Article Ambush Killings of the Police, 1970–2018: A Longitudinal Examination of the “War on Cops” Debate Police Quarterly 0(0) 1–21 ! The Author(s) 2020 Article reuse guidelines: sagepub.com/journals-permissions DOI: 10.1177/1098611120919441 journals.sagepub.com/home/pqx Michael D. White1 Abstract Over the last few years, there has been a series of high-profile, premeditated ambush attacks on police, which has led some to conclude there is a “war on cops.” Unfortunately, prior research has not examined the prevalence of police ambushes over an extended period of time, and the most recent study only analyzed the phenomenon through 2013. Moreover, the “war on cops” thesis implies a very specific motivation for an ambush: hatred of police or desire to seek vengeance in response to police killings of citizens. Prior research has not sufficiently explored the motivations of ambush attacks, or whether recent trends in ambushes are linked to a “war on cops” motive. I investigate ambush killings of police from 1970 to 2018 using data from the Officer Down Memorial Page in an attempt to address these research gaps. I apply a temporal coding scheme of when the attack occurred to isolate killings of police that are consistent with the International Association of Chiefs of Police definition of an ambush. Results from linear regression show that the annual rates of ambush killings of police have declined by more than 90% since 1970. Although ambushes spiked in 2016 and 2018 to the highest rates in 20 years, interrupted time series analysis indicates no statistically significant increase post-2013. Spikes have also occurred in nonambush killings since 2014. Police leaders and researchers should monitor trends in ambush and nonambush killings of police, as the recent spikes may presage the emergence of a chronic problem. 1 School of Criminology and Criminal Justice, Arizona State University Corresponding Author: Michael D. White, School of Criminology and Criminal Justice, Arizona State University, 411 North Central Avenue, Suite 600, Phoenix, AZ 85004, United States. Email: mdwhite1@asu.edu 2 Police Quarterly 0(0) Keywords ambush killings of police, war on cops, violence against police Several decades of research has demonstrated the dangers inherent in police work. Policing has the highest occupational rate of violent victimization (Duhart, 2001; Fridell et al., 2009), and each year, approximately 10% of all police officers in the United States are assaulted on the job (Bierie, 2017). Although the vast majority of police–citizen encounters do not involve violence at all, the sheer volume of encounters, more than 50 million per year, means that violence against police is a daily event. Indeed, data from the Federal Bureau of Investigation (FBI; 2010) indicate that officers are assaulted in approximately 160 encounters per day. The focus on violence against police has intensified since 2014. The increased attention can be traced back to a series of controversial police killings of minority citizens (e.g., Michael Brown, Eric Garner, Walter Scott), which led to public protest, civil disorder, a national movement demanding police reform (https:// blacklivesmatter.com/), and strong anti-police sentiment (White et al., 2019). At the same time, there has been a series of high-profile ambush attacks on police officers, most notably in Brooklyn (two officers killed, December 2014), Dallas (12 officers shot, five officers killed, July 2016), and Baton Rouge (three officers killed, July 2016). The International Association of Chiefs of Police (IACP; 2013) states an ambush attack on police is defined by four characteristics: “an element of surprise; concealment of the assailant, their intentions, or weapon; suddenness of the attack; and a lack of provocation” (p. 3). Notably, it is often difficult to determine the motivation behind an ambush attack, and in fact, motive is not an element of the IACP definition of an ambush. These recent ambush killings of officers has led some police executives and outside observers to claim there is a “war on cops” (Hattem, 2015; Mac Donald, 2016; Safir, 2015). Although results from recent studies show significant declines in felonious killings of police since 1990 (Maguire et al., 2017; White et al., 2019), data from the FBI indicate an increasing percentage of those deaths are classified as ambush killings. From 1990 to 2012, the proportion of ambush killings increased by 33% (IACP, 2013). In 2016, ambush attacks against police reached a 10-year high (Ingraham, 2016). However, research suggests that there are varied motivations behind attacks on police, including ambushes, from revenge for a perceived transgression to attacks that are designed to avoid arrest or facilitate escape (Breul & Keith, 2016; Kercher et al., 2013). Although the IACP (2013) definition of ambush does not include motive as a core element, they do identify two categories of ambushes based on motive: (a) entrapment ambushes which are premeditated White 3 and (b) spontaneous ambushes which involve no planning. Each type meets the four defining characteristics of an ambush: surprise, concealment, suddenness, and a lack of provocation. The critical difference is the premeditation or planning. The “war on cops” thesis implies a very specific motivation for an ambush: hatred of police and/or desire to seek vengeance in response to police killings of citizens. The Brooklyn, Dallas, and Baton Rouge cases mentioned earlier are consistent with a “war on cops” explanation and fit IACP’s first category of a premeditated attack. Other recent ambushes of police, such as those in Florence (South Carolina; seven officers shot, two killed, October 2018), Houston (five officers shot, January 2019), and Philadelphia (six officers shot, August 2019), reflect IACP’s second category, with motives that are not clearly consistent with the “war on cops” thesis. For example, in the Houston and Florence cases, officers were serving search warrants. In the Philadelphia case, officers were serving an arrest warrant. All involved violent felons who reacted to direct police intervention, rather than a premeditated attack on officers. Although all felonious killings of police are tragic, of course, ambushes are conceptually, practically, and empirically significant. This distinction is especially important because ambush attacks on police are shocking to the conscience. Such attacks carry a symbolic weight that highlights the vulnerability of police and threatens a community’s conception of law and order. Perhaps more than any other event, ambush killings of police