From Victims to Dissidents: Legacies of Violence and Popular Mobilization in Iraq (2003 – 2018)

A growing literature links experiences of armed conflict with postwar political behavior. This paper examines how legacies of wartime violence shape dynamics of protest in twenty-first-century Iraq. We argue that experiences of shared violence against civilians generate strong social and organizational ties, as individuals turn to neighbors, friends, and communal organizations or social groups to help them cope. These strengthened social networks endure beyond the end of the conflict, forming important vehicles that can facilitate the organization of protest when new grievances or opportunities arise. Further, we posit that these effects will be strongest when the perpetrator of wartime violence is a clear out-group — e.g., a foreign army or non-coethnic militia — which facilitates network strengthening by creating a sense of collective victimization and in-group solidarity. We support these arguments using an original database of Iraqi protests from 2010 to 2012 and data on civilian casualties during Iraq ’ s 2004 – 2009 conflict. We further test our argument with geo-referenced Arab Barometer surveys. We leverage a case study of Fallujah, based on original interviews and other qualitative data, to unpack mechanisms of network strengthening, endurance, and reactivation during the Iraqi protest wave of 2011.


INTRODUCTION
T he twenty-first century has witnessed a striking degree of mass protest. Citizens the world over have taken to the streets at a scale and frequency that some scholars have called unprecedented in modern history (Beissinger 2022;Chenoweth et al. 2019). In some cases, protest activity has taken the form of small-scale resistance and everyday claim-making, whereas in others, protests have escalated into nationwide uprisings or revolutions. While such uprisings are normally thought to occur in relatively well-consolidated states and stable settings, significant protest movements have also erupted in post-conflict societies. According to one prominent database, for example, nearly a quarter of nonviolent uprisings since 1945 have occurred within 15 years of an armed intrastate conflict. 1 Protesting in post-conflict societies often represents an attempt to remake states and societies still struggling with the effects of war and citizens participating in protests are very likely to have experienced wartime violence, either firsthand or indirectly through families and community members. Yet scholars have paid scant attention to how legacies of armed conflict may shape, constrain, or facilitate unarmed mobilization.
Existing research provides good reason to believe that conflict legacies should meaningfully shape subsequent protests. Armed conflicts leave deep and lasting effects on the societies that experience them. Yet for all their obvious destructive impacts, scholars have also consistently found that experiences of violence may paradoxically generate "pro-social" legacies-i.e., that individuals and communities who have suffered violence may exhibit higher levels of trust, more political engagement, and stronger intracommunal social relations.
Building on this literature, we argue that communities suffering higher levels of wartime violence are more likely to organize protests in subsequent periods. Armed conflicts generate severe forms of loss and grievance and individuals experiencing acute collective violence tend to turn to their local networksfriends, neighbors, communal groups, and social organizations-to help them cope. These coping processes strengthen both informal and organizational social ties, resulting in more robust local networks, which then endure after the conflict ends, potentially for years or decades. During these later periods, when new opportunities or grievances arise, the same communities may leverage their strengthened social networks to overcome the collective action problem and organize protests. Further, we posit that this causal process will be most robust when the violence experienced by communities is attributable to a clear out-group-e.g., a foreign army or a militia from a rival social group. Social ties in this case will be strengthened not just by collective coping, but also by a sense of collective victimization and in-group solidarity.
Our study adopts a multimethod subnational comparative design. We focus on the case of Iraq from 2003 to 2018, during which time the country experienced a foreign invasion and deadly civil war, as well as several waves of mass protest. Iraq forms an instructive case not only because of the close temporal proximity of its armed conflict and subsequent protests, but also because its conflict involved a foreign aggressornamely the U.S.-led coalition-and various domestic militias and insurgents. The multidimensional nature of this conflict allows us to probe the mechanisms linking different relational forms of violence with possibilities for later mobilization.
Given the complexity of the case and question, we analyze several forms of data, which yield an array of findings in support of our theory. First, we draw on an original protest dataset from Iraq's first postwar wave of mass mobilization (2010)(2011)(2012), as well as conflict and civilian casualty data from the Empirical Studies of Conflict (ESOC) project. We find that districts with higher wartime civilian casualties experienced significantly higher levels of protest during this time. Further, this relationship is strongest when casualties were inflicted by U.S.-led Coalition forces or sectarian militias. And, we find a small yet significant negative relationship between casualties and later protests when that violence could not be clearly attributed.
We then draw on survey data from Wave V of the Arab Barometer, which was fielded in Iraq from December 2018 to January 2019, and asked participants about their past protest participation. Analyses of these data yield findings that are strikingly similar to those from the event analyses: individuals were more likely to report having joined a protest if they lived in districts with high civilian casualties inflicted by the U.S.-led Coalition or by sectarian militias.
Finally, we draw on qualitative data, including original interviews, YouTube videos, newspaper articles, and leaflets and banners to conduct a short case study of one district that experienced high levels of both violence and protest: Fallujah. We find that the intense violence experienced in the city during the war bred a sense of collective victimization and led to a strengthening of local networks, which subsequently contributed to the organization of protests around a new set of grievances.
This striking resonance in empirical findings-based on triangulating between different types of data, taken from different historical periods-strongly suggests that the communal experience of violence may exert a powerful and durable effect on subsequent capacities for mobilization. Thus, our study bridges two important comparative debates: one about the long-term legacies of violence, and the other about the social and political factors that spur communities to protest.

