Excessive Force or Armored Restraint? Government Mechanization and Civilian Casualties in Civil Conflict

Abstract

Does increasing counterinsurgent mechanization result in higher levels of unintentional civilian casualties? Existing research on unintentional civilian victimization in recent conflicts has focused on air strikes, but this question remains unexplored in research examining counterinsurgent force structure for ground units. However, a host of counterinsurgency practitioners in Iraq have cited the mechanized forces’ effectiveness in delivering precision fires that limit civilian casualties. We propose an armored restraint theory, suggesting that mechanized crews’ armored protection enhances soldiers’ decision space when making the consequential choice to employ lethal force. When this enhanced decision space is combined with units that systematically respect jus in bello principles and non-combatant immunity norms, it results in armored restraint, which may reduce government-caused civilian casualties in civil conflicts. We test this theory using micro-data from Iraq and find mechanized units are associated with significantly lower civilian casualty levels compared to dismounted units.

Keywords

counterinsurgency mechanization military force structure civilian casualties Iraq

Introduction

Following World War I and the dawn of modern combined arms warfare, military modernization efforts spawned mechanized forces around the world (Biddle 2004, 28). Mechanized units proved decisive in critical World War II battles and brought soldiers increased protection, firepower, and mobility.¹ Militaries continued mechanizing into the Cold War (Sechser and Saunders 2010, 485), yet rates of interstate conflicts decreased while rates of civil conflicts steadily increased (Pettersson, Högbladh, and Öberg 2019, 590). Modern militaries, structured for conventional warfare, were often deployed to these civil conflicts to combat rebellions and insurgencies, frequently operating around civilian populations (Caverley and Sechser 2017; Dell and Querubin 2018; Lyall and Wilson 2009).

Noting these incompatibilities, the prevailing narrative cautions against employing mechanized units as counterinsurgents, instead preferring light infantry who can better access local intelligence and act with increased precision (Galula 2006; Leites and Wolf 1970; Mack 1975; Lyall and Wilson 2009). Some have further suggested that mechanized counterinsurgents’ substantial firepower leads to increasing rates of civilian victimization (Levy 2019; Wallace 2007). However, several counterinsurgency practitioners from recent conflicts have argued that mechanized units were critical to selectively clearing insurgent held locations in Iraq and Afghanistan.² Further, while mechanized firepower has increased militaries’ ability to inflict widespread violence, improved target acquisition technology means mechanized units can be more precise when engaging targets.

When taken together, how do these mechanization trends relate to civilian casualty rates in civil conflicts? To date, research analyzing government mechanization in civil conflicts has consisted of large-N, inter-state studies assessing counterinsurgent effectiveness, and civil conflict duration.³ Separately, recent research examining unintentional civilian casualties has primarily focused on Western reliance on precision air strikes, while neglecting ground forces’ relationship to unintentional civilian deaths (Cronin 2018; Kaempf 2018). There has not been a thorough examination to determine if variation in counterinsurgent force structure (between mechanized, motorized, and light infantry units), impacts local civilian casualty rates.⁴ Omitting ground-dynamics that lead to civilian casualties in recent conflicts is problematic and data collected from Operation Iraqi Freedom (OIF) suggests 62% of government-caused civilian casualty incidents were exclusively linked to ground-forces.⁵ Better understanding the systematic causes behind civilian casualties has important policy and normative implications, and research on this subject may better protect civilians in conflict zones.

Reviewing available evidence, this study provides an initial attempt at exploring how counterinsurgent force structure impacts civilian casualty rates. We propose a crew-level, armored restraint theory, suggesting that mechanized crews’ armored protection enhances soldiers’ decision space when making the consequential choice to employ lethal force. When this enhanced decision space is combined with units that systematically respect jus in bello principles and non-combatant immunity norms, it results in armored restraint, which may reduce civilian casualty rates in civil conflicts. While dismounted units may adhere to the laws of war, they do not have similar protected decision space and may not be able to act with similar restraint during tactical engagements out of self-defense necessities.

OIF offers a unique opportunity to test this hypothesis in a micro-setting. Coalition mechanized, motorized, and light infantry units were stationed across Iraq amongst civilian populations following the 2003 invasion. We employ new, spatially-weighted mechanization variables, derived from Carrie Lee’s Iraq Order of Battle database, which documents Coalition forces’ composition and disposition in OIF from 2004 to 2008.⁶ Drawing on Iraqi district-level civilian casualty data, we empirically analyze the relationship between mechanization and Coalition-caused civilian casualties (Condra and Shapiro 2012). The results suggest that mechanized forces are associated with lower levels of civilian casualties compared to dismounted units. While this research provides an initial examination of previously unexplored dynamics, our findings suggest initial support for the armored restraint theory. We conclude with recommendations for future research and policy implications.

Literature Review: Mechanization and Civilian Casualties

Civilian Victimization in Civil Conflict

Relatively little research has examined unintentional civilian casualties and the contemporary literature has predominately focused on intentional violence against civilians during civil conflicts (Valentino 2014, 89). Cross-national analyses found that governments intentionally target civilians out of desperation or strategic necessity (Downes 2006; Valentino, Huth, and Croco 2006). Others posit that governments may resort to indiscriminate targeting if counterinsurgents lack the detailed intelligence required to identify insurgents living amongst the local population (Kalyvas 2006). These works explain how rational combatants may intentionally target civilian populations during civil conflict (Downes 2007; Fjelde and Hultman 2014; Hultman 2012; Lyall 2009; Salehyan, Siroky, and Wood 2014; Toft and Zhukov 2015; Wood 2010).

Despite ample evidence demonstrating intentional government targeting, it is also clear that some governments adhere to the laws of war and attempt to avoid civilian casualties (Morrow 2014; Slim 2003). Referencing conflicts in Iraq and Afghanistan, Morrow notes: “The recent wars have been characterized by efforts by the U.S. and its allies to restrict their war efforts to comply with their legal obligations regardless of the conduct of the other side and with greater public scrutiny of their conduct” (Morrow 2014, 302). Militaries which abide by the Geneva Conventions and Protocols, are generally only allowed to use lethal force when they positively identified an enemy combatant demonstrating a hostile act and hostile intent (Kahl 2007, 19). However, if civilian casualties may result from the lethal use of force, then jus in bello principles require military leaders to consider the action’s military necessity, distinction of non-combatants, and proportionality of response (Pictet 1958). The distinction and proportionality principles form the basis of noncombatant immunity, which is designed to outlaw intentional civilian targeting and minimize unintentional civilian casualties (Department of the Army 1956). While forces may be adhering to jus in bello requirements, the potential for unintentional civilian casualties still exists – what Crawford refers to as “systemic collateral damage” (Crawford 2013, 232).

Statistics from the Iraq War highlight the phenomenon of systemic collateral damage by rule-following counterinsurgents. Kahl suggests broad U.S. adherence to non-combatant immunity norms during Operation Iraqi Freedom (OIF). However, Kahl also notes that up to 70,000 Iraqi civilians were killed during OIF by 2006, and that up to 9000 of those may have been unintentionally caused by U.S. or coalition forces (Kahl 2007, 13, 42). Recent research suggests that Western reliance on aerially delivered, precision guided munitions has increased civilian casualties in predictable ways (Cronin 2018; Kaempf 2018). These studies provide compelling arguments that Western precision air strikes transfer risks to local non-combatants to protect western combatants. However, this research largely neglects dynamics amongst ground forces that cause unintentional civilian casualties. Omitting ground-based factors is problematic as data from Iraq’s civil conflict between 2004 and 2008 suggests that 38% of Coalition-caused civilian casualties resulted from airstrikes.⁷ Understanding any systematic causes behind these tragedies is not solely a moral imperative – civilian casualties can impact civil conflict dynamics.