epitomize the notion of a “war on cops.” Such ambushes have the potential to dramatically alter police attitudes and behavior toward citizens and to shape the important debate over officer safety (Balko, 2015; Mac Donald, 2016). Unfortunately, prior research on violence against police is primarily descriptive, cross-sectional (Batton & Wilson, 2006), and often fails to examine ambush attacks separately. No studies have examined ambushes over an extended period of time, and the most recent study only analyzed the phenomenon through 2013 (Fachner & Thorkildsen, 2015). In this study, I explore this research gap using data from an open source, web-based clearinghouse of officer line-of-duty deaths to examine all ambush killings of the police over nearly 50 years (1970–2018). I apply a temporal coding scheme based on when the attack occurred to isolate two types of ambushes of police: (a) pre-encounter (officer had no knowledge of the suspect, was not engaged in any sort of police business, and had no opportunity to plan for engagement with the suspect) and (2) anticipation (officer who was engaged in formal police business [e.g., responding to a call for service], was generally aware of a suspect, but the officer had no contact with the suspect before the attack). Each fit the IACP’s (2013) general definition of an ambush, though I do not imply any sort of motive or assessment of premeditation.1 I trace long-term annual trends in both types of ambushes over the five-decade period, controlling for the number of officers working each year. I also compare those trends to rates of other types of felonious killings of police (i.e., nonambush). Last, I test 4 Police Quarterly 0(0) whether the rate of ambush killings of police has increased since 2013, shedding light on whether the recent trends support the “war on cops” thesis. Prior Research Data Sources and Prevalence Several different data sources capture felonious killings of police officers, including ambushes. The FBI’s Law Enforcement Officers Killed and Assaulted program (LEOKA) is published annually for all agencies that report to the Uniform Crime Reporting system (e.g., Bailey & Peterson, 1987; Jacobs & Carmichael, 2002; Kaminski & Stucky, 2009; Kent, 2010; Lester, 1984). Kaminski and Marvell (2002) compiled a database of felonious killings of police from 1903 to 1998, using information from the National Law Enforcement Officers Memorial Fund. The FBI’s National Incident-Based Reporting System also collects data on violence against police (e.g., Fridell et al., 2009; Willits, 2014), as does the National Institute for Occupational Safety and Health (NIOSH, 1996). The Centers for Disease Control and Prevention’s National Violent Death Reporting System captures information on all violent deaths, including police officers, though it does not stratify data by occupation. Researchers have also used unpublished FBI data (Bailey, 1982), obituaries, the National Death Index (e.g., Violanti et al., 1998), administrative data from one or more police departments (e.g., Gibbs et al., 2014; Rabe-Hemp & Schuck, 2007), and the Officer Down Memorial Page (an open source, webbased clearinghouse; Maguire et al., 2017; White et al., 2019). Felonious violence against police is a statistically rare event. Data from the FBI’s (2010) LEOKA program indicate that 57,268 officers were assaulted in 2009. Given police made 13.6 million arrests that same year (Snyder, 2011), violence against police occurred in less than one-half of 1% of all encounters ending in arrest. Moreover, the vast majority of attacks on police result in relatively minor injuries, and only a very small percentage result in death (Bierie, 2017). Although statistically rare, the enormous number of encounters between police and citizens means that assaults on police occur in approximately160 encounters per day (FBI, 2010). In terms of longer term trends in police victimization, data from LEOKA show that rates of police killings declined significantly from the early 1970s to the 1990s (Quinet et al., 1997). White et al. (2019) examined felonious killings of police from 1970 to 2016 and reported an 80% decline in deaths over the nearly five-decade period. The “War on Cops” Since 2014, several high-profile deadly force incidents involving minority citizens have intensified the focus on violence against police. The deaths of Michael White 5 Brown, Freddie Gray, and others led to public protests, civil disorder, a national movement demanding police reform, and in some cases, strong anti-police sentiment. The vitriolic rhetoric against officers, coupled with high-profile ambushes of police, led some police leaders, politicians, and conservative essayists to suggest there is a new “war on cops” (Hattem, 2015; Mac Donald, 2016; Safir, 2015). In 2016, Chuck Canterbury, national president of the Fraternal Order of Police sent an open letter to then-President Obama stating “It is not just talk; it is not just rhetoric . . . There is a very real and very deliberate campaign to terrorize our nation’s law enforcement officers” (Canterbury, 2016, p. 1).Schouten and Brennan (2016) note “Recent events in the United States and elsewhere raise the specter of LEOs [law enforcement officers] increasingly becoming the targets of planned acts of violence by those seeking to advance criminal enterprises, promote extremist causes, and express anti-police hostility” (p. 609). In fact, 58% of citizens surveyed in 2015 agreed that there is a “war on police” in the United States (Rasmussen, 2015). Those supporting the “war on cops” thesis point to a handful of high-profile, premeditated attacks on police as evidence of retaliatory violence. On July 7, 2016 Micah Johnson ambushed and shot 12 Dallas police officers, killing five of them. During negotiations with police after the shooting, Johnson stated “he was upset about Black Lives Matter . . . he was upset about recent police shootings . . . he was upset at white people . . . he wanted to kill white people, especially white officers” (Fernandez et al., 2016, p. 1). A few days later, Gavin Long shot six officers in Baton Rouge, killing three (Visser, 2016). In August 2019, six Philadelphia police officers were shot and wounded (Fortin & Bogel-Burroughs, 2019). A handful of studies have sought to directly test the “war on cops” thesis. Maguire et al. (2017) used interrupted time series analysis to study the prevalence of felonious police line-of-duty deaths from January 2010 through March 2016. They found no evidence that the events in Ferguson (or after) led to an increase in felonious killings of