Fatal outcomes of militarization: Re-examining the relationship between the 1033 Program and police deadly force

Highlights

• Value of 1033 acquisitions has a significant, negative relationship using raw data.

• Number of 1033 acquisitions has a modest, positive relationship using cleaned data.

• 1033 Program effects disappear when examining the relationship by agency size.

• Percentage of Black population predicts fatal force in all models but small agencies.

• Additional measures of militarization and police behavior are recommended.

Abstract

Purpose

Prior literature has examined police militarization by operationalizing the materiel dimension using the 1033 Program – a U.S. military surplus transfer program – and assessing its relationship with police fatal force. Given mixed findings, the present study demonstrates the importance of data considerations while examining this relationship and any assessment of militarization or fatal force.

Methods

The present study uses the 1033 Program data and the crowdsourced Fatal Encounters database to conduct multi-level mixed effects negative binomial regression models using both raw and cleaned data.

Results

While the cost of 1033 Program items was negatively associated with fatal force in the raw models, this relationship disappears using the cleaned data. The cleaned data models demonstrate a moderate, positive relationship between the number of 1033 Program items obtained and fatal force, though this relationship disappears when agencies are disaggregated by size.

Conclusion

These findings demonstrate the importance of data selection, validation, and categorization, given that previous literature examining this relationship vary in their extent of data cleaning and the data sources used. These findings also demonstrate the necessity for militarization research to examine police outcomes beyond fatal force, and to account for additional measures of militarization beyond the 1033 Program.

Introduction

An increasing amount of attention has been directed towards the phenomenon of police militarization, growing with events in the United States such as the Ferguson riots and coming to greater national attention with filmed instances of police responses to Black Lives Matter demonstrations protesting the deaths of George Floyd and Breonna Taylor in 2020. Beyond a growing literature examining public perceptions of militarization (Fox, Moule Jr., & Parry, 2018; Moule Jr., Burruss, Parry, & Fox, 2019a; Moule Jr., Fox, & Parry, 2019b), the phenomenon of police militarization is a relatively under-researched subfield with little established consensus regarding its empirical measurement (Koslicki & Willits, 2018). It is therefore essential that emerging police militarization literature is theoretically sound (Kappeler & Kraska, 2015), attentive to the nuances of the data used (Koslicki, 2017), and examines specific outcomes theoretically linked to increased militarization (Bieler, 2016). Recent research in the fields of political science and economics have addressed Bieler's (2016) call to examine outcomes of indicators of police militarization by exploring the impact of the U.S. Defense Logistics Agency's (DLA) Law Enforcement Support Office (LESO) Program – a military surplus dispersion program more commonly known as the 1033 Program – on the number of civilians killed by United States police (Bove & Gavrilova, 2017; Harris, Park, Bruce, & Murray, 2017; Delehanty, Mewhirter, Welch, & Wilks, 2017). While these authors seek to contribute to the field of police militarization with empirical, outcome-driven research, their findings – taken as a whole – are inconclusive regarding the effect of 1033 Program participation on police use of lethal force. We posit that these inconclusive findings may be due to the problems inherent to the data sources used and the lack of a consistent data handling strategy across studies, which may have led to inaccurate results and conclusions. It is prudent therefore to not only retest the relationship between 1033 Program participation and its effect on police lethal force, but to also engage in a broader discussion regarding the methodological concerns surrounding these data sources as their use becomes more widespread across several academic disciplines. In the current study, we re-examine this relationship using 1033 data and data from fatalencounters.org (FE). We estimate models with both cleaned and uncleaned data, and examine the effect of measuring 1033 acquisitions by cost versus by amount of items obtained, demonstrating the importance of carefully managing and using data sources to study this broader social problem.

While police militarization has quickly become a current point of discussion since the events of Ferguson and the law enforcement response 2020 Black Lives Matter demonstrations across the nation, the study of police militarization in the United States was first publicized in the 1970s, with scholars from the University of California at Berkeley asserting that police militarization and community policing were two coherent strategies, with community policing as the “velvet glove” concealing the “iron fist” of the militarized police institution (Center for Research on Criminal Justice, 1977). The study of police militarization in the United States was further spearheaded by Kraska, Kappeler, and colleagues in the 1990s (see Kraska & Cubellis, 1997; Kraska & Kappeler, 1997; Kraska & Paulsen, 1997).

