In December 2017, Iraqi Prime Minister Haider al-Abadi declared victory over the Islamic State (IS) after a nine-month battle to recapture control of Mosul. Iraq’s second-largest city had endured more than three years of the insurgent group’s harsh rule. Speaking to a crowd of Iraqi troops, al-Abadi congratulated them, thanked the US-led multinational coalition for its support, and alluded to the painful price of victory paid “by the blood of our martyrs.”Footnote 1 The battle, which was described by senior military officials at the time as “the most significant urban combat … since World War II,”Footnote 2 achieved the objective of expelling IS from its last major stronghold in Iraq, but at an enormous cost: at least 11,000 civilians and 8,200 Iraqi forces were killed,Footnote 3 138,000 homes suffered more than USD 6 billion worth of damage,Footnote 4 and irreplaceable cultural heritage sites were destroyed.Footnote 5
Although the battle and its aftermath are well documented, most accounts overlook significant temporal and spatial variation in both the extent of damage and, especially, civilian attitudes toward the Iraqi forces that liberated the city. In East Mosul, where Iraq’s elite Counter-Terrorism Service (CTS, also known as the Golden Division) led the fighting under orders to minimize collateral damage, their efforts to protect civilians were widely recognized and respected. However, just across the Tigris River in West Mosul, where the Iraqi federal police led the fight relying heavily on artillery and air power, civilians who suffered the worst violence by IS—some used as human shields and summarily executed in the final weeks of the battleFootnote 6 —nonetheless questioned the legitimacy of their liberators and of the war itself. What explains these varying perceptions of counterinsurgent forces, and do these differences simply reflect the greater harm caused in West Mosul?
We use an iterative mixed-methods approach to develop a theory of “military legitimacy,” which we define as civilians’ perception of an armed force as wielding violence in a manner that maintains “legal and moral authority” on the battlefield, such that it is deemed subjectively worthy of civilian support.Footnote 7 We argue that civilians assess an armed actor’s military legitimacy by observing its identity, objectives, and actions, as well as the effects those actions have on civilians. These informational inputs then inform civilians’ judgments of whether an armed actor’s cause and conduct are just and, ultimately, whether the actor is militarily legitimate.
We apply and test our theory in the context of the Battle of Mosul, where geographic and other plausibly exogenous features made it necessary for the Iraqi forces and their coalition partners to liberate the city in two distinct phases, first in East Mosul and then in West Mosul. Further factors, unrelated to civilian attitudes, prompted significant shifts in the composition of counterinsurgent ground forces, their rules of engagement, the weapons and tactics they employed, and the overall strategic goals of the two phases—from what military commanders at the time described as a goal of IS “attrition” to one of “annihilation.”Footnote 8 Conceptualizing these changes as a “bundled treatment” of the residents of West Mosul, we measure its effect on civilian perceptions of the Iraqi forces’ military legitimacy.
Our mixed-methods approach leverages both quantitative and qualitative evidence from the Battle of Mosul. We first combine an original household survey of 1,458 residents of Mosul conducted in 2018—approximately eight months after the battle—with data from satellite images to study the effects of the battle and the determinants of military legitimacy. We validate our quantitative results and further explore causal mechanisms with rich qualitative data collected over multiple rounds of careful field research in Mosul before and after the implementation of the 2018 household survey. We use both quantitative and qualitative evidence to validate our research design by showing that the shifts making up our bundled treatment were the likely consequence of plausibly exogenous factors not related to underlying differences in the attitudes or other attributes of civilians in East versus West Mosul.
The results are striking. We present evidence that the shifts in troop composition, rules of engagement, weapons and tactics, and operational and strategic goals contributed to significantly more civilian harm and property destruction in West Mosul than in otherwise similar neighborhoods in East Mosul, as measured by self-reported harm and satellite imagery. We also find strong evidence that the two phases of the battle resulted in markedly different attitudes toward counterinsurgent forces. In West Mosul, respondents are substantially more likely to describe Iraqi forces as at least somewhat likely to kill innocent civilians, which we use as a concrete proxy for our more abstract outcome of interest, “military legitimacy.” Most strikingly, these results persist even after conditioning on respondents’ personal experiences with physical and material harm. Even comparing respondents who were similarly victimized or whose neighborhoods experienced similar levels of collateral damage, those in West Mosul still perceived counterinsurgents as less legitimate than respondents in East Mosul did.
This novel finding suggests that perceptions of military legitimacy are influenced not only by specific incidents of harm caused by combatants but also by beliefs about the morality of armed forces’ conduct and the cause for which they are fighting. Together, our results suggest that conflict-affected civilians are capable of distinguishing between what they perceive as “just” and “unjust” harm, consistent with surveys of Western publics.Footnote 9 Put simply, Moslawis’ perceptions of armed forces appear to be shaped by both how and why they harmed civilians.
