of shooters, we first created a dichotomous variable where single-shooter incidents were assigned scores of 0 and multiple-shooter incidents were assigned scores of 1 and then utilized logistic regression to regress this measure on the possible predictors listed in Table 2.

Table 2.

Incident-, Officer-, and Suspect-Related Variables

Incident-Related Characteristics

Incident type (assault on officer)

0 = gun

1 = non-gun assault

Time of day

0 = 8:00 p.m.-2:00 a.m.

1 = other

Location

0 = indoors

1 = outdoors

How officer became involved

0 = radio call/civilian advised

1 = officer initiated

Job type

0 = other

1 = robbery, man with gun, assault, rape, serving warrant, attempt arrest

Lighting conditions

0 = good

1 = dark/poor

Reason for shooting

0 = other

1 = defend self/others

Who fired first shot?

0 = officer

1 = suspect

Time period

0 = 1970-1978

1 = 1987-1992

Suspect/officer race/ethnicity

0 = White officer/non-White suspect

1 = other

Officer-Related Characteristics

Rank

0 = patrol officer

1 = other

Officer gender

0 = male

1 = female

Officer race/ethnicity

0 = White

1 = non-White

Officer duty status

0 = on duty

1 = off duty

Officer assignment

0 = patrol

1 = specialized unit

Police officer actions prior to shooting

0 = approaching vehicle, confront suspect

1 = other

Officer’s gun

0 = revolver

1 = semiautomatic

Prior knowledge suspect was armed

0 = no

1 = yes

Officer used cover

0 = no

1 = yes

Officer called for back-up prior to shooting

0 = no

1 = yes

Officer time to aim

0 = no

1 = yes

Officer injury

0 = no

1 = yes

Suspect-Related Characteristics

Suspect race/ethnicity

0 = White

1 = non-White

Suspect gender

0 = male

1 = female

Suspect actions at time of shooting

0 = fighting/attacking

1 = other

Suspect position

0 = standing, crouching, advancing

1 = other

Suspect weapon displayed

0 = no

1 = yes

The results are shown in Table 3.¹³ The predictors of multiple-officer shootings are as follows:

Table 3.

Logistic Regression Results Predicting the Number of Shooters: Multiple Shooter (no, yes)

Predictor Variable	B	SE	Wald	OR	p Value
Time period (1970-1978, 1987-1992)	−1.077	.310	12.094	.341	.001
Assault type on officer (gun, not a gun)	−.717	.254	7.997	.488	.005
Officer called for back-up (no, yes)	1.048	.256	16.793	.351	.000
Job type^a	.579	.227	6.478	1.784	.011
Who fired first shot? (officer, suspect)	.930	.326	8.167	2.536	.004
Log Likelihood	476.555
R² (Cox and Snell)	.152
Chi-square	65.160
df	5
Significance	.000
N	394

Job type is coded as 0 = other, 1 = robbery, man with gun, assault, rape, serving warrant, attempting arrest.

the time period 1970-1978

gun assault on the officer

the officer called for back-up prior to the shooting

the call involved a high-risk event such as a robbery, man with a gun, rape, and so on (hereafter referred to as a high-risk call)¹⁴

the suspect fired a shot at the officer first

The odds ratios suggest that the variables vary notably in their predictive power. In particular, the type of call and the suspect firing first increase the likelihood of a multiple-shooter event by 1.7 times and 2.5 times, respectively.

We next explored the multiple-shooter precondition using a continuous-level variable to identify factors that predicted the number of officers who fired in those cases where at least two officers were present. To do this, an ordinary least squares (OLS) model was estimated with the same predictors included in the logistic regression model. The results of this procedure are displayed in Table 4, illustrating that the independent variables are equally effective in predicting the continuous-level version of the multiple-shooter outcome.

Table 4.