To date, only a few existing works deliberately examine ground-based counterinsurgents’ role in generating civilian casualties. Leveraging case studies of the U.S. military in OIF, Levy argues that U.S. reliance on conventional, mechanized units entailed excessive lethality and violated non-combatant immunity norms (Levy 2019, 108-109). Condra and Shapiro provide one of the only quantitative studies analyzing the operational impact of government-caused civilian casualties (Condra and Shapiro 2012). They diverge from Levy, and suggest mechanized units did not cause additional civilian casualties, compared to light infantry and motorized counterinsurgents (Condra and Shapiro 2012, 181-182). Their findings suggest that Coalition-caused civilian casualties are associated with increasing levels of insurgent attacks on government forces in Iraq, as civilians punish the government through increased collaboration with insurgents and decreased intelligence sharing. Yet they still leave open the question of how and why those civilian casualty incidents occurred. Other research broadly supports the argument that the government’s unintentional killing of civilians is counterproductive, leading to increased support for insurgents.⁸ While these studies have examined the strategies that lead to civilian casualties and their second order impacts, civil conflict scholars have not yet systematically studied if and how distinct counterinsurgent force structures (mechanized, motorized, or light infantry units) impact unintentional civilian victimization.

Though largely unexamined in academic research, narrative evidence from counterinsurgent practitioners suggest that mechanized units may limit civilian casualties (Gentile et al. 2018; Haight, Laughlin, and Bergner 2013; Smith 2008). Writing about the mechanized units’ importance during insurgencies, Johnson notes, “Heavy armor enables friendly forces to survive the initial engagement and respond with precise, timely, direct fire that generally generates less collateral damage than do artillery or airstrikes” (Johnson 2011). H.R. McMaster, a former armored regiment commander, and his subordinates suggest that “armored vehicles […] can take greater risks to secure the population, holding fire until the enemy reveals its hostile intent” (McKinney, Elfendahl, and McMaster 2013). Extending beyond the U.S. military, the Canadian Army found similar utility in mechanized counterinsurgents when tank units were deployed to Southern Afghanistan (Cadieu 2008). Though not underpinned by a cohesive theory, these narrative accounts suggest a common theme where armored protection enables crew restraint and precision fires when operating in populated areas.

In review, several studies convincingly argue that unintentional, government-caused civilian casualties hinders counterinsurgent effectiveness. Other studies have carefully analyzed how precision-air strikes in contemporary conflicts have increased civilian casualty rates. However, existing scholarly research has not yet examined how ground-based counterinsurgents may systematically impact civilian casualty trends. While narrative evidence from counterinsurgent practitioners suggests that unit force structure may play an important role in reducing civilian casualties, these observations lack a comprehensive theoretical framework or widespread empirical support. To better understand systemic, ground-based civilian casualty dynamics, we proceed with initial theorizing to explicate how counterinsurgent force structure affects civilian victimization outcomes.

Armored Restraint Theory

Building on narrative evidence from counterinsurgency practitioners, we propose a crew level armored restraint theory detailing how well-disciplined, mechanized units may lower civilian casualties, when compared to other unit types. Nesting in the broader civil conflict literature, this theory specifically focuses on irregular conflicts where government forces enjoy an asymmetric military power advantage over insurgents (Kalyvas and Balcells 2010). This theory presumes that government forces broadly adhere to the jus in bello principles and respect non-combatant immunity norms. As detailed above, this is often not the case in civil conflicts where intentional, government violence against civilians is a common occurrence. This assumption narrows this research’s external validity to counterinsurgents who systematically respect non-combatant immunity norms. Despite its narrow focus, there are numerous militaries adhering to this norm, which are currently active in several civil conflicts around the globe.

This armored restraint theory is driven by three crew-level factors. First, armored crews can employ low-caliber, precision weapons that allow for increased accuracy. Second, hierarchical decision-making processes in mechanized units’ chain of command may lead to greater target analysis and discrimination for armored crews, compared to light infantry units. Finally, and most importantly, armored vehicles’ increased protection increases crews’ tactical patience, providing them the opportunity to avoid or delay engagements that might endanger nearby civilians. Below, we explore these three mechanisms in detail and propose a testable hypothesis stemming from these factors.

Increased Accuracy: Mechanized units’ firepower is sometimes assumed to increase civilian casualties.⁹ However, while armored vehicles have the potential for increased firepower, crews have access to low-caliber, precision weapons.¹⁰ Armored vehicles’ low-caliber weapons are mechanically stabilized, enhanced with advanced optics, and integrated into the vehicle’s fire control system, significantly increasing accuracy compared to their dismounted variants. Therefore, when armored crews respect non-combatant immunity norms, then they have the means available to limit civilian casualties.

Hierarchical Decision-making: Beyond technical differences in precision firepower, there are key differences between how mechanized and dismounted units decide to employ lethal force and train to fight. In a U.S./U.K. armored vehicle, the authority to employ lethal force rests solely with the vehicle commander (Department of the Army 2009). For example, U.S./U.K. tank crews have four members-the commander, gunner, driver, and loader. The commander, gunner, and loader each have access to their own mounted weapon system, and the gunner and commander jointly control the tank’s main cannon. However, the tank commander is the only individual who can authorize any crew-member to fire their weapon. When a tank crew member spots an enemy threat, he announces a description of the threat over the tank’s intercom. Once the tank commander has identified the suspected enemy, she determines if the enemy displaying a hostile act and hostile intent and should be killed. If the commander decides to use lethal force, she gives a fire command that directs which crew member will fire and which weapon system they will use (i.e. main cannon or mounted machine gun) (Department of the Army 2000). The commander then supervises the engagement and determines when her crew will cease firing. Though this target identification process may only take 10 seconds, it provides the vehicle commander with a degree of control during the chaos of combat.

The armored crew’s hierarchical decision-making stands in sharp contrast with a dismounted infantry platoon. If a dismounted platoon received direct fire from insurgents, then platoon members receiving enemy fire would “immediately return fire on known or suspected enemy positions and assault,” to kill the enemy (Department of the Army 2007, J-13). Given the rules of engagement’s self-defense imperative, these dismounted soldiers can immediately return fire without the approval of their immediate chain of command (Grunawalt 1997). The infantry platoon’s immediate suppressing fire against “known and suspected” insurgent positions presents high risk if civilians are nearby, as is often the case in insurgency conflicts.¹¹

If dismounted infantry units are more reliant on air-strikes and artillery during kinetic engagements, then risks to local civilians will further increase. These divergent engagement techniques are inculcated in light infantry soldiers and armored crews starting in basic training. After rehearsing “react to enemy contact” drills thousands of times, controlled fire commands become second nature to mechanized crews, just as immediately returning fire at known and suspected enemy locations is second nature for light infantry (Department of the Army 2007; Department of the Army 1998). It is important to emphasize that this dismounted platoon’s reaction to an insurgent attack is not necessarily reckless or in violation of the laws of war. Rather, the platoon’s immediate suppressive fire is the unit’s main mechanism to survive the insurgent’s attack and is driven by self-defense and survival imperatives, based on their relative lack of armored protection. This stands in sharp contrast to the degree of protection afforded to armored vehicles crews and leads to the final factor.¹²