Kraska (2007) has developed a four-dimension framework for the assessment of police militarization, with militarization examined through (1) materiel indicators (examining the extent to which the police use military weapons and equipment), (2) cultural indicators (the extent of military language, appearance, and values used by the police), (3) organizational indicators (the extent to which military arrangements are used, such as paramilitary policing units [PPUs], or SWAT teams), and (4) operational indicators (the extent to which police operations are patterned after those used by the military). While the majority of police militarization scholars agree that the police have always been militarized to some degree due to their paramilitary roots and state-sanctioned ability to use force (Kappeler & Kraska, 2015; Koslicki, 2020; Kraska, 2007; Maguire & King, 2004), Kraska's (2007) four indicators assist in determining the extent to which the police are militarized upon a continuum of low to high police militarization.

In a state-of-the-field review of police militarization literature, Bieler (2016) observes that while some of the indicators in Kraska's (2007) framework are easily quantifiable – such as numbers of military equipment transfers or the presence of PPUs in departments – few studies empirically examine whether these quantifiable indicators lead to observable changes in police behavior. Indeed, while a number of scholars have examined the transfer of military equipment to the police (see D'Esposito, 2016; Endebak, 2014; Haggerty & Ericson, 1999; Hall & Coyne, 2013; Johnson & Hansen, 2016), few have assessed whether the acquisition of this equipment affects real change in the behavior of recipient departments. Recent research has assessed factors that increase an agency's likelihood of acquiring specific categories of equipment through the 1033 Program, with Burkhardt and Baker (2018) finding that agencies with “warrior” tendencies (measured through agencies' body armor policies and special units) and that use asset forfeiture were significantly more likely to acquire a mine-resistant ambush-protected (MRAP) vehicle. Ramey and Steidley (2018) likewise assessed the influence of violent crime rates, drug arrest rates, and proportion of minority population on agencies' participation in the 1033 Program, finding that high violent crime rates and high proportion of Black population increased an agency's likelihood to obtain any equipment from the 1033 Program.

The few studies that currently exist that measure the effect of 1033 Program participation on agency behaviors (and other outcomes) produce mixed findings. For example, Harris et al. (2017) found that acquiring tactical items (or that which the police militarization literature would consider militarized; discussed at length in the following section) led to decreases in citizen complaints, as well as decreases in several crime categories (robbery, assaults, and vehicle thefts) and increases in drug-related arrests. Harris et al. (2017) also found no effect on suspect deaths, which the authors argue is indicative of a null effect of tactical military gear on officer violence. Similarly, Bove and Gavrilova (2017) found military vehicles and equipment such as special clothing and night vision (what Bove and Gavrilova categorize as “gears”) obtained from the 1033 Program to have a significant, negative impact on most crime rates, as well as a null effect on suspect deaths. The authors of both studies argue that tactical or visibly military items obtained from the 1033 Program may have a deterrent effect on local crime rates and no appreciable affect on police use of force. It must be noted, however, that Bove and Gavrilova (2017) claim to have obtained data on suspect deaths from the Federal Bureau of Investigation's (FBI) Law Enforcement Officers Killed and Assaulted (LEOKA) data, which does not include information on suspect deaths (only those of law enforcement officers). Harris et al. (2017) claim to have obtained data of suspect deaths from Bove and Gavrilova. As such, it is difficult to assess the validity of these results without knowing how suspect deaths were measured or where the data were obtained.

Contrary to the findings of Bove and Gavrilova (2017) and Harris et al. (2017), research by Delehanty et al. (2017) has found a positive relationship between participation in the 1033 Program and police use of lethal force. Delehanty et al. (2017) argue that this military equipment (Kraska's materiel dimension) will introduce martial language, lead to a willingness to engage in high-risk operations, and lead to a higher likelihood to create military arrangements like PPUs, thus affecting all aspects of Kraska's dimensions of police militarization (and resulting, ultimately, in higher levels of police use of force and violence). Mirroring this concept and approach, Lawson Jr. (2019) likewise found a significant, positive relationship between 1033 Program participation and police lethal force when examining a larger sample at the agency level (no significance was found, however, in Lawson's zero-inflation model).

While these four recent studies provide a considerable step towards understanding the effects of the materiel dimension on police militarization, their inconclusive findings demand further assessment of the 1033 Program on police behavioral outcomes – particularly the use of lethal force. The disagreement of these studies' findings regarding suspect deaths likewise demands further examination into data sources and handling. Specifically, we anticipate that the inconclusive findings may be explained in part by these studies' differences in data sources used, as well as their treatments of these data (including their categorization and cleaning considerations).

The 1033 Program allows law enforcement agencies to obtain surplus equipment from the military. However, it is important to note that not all equipment provided in the 1033 program is militaristic in nature. Indeed, many of the items obtained by law enforcement agencies from this program are not what would be commonly considered as militarized gear when comparing these items to Kraska's (2007) definition of the materiel dimension (Johnson & Hansen, 2016; Koslicki & Willits, 2018; Ramey & Steidley, 2018). These non-military, or general, items available through the 1033 Program include categories such as furniture, office supplies, exercise equipment, first aid and medical equipment, and miscellaneous tools and industrial parts (Else, 2014; Koslicki & Willits, 2018; Ramey & Steidley, 2018).