Ordinary Least Squares Regression Results Predicting the Number of Shooters: Multiple Shooter (continuous)

Predictor Variable	B	SE	Beta	T	p Value
Time period (1970-1978, 1987-1992)	−.410	.117	−.161	−3.513	.000
Assault type on officer (gun, not a gun)	−.208	.104	−.103	−1.992	.047
Officer called for back-up (no, yes)	.673	.102	−.298	−6.611	.000
Job type^a	.250	.094	.123	2.661	.008
Who fired first shot? (officer, suspect)	.651	.129	.253	5.047	.000
R² (adjusted)	.223
Sum of squares (total)	397.343
F	23.587
Significance	.000
N	393

Job type is coded as 0 = other, 1 = robbery, man with gun, assault, rape, serving warrant, attempting arrest.

The Number of Shots Fired

The next component of our analysis included all cases and examined the average number of shots fired per shooter, per incident, calculated as follows: the total shots fired per incident / number of shooters per incident. As noted above, implicit in the contagious fire thesis is the notion that the average number of shots fired per officer should be higher when multiple officers shoot (as compared with situations in which just one officer shoots). As reported above, this is, in fact, the case; the average number of shots that officers fire when they are sole shooters is 2.39 and 3.13 when there are two or more shooters. But it is also implicit in the contagious fire thesis that the difference in number of shots fired should remain when other features of shooting incidents are taken into account.

To explore this matter, we first estimated a one-predictor OLS model with the average number of shots fired per officer (per incident) as the dependent variable and the number of officers who fired as the predictor to simply establish the effect size of the previously reported difference between single- and multiple-shooter incidents. Table 5 (top section) shows that the number of shooters is statistically significant (B = .430, p < .000). With this baseline regression model in hand, we then estimated an OLS model that includes the number of shooters as well as the other officer-, suspect-, and incident-related variables that were included in the above-reported logistic and OLS models to assess whether the relationship between the number of shooters and the average number of shots fired holds when controlling for other situational factors.

Table 5.

Ordinary Least Squares Regression Results Predicting the Average Number of Shots Fired per Officer per Incident

Predictor Variable	B	SE	Beta	T	p Value
Model 1
Number of shooters	.430	.083	.182	5.218	.000
R² (adjusted)	.032
Sum of squares (total)	2964.644
F	27.227
Significance	.000
N	794
Model 2
Number of shooters	.159	.115	.060	1.385	.167
Officer injured (no, yes)	.402	.199	.086	2.025	.043
Who fired first shot? (officer, suspect)	1.547	.260	.261	5.941	.000
Job type^a	.448	.182	.106	2.464	.014
Gun type (revolver or semiautomatic)	1.644	.292	.236	5.626	.000
R² (adjusted)	.182
Sum of squares (total)	2131.991
F	22.799
Significance	.000
N	491

Job type is coded as 0 = other, 1 = robbery, man with gun, assault, rape, serving warrant, attempting arrest.

The bottom portion of Table 5 shows the results of this model, indicating that the number of officers who shoot is no longer a significant predictor of the average number of shots fired. This finding suggests that the number of officers shooting does not independently affect the number of shots that officers fire. This model also identifies four other variables as the factors that drive the number of rounds that officers discharge: the degree of risk in the call officers were handling, who fired the first shot in gunfights, whether officers suffered injuries, and the type of gun officers carried. The number of shots fired increased when officers handled high-risk calls, when suspects fired the first shot, when officers were injured, and when officers carried semiautomatic handguns.¹⁵

Multi-shooter, Multi-shot Incidents

The analyses in the previous sections represent an indirect examination of the contagious fire thesis, as we explored each of the two requisite preconditions separately. To conduct a more direct assessment of contagious fire, we created a third outcome variable that isolates shootings that meet both preconditions: those with multiple shooters and a high number of rounds fired. This final analysis takes a conservative approach, as we define cases where multiple officers together fire 10 or more rounds as possible contagious shootings, to explore what might predict such cases.