officers. White et al. (2019) reported similar results in their study of officer deaths from 1970 to 2016. Ambush Attacks on Police Neither of those studies examined ambush killings of police separately. Ambushes, particularly premeditated attacks, on police arguably represent the most direct measure of a “war on cops.” Although the FBI’s LEOKA records ambushes separately, there has been very little research on the phenomenon. Young (1990) examined all ambush killings of police from 1970 to 1987 (n ¼ 170) and tested whether ambushes were related to national murder rates or police line-of-duty deaths by other felonious circumstances. Noting the lack of a clear relationship, Young concluded “police deaths by ambush must be seen as a phenomenon distinct from national murder trends or other line-of-duty 6 Police Quarterly 0(0) deaths by murder” (p. 206). The IACP, U.S. Department of Justice, and CNA published a “fact sheet” in 2013 which defined two types of ambushes—entrapment and spontaneous—and examined LEOKA data from 1990 to 2012 (IACP, 2013). During that time period, spontaneous ambushes were twice as common as entrapment ambushes (68% vs. 32%). Although felonious killings of police, including ambushes, generally declined from 1990 to 2012, the report concludes, “officers murdered in the line of duty are increasingly likely to have been ambushed” (IACP, 2013, p. 5). Schouten and Brennan (2016) similarly classified ambushes of police into affective (or impulsive) violence and predatory (or planned) violence (see also Meyer et al., 1986). Gruenewald et al. (2016) used a mixed-method design to study attacks on police by far-right extremists. The authors identified four different motives (avoiding arrest, mission offense, defending property, and defending family), each of which could include ambushes of police. Prior research on police officer deaths consistently shows a wide range of motives for the attacks. Margarita (1980) examined New York Police Department (NYPD) officer deaths over a 100-year period and found that most killings involved rational suspects engaged in robberies who were fleeing the scene. Stone (2015) also found that most assailants were actively involved in the commission of another crime when they killed a police officer. Kercher et al. (2013) reported that about 15% of law enforcement officer homicides from 1996 to 2010 occurred during domestic disturbance calls. In 2016, when high-profile ambushes occurred in Dallas and Baton Rouge, approximately 75% of felonious killings occurred during nonambush encounters, such as domestic disturbance calls, traffic pursuits/stops, arrests, investigative activities, and tactical situations (e.g., serving warrants, barricaded persons; FBI, 2017). Fachner and Thorkildsen (2015) examined ambushes from 2007 to 2011 to explore whether “community, criminogenic, and organizational environments of police agencies influence the prevalence of ambushes against their officers” (p. 21). Notably, agencies that require some college education as a prehiring requirement and those that have in-car cameras experienced significantly fewer ambushes of their officers (Fachner & Thorkildsen, 2015). In a separate analysis of ambushes from 1990 to 2013, Fachner and Thorkildsen (2015, p. 3) concluded: Examining this data from the past 25 years shows that, like crime, reported ambush attacks against the police have decreased, falling steadily throughout the 1990s and remaining at a level of approximately 200 ambushes per year for the 2000s. This figure includes ambushes that were fatal, injurious, and noninjurious . . . None of these studies capture ambush attacks on police after 2013, and as a result, prior research on ambushes offers no insights on the current contentious White 7 dialogue regarding a potential “war on cops.” This study examines ambush killings of police from 1970 to 2018 to investigate three research questions: 1. What is the prevalence of ambush killings over the last 49 years? 2. How does the rate of ambush killings of police compare with rates of other types of felonious killings of officers? 3. Do recent trends in ambush killings of police support the “war on cops” thesis? Methods The Officer Down Memorial Page (ODMP; https://www.odmp.org/) includes information on all officer line-of-duty deaths dating back to the 1790s. Multiple student coders completed an abstraction form for each officer’s death from January 1, 1970 to December 31, 2018 (8,553 deaths). Coding occurred from February 2015 to May 2019. A 10% reliability check was conducted for each year to ensure coder consistency, and errors never exceeded 2% of variables coded in any given year. Inconsistencies in the data were resolved prior to analyses. The ODMP records 37 different causes of death, and felonious causes include assault, bomb, explosion, gunfire, poisoned, stabbed, terrorist attack, 9/11-related illness, vehicle pursuit, and vehicle assault. Given the focus of this study, several categories of deaths were removed from the data. First, all nonfelonious deaths were excluded from the analysis (e.g., car accidents, heart attacks). Also, suicides are not collected in the ODMP data. Second, the ODMP provides data on officer deaths for all U.S. territories (e.g., Puerto Rico, Guam). These line-ofduty deaths are excluded to remaining consistent with prior research (Maguire et al., 2017), and because these U.S. territories are not included in the Bureau of Justice Statistics (BJS) calculations of the total number of police officers. Police K-9 deaths are also excluded. The final data set includes 3,868 felonious killings of police from 1970 to 2018. Identification of Ambush Killings The ODMP does not record ambushes separately, nor does it consistently report assailant motive or planning. In order to identify ambush killings, coders recorded the timing of the felonious attack using a well-established temporal classification of police–citizen encounters (Binder & Scharf, 1980; Fyfe, 1986; Terrill, 2005). Each death was assigned one of the following temporal codes: 0 ¼ Pre-encounter [ambush] Ambush attack of an officer who had no knowledge of the suspect, was not engaged in any sort of police business and had no opportunity to plan for 8 Police Quarterly 0(0) engagement with the suspect (e.g., an officer sitting in a patrol car eating lunch; suspect approaches the car window and shoots the officer). 1 ¼ Anticipation [ambush] Ambush attack of an officer who was engaged in formal police business (e.g., responding to a call for service), was generally aware of a suspect, but the officer had no contact with the suspect before the attack (e.g., an officer responds to a call about a suspicious person in a park. The officer is searching the park when the suspect jumps out from behind a tree and shoots the officer). 