We build on an extensive literature on the determinants of civilian attitudes toward wartime actors.Footnote 10 Our work is most similar to that of Dell and Querubin, who compare the military strategies of “overwhelming firepower” and “hearts and minds” during the Vietnam War.Footnote 11 We extend that scholarship by connecting it to the concept of military legitimacy and by providing a more complete theoretical framework with which to analyze such strategic shifts in combatant behavior. Our findings also complement previous findings on individual-level variables (such as co-ethnicity between perpetrators and victims) that shape civilians’ subjective judgments about harm.Footnote 12 We also build on a separate literature on the determinants of legitimacy of government institutions and authorities, including police and courts.Footnote 13
Although governments and militaries have increasingly recognized the strategic benefit of military legitimacy for battlefield effectiveness and national security, we argue that it also provides important benefits beyond the state. Military legitimacy enhances civilian well-being, promotes respect for human rights and compliance with international law, and improves the prospects for lasting peace. All together, we draw on the Battle of Mosul to develop a theory of military legitimacy with broader implications for the study of war, peace building, and postconflict governance.
Context: The Battle of Mosul
The Battle of Mosul was the apex of the Iraqi state’s nine-month fight to liberate Mosul from IS control in 2016 and 2017. This section describes the distinct phases of fighting in East and West Mosul, which we leverage to study the determinants of military legitimacy. Since Mosul is divided by the Tigris River, the battle to retake the city necessarily unfolded in two phases. The key differences between the two phases can be characterized as a bundled treatment with four primary components:
-
1. Changes in the composition of Iraqi ground forces resulting from heavy casualties sustained by the elite multiethnic and cross-sectarian US-trained CTS in East Mosul, which necessitated greater reliance on the less disciplined and predominately Shia Federal Police in West Mosul, who were less trusted by Mosul’s majority Sunni population;Footnote 14
-
2. An important change in the coalition’s rules of engagement, Tactical Directive 1, which decentralized authority over use of force, enabling lower-level Iraqi commanders to call in air strikes and artillery faster and with less oversight;Footnote 15
-
3. Heavier use of air strikes, artillery, and other wide-area munitions in West Mosul, as compared with the careful house-to-house urban warfare that was credited with winning the battle for East Mosul; and
-
4. A shift in the operation’s overall objective from “attrition,” aimed at degrading IS, to outright “annihilation,” by trapping and killing all remaining IS fighters in West Mosul.
Let us look at the phases of fighting and the elements of our bundled treatment in more detail.
Phase 1: East Mosul, October 2016 to January 2017
The battle in East Mosul, which began in October 2016, was led by Iraq’s CTS, which sought to clear East Mosul neighborhood by neighborhood and house by house.Footnote 16 The CTS was widely perceived as the most professional and disciplined of the Iraqi forces. According to one interviewee, “There was no looting in this neighborhood because the Golden Division was here, but I did hear about looting in other neighborhoods.”Footnote 17 Relative to other components of Iraqi armed forces, the CTS was also more ethnically diverse and more representative of Mosul’s predominantly Sunni population.Footnote 18
During the initial months of the battle, the CTS and other ground forces received minimal air and artillery support, in part due to relatively restrictive rules of engagement and targeting protocols,Footnote 19 which were subsequently relaxed by Tactical Directive 1, part of the bundle of changes that occurred in December 2016. Before this change, all coalition targets had to be approved by senior US commanders from a centralized strike cell in Baghdad. Targets were carefully assessed for probable collateral damage, and legal concerns regarding collateral-damage estimates could slow the process further. Furthermore, the CTS forces were told, by commanders who hoped to win over civilians and turn them against IS, to minimize civilian casualties.Footnote 20 Air strikes were largely limited to precision-guided munitions in this first phase of the battle.Footnote 21
The coalition’s strategy in East Mosul was commended for limiting civilian harm, but at enormous cost to the CTS, who suffered massive casualties that depleted its fighting force by 75 percent.Footnote 22 An Iraqi commander said of this trade-off between civilian protection and force protection, “Our soldiers have to be very careful. We can’t just bomb a neighborhood and then go clear it, we have to fight from house to house and that is costing us men.”Footnote 23
Phase 2: West Mosul, February 2017 to July 2017
After East Mosul was liberated, in January 2017, the coalition took a three-week “operational pause” to rest, bring in reinforcements, and repair equipment. Iraqi forces began their offensive to retake West Mosul in February.
With the CTS largely incapacitated by the East Mosul operation, the coalition was forced to rely on the Federal Police, a predominately Shia paramilitary force that was widely recognized as the least professional and least effective of the various Iraqi armed forces involved in the battle. Federal Police interviewed during the battle admitted to having no knowledge of rules of engagement or the laws of war—said one, “I did not receive any instructions on who to shoot and not to shoot.”Footnote 24 A CTS commander said of the Federal Police, “they are acting with recklessness and madness,” referring to their heavy use of rockets and artillery in West Mosul.Footnote 25 There were widespread allegations of looting, property destruction, and sexual violence against civilians by Federal Police.Footnote 26 According to one of our interviewees, “The Federal Police were looting and stealing a lot from civilians, and they were watching other people stealing and doing nothing.”Footnote 27
Another important change was in the rules of engagement. In late 2016, just before the end of Phase 1, the coalition issued Tactical Directive 1, which dropped the requirement that all air strikes be approved by a “strike cell” and empowered lower-ranking commanders to call in air strikes faster and more easily than they had been able to in East Mosul.Footnote 28 This change was partially driven by the coalition’s concern that “the centralization of [target engagement authority] … limited the effectiveness of airpower and forced it to be employed in a restrained manner” during Phase 1.Footnote 29 Around the same time, military sources reported that the “noncombatant casualty cutoff value” “was raised slightly” from its previous level of zero,Footnote 30 meaning that the coalition was willing to risk some nonzero number of civilian casualties without the approval of a senior commander.