Protection: Armored vehicles’ composite metal protects crews and provides them with relative safety before returning fire during an attack. This armored protection is maximized during asymmetric conflicts, where insurgents lack access to extensive anti-armor weaponry. This in-creased protection enables crews to exercise ‘tactical patience’ and provides them with additional decision-making time.¹³ When combined with the deliberate chain of command target analysis and precision weapon systems, this tactical patience enables the armored crew to accurately assess a threat, consider nearby risks of civilian casualties, and select the most appropriate weapon system for the engagement. Similar tactical patience dynamics may not exist for more-vulnerable light infantry who need to immediately return fire against “known and suspected enemy positions” out of simple self-defense realities.¹⁴

Psychology research on stress and decision-making suggests that mechanized forces’ increased protection may enable armored crew members to make sounder decisions, resulting in lower levels of civilian casualties. Larsen found that military service members under high levels of stress were less effective at distinguishing between different targets (Larsen 2001). Keinan suggests that humans who make decisions under stress by reducing decision-making time to a few seconds fail to fully consider alternate courses of action (Keinan 1987). If mechanized unit’s increased armored protection reduced crew stress levels during kinetic engagements-compared to more vulnerable dismounted units, then these psychology studies suggest that mechanized units can better distinguish between civilians and insurgent fighters, and select the most appropriate weapon system for a given engagement.

What Armored Restraint Looks Like in Practice: While this theory focuses on crew-level dynamics, these mechanisms extrapolate up and impact tactical force employment at the platoon and company level. Consider a Coalition mechanized platoon conducting a routine security patrol in Baghdad. The platoon of two M1 Abrams tanks and two M2 Bradley fighting vehicles enter a complex ambush, triggered by an improvised explosive device (IED), and followed with small arms and rocket fire. The platoon’s 28 soldiers, operating from the relative safety of their armored vehicles, can assess the situation and identify the enemy positions. If the armored crew is aware that civilian casualties may result from using their 25 mm and 120 mm main cannons, they have the option to postpone the engagement or return fire with precision, low caliber systems, like 0.50 caliber and

7.62 mm machine guns. From the relative safety of their vehicles’ protected hull, crew members work together to positively identify hostile insurgents. After vehicle commanders confirm insurgent positions and assess the local potential for civilian casualties, they issue fire commands to their crew, identifying the most appropriate weapon system to return fire and minimize the potential for civilian casualties. The platoon’s two infantry squads can then dismount from their Bradleys to secure the scene.

Contrast this by considering a dismounted light infantry platoon of 39 soldiers in the same complex ambush. This unit must immediately return fire to suppress known and suspected insurgent positions to cover an assault or withdrawal. The dismounted unit does not enjoy mobile armored protection, secure decision space, or precision mounted machine guns like the tank platoon. To prevent friendly casualties, the light platoon’s suppressing fire is immediate out of necessity, with little time to react. The dismounted platoon’s response to the insurgent attack has the potential to inadvertently cause civilian casualties, which may further increase if the unit requests close air support or artillery to aid in their efforts to suppress enemy positions. In this sense, the dismounted platoon’s relative vulnerability through its absence of armored protection results in a trade-off between civilian and friendly casualties. This dilemma is exacerbated when operating in densely populated areas and against insurgents seeking to de-legitimize counterinsurgents by instigating government caused collateral damage.

Alternative Explanations

As noted above, civil conflict scholars have long cautioned against employing mechanized counterinsurgents. (Galula 2006; Leites and Wolf Jr 1970; Lyall and Wilson 2009; Mack 1975; Wallace 2007). While these scholars provide compelling causal logic highlighting the importance of dismounted patrolling that leverages local intelligence, they do not explicitly consider how mechanization impacts civilian casualty rates. Lyall and Wilson provide one of the only empirical studies analyzing how government mechanization impacts civil conflict dynamics. In a large-N, interstate analysis, they introduce the “mechanization hypothesis” which argues that “increasing mechanization of state militaries has steadily undercut their effectiveness by truncating their ability to collect local information” (Lyall and Wilson 2009, 102). Civilian casualties are not their work’s focus, and their outcome of interest is conflict outcomes. However, they cite narrative evidence in an OIF case study noting that mechanized forces are more likely to cause civilian casualties (Lyall and Wilson 2009, 100). Some have critiqued their analysis, noting their case study focuses on a U.S. Army mechanized unit that relies on excessive force, while omitting other mechanized units that practiced armored restraint (Smith and Toronto 2010, 523-525). More importantly the mechanization hypothesis’ causal logic depends on tactical patrolling practices, while their quantitative tests employ strategic variables (military wide mechanization and conflict outcomes), potentially obscuring critical tactical dynamics.

Some psychology research suggests that mechanization increases civilian casualties. Grossman argues that planes, artillery, and tanks – with increasing physical and mechanical distance – lessens the psychological and physiological restraints that inhibit humans from killing (Grossman 2009, 108). A large survey finds that US-based Air Force remotely piloted aircraft operators (drone pilots) have significantly lower rates of post-traumatic stress disorder than U.S. service-members returning from overseas deployments (Chappelle et al. 2014). This study suggests that the remote nature of the drone pilots’ kinetic engagements lessens the psychological burden of killing other humans. These findings imply that even units adhering to non-combatant immunity norms may cause increased civilian casualty rates when armored vehicles’ technological remoteness is paired with excessive firepower. If these alternative explanations are correct, then mechanized counterinsurgents should be associated with increasing civilian casualty rates.

Testable Hypothesis

A testable hypothesis emerges from our existing research, counterinsurgency practioners’ accounts, and our armored restraint theory:

Hypothesis (H1). Mechanized counterinsurgent units that adhere to non-combatant immunity norms will be associated with lower civilian casualty rates, compared to light and motorized counterinsurgent units.

If H1 is accurate, then empirical tests should demonstrate that mechanized counterinsurgents are systematically associated with lower civilian casualty rates, compared to light and motorized infantry. If H1 is false, then mechanized counterinsurgents will be systematically associated with increasing civilian casualty rates, compared to light and motorized infantry, lending support to the alternative explanations listed above. Given observations from counterinsurgency practitioners suggesting mechanized units’ armored restraint and precision, and U.S. and U.K. adherence to the rules of engagement during OIF, we expect H1 to obtain under empirical tests. OIF offers a unique opportunity to study how and if mechanization impacts local civilian casualty levels. Thirty seven percent of the U.S. forces which fought in Iraq between 2004-2008 were mechanized units. Moreover, these units routinely operated in areas densely populated with civilians. Though hindered with identification and operationalization challenges, these data provide the basis to perform an initial exploration of the armored restraint theory.

Validity and Scope Conditions

We acknowledge that the armored restraint theory has a relatively narrow scope and is limited to civil conflicts where counterinsurgents are committed to following the laws of war and respecting non-combatant immunity norms.¹⁵ Additionally, insurgents possessing advanced anti-tank guided missiles (ATGMs) will likely reduce armored restraint, diminishing crews’ protection, lessening their tactical patience, while increasing the need to suppress possible ATGM ambush locations or conduct dismounted clearances. Further, this theory does not suggest that mechanized units are better counterinsurgents than dismounted units. Lowered civilian casualty rates cannot be conflated with overarching counterinsurgent success. This theory does not disprove the mechanization hypothesis’ causal logic, which argues that mounted patrols are insulated from interactions with local civilians, unlike dismounted patrols (Lyall and Wilson 2009). Beyond interactions with civilians, dismounted counterinsurgents are needed to clear buildings, interact with local civilians, and maintain 360-degree situational awareness.