2 ¼ Entry/Initial Confrontation [non-ambush] Officer has entered the immediate physical scene and is aware of the specific suspect. There is visual contact. There may also be an initial exchange of verbal commands before the attack occurs (e.g., an officer pulls a car over and is approaching the vehicle on foot when the suspect gets out and shoots the officer). 3 ¼ Dialogue/Information Exchange [non-ambush] The encounter has progressed beyond the initial entry/confrontation. There has been communication between the officer and the suspect (e.g., an officer stops and questions a pedestrian. During the questioning, the suspect pulls out a knife and fatally stabs the officer). Values 0 and 1 both capture ambush killings and are consistent with the IACP’s (2013) four defining criteria of an ambush: surprise, concealment, suddenness, and a lack of provocation. For example, the July 2016 attack on Dallas police officers is coded as a pre-encounter ambush. The July 2016 Baton Rouge ambush is a good example of an anticipation ambush: The officers were ambushed as they responded to a call for service about a man with a gun. The ODMP data rarely specifies motive and whether the assailant engaged in planning the attack on the officer. As a result, I cannot apply the ODMP’s (2013) dual classification of ambushes (premeditated, spontaneous) to these data. That said, there may be some degree of alignment between the IACP’s premeditated designation and this study’s pre-encounter ambush. The July 2016 Dallas case fits the criteria for both. The Baton Rouge case, an anticipation ambush, does not fit as evidence suggests the assailant planned the attack.2 As a result, the extent of that alignment is unknown. Two coders independently recorded and verified the assigned temporal code of each death using a double blind protocol. Whenever the coders disagreed on a temporal code, they met in-person to resolve the inconsistency. If they could not come to an agreement, the two coders and author met to discuss the case and White 9 determine the appropriate temporal code. If no agreement could be reached, the case was coded as missing. All other inconsistencies in the temporal code were resolved prior to analyses. Disputes of ambushes coded as 0 (pre-encounter) and 1 (anticipation) were rare. The majority of disagreements involved nonambush cases, often because the details provided were vague regarding the degree of contact and communication between the officer and suspect. Over the entire study period, the temporal classification of felonious deaths is as follows: 0 1 2 3 (pre-encounter ambush): 6.1% (n ¼ 206) (anticipation ambush): 20.9% (n ¼ 707) (entry/initial confrontation): 22.0% (n ¼ 745) (dialogue/information exchange): 50.9% (n ¼ 1,721) From 1970 to 2018, just over half of officer killings occurred at the dialogue/ information exchange of the encounter, and 22% occurred at the entry/initial confrontation stage. The remaining 27.0% of felonious killings were ambushes: either pre-encounter (6.1%) or at the anticipation phase (20.9%). Last, approximately 13% of the deaths (n ¼ 489) are coded as missing because the data did not include enough detail to make an accurate determination of the temporal classification. Older cases were more likely to be coded as missing, as the level of detail declines in past years. For example, 44% of missing cases occurred in 1980 or before. Only 12% of missing cases occurred after 2000. The missing data issue a limitation of the data source and should be considered when interpreting the results. Standardization of Ambushes by the Number of Officers Ambush killings are standardized as an annual rate of the number of police officers per year (i.e., annual rate of pre-encounter ambushes per 100,000 officers). Separate rates are calculated for pre-encounter and anticipation ambushes. The rationale for standardization by the number of officers is straightforward: The number of killings will be affected by the number of officers working and that number has varied significantly. In 1970, there were 276,056 officers in the United States. In 2017, there were 670,279. Standardization by the number of officers assures that any change in the number of ambush killings is not simply the result of more officers on the job. Standardized annual rates were calculated using data from the FBI (https://ucr.fbi.gov/crime-in-the-u.s) and BJS (https://www.bjs.gov/index.cfm? ty=tp&tid=7). Although recent data on the number of police officers are reliable, there is some concern about those measures historically (see Maguire et al., 1998). This is a limitation of this study, but the calculation of rates still represents an important step toward standardization. Also, at the time of this writing, 10 Police Quarterly 0(0) the number of officers working in 2018 was not available. The 2018 figure was calculated as an average of the prior three years. I also standardized officer deaths by the number of violent crimes per year, as prior research has demonstrated a link between community violence and attacks on police (Jacobs & Carmichael, 2002). The results did not vary. Analysis The current analysis focuses primarily on the 206 pre-encounter ambush killings and the 707 ambushes occurring at the anticipation phase. First, I use linear regression to examine longitudinal trends in the annual rate of each type of ambush killing of police per 100,000 officers. The annual rates of ambush killings (pre-encounter, anticipation phase) are the dependent variables and time is the independent variable. Second, I do the same analysis with nonambush killings of police (temporal codes of 2 and 3) as a comparison to the ambush rates. Last, I conduct AutoRegressive Integrated Moving Average (ARIMA) analyses to test whether the rate of ambush killings of police has increased since 2013. ARIMA is a quasi-experimental time series design that allows for comparison of preintervention and postintervention values of an outcome (in this case, ambush killings). ARIMA overcomes several threats to internal validity and violations of the independence assumption (e.g., serial correlation; McDowall & McCleary, 2014; McDowall et al., 1980). For this study, I tested interventions of varying onsets (abrupt, gradual) and durations (temporary, permanent), with both July 2013 and August 2014 as starting points. July 2013 is the month the Black Lives Matter was formed, following