Single-officer shootings have again been set aside, and we focus on the 437 incidents in which at least two officers were present. Among these cases, 53 or 12.1% involved more than one shooter and 10 or more rounds fired. We assigned a score of 1 to each of these cases and 0 to the rest and used logistic regression to see whether specific features of shooting incidents were significant predictors of multi-officer, high-round shootings. Table 6 shows that there are four predictors of multi-shooter, high-round incidents: a gun assault on the officer, a high-risk call, officer called for back-up prior to the shooting, and the suspect fired the first shot. Despite the small number of cases of interest (n = 53), we are able to model multishooter, multi-shot incidents quite successfully, and it is, perhaps, not surprising that the predictors of the combination outcome variable mirror those that emerged in the earlier individual precondition analyses.

Table 6.

Logistic Regression Results Predicting Multi-shooter, Multi-shot Incidents (10+ for high round)

Predictor Variable	B	SE	Wald	OR	p Value
Assault type on officer (gun, not a gun)	−1.408	.479	8.640	.245	.003
Job type^a	.897	.354	6.416	2.453	.011
Officer called for back-up (no, yes)	1.509	.344	19.283	.221	.000
Who fired first shot? (officer, suspect)	.963	.378	6.487	2.619	.011
Log likelihood	231.951
R² (Cox and Snell)	.151
Chi-square	65.288
df	4
Significance	.000
N	399

Job type is coded as 0 = other, 1 = robbery, man with gun, assault, rape, serving warrant, attempting arrest.

Discussion and Conclusion

The foregoing findings shed considerable light on the contagious fire thesis by presenting evidence bearing on several hypotheses derived from it. Consideration of the implications of these findings begins with a discussion of those variables that turned out to be unrelated to the outcomes of interest. First, officer and suspect race—specifically, White officer and non-White suspect—were not associated with any of the three outcomes examined (i.e., multiple officer shootings, number of shots fired, and the combination of the two). That our analysis did not uncover any race effects places our research on a long list of empirical studies of police action (including the use of deadly force) that have found that race does not matter when it comes to explaining why officers act as they do. It should be noted in this connection, however, that there also exist a number of studies of police action that have found race effects and that the role that race plays in police behavior appears to be contingent on other factors (such as time and place; see Skogan & Frydell, 2004). Given this, it is important to keep in mind that our study was limited to a single police agency and that although our data cover a substantial time period, they are not temporally comprehensive.

The remaining nonsignificant factors are more closely tied to tactical aspects of deadly force encounters. That neither lighting conditions, the nature of the officers’ assignment (i.e., plainclothes vs. uniform), nor officers’ approach, preparedness, and position at the time of the shooting were associated with any of our outcome measures when controlling for other variables suggest that these factors might not be important when considering how to control officers’ use of deadly force. If any of them had been significant predictors, it would suggest that police trainers and policy makers may be able to reduce the likelihood of multi-officer, multi-shot shootings by providing policy guidance and/or training about matters such as light control, field tactics for specific assignments, and so on.

Where factors that were significant determinants of police shooting behavior are concerned, several important themes emerged. First, the predictors of multi-shooter incidents are similar to the predictors of average number of shots fired per officer, which logically, then, are similar to the predictors of the combination variable. Two of the five predictors are common across all outcome measures: suspect firing the first shot and high-risk call. Other shared predictors include assault type on the officer (gun assault) and officer called for back-up. The overlap among predictors is explained, at least in part, by the shared relationship between the outcomes of interest: The number of shooters and the number of shots fired are components of the combination shooting dependent variable. And the number of shots fired is related to the number of shooters (e.g., as the number of shooters increases, so does the number of shots fired).

Second, there are a few variables whose association with the shooting outcomes makes intuitive sense because of their relationship to the study site or police work in general. The time period variable likely emerges as significant in the multi-shooter model because of the well-documented changes in the administrative oversight of police shooting decisions in the PPD. The restrictive administrative policy instituted in 1980 by the PPD was followed by a 60% decrease in deadly force incidents (Fyfe, 1988), including those involving multiple shooters.