As the battle for West Mosul unfolded, journalists and human rights organizations observed a sharp increase in the frequency of air strikes and a corresponding increase in civilian casualties. A Federal Police colonel in West Mosul admitted that Iraqi forces were frequently calling in air strikes against IS snipers on rooftops without knowing whether those buildings contained civilians. “It was impossible to know who was in homes. We had to advance, so when ISIS snipers would attack us, we had to call in airstrikes. Entire families were likely killed.”Footnote 31 Changes in munitions on the ground also contributed to more collateral damage in West Mosul. In East Mosul, ground forces had used relatively precise gun rounds and antitank guided missiles; but in West Mosul, narrower streets forced them to rely on unguided “wide impact area” weapons, including close-range AT-4 rockets, rocket-propelled grenades, mortars, and artillery projectiles.Footnote 32
Also changing between Phases 1 and 2 were counterinsurgents’ broader strategic and operational goals. During the operational pause, counterinsurgents weighed their ultimate objective—the liberation of Mosul versus the complete destruction of IS—and chose the latter. In May 2017, amid ongoing fighting in West Mosul, US Secretary of Defense Mattis described this new approach as a shift “from shoving ISIS out of safe locations in an attrition fight to surrounding the enemy strongholds, so we can annihilate ISIS.”Footnote 33 Although an earlier proposal had recommended an exit corridor that would have allowed IS fighters to withdraw from Mosul westward into the desert, where they could be surrounded in a “kill box” far from civilians,Footnote 34 the coalition ultimately decided to trap and kill IS fighters in the Old City of West Mosul. By July 2017, the coalition had achieved its goal of “annihilating” IS, but at the cost of heavy civilian casualties and property destruction in West Mosul.
Quantitative Validation
While journalists and humanitarian actors have detailed the differences in harm between East and West Mosul, this study is the first to systematize and quantify these differences. First, using satellite data and an original survey, we confirm that West Mosul had more collateral damage. Figure 1 visualizes the disparity in battle-related damage between the two sides of the city. West Mosul respondents were eight percentage points more likely than those in East Mosul to report that a member of their household had been killed during the battle (see Table A-7 in the online appendix). Second, we confirm qualitative differences in harm: West Mosul residents were more likely to report looting by Iraqi forces, reflecting a lack of professionalism among ground forces (Table A-7). Analysis of survey data from Lafta, Al-Nuaimi, and Burnham also supports greater reliance on air power and indiscriminate violence in West Mosul, where respondents were more likely to attribute deaths to air strikes (Figure A-4 and Table A-8).Footnote 35
Satellite-assessed damage in Mosul
A Theory of Military Legitimacy
To unpack the consequences of the Battle of Mosul for civilian attitudes, we first offer a definition of our outcome of interest, military legitimacy. Focusing on this concept both complements and deepens earlier work on counterinsurgency and harm. We then develop a theory of the determinants of military legitimacy, from which we derive testable hypotheses in the context of the Battle of Mosul.