Despite the enduring importance of dismounts during counterinsurgency operations, civilian casualties are most likely to occur during firefights and mechanized units’ precision firepower and armored protection provide a comparative advantage over light infantry units during tactical engagements. As we suggest in other research, these mechanized and dismounted units can serve as important complements, rather than substitutes (Van Wie and Walden 2022). In practical application, this matters because unintentional civilian casualties have been shown to alter strategic dynamics in insurgency conflicts (Condra and Shapiro 2012). More importantly, limiting civilian casualties is a moral imperative. It is important for scholars, policymakers, and practitioners to understand the tactical dynamics that potentially increase civilian casualties during civil conflict.

Micro-Analysis: Mechanization and Civilian Casualties at the Iraq District Level

Data Overview

We use existing OIF data-sets as the basis for empirical exploration of counterinsurgent mechanization and the armored restraint theory. To approximate for counterinsurgent mechanization, we introduce the Dismount Ratio and Armored Units variables, capturing Coalition units’ mechanization levels at the district-battalion-week level.¹⁶ Remaining data are derived from the Iraq Body Count Project and the Empirical Studies of Conflict research group’s Iraq Civil Conflict dataset which include Iraq district-week data from 2004 to 2008 (Condra and Shapiro 2012). Like many studies analyzing quantitative data on civilian victimization, these data suffer from operationalization challenges and biases toward urban and mass-casualty events (Balcells and Stanton 2021, 49). We are careful to point out the data’s shortcomings, and consider this when analyzing results and making policy recommendations.

Dependent Variable: Coalition-Caused Civilian Casualties

Coalition-caused civilian casualty events are this study’s outcome of interest. We use civilian casualty data in two formats: one, as a count data of civilian casualties events aggregated at the district-week level, and two, as per capita collateral damage rate within each district (Condra and Shapiro 2012) The data come from the Iraq Body Count (IBC) project, which tracked civilian casualty events by monitoring a variety of media reports, government, hospital and morgue reports across Iraq. As coded in Condra and Shapiro, these data include 19,961 civilian casualty events and over 59,000 civilian deaths. Complete figures include civilian casualties caused by Coalition forces, Iraqi Security Forces, insurgents, sectarian violence, and unknown factors. We exclude attacks that were caused by insurgents, sectarian strife, and Iraqi Security Forces, leaving 4926 civilian casualty events caused by Coalition security forces between 2004 and 2008.¹⁷ Condra and Shapiro provide a detailed review of potential shortcomings with these IBC data and conduct robustness checks suggesting that IBC’s collection methods do not suffer from widespread measurement error. (Condra and Shapiro 2012, 171). However, Donnay and Filimonov argue that these IBC data suffer from limited coverage outside of major urban centers and have accuracy issues with event time-stamps, when compared to the Multinational Forces Iraq, Significant Actions (SIGACTs) database which catalogues violent events (Donnay and Filimonov 2014, 25-26). Since we use IBC data, aggregated at the district-week level, we cautiously proceed understanding these data’s shortcomings. Figure 1 denotes the geographic distribution of Coalition-caused civilian casualties from 2004 to 2008.¹⁸ In an average week, there were 0.53 Coalition-caused civilian casualty incidents in each observed district.

Figure 1.

Mean weekly civilian casualty incidents by district, 2004-2008.

Measuring Counterinsurgent Mechanization

Combining U.S. and U.K. force structure data with Carrie Lee’s Iraq Order of Battle data, we introduce new variables representing the district-level mechanization of U.S. and U.K. combat battalions. Specifically, the Dismount Ratio divides the total number of Coalition combat soldiers by Coalition armored vehicles in a given Iraqi district to create a ratio of troops to armored vehicles.¹⁹ Districts that do not include U.S. or U.K. forces are dropped, leaving 9,222 district-week observations. To account for districts that contain multiple battalions, we apply a unit distribution weighting scheme that considers a district’s overall counterinsurgent composition for a given week and accounts for units that are split between two or more districts. The resulting district-level Dismount Ratio provides a linear measure reflecting the mechanization levels of all U.S. and U.K. combat units deployed to a given district. See the Supplementary Appendix for a more detailed list of the assumptions and resources used to operationalize the Dismount Ratio variable.²⁰ Figure 2 displays the overall Dismount Ratio distribution across all 9222 district-week observations. Clusters can be observed at U.S. Army armored battalions (9), U.S. Army Stryker battalions (13), U.S Army light infantry battalions (39), and U.S. Marine Corps rifle battalions (41). The remaining samples indicate a variety of light, motorized, and mechanized U.K. combat units and districts with a combination of mechanized and dismounted units. For the entire sample, the average dismount ratio is 19.38 soldiers per armored vehicle.

Figure 2.

Dismount Ratio distribution, 2004-2008.

Independent Variable: Armored Units

Since the Dismount Ratio represents the pooled district mechanization level, it may include a mixture of mechanized and light infantry units with a resulting district Dismount Ratio that reflects a motorized unit. Considering U.S. Army Strykers, U.S. Marine Corps Light Armored Vehicles, U.S. light infantry, and U.K. equivalents all fought with similar levels of firepower and vehicle protection, this limits the Dismount Ratio’s utility in testing H1.²¹ These vehicles all have relatively light protection levels that may not stop heavy caliber machine gun rounds from penetrating the vehicles’ hull. To overcome these challenges and to directly test the proposed hypothesis, a new variable is introduced to accurately distinguish districts which are dominated by mechanized units with armored protection and enhanced firepower. This dichotomous Armored Unit variable captures all districts with weighted Dismount Ratios under 11 troops/armored vehicle and includes units equipped with M1 Abrams Tanks, M2 Bradley Fighting Vehicles, and U.K. equivalents. Based on H1 outlined above, if mechanized units demonstrate armored restraint during engagements with insurgents, then districts coded by Armored Units should be associated with lower civilian casualty rates, compared to light infantry and motorized units. There are 3,491 Armored Unit district-week observations in the data set, indicating that over one third of Coalition-controlled territory was predominately patrolled by mechanized forces.

We approximate concerns of unit training and culture among the predominant counterinsurgent forces: U.K., U.S. Marines Corps, and U.S. Army controlled districts.²² The U.S. Army is the regression baseline group. Kahl argues that U.S. and U.K. militaries both adhered to non-combatant immunity norms during OIF, while some have suggested the U.K. military acted with more restraint than the U.S. military in Iraq (Dixon 2009; Friesendorf 2019). If different military organizations are significantly associated with civilian casualties, this would challenge Kahl’s findings and our assumption that U.S. and U.K. units generally adhered to non-combatant immunity norms.