Officer gun type—specifically use of semiautomatic handguns—emerges as a significant predictor of the average number of shots fired when controlling for the number of shooters and other relevant variables. Simply put, when controlling for other factors including the number of shooters, officers with semiautomatic handguns fired significantly more shots on average than officers armed with revolvers.¹⁶ Given the nonsignificance of gun type in the multi-shooter models and the nonsignificance of number of shooters in the OLS model, this finding is likely a result of the increased firepower of semiautomatics and not the result of a contagion effect. This finding should also be considered in the context of the data. During the 1970s, the PPD standard-issue firearm was a .38 caliber revolver. During the late 1980s, the PPD began making semiautomatic firearms available, although the revolver continued to be the firearm of choice among Philadelphia police officers throughout the study period. Given the relationship between semiautomatics and an increase in shots fired, the potential for multi-shooter, multi-shot—or contagious fire—incidents may have been limited in this study. Additional research with more recent data is needed to further explore the relationship between gun type and the potential for contagious fire.

Last, and perhaps most important, the remaining significant predictors in the regression models represent or are indicative of an increased threat or lethality to the police. Nearly all of the remaining predictors intimate an armed and dangerous suspect:

gun assault against the officer

officers called for back-up prior to the shooting

the suspect fired a shot at officers first

the incident involved a high-risk call for service (robbery, man with gun, etc.)

at least one officer was injured

Taken together, all of these findings suggest that there may not be much that police policy makers and trainers can do to reduce multi-officer, multi-shot shootings, as these events appear to involve situations where those types of outcomes might be expected: Armed and dangerous suspects present an imminent threat that leads officers to use deadly force to protect their own lives and/or the lives of others. Moreover, the nonsignificance of the number of shooters variable in the second OLS model suggests that the number of additional officers using deadly force at these highly dangerous incidents is unrelated to the average number of shots fired by each officer. This finding represents strong evidence against the existence of a persistent contagion effect in these officer-involved shootings.

This is by no means the last word on the contagious shooting thesis, however. Although it seems clear from this research that the rarely occurring multi-officer, high-shot shootings were a function of forces outside of the involved officers (i.e., forces that indicated a heightened degree of danger), it is still possible that some small number of officer-involved shootings involve contagious fire. Recall that the contagious shooting thesis maintains that a high volume of shots in multiple-officer shootings are the product of a panic response on the part of those officers who shoot after an initial officer discharges his or her weapon. As we noted previously, these data cannot speak to the critical question of how subsequent officers come to decide to fire their weapons. In short, we were only able to empirically address two of the three key facets of the contagious fire thesis. Further research that examines the decision-making processes in which officers involved in multi-officer, multi-shot shootings engage in is needed to examine the possibility that some small number of police shootings involve contagious fire. Such an inquiry would require data from interviews with officers following shootings—data that, as mentioned above, do not presently exist. In the meantime, this research indicates that even if some shootings do involve contagious fire, such shootings are quite rare and multi-officer, multi-shot shootings seem to be a product of the increased danger that officers typically face in such incidents.

Footnotes

The authors declared no potential conflicts of interests with respect to the authorship and/or publication of this article.

The authors received no financial support for the research and/or authorship of this article.

Notes

References

50 shots fired, and experts offer a theory. (2006, November 27). New York Times. Retrieved February 17, 2007, from http://www.nytimes.com

Bayley

D. H.

(1986). The tactical choices of police patrol officers. Journal of Criminal Justice, 14, 329-348.

Binder

Fridell

(1984). Lethal force as a police response. Criminal Justice Abstracts, 16(2), 250-280.

Binder

Scharf

(1980). The violent police–citizen encounter. Annals of the American Academy of Political and Social Science, 452, 111-121.