We define “military legitimacy” as civilians’ perception of an armed force as wielding violence in a manner that maintains “legal and moral authority” on the battlefield such that the military is deemed worthy of civilian support.Footnote 36 Concepts of legitimacy have long been used to describe states,Footnote 37 legal authorities and institutions,Footnote 38 and organizational actors more generally.Footnote 39 Prominent research has recently examined the consequences and determinants of police legitimacy,Footnote 40 where legitimacy is often linked to “procedural justice”—the perception that authorities follow procedures that people see as fair.Footnote 41 At its most basic, legitimacy is a condition that inheres “when people are influenced by an authority or institution not by means of the use of power but because they believe that the decisions made and rules enacted by that authority or institution are in some way ‘right’ or ‘proper’ and ought to be followed.”Footnote 42 Legitimacy therefore tends to increase the likelihood of obedience to authorities, which can be observed in behavioral outcomes, including compliance with tax obligations, court decisions, and public health regulations.Footnote 43
In wartime settings, military legitimacy is believed to be important for similar behavioral outcomes of civilian obedience and support, but legitimacy and support are not coterminous. Decisions to support an armed actor may be the result of rational choices rooted in the actors’ exercise of coercive power, but civilian support can also result from emotional or psychological mechanisms.Footnote 44 We propose perceptions of military legitimacy as one such mechanism. Military legitimacy is an attitudinal precursor to behavioral support, though the relationship is not necessarily automatic or deterministic. Individuals who aid armed actors may believe that they are legitimate, or they may be acting instrumentally, driven by perceived opportunity costs or safety concerns. And legitimacy is not necessarily “zero sum”; civilians living in areas fought over by state and nonstate actors may perceive more than one authority as legitimate.Footnote 45 In these ways, our concept of military legitimacy builds on recent scholarship emphasizing civilians’ agency and sophistication in responding to armed conflict.Footnote 46
Legitimacy matters for reasons beyond its relationship to behavioral support. Our concept of military legitimacy is based not only on whether armed forces receive civilian support but also on the degree to which they are perceived as worthy of support. While this distinction is subtle, its implications are significant. Much of the policy on counterinsurgency takes an instrumental approach to “winning hearts and minds” from the perspective of combatants, who value civilians as sources of information, material support, and other strategic benefits. In contrast, we take a civilian-centric approach. While military legitimacy does have strategic benefits for armed forces, because it induces voluntary cooperation, it is also intrinsically valuable for civilians, “because it reflects an individual’s own values rather than a reliance on outcomes to regulate behavior.”Footnote 47 From a normative perspective, legitimacy reflects values of individual dignity, consistent with the laws of war. Further, legitimacy is associated with longer-term outcomes, providing a “solid foundation for transitioning from war to peace.”Footnote 48
Our central theoretical contribution is not just to define military legitimacy and defend its conceptual importance but also to hypothesize the process by which civilians assess it. We theorize that, in assessing a military’s legitimacy, civilians rely on three distinct informational inputs, each realized over the course of a conflict (Figure 2).
Theory of military legitimacy
First, over the course of a conflict—and perhaps even earlier—civilians gradually learn who the various armed actors are, including their identities and distinct objectives. Second, as armed actors engage in conflict, civilians observe and develop beliefs about the types of actions that various actors take or might take to achieve their objectives. Third, civilians learn about and experience the effects of armed actors’ actions on themselves, their households, and their communities. From these informational inputs, civilians then develop common-sense intuitions about whether an armed actor’s cause and conduct are just. At this stage, civilians may use heuristics to fill in informational gaps, and they may also exhibit biases. For example, research on in-group favoritism predicts that civilians will discount the unjust conduct or cause of a co-ethnic armed actor while judging out-group members more harshly.Footnote 49
We theorize that military legitimacy depends on tactics and strategies that civilians experience as being fair and situationally appropriate, including by following the international humanitarian law principles of just cause (jus ad bellum) and just conduct (jus in bello). In doing so, we build on previous findings that civilians distinguish between “just harm” and “unjust harm,” where the former results from just conduct or is necessary to achieve a just cause.Footnote 50 As a result, military legitimacy in most contexts will require that civilians perceive combatants and commanders as exercising “due care”Footnote 51 to avoid harming innocent civilians,Footnote 52 and due care will often require militaries to act with restraint. Indeed, the US military itself has adopted a definition of legitimacy as “a condition based on the perception of … the legality, morality, or rightness of a set of actions … [that] may be reinforced by restraint in the use of force, the type of forces employed, and the disciplined conduct of the forces involved.”Footnote 53
Another way combatants can demonstrate just conduct and thereby enhance their perceived legitimacy is by assuming risk and in some cases visibly putting themselves in harm’s way, especially where it is necessary to reduce the risk of harm to civilians. By having “skin in the game,”Footnote 54 combatants align their incentives with those of the civilian community and demonstrate a credible commitment to acting with restraint and exercising due care. This is more easily achieved in ground operations, where the conduct of individual combatants is visible to civilians, than in operations that rely heavily on air strikes or artillery fired from the relative safety of fighter jets or armored vehicles. In short, perceptions of military legitimacy depend not only on what civilians experience at the hands of armed actors but also how they experience it—whether or not combatants’ conduct is just—and why they experience it—whether or not they approve of the cause for which combatants are fighting.
Applying this theory, we generate two predictions for patterns of military legitimacy following the Battle of Mosul. First, because of the differences in the identities and objectives of counterinsurgent forces in East and West Mosul, the differences in their strategies and tactics, and the much greater collateral damage in West versus East Mosul, we hypothesize:
H1 Relative to respondents in East Mosul, respondents in West Mosul should view counterinsurgent forces as less legitimate.
This is a necessary observable implication of our theory of military legitimacy. However, because collateral damage was much worse in West Mosul than in East Mosul, it is not on its own sufficient to support our theory that military legitimacy is also determined by civilian perceptions of just conduct and just cause. To address this limitation, our second hypothesis helps rule out the possibility that military legitimacy is determined solely by the amount of conflict-related harm that civilians experience. If just conduct and just cause indeed play a role, we would also expect:
H2 Relative to respondents in East Mosul, respondents in West Mosul should view counterinsurgent forces as less legitimate, holding fixed a respondent’s personal exposure to conflict-related harm.