Regardless of counterinsurgent mechanization levels, district-level troop density may also influence civilian casualty rates and increasing numbers of troops may result in additional civilian casualties. To control for this possibility, we use a Coalition troop density variable that represents the number of coalition troops per 1,000 residents.²³ Other controls come from Condra and Shapiro’s “Who Takes the Blame,” as this is the only existing study that has empirically explored micro-level civilian casualty rates in Iraq. We follow this research model and draw control variables from the Empirical Studies of Conflict Iraq (ESOC) Civil War dataset. Based on Berman, Callen, et al., we use measures for district level unemployment and population density. Berman et al. argue that increasing unemployment rates are associated with lower rates of insurgent violence, and their logic suggests a similar trend should be observed with civilian casualties (Berman et al. 2011a). Following Berman, Shapiro, and Felter, we proxy ethno-sectarian divisions using governorate level vote returns from the December 2005 elections.²⁴ Sunni, Shia, and mixed districts, which typically experienced more fighting, should be positively associated with civilian casualties, compared to Kurdish districts. In Table 1, descriptive statistics are listed for these controls, civilian casualties, and our new mechanization variables.

Table 1.

Descriptive Statistics of Selected Variables.

	Count	Mean	Std. Dev.	25%	50%	75%	Max.
Government CIVCAS	4,926	0.53	9.37	0	0	2	655
Insurgent CIVCAS	8,794	0.95	3.66	0	0	5	74
Sectarian CIVCAS	32,766	3.55	16.1	0	0	17	972
Unknown CIVCAS	2,341	0.25	1.85	0	0	1	60
SIGACTs	152,986	16.6	25.8	2	7	20	293
6 months. SIGACTs lag	145,590	17.5	26.8	2	8	68	293
Dismount Ratio	9,222	19.4	12.8	8.88	12.4	41.0	41.0
Armored Unit	3,491	0.38	0.49	0	0	1	1
British districts	1,225	0.13	0.34	0	0	1	1
US marine districts	1,527	0.17	0.37	0	0	1	1
Troop density	9,222	4.80	7.39	1.44	2.32	18.1	75.8
Unemp. Rate	9222	0.097	0.068	0.042	0.090	0.20	0.51
Pop. Density	9,222	0.94	2.56	0.037	0.10	7.68	13.9
Sunni share	9,222	0.40	0.31	0.21	0.25	0.92	0.92
Shia share	9,222	0.36	0.33	0.074	0.22	0.89	0.90
Kurd share	9,222	0.069	0.13	0	0.011	0.53	0.53
District week observations	9,222

Nonrandom Unit Deployment

We conduct preliminary empirical testing and isolate relevant narrative evidence from OIF in addressing endogeneity concerns with unit type and unit assignment. Both offer significant support for an assumption of as-if-randomness of unit assignment at a macro level. Still, individual cases contradict these patterns where mechanized units were placed in more violent areas. This constrains our findings to suggestive, rather than conclusive, results, while showcasing patterns consistent with armored restraint theory.

The deployment of mechanized units during OIF was not always random and in certain instances, Coalition commanders sent these units to the most violent districts to support clearing operations, as was the case in the 2nd Battle for Fallujah when two U.S. Army tank battalions were temporarily attached U.S. Marines, lacking organic armored vehicles (Rayburn and Sobchak 2019b, 349). A bivariate regression between insurgent violence and Coalition-caused civilian casualties indicates a positive and significant association. If armored units are endogenously deployed to the most violent districts, this introduces selection effects that would bias mechanized units towards higher levels of civilian casualties. Another possible concern is that Coalition commanders kept armored units away from urban areas, understanding the increased potential for these units to cause civilian casualties. If this trend was prevalent, then dismounted units would be associated with higher levels of civilian casualties as a result of operating around large civilian populations.

Narrative evidence shows that Coalition leadership typically treated incoming combat brigades as interchangeable units.²⁵ The U.S. Army’s official Iraq War history describes how senior leaders were often forced to determine incoming unit assignments due to outgoing unit rotation schedules, rather than considerations such as unit type (Rayburn and Sobchak 2019b, 303). During the height of insurgent violence during 2006-07, light infantry units replaced tank units, and vice versa. As noted in Small Wars, Big Data, “the lead Coalition force planner for the development and implementation of the Baghdad Security Plan…told [the authors] that there was no deliberate effort to match more mechanized units to more violent areas.”²⁶ Primary evidence from the recently declassified CENTCOM Iraq papers confirms these narrative accounts (Multi-National Forces Iraq 2007).

Condra and Shapiro provide quantitative support for this perspective, suggesting that counterinsurgent mechanization is not endogenously correlated with local violence levels and urban terrain (Condra and Shapiro 2012, 181-182). Building on their quantitative findings, and lever-aging our new mechanization variables, we empirically assess for selection effects by implementing a series of fixed effects and logit models using our mechanization variables as the dependent variable. If these narrative accounts were inaccurate, then the quantitative results would indicate insurgent violence is significant and positively associated with counterinsurgent mechanization. The key independent variables in these models are several temporal SIGACTs indicators to include: current district-week SIGACTs, 2-month lags, and 6 month lags.²⁷ Following Condra and Shapiro, Models 1-3, employ sectarian-half year fixed effects regressing the continuous Dismount Ratio against insurgent violence indicators and controls. Models 4-6 employ random effects logit models, regressing the dichotomous Armored Unit variable against insurgent violence indicators and controls (Table 2).

Table 2.

Determinants of Armored Unit Deployments, Fixed Effects and Logit Models.

Model Number	(1)	(2)	(3)	(4)	(5)	(6)
Dependent Variable	DisRat	DisRat	DisRat	Arm. Unit	Arm. Unit	Arm. Unit
Model Type	FE	FE	FE	RE Logit	RE Logit	RE Logit
Current SIGACTs	−0.000 (0.03)	—	—	0.006 (0.02)	—	—
2 months. SIGACTs lag	—	−0.007 (0.03)	—	—	0.007 (0.04)	—
6 months. SIGACTs lag	—	—	−0.008 (0.03)		—	−0.002 (0.06)
N	9222.000	8954.000	8296.000	9222.000	8954.000	8296.000
AIC	70622.4	68677.6	63858.2	8528.6	8167.8	7361.8

Values in parenthesis are robust standard errors clustered on district. Models 1-3 control for sectarian*half year fixed effects. All models control for district troop density, population density and unemployment rate. Models 4-6 control for Sunni and Shia vote shares. These unlisted coefficients were not significant. N is decreased.

In models 2, 3, 5, and 6 based on including 2-month and 6-month SIGACTs lags.

*p < .10, **p < .05, ***p < .01.

Across all six model specifications, insurgent violence indicators and lags are not significant predictors of a district mechanization levels. Control variables, including district unemployment rate, population density and all other control variables are also generally insignificant, supporting Condra and Shapiro earlier findings. These results suggest empirical support for the narrative accounts arguing that mechanized units were not systematically deployed to the most violent districts. Given this contextualizing analysis on the endogeneity of Coalition unit employment patterns, we proceed leveraging the variation in district mechanization levels.

Empirical Analysis

Modeling Strategy

We tested the relationship of mechanization and civilian casualties across four main model specifications with fixed and random effects.²⁸ These models test both per capita and count civilian casualties, and incorporate district-month data with clustered standard errors.

Our four models use many of the same specifications as Condra and Shapiro, building from a simple regression of Armored Unit against violence and controlling for Troop Density in Model 1. In each model, we are focused on the relationship of our Armored Unit binary variable with our civilian casualties per 100,000 dependent variable. In every model, we also incorporate a control measure for U.S. and U.K. troop density. In Models 1-3, we also include Condra and Shapiro’s sectarian x half-year fixed effect, to account for baseline changes in sectarian area violence due to political realignment.²⁹ The first three models leverage sectarian x half-year fixed effects.³⁰ In Model 2, we add in controls for population density and unemployment rate by district, two factors that may affect the utility of mechanization and propensity for violence, respectively. In Model 3, we add in a 6-month SIGACTs lag to account for baseline violence trends by district.