the acquittal of George Zimmerman for the killing of Trayvon Martin (https://blacklivesmatter.com/). August 2014 is the month that Michael Brown was in killed in Ferguson, Missouri by officer Darren Wilson. Prior studies have also used this data in ARIMA analyses of officer killings (Maguire et al., 2017; White et al., 2019). I first conducted ARIMA for pre-encounter ambushes only as they may be most consistent with “war on cops”-style attacks (again, without the ability to discern any motive or planning). The analysis is conducted with both annual (49 data points) and quarterly (196 data points) rates of pre-encounter ambushes. I also conducted ARIMA with both pre-encounter and anticipation ambushes combined (i.e., temporal codes of 0 and 1 combined). This analysis in conducted with annual, quarterly, and monthly rates of all ambushes.3 Results Figure 1 shows the annual rate of both types of ambushes killings of police per 100,000 officers. At the beginning of the study period, ambushes at the anticipation stage (solid line) far exceeded pre-encounter ambush attacks (dotted line). Notably, these trends are consistent with prior work by IACP (2013), White 11 Figure 1. Annual Rate of Felonious Ambush Killings of Police, 1970 to 2018, per 100,000 Police Officers. which reported that spontaneous ambushes outnumber premeditated attacks by two to one. Although motivation and planning are not accounted for in this study, the temporal classification of ambushes may track roughly with the IACP’s dual classification. The rate of pre-encounter ambushes declines precipitously during the 1970s and 1980s, by 94% (from 3.62 in 1970 to 0.21 in 1988). The rate increases slightly during the 1990s before declining to under 0.6 per 100,000 officers for most of the 2000s. There are two spikes near the end of the study period (2014 and 2016)—both are the highest rate since 1997. The rate of ambushes at the anticipation phase follows a similar pattern on a larger scale, declining by 93% between 1970 and the late 1980s. There are spikes in 2016 and 2018, both the highest rates in 20 years. Results from regression analysis show the decline is statistically significant for both pre-encounter (t ¼ 4.745; p < .001) and anticipation (t ¼ 6.724; p < .001) ambushes (see Table 1). Figure 2 shows the rates of nonambush killings during the same time period. These rates provide an important baseline for comparison with ambush killings. The rates of killings at both the entry/initial confrontation (dotted line) and dialogue/information exchange (solid line) stages follow a pattern similar to ambush attacks during the first part of the study period (1970–1971 through the late 1980s–early 1990s), declining by from 79% (dialogue/information exchange) to 88% (entry/initial confrontation). The rate of killings at the entry/initial confrontation stage increases slightly through 1994 before trending downward the remainder of the study period. Like ambushes, there are spikes in 12 Police Quarterly 0(0) Table 1. Results From Linear Regression and ARIMA. Linear regression Pre-encounter ambush (0) Anticipation ambush (1) Entry/initial confrontation (2) Dialogue/information exchange (3) ARIMA—pre-encounter onlya Annual analysis (1,1,0) Quarterly analysis (0,1,1) ARIMA—all ambushesa Annual analysis (2,0,0) Quarterly analysis (0,1,1) Monthly analysis (0,1,1) Standard error R2 B 0.309 0.479 0.569 0.627 0.033 0.135 0.131 0.211 0.007 0.020 0.016 0.023 4.745 6.724 8.024 9.029 0.275 0.136 0.074 0.003 0.196 0.011 0.377 (.708) 0.266 (.790) 0.447 0.629 0.178 0.332 0.004 0.004 1.139 0.034 0.014 0.291 (.772) 0.129 (.898) 0.256 (.798) t (p value) (>.001) (>.001) (>.001) (>.001) Note. ARIMA ¼ AutoRegressive Integrated Moving Average. a All models shown here were run with an abrupt, permanent impact beginning in August 2014 (when Michael Brown was killed in Ferguson, MO). Figure 2. Annual Rate of Felonious Non-Ambush Killings of Police, 1970 to 2018, per 100,000 Police Officers. White 13 2016 and 2018, but there is also a spike in 2011. The rate of killings at the last encounter stage trends downward from the 1990s through 2018, with the exception of 2001.4 Like ambushes, there are spikes at the end of the study period following a notable drop in 2014. There is also a spike in 2011. Results from regression analysis show the decline is statistically significant for both initial entry/confrontation (t ¼ 8.024; p < .001) and dialogue/information exchange (t ¼ 9.029; p < .001) killings (see Table 1). Interrupted Time Series Analysis I first carried out ARIMA with pre-encounter ambushes using both the annual rate (see Figure 1, dotted line) as well as the quarterly rate (see Figure 3).5 The first stage of ARIMA is model identification. For the annual analysis, the data support a model with a first-order autoregressive process (differenced; [1,1,0]).6 I tested interventions of varying onsets (abrupt, graduation) and durations (temporary, permanent), with both 2013 and 2014 as starting points. None of the models indicate a statistically significant change in the annual rate of preencounter ambush killings at or after 2013 and 2014. As an illustration, Table 1 shows the results of the model with an abrupt, permanent impact starting in Figure 3. Quarterly Rate of Pre-encounter Ambush Killings, January 1970 to December 2018. 14 Police Quarterly 0(0) 2014 and continuing through the end of the study period. The annual rate of preencounter ambush killings of police has not increased significantly since 2013. For the quarterly analysis, I employed the same approach with model identification (0,1,1)7 and the intervention, and again, none of the models were statistically significant. Table 1 shows the August 2014 abrupt, permanent intervention. I also ran ARIMA for all ambushes combined (pre-encounter and anticipation). The results for ARIMA with all ambushes, calculated as annual (model 2,0,0), quarterly (model 0,1,1), and monthly (model 0,1,1) rates, are shown at the bottom of Table 1.8 Interventions of varying onsets and durations were included in the models, and none reached statistical significance. Overall, there is no statistically significant increase in ambushes associated with the formation of Black Lives Matter or the death of Michael Brown in Ferguson. Discussion This study sought to determine the prevalence of ambushes of police over time and to assess whether recent trends in the phenomena support the “war on cops” thesis. Using a well-established temporal classification of police–citizen encounters that adheres to the IACP’s (2013) four-criteria definition of ambushes, I