To be sure, tests of H1 and H2 still are not sufficient to confirm all aspects of our theory. Ideally, our design and data would support individual tests of the causal pathways outlined in Figure 2—for example, how the identity of actors informs just cause, or how an actor’s actions affect just conduct. Although limitations in our survey data and the bundled nature of our treatment do not support such granular testing, we can nonetheless use our two hypotheses to confirm the core logic of our theory and help rule out that military legitimacy is driven by exposure to harm alone. In the “Further Evidence” section we supplement our main analysis with qualitative interviews and some additional quantitative tests in order to probe the inner workings of our theory.
Our theory has an important scope condition. For the theory to apply, civilians must be able to evaluate the three informational inputs depicted in Figure 2. Thus, it is likely to be most applicable in high-intensity urban, civil conflict where civilians can directly observe the identities and behaviors of combatants. Still, we believe it may be generalizable beyond the context of counterinsurgency, and may also help explain civilian perceptions of military legitimacy (or lack thereof) in more conventional battles away from urban centers and air wars where combatants are less visible to civilians.
Research Design: Comparing the Two Phases in the Battle of Mosul
To learn from the Battle of Mosul and its aftermath, we combine quantitative analysis of survey and satellite data with qualitative interviews in an iterative, mixed-methods approach.Footnote 55 Our quantitative research design is straightforward: to assess the relationship between counterinsurgent strategy and perceived military legitimacy, we compare the self-reported perceptions and lived experiences of civilians who were living in East and West Mosul during the battle against IS. Here we detail our data sources and research design.
Quantitative and Qualitative Data
Our study is based on an original household survey of a random sample of 1,458 residents of East and West Mosul (see Figure 3 for locations), along with qualitative data from field research in Mosul and other areas of northern Iraq. The survey was conducted by a gender-balanced team of Iraqi enumerators in March 2018, approximately eight months after Mosul was recaptured from IS by Iraqi forces. Further details on survey methodology and fieldwork, including ethical considerations, are given in section B of the online appendix.
Approximate survey locations in Mosul, Iraq
We supplement our original survey data with observational, spatial data. The United Nations Satellite Centre (UNITAR–UNOSAT) acquired and cleaned satellite imagery of Mosul in August 2017, at the conclusion of the battle. A total of 19,888 battle-affected structures were identified within the city. By matching the coordinates of damaged buildings to a respondent’s location, we can detect whether there was any significant structural damage to the residential unit associated with the geographic coordinates.
We complement our quantitative data sources with qualitative semi-structured interviews and in-person observation of Mosul’s urban geography. The design of the survey was informed by in-depth interviews conducted in recently liberated neighborhoods in the outskirts of East Mosul over the course of several research trips in 2017, while West Mosul was still occupied by IS, and in 2018, after IS’s defeat, when it was possible to conduct interviews in both East and West Mosul.Footnote 56 In August 2023, after quantitative analysis of the 2018 survey data and spatial analysis of the satellite data raised new questions about the determinants of military legitimacy that could not be adequately addressed with quantitative data alone, we returned to Mosul as a team to conduct follow-up interviews.Footnote 57 This second round of field research provided further evidence for the assumptions of our design and the underlying mechanisms of our theory. Excerpts of key quotes from these interviews and photos of the neighborhoods are provided in section A of the appendix.
Empirical Strategy
We test H1 (predicting differences between East and West Moslawis’ perceptions of the military legitimacy of counterinsurgent forces) with a simple regression:
$${\textrm{MILITARY LEGITIMACY}_{ij}} = \alpha + \beta {\textrm{WEST}_j} + {\bf{X}}{{\rm{^\prime}}_i}{\rm{\Gamma }} + {\bf{N}}{{\rm{^\prime}}_j}{\rm{\Pi }} + {\varepsilon _{ij}}$$
The dependent variable is constructed from a series of Likert-scale survey questions asking respondents to separately assess how likely each of the three primary components of the Iraqi armed forces (the Iraqi Army, Federal Police, and CTS) are to kill innocent civilians in a hypothetical conflict.Footnote 58 We average the three responses to obtain a proxy for each respondent’s overall assessment of the Iraqi security forces.Footnote 59 The choice to average across the Iraqi armed forces reflects findings from our qualitative interviews. Although Iraqi civilians are generally well informed regarding distinctions between various units of the Iraqi security forces, they were not always able to distinguish between them amid the chaos of the battle, and they tend to view them as representative of an overall Iraqi security and defense apparatus.Footnote 60 The survey questions were carefully worded to elicit forward-looking and subjective views of how likely the various armed actors were to kill innocent civilians in the future, rather than merely reflecting the amount of killing respondents witnessed in the Battle of Mosul itself.
We consider this measure to be a plausible proxy for military legitimacy, capturing Ayres and Thurnher’s description of legitimacy as the “actual and perceived righteousness of [combatants’] conduct.”Footnote 61 Our measure is similar to Condra and Wright’s survey question, which gauges the “perceived level of effort that the government and insurgents exert to avoid civilian casualties.”Footnote 62 Admittedly, it captures only one facet of military legitimacy (efforts to avoid harming civilians), and scholars of legitimacy in other contexts have used different questions to measure other observable indicators of legitimacy, including trust and obedience. However, we believe that our question was the best and most concrete way to operationalize military legitimacy in a postbattle survey of civilians, and our qualitative data provide additional evidence on alternative dimensions of military legitimacy from civilians’ own words.