While it could be argued that these general items should still be considered militarized, as they were obtained from a direct military source, prior literature has operationalized Kraska's (2007) materiel dimension largely by examining indicators such as armored vehicles, firearms (often long rifles, though some researchers include all firearms), uniforms of military appearance or origins, and body armor or other wearable gear of paramilitary appearance (such as night vision, riot gear and ballistic helmets) (Bieler, 2016; Burkhardt & Baker, 2018; Carriere & Encinosa, 2017; Institute for Intergovernmental Research, 2015; Johnson & Hansen, 2016; Koslicki & Willits, 2018; Kraska & Kappeler, 1997; Maguire & King, 2004; Ramey & Steidley, 2018). We therefore propose that the inclusion of the more non-military or general items obtained from the 1033 Program into an analysis determining the effects of military equipment on police use of fatal force may obfuscate the effects of gear more commonly identified by the literature as militaristic on police use of fatal force. The aforementioned study conducted by Delehanty et al. (2017) included all items obtained from the 1033 Program in the models, while Lawson Jr. (2019) likewise assessed 1033 Program materiel based on US dollar (USD) value, reasoning that the more expensive items are those that are higher in technological sophistication and more visibly militaristic. Therefore, it is prudent to retest their research question after categorizing the 1033 Program data to parse out the effects of an LEA receiving equipment commonly considered to be militarized on their use of fatal force, as opposed to their acquisition of any items through the 1033 Program.

Police lethal force data is equally difficult to measure. As previously discussed, two of the aforementioned studies of the 1033 Program's effect on suspect deaths utilized LEOKA data; however, the LEOKA dataset does not contain information on assaults or deaths beyond those experienced by law enforcement officers (though these data can still be useful in examining other policing outcomes, like officer safety, as also examined by Bove and Gavrilova). Historically, most researchers have made use of one of three data sources to study police lethal force: the National Vital Statistics System (NVSS), Arrest Related Deaths (ARD) (a subprogram of the Death in Custody Reporting Program [DCRP]), or Supplementary Homicide Report (SHR) (Banks, Couzens, Blanton, & Cribb, 2015; Loftin, Wiersema, McDowall, & Dobrin, 2003; Mumola, 2007). It is important to note, however, that there is currently no federal policy mandating U.S. law enforcement agencies to report suspect or arrest-related deaths; agency reporting of these statistics to the U.S. Bureau of Justice Statistics is currently voluntary (President's Task Force on 21st Century Policing, 2015). Perhaps unsurprisingly, there is substantial within-agency variation in police lethal force rates across these sources (Loftin et al., 2003; Mumola, 2007). Furthermore, Klinger (2012) has demonstrated that SHR estimates do not match internal agency statistics and, consequently, has argued that these data are ill-suited to the study of police lethal violence.

Due to the lack of a national, comprehensive reporting and data collection policy, some scholars, such as Delehanty et al. (2017), Lawson Jr. (2019), and Nix, Campbell, Byers, and Alpert (2017), have turned to crowdsourced datasets such as the Washington Post's Killed by Police database, Mapping Police Violence, and Fatal Encounters. Fatal Encounters (FE) provides one of the most comprehensive databases that includes descriptive information on incidents involving police use of lethal force, as the Washington Post data only counts shooting-related fatalities, and FE includes more fatalities per year than Mapping Police Violence. However, even a cursory examination of the FE data reveals that it captures incidents which many researchers would not typically count as instances of police use of force. As such, while these data likely do a better job of tracking lethal police force than prior official data sources, it is clear that these data must be verified and cleaned prior to use. Indeed, the inclusion criteria that Fatal Encounters adheres to is unclear, making it crucial for researchers to understand the type of incidents included.

Put simply, there are major issues related to both the measurement of materiel police militarization and police lethal force. Neither the 1033 data, nor newly available crowdsourced datasets (like FE) are clean and ready for analysis upon download. As such, we argue that very little is actually known about the relationship between police militarization and police use of force. Here, we describe how we manage both of these data sources, then provide an analysis of the raw (uncleaned) and cleaned data, examining the data both in terms of USD value and amount of 1033 equipment obtained. Our analysis has two goals: 1) to demonstrate the importance of data cleaning and considerations and 2) to provide robust empirical evidence regarding the link between 1033 Program acquisitions and police lethal force.