found the annual rate of ambush killings has declined by more than 90% over the last 50 years. The rate of pre-encounter ambush killings has remained low and stable since the late 1990s though the rate spiked in 2014 and 2016 to the highest levels witnessed since 1997. The trend in anticipation ambushes closely matches the pre-encounter ambush pattern (though they are more prevalent), with an additional spike in 2018. Results from ARIMA show the rate of ambush killings has not increased significantly since 2013. Recall the “war on cops” thesis implies a very specific motivation for attacks on police, and this study admittedly does not account for motivation. Nevertheless, the results presented here contradict the “war on cops” thesis and are consistent with prior research investigating post-Ferguson trends in attacks on police (Maguire et al., 2017; White et al., 2019). A review of the monthly data shows the 2016 and 2018 spikes are explained by three especially violent months: July 2016 (nine ambushes [five pre-encounter]), November 2016 (seven ambushes [two pre-encounter]), and February 2018 (six ambushes [no pre-encounter]). Moreover, the rates of nonambush felonious killings of police follow a similar pattern to ambush rates, with large declines over time and spikes in 2016 and 2018. The consistent pattern between ambush and nonambush killings of police post-2013 also refutes the “war on cops” thesis. In other words, the spikes in ambushes at the end of the study period are not unique; rather they co-occur with spikes in other types of felonious killings of police. Although there is no evidence supporting the “war on cops” thesis, there are notable spikes since 2014 in killings at all four temporal stages. Pre-encounter ambushes spiked in 2014 and 2016. Ambushes at the anticipation phase spiked White 15 in 2016 and 2018. There are also spikes in non-ambush killings at the entry/ initial confrontation (2016 and 2018) and dialogue/information exchange stages (2015, 2017, and 2018). A comparison of percent-change in average annual rates from 2009–2013 to 2014–2018 also demonstrates this point, as rates for killings at three of the four temporal stages increase notably: • • • • Pre-encounter ambushes: Anticipation ambushes: Entry/initial confrontation: Dialogue/information exchange: 25% increase 9% increase 11% decrease 22% increase It is unclear whether the recent trends in killings of police represent a short-term spike, or if the spikes are the start of a chronic problem. Ratcliffe (2019) describes spikes and chronic problems in the context of crime, but the scheme works equally well for other phenomena, including felonious killings of police. Chronic problems are long-term, “persistent challenges” while spikes are “a sudden and often surprising increase in, or appearance of, a . . . problem” (p. 40). Ratcliffe also highlights “false panics,” which may be legitimate spikes or they may be “false” and are particularly worrying because they disproportionately and “dramatically skew the public’s [and police officers’] sense of risk and concern” (p. 42). Are the recent increases in felonious killings of police an indicator of a chronic problem, a short-term spike, or a false panic? Regression and ARIMA results do not support the existence of a chronic problem. The data show large declines in killings of police over time, especially ambushes, with no statistically significant increase in the trend since 2013. The data also do not support a false panic. Killings of police officers remain a terrible reality in American policing, and the prevalence of ambush killings of police has reached levels not seen since the early 1990s. Figures 1 to 3 illustrate this fact. Spikes warrant close observation because they can presage the emergence of a chronic problem (i.e., will ambush killings increase in 2019–2020?). Two recent examples demonstrate how quickly a spike can become chronic problem. The first is the current opioid crisis. From July 2016 to September 2017, opiod overdoses increased by 30% across 45 different states. The Midwest was especially hardhit, as overdoses increased by 70% during the time period (VivoloKantor et al., 2018). In 2017, nearly 50,000 Americans died from an opioid overdose (https://www.drugabuse.gov/drugs-abuse/opioids/opioid-overdose-cri sis#three). The second example involves police officer suicides. In each of the last 3 years, officer suicides have outnumbered line-of-duty deaths (Lawrence, 2019; PoliceOne, 2019). There is a mental health crisis in policing (PoliceOne, 2019). Spikes warrant close observation because they may also disappear, as has sometimes occurred with crime. “Police can expend considerable effort dealing with a crime spike that, if left unattended, would likely return back to the normal level of crime after a short while” (Ratcliffe, 2019, p. 43). 16 Police Quarterly 0(0) There are several limitations to this study that warrant mention. First, the data are drawn from an open-source, public website. Deaths not reported on this website are not included in the data. However, several recent empirical studies have relied on the ODMP (Eliason, 2011; Johnson, 2013; Maguire et al., 2017; Varvarigou et al., 2014; White et al., 2019). Second, the available information on 12.6% of felonious deaths was not detailed enough to assign a temporal classification code, and these cases were excluded. Moreover, older cases tended to have more missing information than recent deaths. Third, “war on cops” attacks on police imply a very specific motivation, and unfortunately, the ODMP data does not routinely report assailant motive, nor does it report on premeditation. This limitation highlights the difficulty in determining the very specific motive behind “war on cops” attacks of police. Rather, this study employs a temporal classification that isolates two types of ambushes based on when during the encounter the attack occurred. This temporal classification represents an alternative mechanism for identifying deaths that meet IACP’s (2013) general definition of ambushes, without having to assess motivation. Last, this study focused only on ambushes resulting in officer deaths. Nonfatal ambush attacks are not included in this study. Much of the decline in fatal ambush attacks since 1970 may be explained by changes in police policy and practice, including better training and advances in body armor, as well as improvements in emergency medical treatment. The long-term and recent trends in nonfatal ambush attacks on police are not known. Conclusion Ambush killings represent the most severe, traumatic form