Brown

L. P.

(1992). The choice of handguns for police officers: Revolvers or semi-automatics. New York: New York City Police Department.

Bureau of Justice Statistics. (2001). Policing and homicide: 1976-1998. Washington, DC: U.S. Department of Justice.

Cerar

J. C.

(1990). Firearms discharge assault report, 1990: New York City Police Department Academy Firearms and Tactics Section. Unpublished annual report by NYPD.

Copay

A. G.

Charles

M. T.

(2001). Handgun shooting accuracy in low light conditions: The impact of night sights. Policing: An International Journal of Police Strategies and Management, 24(4), 595-604.

Donahue

M. E.

Horvath

F. S.

(1991). Police shooting outcomes: Suspect criminal history and incident behaviors. American Journal of Police, 10(3), 17-34.

10.

Ederheimer

Fridell

(2005). Chief concerns: Exploring the challenges of police use of force. Washington, DC: Police Executive Research Forum.

11.

Fyfe

J. J.

(1978). Shots fired: An analysis of New York City police firearms discharges. Doctoral dissertation, State University of New York at Albany. (University Microfilms No. AAT78-14335)

12.

Fyfe

J. J.

(1979). Administrative interventions on police shooting discretion: An empirical examination. Journal of Criminal Justice, 7, 309-324.

13.

Fyfe

J. J.

(1980). Philadelphia police shootings, 1975-78: A system model analysis. Report for the Civil Rights Division, U.S. Department of Justice.

14.

Fyfe

J. J.

(1981a). Toward a typology of police shootings. In Fyfe

J. J.

(Ed.), Contemporary issues in law enforcement (pp. 136-151). Beverly Hills, CA: Sage.

15.

Fyfe

J. J.

(1981b). Who shoots? A look at officer race and police shootings. Journal of Police Science and Administration, 9, 367-382.

16.

Fyfe

J. J.

(1986). The split-second syndrome and other determinants of police violence. In Campbell

A. T.

Gibbs

J. J.

(Eds.), Violent transactions (pp. 583-598). Oxford, UK: Basil Blackwell.

17.

Fyfe

J. J.

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

18.

Fyfe

J. J.

(1989). Police/citizen violence reduction project. FBI Law Enforcement Bulletin, 58, 18-25.

19.

Fyfe

J. J.

(1996). Methodology, substance and demeanor in police observational research. Journal of Research in Crime and Delinquency, 33(3), 337-448.

20.

Fyfe

J. J.

(2002). Too many missing cases: Holes in our knowledge about police use of force. Justice Research and Policy, 4, 87-102.

21.

Gain

(1971, December 23). Discharge of firearms policy: Effecting justice through administrative regulation. Unpublished memorandum.

22.

Geller

Karales

(1981a). Shootings of and by Chicago police: Uncommon crises, Part I: Shootings by Chicago police. Journal of Criminal Law and Criminology, 72(4), 1813-1866.

23.

Geller

Karales

(1981b). Split-second decisions: Shootings of and by Chicago police. Chicago: Chicago Law Enforcement Study Group.

24.

Geller

Scott

M. S.

(1992). Deadly force: What we know. Washington, DC: Police Executive Research Forum.

25.

How common is contagious shooting? (2006, November 27). ABC News. Retrieved February 17, 2007, from http://www.abcnews.com

26.

In a flash, a tragic turn: Cop started shooting as man reached for waistband. (2006, November 28). New York Daily News. Retrieved April 11, 2007, from http://www.nydailynews.com

27.

Jacobs

O’Brien

R. M.

(1998). The determinants of deadly force: A structural analysis of police violence. American Journal of Sociology, 103, 837-862.

28.

Kania

R. E.

Mackey

W. C.

(1977). Police violence as a function of community characteristics. Criminology, 15, 27-48.

29.