Our fourth model replaces a fixed effects linear regression with a negative binomial count model designed to compensate for the over-dispersion of 0-count violence events in SIGACTs – most of our observed SIGACTs data show no violence. This model also includes controls for all of the prior controls in Model 3, but instead of the sectarian-half year fixed effects, we use variables capturing local sectarian composition. In Model 4, we also add dummy binary variables for United States Marine Corps and U.K. controlled districts, to assess any organizational culture effects, based on training differences.

Results

Table 3 displays our primary empirical results. These linear models point to a negative association between Armored Unit presence and civilian casualties across all four models, with marginal significance. Coefficient estimates in each are very similar, and range from −0.078 to −0.094. In our most specified count model, Model 4, including controls for ethnic composition, troop type, and year, Armored Unit has a significant negative relationship with civilian casualties, with a coefficient estimate of −0.428.

Table 3.

Regression Results of Civilian Casualties on Armored Unit and Controls.

Model Number	(1)	(2)	(3)	(4)
Dependent Variable	CIVCAS/100,000	CIVCAS/100,000	CIVCAS/100,000	CIVCAS Count
Model Type	FE-1	FE-2	FE-3	NBRM
Armored unit	−0.094* (0.05)	−0.093* (0.05)	−0.078* (0.04)	−0.428**(0.20)
Troop density	0.018 (0.02)	0.018 (0.02)	0.018 (0.02)	−0.005(0.03)
Pop. Density	—	−0.012 (0.01)	−0.008 (0.01)	0.102 (0.08)
Unemp. Rate	—	0.305 (0.29)	0.369 (0.30)	0.597 (2.84)
6 months. SIGACTs lag	—	—	0.000 (0.00)	0.013** (0.01)
Sunni share	—	—	—	3.084*** (0.91)
Shia share	—	—	—	2.004** (0.83)
2005	—	—	—	−1.905*** (0.70)
2006	—	—	—	−1.845*** (0.67)
2007	—	—	—	−1.548** (0.66)
2008	—	—	—	−1.606**(0.71)
Constant	0.108* (0.06)	0.091 (0.06)	0.058 (0.05)	−1.493*(0.85)
log(α)	—	—	—	2.878*** (0.19)
District-week observations	9222.000	9222.000	8296.000	8296.000
AIC	42,324	42,326	35,882	9072
_R2	0.008	0.008	0.012	-

Values in parenthesis are robust standard errors clustered on district. Models 1-3 control for sectarian*half-year fixed effects. Model 4 controls for USMC and U.K. controlled districts, but coefficients were not significant.

Models 3-4 observations shrink to 8296 district-week observations by controlling for 6-month SIGACTs lag.

*p < .10, **p < .05, ***p < .01.

The consistency of our negative Armored Unit coefficient across our fixed effects models, and strong support for a negative relationship in Model 4 is suggestive of support for our hypothesis that mechanized units do not increase civilian casualties. Here, the evidence suggests mechanized units cause fewer civilian casualties than other unit types.

Translating the Armored Unit coefficient of Models 1-3 into real terms, the fixed effects numbers of −0.094 to −0.078 suggest that Armored Unit presence in a district on a given week reduces the likelihood of a civilian casualty events by about 0.1 per 100,000 population, or a little less than one casualty event per million Iraqis per week. Translated over the scope of years, and a population of approximately 25 million, this suggests the reduction of many hundreds of civilian casualty events a year through the employment of armored equipment.

Although we believe the models used here best approximate the relationship between mechanization and civilian harm, we performed robustness checks across other model specifications. We find that this negative relationship between mechanization and civilian casualties is consistent across model types. In the Supplementary Appendix, we report model results for other fixed-effects specifications with different time fixed effects, and with negative binomial and zero-inflated models. In those, Armored Unit is consistently negative, and often significant.

Conclusion

Mechanized military forces are prevalent in modern militaries and they are routinely operating in civil conflicts around the world (Caverley and Sechser 2017; Sechser and Saunders 2010). Fully understanding the downstream impacts these forces have on civilian populations is crucial for policy makers and military leaders. Given data operationalization issues discussed above, we cautiously note that our findings only suggest initial support for the armored restraint theory and do not provide clear evidence. With data limitations for statistical inference in mind, empirical findings suggest that disciplined mechanized forces, adhering to non-combatant immunity norms, can leverage their armored protection and precision firepower to operate with restraint and protect local civilians while targeting insurgents. This article does not claim that mechanized units are more effective counterinsurgents than dismounted units. Rather, it primarily explores how mechanization impacts civilian casualty rates. However, these findings do suggest that assumptions about mechanized units’ limitations as counterinsurgents may overlook their utility and precision when operating around civilians.

This work is predicated on the assumption that U.S. and U.K. units generally followed jus in bello principles and we acknowledge this limits our findings external validity. Variant scope conditions include differences in military culture and strategy. It is plausible that states with lower adherence to jus in bello principles, and states that deliberately target civilian populations, would not be expected to follow our theorized set of predicted outcomes based on armored protection. For example, Toft and Zhukov note how variant unit cultures between rebel and state-based groups led to differences in targeting strategy and adherence to the laws of war (Toft and Zhukov 2015). Future research could explore if these trends hold outside of the OIF context, and how mechanized counterinsurgents impact civilian casualties when the state does not adhere to jus in bello principles. Several important policy implications emerge from this study. First, politicians, defense policy-makers, and military leaders face tradeoffs when designing military force structures to address unknown, future threats (Betts 1995, 43). David Johnson highlights this tension in the U.S. context: “The Army will be increasingly challenged to justify the number size and cost of its heavy formations” (Johnson 2011). While mechanized units clearly have their place in high-intensity, inter-state conflict, these findings suggest that mechanized forces can complement light infantry units and play a role in tactically combating insurgents while protecting counterinsurgent forces and without needlessly endangering local civilians. A combined arms force, with flexibility to employ mechanized, motorized, and light infantry units still appears to offer an ideal force structure to respond to a host of likely threats, be they conventional state threats or unconventional non-state threats.

Second, recent technology developments, like the Stryker’s Remote Weapon System (RWS) and the Common Remotely Operated Weapon Station (CROWS), which allow vehicle crews to remain protected during engagements, should increase tactical patience and protected decision space (Dunn 2016). Third, despite the widespread use of mechanized counterinsurgents during OIF and more limited employment in Southern Afghanistan, modern counterinsurgency doctrine fails to address how mechanized forces should be employed. For example, the U.S. Army and Marine Corps’ latest version of Field Manual 3-24, Insurgencies and Countering Insurgencies completely omits any mention of tanks, armored vehicles, or mechanized units-let alone how these forces can be best leveraged to complement infantry units during COIN operations. Doctrine should be revised to reflect how mechanized units can be responsibly employed as counterinsurgents.

Finally, thoroughly training all vehicle crews on deliberate target identification, controlled crew commands, non-combatant immunity norms, and rules of engagement is critical for inculcating counterinsurgents with respect for jus in bello principles. This can be directly improved by incorporating distinguishability tasks as evaluated criterion during weapons qualification training. This would force soldiers to practice positively identifying potential targets as combatants or non-combatants; a technique not currently incorporated in U.S. Military weapons training. These policies will help ensure that counterinsurgents further minimize civilian suffering in future in civil conflicts.