of violence against police. There has been a long-term, large decline in ambush killings of police, and there is no evidence to support the recent “war on cops” thesis. However, there are post-2013 spikes in killings of police at three of the four temporal stages, including both types of ambushes. These spikes represent a troubling trend that warrants close monitoring by police practitioners and researchers. The persistence of these spikes in 2019 and beyond may have significant policy, training, operational, and health implications for police, as well as for the communities they serve. Acknowledgments The author would like to thank Dr. Lisa Dario and Dr. John Shjarback for their contributions to the larger Assessing Dangerousness in Policing project. The author would also like to thank Michaela Flippin and Connor Stewart for their data collection efforts, and Dr. Aili Malm and three anonymous reviewers for their comments on the paper. White 17 Declaration of Conflicting Interests The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author received no financial support for the research, authorship, and/or publication of this article. Notes 1. It may be that the two types of ambushes examined here roughly align with the IACP’s (2013) dual classification of premeditated and spontaneous. However, motive or degree of planning is rarely evident in the Officer Down descriptions of cases. 2. Baton Rouge was in turmoil from the police killing of Alton Sterling on July 5, 2016, just 2 weeks prior to the ambush attack on officers. Investigators later discovered that the attacker, Gavin Long, had searched online for the home addresses of the officers who killed Sterling (Elliott, 2017). He made several online posts leading up to the attack calling for violence against police, and he left a suicide note describing the “destruction I must inflict upon cops” (Elliott, 2017, p. 2). 3. There are not enough pre-encounter ambushes to support a monthly analysis (206 ambushes over 588 months). With both types of ambushes combined, the monthly prevalence is still quite low—913 ambushes over 588 months—but I conducted the analysis to be exhaustive. 4. Officers killed on September 11, 2001 were coded as a “3” given the deaths occurred during the collapse of the World Trade Center buildings (or after), as officers responded to the terrorist attack (or died in the months or years after as a result being on scene). 5. For the quarterly rate, the same annual number of police officers is used as the denominator for each quarter of a given year. 6. The autocorrelation function (ACF) has a decaying trend and the partial autocorrelation function (PACF) has a single spike. 7. The ACF has a single spike and the PACF has a decaying trend. 8. For the annual data, the ACF has a decaying trend and the PACF has a second-order spiking trend. For the quarterly and monthly data, the ACF has a single spike and the PACF has a decaying trend. ORCID iD Michael D. White https://orcid.org/0000-0002-6151-6399 References Balko, R. (2015). Once again: There is no “war on cops.” And those who claim otherwise are playing a dangerous game. Washington Post. https://www.washingtonpost.com/ news/the-watch/wp/2015/09/10/once-again-there-is-no-war-on-cops-and-those-whoclaim-otherwise-are-playing-a-dangerous-game/ 18 Police Quarterly 0(0) Bailey, W.C. (1982). Capital punishment and lethal assaults against police. Criminology, 19, 608–625. Bailey, W.C., & Peterson, R.D. (1994). Murder, capital punishment, and deterrence: A review of the evidence and an examination of police killings. Journal of Social Issues, 50, 53–73. Batton, C., & , & Wilson, S. (2006). Police murders: An examination of historical trends in the killing of law enforcement officers in the United States, 1947 to 1998. Homicide Studies, 10(2), 79–97. Bierie, D. M. (2017). Assault of police. Crime and Delinquency, 63(8), 899–925. Binder, A., & , & Scharf, P. (1980). The violent police–citizen encounter. Annals of the American Academy of Political and Social Science, 452, 111–121. Breul, N., & Keith, M. (2016). Deadly calls and fatal encounters: Analysis of U.S. law enforcement line of duty deaths when officers responded to dispatched calls for service and conducted enforcement (2010–2014). Washington, DC: National Law Enforcement Officers Memorial Fund. Canterbury, C. (2016). Letter from National FOP President Canterbury to President Obama. Washington Examiner. Feb. 15. Retrieved November 20, 2019 from washingtonexaminer.com/obama-silence-on-cop-deaths-ripped-ill-save-you-a-spot-atfunerals-mocks-sheriff/article/2583289. Duhart, D. T. (2001). Violence in the workplace, 1993-99. US Department of Justice, Office of Justice Programs. Eliason, S. L. (2011). Death in the line of duty: Game warden mortality in the United States, 1886–2009. American Journal of Criminal Justice, 36(4), 319–326. Elliott, D. (2017). Gunman who ambushed Baton Rouge officers searched for police home addresses. NPR. https://www.npr.org/sections/thetwo-way/2017/06/30/535093321/ gunman-who-ambushed-baton-rouge-officers-searched-for-police-home-addresses Fachner, G., & , & Thorkildsen, Z. (2015). Ambushes of police: Environment, incident dynamics, and the aftermath of surprise attacks against law enforcement. Office of Community Oriented Policing Services. Federal Bureau of Investigation. (2010). Law enforcement officers killed and assaulted, 2009. U.S. Department of Justice. Federal Bureau of Investigation. (2017). Law enforcement officers killed and assaulted, 2016. U.S. Department of Justice. Fernandez, M., Perez-Pena, R., & , & Bromwich, J. E. (2016). Five Dallas officers were killed as payback, Police Chief says. New York Times. https://www.nytimes.com/2016/ 07/09/us/dallas-police-shooting.html Fortin, J., & , & Bogel-Burroughs, N. (2019). Philadelphia police shooting: What we know about the standoff and the suspect. New York Times. https://www.nytimes.com/ 2019/08/15/us/philadelphia-shooting-maurice-hill.html Fridell, L., Faggiani, D., Taylor, B., Brito, C. S., & , & Kubu, B. (2009). The impact of agency context, policies, and practices on violence against police. Journal of Criminal Justice, 37(6), 542–552. Fyfe, J. J. (1986). The split-second syndrome and other determinants of police violence. In A. T. Campbell & & J. J. Gibbs (Eds.), Violent transactions (pp. 207–225). Basil Blackwell. Gruenewald, J., Dooley, K. M. G., Suttmoeller, M. J., Chermak, S. M., & , & Freilich, J. D. (2016). A mixed method analysis of fatal attacks on police by far-right extremists. Police Quarterly, 19(2), 216–245. White 19 Gibbs, J.C., Ruiz, J., & Klapper-Lehman, S.A. (2014). Police