Our treatment variable,
$${\textrm{WEST}_j}$$
, indicates that the respondent was living in West Mosul during the battle;Footnote
63
${{\bf{X}}_i}$
is a vector of individual-level demographic characteristics; and
${{\bf{N}}_j}$
is a vector of neighborhood-level geographic characteristics.Footnote
64
Our empirical strategy thus boils down to a comparison of survey responses among observably similar respondents living in East and West Mosul.
For this design to yield credible estimates of the effects of the bundled treatment on perceived military legitimacy, we must make two assumptions: that exposure to the bundled treatment is exogenous; and that we are able to control for all systematic differences between East and West Mosul. In other words, after controlling for individual- and neighborhood-level observable characteristics, we have individuals living on the two sides of Mosul that are similar along the key dimensions that could shape perceived military legitimacy except for what they experienced during the Battle of Mosul. Three key pieces of evidence support this. First, East and West Mosul are balanced on most observables, and all covariates are only marginally jointly significant (Tables 1 and A-4). Second, evidence strongly suggests that the shift in strategy and tactics between East and West Mosul was driven by plausibly exogenous factors (section C.1 of the appendix). Third, Oster’s bounds,Footnote 65 reported in Table 3, and Cinelli and Hazlett sensitivity analysis,Footnote 66 presented in Table A-5, suggest that confounding from unobservables would need to be far larger than from our carefully chosen set of observable covariates before it would induce meaningful bias. Although the ever-present potential for omitted-variable bias cautions against a narrowly causal interpretation of our quantitative findings, we nonetheless interpret them—particularly in combination with our findings from qualitative interviews—as suggestive of the determinants of military legitimacy.
Balance on covariates
Notes: HC1 robust, neighborhood-clustered standard errors; *
$p \lt .10$
; **
$p \lt .05$
; ***
$p \lt .01$
.
Difference in means of military legitimacy
Notes: Standard deviations in parentheses; *
$p \lt .10$
; **
$p \lt .05$
; ***
$p \lt .01$
.
Effect of the bundled treatment on military legitimacy
Notes: HC1 robust, neighborhood-clustered standard errors in parentheses. Demographic controls are gender, education level, age, pre-IS household income, and primary identity. Additional individual-level controls are reported harm during IS rule (pre-battle), experienced various grievances with the Iraqi government (pre-IS), voted in the 2014 parliamentary elections, and paid taxes to IS. Neighborhood controls are residential-unit density, population density, and street density; *
$p \lt .10$
; **
$p \lt .05$
; ***
$p \lt .01$
.
Controlled direct effect of the bundled treatment on military legitimacy
Note: Demographics controls are gender, education level, age, and pre-IS household income. Additional individual level controls are reported harm during IS rule (pre-battle), experienced various grievances with the Iraqi government (pre-IS), voted in the 2014 parliamentary elections, and paid taxes to IS. Neighborhood controls are residentialunit density, population density, and street density. Intermediate controls are primary identity, support for sharia law, attendance at Friday prayer, preferences for IS governance, evaluation of IS corruption, fairness of IS taxation, and evaluation of Iraqi government corruption. *
$p \lt .10$
; **
$p \lt .05$
; ***
$p \lt .01$
; standard errors from Acharya, Blackwell, and Sen’s consistent variance estimator errors in parentheses.
We turn now to a test of H2—that respondents from West Mosul will systematically view the Iraqi military as less legitimate than respondents from East Mosul even after controlling for personal exposure to conflict-related harm. The most straightforward test of H2 would be a regression similar to Equation (1) where we simply control on the right-hand side for respondents’ individual experiences of harm.Footnote 67 However, because controlling for post-treatment variables like actual harm from the battle can induce bias, we instead use a two-step regression method, proposed by Acharya, Blackwell, and Sen,Footnote 68 where we first “demediate” our dependent variable (military legitimacy) by residualizing the effect of actual harm, and then estimate the effect of our treatment variable on our demediated version of military legitimacy.Footnote 69
Under certain assumptions,Footnote 70 this approach estimates the average controlled direct effect (ACDE) of the treatment (whether a respondent resided in East or West Mosul) on the outcome (our survey measure of military legitimacy), which effectively holds constant whether a respondent directly experienced collateral damage during the Battle of Mosul—measured either by self-reports of damage to their home or injury or death to members of their household, or by satellite measurements of damaged buildings nearby.Footnote 71 This approach does not elucidate the precise pathways by which our treatment affects military legitimacy, but it does allow us to rule out a scenario in which military legitimacy is determined exclusively through a mechanism of collateral damage.Footnote 72
Results
Main Results
We begin the results with our tests of H1, which predicts that respondents from West Mosul will perceive Iraqi armed forces as less legitimate than those from East Mosul. Table 2 provides a “naive” test of this hypothesis by simply comparing respondents on the two sides of the river through a standard difference-in-means t-test. In West Mosul, civilians perceive Iraqi forces as more likely to kill innocent civilians in a hypothetical future scenario, and thus less legitimate.