Kindrick

R. E.

(1992, March 11). Officer-involved deadly force incidents report. Unpublished memorandum from Portland, Oregon, Police Bureau Training Division Captain Kindrick to Assistant Chief Inman.

30.

Klinger

(2004). Into the kill zone: A cop’s eye view of deadly force. San Francisco: Jossey-Bass.

31.

Klinger

D. A.

(2007). Impact munitions: A discussion of key information. Policing: An International Journal of Police Strategies and Management, 30(3), 385-397.

32.

Liska

A. E.

(1992). Specifying and testing the threat hypothesis. In Liska

A. E.

(Ed.), Social threat and social control (pp. 53-68). Albany: State University of New York Press.

33.

Matulia

K. J.

(1982). A balance of forces: A study of justifiable homicide by the police. Gaithersburg, MD: International Association of Chiefs of Police.

34.

Matulia

K. J.

(1985). A balance of forces: Model deadly force policy and procedure (2nd rev. ed.). Gaithersburg, MD: International Association of Chiefs of Police.

35.

Meyer

M. W.

(1980). Police shootings of minorities: The case of Los Angeles. Annals of the American Academy of Political and Social Science, 452, 98-110.

36.

Milton

Halleck

J. S.

Lardner

Albrecht

G. L.

(1977). Police use of deadly force. Washington, DC: Police Foundation.

37.

Morrison

Vila

(1998). Police handgun qualification: Practical measure or aimless activity? Policing: An International Journal of Police Strategies and Management, 2(3), 510-533.

38.

Mumola

C. J.

(2007). Arrest-related deaths in the United States, 2003-2005. Washington, DC: Bureau of Justice Statistics.

39.

New York City Police Department. (1988). Firearms discharge assault report for 1987. Unpublished report.

40.

Plant

E. A.

Peruche

B. M.

(2005). The consequences of race for police officers? Responses to criminal suspects. Psychological Science, 16, 80-183.

41.

Police fire 80 rounds in killing. (2007, July 10). Philadelphia Inquirer. Retrieved August 16, 2007, from http://www.philly.com

42.

Police killing draws national notice. (1999, February 8). New York Times. Retrieved February 17, 2007, from http://www.nytimes.com

43.

Schade

T. V.

Bruns

G. H.

(1989). Police shooting performance in threatening environments. American Journal of Police, 8(2), 31-48.

44.

Sherman

L. W.

Langworthy

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

45.

Skogan

Frydell

(Eds.). (2004). Fairness and effectiveness in policing: The evidence. Washington, DC: National Academies Press.

46.

Skolnick

J. H.

Fyfe

J. J.

(1993). Above the law: Police and the excessive use of force. New York: Free Press.

47.

Snell

(1992, April 26). Fear a factor when police pull trigger: Reality defies classic image of police officer. The Oregonian, p. A22.

48.

Tennenbaum

A. N.

(1994). The influence of the Garner decision on police use of deadly force. Journal of Criminal Law and Criminology, 85, 241-260.

49.

Tennessee v. Garner, 471 U.S. 1 (1985).

50.

Verhovek

S. H.

(1992, May 30). Foes of police-gun bill say blocking vote may be only hope. New York Times, p. 12.

51.

Vila

Morrison

(1994). Biological limits to police combat shooting accuracy. American Journal of Police, 13(1), 1-30.

52.

White

M. D.

(1999). Police shootings in Philadelphia: An analysis of two decades of deadly force. Doctoral dissertation, Temple University. (University Microfilms No. AAT99-55871)

53.

White

M. D.

(2001). Controlling police decisions to use deadly force: Reexamining the importance of administrative policy. Crime & Delinquency, 47(1), 131-151.

54.

White

M. D.

(2003). Examining the impact of external influences on police use of deadly force over time. Evaluation Review, 27(1), 50-78.

55.

White

M. D.

(2006). Hitting the target (or not): Comparing characteristics of fatal, injurious and non-injurious police shootings. Police Quarterly, 9(3), 303-330.