officers killed on duty: Replicating and extending a unique look at officer deaths. International Journal of Police Science and Management, 16, 277–287. Hattem, J. (2015). Killings of police officers on the rise. The Hill. https://thehill.com/ policy/national-security/241594-fbi-officer-killings-on-the-rise Ingraham, C. (2016). Ambush killings of police officers has hit a 10-year high. Washington Post. https://www.washingtonpost.com/news/wonk/wp/2016/11/21/ ambush-killings-of-police-officers-has-hit-a-10-year-high/?noredirect=on International Association of Chiefs of Police. (2013). Ambush fact sheet. Jacobs, D., & , & Carmichael, J. T. (2002). Subordination and violence against state control agents: Testing political explanations for lethal assaults against the police. Social Forces, 80(4), 1223–1251. Johnson, R. R. (2013). A longitudinal examination of officer deaths from vehicle pursuits. International Journal of Police Science and Management, 15, 77–94. Kaminski, R.J. & Marvell, T.B. (2002). A comparison of changes in police and general homicides: 1930–1998. Criminology, 40, 171–190. Kaminski, R.J., & Stucky, T.D. (2009). Reassessing political explanations for murders of police. Homicide Studies, 13, 3–20. Kercher, C., Swedler, D.I., Pollack, K.M., & , & Webster, D. (2013). Homicides of law enforcement officers responding to domestic disturbance calls. Injury Prevention, 19, 331–335. Kent, S.L. (2010). Killings of police in US cities since 1980: An examination of environmental and political explanations. Homicide Studies, 14, 3–23. Lawrence, E. (2019). Police officer kills himself, the 9th NYPD death by suicide this year. https://www.usatoday.com/story/news/nation/2019/08/15/nypd-suicide-9th-police-offi cer-dies-amid-mental-health-crisis/2017119001/ Lester, D. (1984). The murder of police officers in American cities. Criminal Justice and Behavior, 11, 101–113. Mac Donald, H. (2016). The war on cops: How the new attack on law and order makes everyone less safe. Encounter Books. Maguire, E. R., Nix, J., & , & Campbell, B. A. (2017). A war on cops? The effects of Ferguson on the number of U.S. police officers murdered in the line of duty. Justice Quarterly, 34(5), 739–758. Maguire, E., Snipes, J., Uchida, C. & & Townsend, M. (1998). Counting cops: Estimating the number of police departments and police officers in the USA. Policing: An International Journal, 21(1), 97–120. Margarita, M. (1980). Killing the police: Myths and motives. Annals of the American Academy of Political and Social Science, 452, 63–71. McDowall, D., & , & McCleary, R. (2014). Interrupted time series models. In G. Bruinsma & & D. Weisburd (Eds.), Encyclopedia of criminology and criminal (pp. 2653–2665). Springer. McDowall, D., McCleary, R., Meidinger, E. E., & , & Hay, R. A., Jr. (1980). Interrupted time series analysis. SAGE. Meyer, C. K., Magedanz, T. C., Feimer, S. H., Chapman, S. G., & , & William Pammer, W. J. (1986). Ambush-related assaults on police: Violence at the street level. Charles C. Thomas. 20 Police Quarterly 0(0) National Institute for Occupational Safety and Health. (1996). Violence in the workplace, risk factors and prevention strategies. Current Intelligence Bulletin 57. U.S. Department of Health and Human Service, CDC. PoliceOne. (2019). Report: More cops died by suicide than in line of duty in 2018. https:// www.policeone.com/lodd/articles/482561006-Report-More-cops-died-by-suicide-th an-in-line-of-duty-in-2018/ Quinet, K.D., Bordua, D.J., & Lassiter, W. (1997). Line of duty police deaths: A paradoxical trend in felonious homicides in the United States. Policing & Society, 6, 283–296. Rabe-Hemp, C.E., & Schuck, A.M. (2007). Violence against police officers: Are female officers at greater risk? Police Quarterly, 10, 411–428. Rasmussen. (2015). 58% think there’s a war on police in America Today [Press release]. http://www.rasmussenreports.com/public_content/lifestyle/general_lifestyle/august_ 2015/58_think_there_s_a_war_on_police_in_america_today Ratcliffe, J. H. (2019). Reducing crime: A companion for police leaders. Routledge. Safir, H. (2015). Police commissioner: Bad old days of crime may be back. Time Magazine. https://time.com/3908254/howard-safir-crime/ Schouten, R., & , & Brennan, D. V. (2016). Targeted violence against law enforcement officers. Behavioral Sciences and the Law, 34, 608–621. Snyder, H.N. (2011). Arrest in the United States, 1980-2009. Washington, DC: Bureau of Justice Statistics. Stone, M. H. (2015). Men who kill policemen. Violence and Gender, 2(1), 87–97. Terrill, W. (2005). Police use of force: A transactional approach. Justice Quarterly, 22(1), 107–138. Varvarigou, V., Farioli, A., Korre, M., Sato, S., Dahabreh, I. J., & , & Kales, S. N. (2014). Law enforcement duties and sudden cardiac death among police officers in United States: Case distribution study. BMJ, 349, g6534. Visser, S. (2016). Baton Rouge shooting: 3 officers dead; shooter was Missouri man, sources say. CNN. https://www.cnn.com/2016/07/17/us/baton-route-police-shooting/ index.html Violanti, J.M., Vena, J.E., & Petralia, S. (1998). Mortality of a police cohort: 1950–1990. American Journal of Industrial Medicine, 33, 366–373. Vivolo-Kantor, A. M., Seth, P., Gladden, R. M., Mattson, C. L., Baldwin, G. T., KitePowell, A., & , & Coletta, M. A. (2018). Vital signs: Trends in emergency department visits for suspected opioid overdoses–United States, July 2016-September 2017. Centers for Disease Control and Prevention. Willits, D.W. (2014). The organizational structure of police departments and assaults on police officers. International Journal of Police Science and Management, 16, 140–154. White, M. D., Dario, L. M., & , & Shjarback, J. A. (2019). Assessing dangerousness in policing: An analysis of law enforcement officer deaths in the United States, 1970–2016. Criminology and Public Policy, 18(1), 11–35. Young, T. J. (1990). Felonious murder of police officers by ambush. Psychological Reports, 67(1), 206–206. Author Biography Michael D. White is a professor in the School of Criminology and Criminal Justice at Arizona State University (ASU) and is an associate director of White 21 ASU’s Center for Violence Prevention and Community Safety. He received his PhD in Criminal Justice from Temple University in 1999. Prior to entering academia, he worked as a deputy sheriff in Pennsylvania. His primary research interests involve the police, including use of force, technology, and misconduct. His recent work has been published in Criminology, Police Quarterly, Criminology and Public Policy, and Criminal Justice and Behavior.