Table 3 reports results from our primary estimation strategy in Equation (1), where we average each respondent’s responses across the three primary components of the Iraqi armed forces to assess their military legitimacy as a whole. Still using the likelihood of killing civilians as our proxy for legitimacy, we find that respondents from West Mosul perceived Iraqi security forces as less legitimate relative to respondents from East Mosul, consistent with H1 (Table 3). On average, and after controlling for individual- and neighborhood-level characteristics, respondents from West Mosul gave Likert-scale responses that were approximately 0.17 points lower than respondents from East Mosul. Interestingly, the primary margin of response is whether respondents describe Iraqi forces as anything more than “very unlikely” to kill innocent civilians—something that happens almost 12 percent more frequently among respondents from West Mosul.
Table 4 presents the results for H2. Specifically, Table 4 presents two-stage regression estimates of the ACDE of the treatment after controlling for the effect of personal exposure to conflict-related harm, which we measure using both survey and satellite data. Column (1) controls for self-reported household damage, column (2) controls for self-reported death or injury to a household member, column (3) controls for satellite-detected building damage within 10 meters of the respondent’s home, and column (4) controls for all three variables simultaneously.
The uniformly negative and relatively sizeable estimated coefficients—extremely similar in magnitude to the estimates in Table 3, where we do not condition on household exposure to harm—provide support for H2: that civilian attitudes are not driven by exposure to physical and material harm alone. Indeed, not even the scale of neighborhood-level harm can by itself explain the lower perceptions of military legitimacy in West Mosul. These ACDE estimates are robust to controlling for satellite-detected harm within a variety of distances (10 to 500 meters) from the respondent’s home (Table A-14). We interpret these results as suggesting that civilian attitudes toward armed forces are influenced by their perceptions of why and how an actor fights, including the casualty-permissiveness of their tactics and the efforts they take to prevent civilian harm.
Further Evidence
Here we leverage qualitative data collected from semi-structured interviews with residents of Mosul, and additional quantitative results, to shed more light on the effects of the Battle of Mosul and our theory of military legitimacy. Due to constraints in our survey data and the nature of the battle itself, our primary quantitative results are limited in two key ways. First, whereas our theory predicts the individual importance of various informational inputs, the Battle of Mosul bundled these inputs into a single source of variation. And second, whereas our theory predicts the individual importance of both just cause and just conduct, our survey data do not support separate tests of each concept, and our outcome variable (whether Iraqi forces are perceived as likely to kill innocent civilians in the future) reflects just conduct more than it does just cause. The following qualitative and quantitative evidence helps address both shortcomings.
First, the qualitative interview evidence helps to establish the individual importance of each of our theorized informational inputs, including the identity and apparent objectives of the armed actors, how they conducted themselves, and the effects of their actions on civilians. For example, several interviewees highlighted differences in the professionalism and attitudes of the CTS relative to the Iraqi police and army. And echoing findings that community-oriented policing can improve police legitimacy in domestic law enforcementFootnote 73 and on the role of intergroup ethnic bias during wartime,Footnote 74 interview subjects specifically cited the Golden Division’s ethnic diversity (and the Iraqi police’s lack thereof) as shaping their experiences of the battle.Footnote 75 Interviewees also emphasized aspects of counterinsurgents’ conduct, especially the excessive and indiscriminate air strikes and wide-impact-area munitions relied on in the west, versus the care counterinsurgents took to avoid civilian casualties in the east.Footnote 76
Our quantitative data are also at least suggestive of many of these same dynamics. For example, we theorize that force composition is one determinant, and indeed, Moslawis on either side of the river tend to view the CTS as more legitimate than the Iraqi Army and the Iraqi Federal Police as less legitimate than either (Table 2).Footnote 77 In addition, survey respondents (from either side of Mosul) who witnessed looting by Iraqi forces tend to view those same forces as less legitimate than respondents who saw no looting (Table A-11). While perhaps not causal, this suggests at least an association between a military’s professionalism and civilians’ perceptions of its legitimacy. Similarly, respondents tend to perceive the Iraqi forces as less legitimate when they live in a neighborhood with high satellite-detected collateral damage (Table A-12), suggesting the importance of both military conduct and the effects of that conduct.
Finally, our qualitative interviews suggest that just cause is an important determinant of military legitimacy, in addition to just conduct. Our respondents overwhelmingly viewed the coalition’s objectives in East Mosul, which they understood to be the liberation of the city with minimal collateral damage, as more justified than the “annihilation” objective used in West Mosul.Footnote 78 As one interviewee said, “there is no such thing as a harmless war,”Footnote 79 but some types of harm were perceived as morally worse than others, consistent with findings from Afghanistan.Footnote 80 Respondents clearly differentiated between intentional and unintentional harms, between harms caused in pursuit of more and less just objectives, and between harms that were more or less necessary to achieve those objectives. For example, several interviewees said that much of the harm in West Mosul could have been avoided if the coalition had allowed IS to retreat into the desert rather than trapping them in the Old City, and that the defeat of IS “was not worth” the destruction it caused.Footnote 81 These findings suggest that civilians evaluate both the actions combatants take and the overall purpose for which they were fighting.