Supplemental Material

Supplemental Material - Excessive Force or Armored Restraint? Government Mechanization and Civilian Casualties in Civil Conflict

Supplementary Material for Excessive Force or Armored Restraint? Government Mechanization and Civilian Casualties in Civil Conflict by Ryan Van Wie and Jacob Walden in Journal of Conflict Resolution.

Supplemental Material

Supplemental Material - Excessive Force or Armored Restraint? Government Mechanization and Civilian Casualties in Civil Conflict

Supplementary Material for Excessive Force or Armored Restraint? Government Mechanization and Civilian Casualties in Civil Conflict by Ryan Van Wie and Jacob Walden in Journal of Conflict Resolution.

Footnotes

Acknowledgments

A special thanks to Raghnild Noordas, Yuri Zhukov, James Morrow, Jordan Becker, Scott Limbocker, Hannah Smith, Carrie Lee, Jake Shapiro, Iain Osgood, John Ciorciari, John Hanson, Yusuf Neggers, Hojung Joo, Alton Worthington for helpful suggestions that improved this paper. Participants in the “Civilians in Conflict” course, the Conflict and Peace, Research and Development group, and the Ford Security Seminar at the University of Michigan, Ann Arbor and the Social Sciences Research Lab at the United States Military Academy, West Point, NY, also provided helpful feedback.

Declaration of Conflicting Interests

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

Funding

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

ORCID iDs

Ryan Van Wie

Jacob Walden

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Balcells

Laia

Stanton

Jessica A.

. 2021. “Violence Against Civilians During Armed Conflict: Moving Beyond the Macro- and Micro-Level Divide.” Annual Review of Political Science 24 (1): 45-69.

Berman

Eli

Callen

Michael

Felter

Joseph H.

Shapiro

Jacob N.

. 2011a. “Do Working Men Rebel? Insurgency and Unemployment in Afghanistan, Iraq, and the Philippines.” Journal of Conflict Resolution 55 (4): 496-528.

Berman

Eli

Felter

Joseph H.

Shapiro

Jacob N.

. 2018. Small Wars, Big Data: The Information Revolution in Modern Conflict. Princeton, NJ: Princeton University Press.

Berman

Eli

Shapiro

Jacob N.

Felter

Joseph H.

. 2011b. “Can Hearts and Minds Be Bought? The Economics of Counterinsurgency in Iraq.” Journal of Political Economy 119 (4): 766-819.

Betts

Richard K.

1995. Military Readiness: Concepts, Choices, Consequences. Washington, DC: The Brookings Institution.

Biddle

Stephen D.

2004. Military Power: Explaining Victory and Defeat in Modern Battle. Princeton, NJ: Princeton University Press.

Cadieu

Trevor

. 2008. “Canadian Armour in Afghanistan.” Canadian Army Journal 104: 5-25.

Caverley

Jonathan D.

Sechser

Todd S.

. 2017. “Military Technology and the Duration of Civil Conflict.” International Studies Quarterly 61 (3): 704-720.

Chappelle

Wayne

Goodman

Tanya

Reardon

Laura

Thompson

William

. 2014. “An Analysis of Post-traumatic Stress Symptoms in United States Air Force Drone Operators.” Journal of Anxiety Disorders 28 (5): 480-487.

10.

Condra

Luke N.

Shapiro

Jacob N.

. 2012. “Who Takes the Blame? The Strategic Effects of Collateral Damage.” American Journal of Political Science 56 (1): 167-187.

11.

Crawford

Neta C.

2013. “Bugsplat: US Standing Rules of Engagement, International Humanitarian Law, Military Necessity, and Noncombatant Immunity.” In: Just War: Authority, Tradition, and Practice, 231-250. Washington, DC: Georgetown University Press.

12.

Cronin

Bruce

. 2018. Bugsplat: The Politics of Collateral Damage in Western Armed Conflicts. Oxford, UK: Oxford University Press.

13.

Dell

Melissa

Querubin

Pablo

. 2018. “Nation Building Through Foreign Intervention: Evidence from Discontinuities in Military Strategies.” The Quarterly Journal of Economics 133 (2): 701-764.

14.

Department of the Army . 1956. FM 27-10, The Law of Land Warfare. Washington, DC: GPO.

15.

Department of the Army . 1998. FM 71-1 Tank and Mechanized Infantry Company Team. Washington, DC: GPO.

16.

Department of the Army . 2000. FM 17-12-7 Tank Gunnery Training Devices and Usage Strategies. Washington, DC: GPO.

17.

Department of the Army . 2007. FM 3-21.8 The Infantry Rifle Platoon and Squad. Washington, DC: GPO.

18.

Department of the Army . 2009. FM 3-20.21, Heavy Brigade Combat Team Gunnery. Washington, DC: GPO.

19.

Dixon

Paul

. 2009. “‘Hearts and Minds’? British Counter-Insurgency from Malaya to Iraq.” Journal of Strategic Studies 32 (3): 353-381.

20.

Donnay

Karsten

Filimonov

Vladimir

. 2014. “Views to a War: Systematic Differences in Media and Military Reports of the War in Iraq.” EJP Data Science 3 (25): 1-29.

21.

Downes

Alexander B.

2006. “Desperate Times, Desperate Measures: The Causes of Civilian Victimization in War.” International Security 30 (4): 152-195.

22.

Downes

Alexander B.

2007. “Draining the Sea by Filling the Graves: Investigating the Effectiveness of Indiscriminate Violence as a Counterinsurgency Strategy.” Civil Wars 9 (4): 420–444.

23.

Dunn

Brian

. 2016. “Reachback for the Squad.” Infantry Magazine. August-December. https://www.benning.army.mil/infantry/magazine/issues/2016/AUG-DEC/pdf/6

24.

Fjelde

Hanne

Hultman

Lisa

. 2014. “Weakening the Enemy: A Disaggregated Study of Violence against Civilians in Africa.” Journal of Conflict Resolution 58 (7): 1230-1257.

25.

Friedman

Jeffrey A.

2011. “Manpower and Counterinsurgency: Empirical Foundations for Theory and Doctrine.” Security Studies 20 (4): 556–591.

26.

Friesendorf

Cornelius

. 2019. “British Operations Among the People and Civilian Risk.”. Small Wars & Insurgencies 30 (3): 615–640.

27.

Galula

David

. 2006. Counterinsurgency Warfare: Theory and Practice. Westport, CT: Greenwood Publishing Group.

28.

Gentile

Gian

Johnson

David E.

Saum-Manning

Lisa

Cohen

Raphael S.

Williams

Shara

Lee

Carrie

Shurkin

Michael

Allen

Brenna

Lovell

Sarah

Doty

James L.

III . 2018. Reimagining the Character of Urban Operations for the US Army: How the Past Can Inform the Present and Future. Santa Monica, CA: RAND Corporation.

29.

Grossman

David

. 2009. On Killing: The Psychological Cost of Learning to Kill in War and Society. New York, NY: Back Bay Books.

30.

Grunawalt

Richard J.

1997. The JCS Standing Rules of Engagement: A Judge Advocate’s Primer.” Air Force Law Review, 245-258.

31.