Discussion and Conclusion
Our mixed-methods study yields strong evidence that the coalition’s changes in composition of ground forces, rules of engagement, weapons and tactics, and broader strategic and operational goals (particularly from “attrition” to “annihilation”) was associated with lower perceived military legitimacy in West versus East Mosul. This effect persists even after conditioning on personal exposure to harm, suggesting that civilians’ attitudes to combatants are driven not just by the physical and material harms caused by war, but also by their perceptions of how armed forces conduct themselves, including the extent to which they exercise due care to minimize civilian harm, and why armed forces fight (that is, whether military objectives are just or unjust).
Military policymakers widely recognize the importance of military legitimacy for battlefield effectiveness and national security,Footnote 82 but we argue that it is also intrinsically valuable for other outcomes, including respect for human rights and international law, conflict resolution, peace building, and laying the foundations for effective postwar civilian governance.Footnote 83 Related scholarship on “police legitimacy” suggests that misconduct by state security forces, including excessive violence against civilians, can undermine the legitimacy of other state institutions, potentially eroding trust in democracy itself.Footnote 84 Our findings also have important implications for militaries that provide training and funding for partners and proxies that may be perceived as locally illegitimate (such as US support for Israel’s operations in Gaza and the UAE’s support for the Rapid Support Forces in Sudan). Our finding that perceptions of military legitimacy depend not only on personal experiences with direct harm but also on how and why the harm was caused speak to a growing literature on “cumulative”Footnote 85 and “reverberating effects”Footnote 86 of conflict that often go unaccounted-for in measures of civilian harm, including moral injuries and indirect harm to surrounding economies, the environment, infrastructure, and cultural heritage objects.Footnote 87
Future research can extend and refine this theory, and test additional implications. We hope that other scholars will take up our invitation to study the microfoundations of military legitimacy in contexts beyond Iraq and across different types of insurgencies and conventional wars. An important next step in this research agenda is to bridge insights from micro-level case studies with macro-level patterns to build generalizable knowledge.Footnote 88 Although our research design does not allow us to fully disaggregate the components of our bundled treatment, future research should explore the individual effects of changes in personnel, tactics, technology, type of harm, identity of perpetrators (co-ethnicity, local or foreign), and timing (earlier or later in the battle) on civilian attitudes and behaviors, as well as interactions between these attributes. Survey experiments are one possible tool for doing so.
We also insist on the importance of studying critically important cases like the Battle of Mosul that are worthy of “mere description” in and of themselves,Footnote 89 regardless of our ability to precisely identify causal effects. The Battle of Mosul is already having a profound effect on military doctrine, education, and training, as evidenced by numerous “lessons learned” documents and course materials,Footnote 90 with implications for the complex urban warfare that is unfolding in Gaza and Ukraine at the time of writing.Footnote 91
Data Availability Statement
Replication files for this research note may be found at <https://doi.org/10.7910/DVN/IWTHDK>.
Supplementary Material
Supplementary material for this research note is available at <https://doi.org/10.1017/S0020818325000098>.
Acknowledgments
For helpful comments and suggestions on earlier drafts, we thank the three anonymous reviewers along with Kyle Beardsley, Anu Bradford, Adam Chilton, Darin Christensen, Janina Dill, Alexander Downes, Maj. Gen. Charles J. Dunlap (Ret.), Chelsea Estancona, Peter Feaver, Michael Frakes, Connor Huff, Elena Kantorowicz-Reznichenko, Andrew Kenealy, Alexander Kustov, Gabriella Levy, Jason Lyall, Edmund Malesky, Rachel Myrick, J.J. Prescott, Stephen Rangazas, Margaux Repellin, Jennifer Robbennolt, Katy Robinson, Regine Schwab, Jack Snyder, Megan Stevenson, Joshua Teitelbaum, James Igoe Walsh, Matthew Waxman, and Elisabeth Jean Wood. We are also grateful to participants in the Carolina Conflict Consortium Conference, the Conference on Empirical Legal Studies, the annual meeting of the American Law and Economics Association, Duke’s Security Peace and Conflict Workshop, George Washington University’s Institute for Security and Conflict Studies Workshop, Columbia Law School’s Comparative and International Law Workshop, and law and economics workshops at the University of Michigan, University of Virginia, Georgetown, and George Mason University. We also thank Robert Cerise for extraordinary research assistance and Stella Martany for providing logistical support and local expertise for our fieldwork in 2023.
Funding
This research was supported by the UK’s Economic and Social Research Council UK (grant ref. ES/X01097X/1; SBE-UKRI: Cumulative Civilian Harm in War: Addressing the Hidden Human Toll of the Law’s Blind Spot), the National Science Foundation (award no. 2336310), and the US Institute of Peace’s Jennings Randolph Peace Scholarship Dissertation Program (2017–2018).
Open access