Haight

David

Paul

Laughlin

Bergner

Kyle

. 2013. “Armor: Key to the Future Fight.” Armed Forces Journal. http://armedforcesjournal.com/armor-key-to-the-future-fight/

32.

Hultman

Lisa

. 2012. “Attacks on Civilians in Civil War: Targeting the Achilles Heel of Democratic Governments.” International Interactions 38 (2): 164-181.

33.

Johnson

David E.

2011. Heavy Armor in the Future Security Environment. Santa Monica, CA: RAND Corporation.

34.

Kaempf

Sebastian

. 2018. “Saving Soldiers or Civilians?” In: Casualty Aversion Versus Civilian Protection in Asymmetric Conflicts. Cambridge, UK: Cambridge University Press.

35.

Kahl

Colin H.

2007. “In the Crossfire or the Crosshairs? Norms, Civilian Casualties, and U.S. Conduct in Iraq.” International Security 32 (1): 7-46.

36.

Kalyvas

Stathis N

2006. The Logic of Violence in Civil War. New York, NY: Cambridge University Press.

37.

Kalyvas

Stathis N

Balcells

laia

. 1987. “International System and Technologies of Rebellion: How the End of the Cold War Shaped Internal Conflict.” The American Political Science Review 104 (3): 415-429. https://www-jstor-org.web.bisu.edu.cn/stable/40863761

38.

Keinan

Giora

. 1987. “Decision Making Under Stress: Scanning of Alternatives Under Controllable and Uncontrollable Threats.” Journal of Personality and Social Psychology 52 (3): 639.

39.

Kocher

Matthew Adam

Pepinsky

Thomas B.

Kalyvas

Stathis N.

. 2011. “Aerial Bombing and Counterinsurgency in the Vietnam War.” American Journal of Political Science 55 (2): 201-218.

40.

Larsen

Rolf P.

2001. “Decision Making by Military Students Under Severe Stress.” Military Psychology 13 (2): 89-98.

41.

Lee

Carrie Ann

(n.d.). “Counterinsurgency Myths and Methods: Evidence from the Iraq War”. Working Paper.

42.

Leites

Nathan

Wolf

Charles

. 1970. Rebellion and Authority: An Analytic Essay on Insurgent Conflicts. Santa Monica, CA: RAND Corporation.

43.

Levy

Yagil

. 2019. Whose Life is Worth More? Hierarchies of Risk and Death in Contemporary Wars. Stanford, CA: Stanford University Press.

44.

Lyall

Jason

. 2009. “Does Indiscriminate Violence Incite Insurgent Attacks: Evidence from Chechnya.” Journal of Conflict Resolution 53 (3): 331-362.

45.

Lyall

Jason

Wilson

Isaiah

. 2009. “Rage Against the Machines: Explaining Outcomes in Counterinsurgency Wars.” International Organization 63 (1): 67-106.

46.

MacFarland

Sean

Smith

Niel

. 2008. “Anbar Awakens: The Tipping Point.” Australian Army Journal 5 (2): 75.

47.

Mack

Andrew

. 1975. “Why Big Nations Lose Small Wars: The Politics of Asymmetric Conflict.” World Politics 27 (2): 175-200.

48.

Mansoor

Peter R.

2008. Baghdad at Sunrise: A Brigade Commander’s War in Iraq. New Haven, CT: Yale University Press.

49.

McKinney

Chris

Elfendahl

Mark

McMaster

H. R.

. 2013. “Why the U.S. Army Needs Armor: The Case for a Balanced Force.” Foreign Affairs 92 (3): 129-136.

50.

Morrow

James D.

2014. Order Within Anarchy: The Laws of War as an International Institution. New York, NY: Cambridge University Press.

51.

Multi-National Forces Iraq . 2007. Iraq Security Forces Order of Battle. Carlisle, PA: U.S. Army Heritage and Education Center. https://ahec.armywarcollege.edu/CENTCOM-IRAQ-papers/0538-TOGETHER-FORWARD-BCT-Employment-Options-Plans-Brief-29-JAN-07.pdf

52.

Pettersson

Therése

Högbladh

Stina

Öberg

Magnus

. 2019. “Organized violence, 1989–2018 and peace agreements.” Journal of Peace Research 56 (4): 589-603.

53.

Pictet

Jean

. 1958. The Geneva Conventions of 12 August 1949: Geneva Convention Relative to the Protection of Civilian Persons in Time of War, Vol. 4. Geneva, Switzerland: International Committee of the Red Cross.

54.

Rayburn

Joel D.

Sobchak

Frank K

. 2019a. The US Army in the Iraq War, Volume 1: Invasion, Insurgency, Civil War, 2003-2006. Carlisle Barracks, PA: US Army War College.

55.

Rayburn

Joel D.

Sobchak

Frank K

. 2019b. The US Army in the Iraq War, Volume 2: Surge and Withdrawal, 2007-2011. Carlisle Barracks, PA: US Army War College.

56.

Salehyan

Idean

Siroky

David

Wood

Reed M.

. 2014. “External Rebel Sponsorship Civilian Abuse: A Principal-Agent Analysis of Wartime Atrocities.” International Organization 68, 3: 633-661.

57.

Sechser

Todd S.

Saunders

Elizabeth N

. 2010. “The Army You Have: The Determinants of Military Mechanization, 1979-2001.” International Studies Quarterly 54 (2): 481-511.

58.

Slim

Hugo

. 2003. “Why Protect Civilians? Innocence, Immunity and Enmity in War.” International Affairs 79 (3): 481-501.

59.

Smith

Niel

. 2008. “The Armor Effect: The Surprising Utility of Armored Forces in Urban and Irregular Warfare”. Air Land Sea Bulletin, 6–9.

60.

Smith

Niel A.

Toronto

Nathan W.

. 2010. “It’s all the Rage: Why Mechanization Doesn’t Explain COIN outcomes.” Small Wars & Insurgencies 21 (3): 519-528.

61.

STATA . 2021. areg — Linear regression with a large dummy-variable set. Publication Title: STATA Online Manuals Type: Software Reference Guide. https://www.stata.com/manuals13/rareg.pdf

62.

Toft

Monica Duffy

Zhukov

Yuri M

. 2015. “Islamists and Nationalists: Rebel Motivation and Counterinsurgency in Russia’s North Caucasus.” American Political Science Review 109 (2): 222-238.

63.

U.S. Marine Corps . 1998. Washington, DC: Organization of Marine Corps ForcesGPO.

64.

Valentino

Ben A.

2014. “Why We Kill: The Political Science of Political Violence against Civilians”.” Annual Review of Political Science 14: 89-103.

65.

Valentino

Benjamin

Huth

Paul

Croco

Sarah

. 2006. “Covenants without the Sword: Inter- national Law and the Protection of Civilians in Times of War.” World Politics 58 (3): 339-377.

66.

Van Wie

Ryan C.

Walden

Jacob A.

. 2022. “Troops or Tanks? Rethinking COIN Mechanization and Force Employment.” Small Wars & Insurgencies 33 (6): 1032-1058.

67.

Wallace

Michael

. 2007. “Leopard Tanks and the Deadly Dilemma of the Canadian Mission in Afghanistan.” Canadian Foreign Policy Journal 2: 1-7.

68.

Wood

Reed M.

2010. “Rebel Capacity and Strategic Violence Against Civilians.” Journal of Peace Research 47 (5): 601-614.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

1.76 MB

0.00 MB

0.52 MB