ARMED CONFLICT AND MASS PROTEST IN COMPARATIVE PERSPECTIVE Explaining Protest and Mobilization
Collective, public protests represent a distinctive and important form of political expression. Citizens develop and deploy "repertoires" of protest (Tilly 1986) in order to oppose public figures, decisions, or institutions, and to advance collective claims for an alternative future. Protests may coalesce into broader uprisings or revolutions (Beissinger 2013;Chenoweth et al. 2019) or they make take the form of smaller-scale or routinized acts of resistance (Schwedler 2022). Protests may be distinguished from institutionalized forms of political participation (such as voting) and from other noncontentious forms of associational life (such as participating in a sports league or a religious service). Particularly where formal political institutions do not serve as reliable channels for popular representation, protests are a primary way that citizens seek to influence politics and redress their urgent grievances.
Among the most consequential waves of twentyfirst-century protest were the "Arab Uprisings" of 2011. Social scientists investigating these uprisings have posited a range of factors to explain patterns of mobilization, including the advent of new internet technologies (Clarke and Kocak 2018;Howard and Hussain 2013), long-term demographic and economic pressures (Cammett and Salti 2018), the damaging effects of neoliberal capitalism (Bayat 2017), and the mobilizing capacity of civil society organizations and parties (Clarke 2014;Yousfi 2017). But, curiously, the literature on the Arab Uprisings-like the field of social movements studies more broadly-has had little to say about how legacies of armed conflict might shape patterns of protest, even as the region is often characterized as "conflict-prone." If anything, many prominent theoretical frameworks for understanding where and when protests emerge tend to presuppose a wellconsolidated state devoid of the kinds of challenges that armed conflicts often produce.
And yet, many protests and uprisings do emerge in countries that have recently experienced armed conflict. Using data from the COW project and the NAVCO campaign database, Figure 1 shows the proportion of nonviolent mass protest campaigns that occurred within 15 years of an "intrastate conflict," broken down by decade. The figure reveals that since 1945, nearly a quarter of nonviolent uprisings (which, of course, represent only one manifestation of protest) have occurred within 15 years of a domestic war. This proportion has actually increased in recent decades, with nearly half of nonviolent campaigns between 2000 and 2009 emerging within 15 years of a conflict. As in the case of Iraq, these Chantal Berman, Killian Clarke, and Rima Majed nonviolent uprisings generally represent popular attempts to reform states and societies still coping with the effects of earlier violence. 2 Given their apparent frequency, and given the wealth of studies on the connections between different waves of violent conflict (e.g., Daly 2012;Grandi 2013;Walter 2004), it is striking that so little research has sought to unpack the potential connections between the prior armed conflicts and later unarmed uprisings. 3

Pro-Social Legacies of Violence
One scholarship we may turn to for insights on the potential relationships between conflict and mobilization is the growing field of quantitative studies that have investigated the individual-and community-level legacies of violence for later political and social outcomes. Despite the wrenching impact that wars inevitably have on societies and populations, scholars in this field have converged upon a finding that violence can exert paradoxically "pro-social" effects on communities and individuals. These effects have been documented following violent conflicts in a broad range of cases, including Sierra Leone, Liberia, Uganda, Burundi, Sudan, Georgia, Nepal, Bosnia and Herzegovina, Sri Lanka, and Colombia. And, they have been found to manifest in a variety of tangible outcomes, including individual attitudes like trust, empathy, and altruism (De Luca and Verpoorten 2015a; Grosjean 2014; Hartman and Morse 2018; Voors and Bulte 2014) and political behaviors like participation in associations, voting, and community leadership (Bellows and Miguel 2006;2009;Blattman (2009);Calvo et al. 2020;De Luca and Verpoorten 2015b).
The finding that violence may have pro-social effects has obvious implications for understanding the relationship between prior conflict and protest. Indeed, many of the studies above explain their results by citing theories of collective action, arguing that experiences of violence help communities come together and act in unison (e.g., Bellows and Miguel 2006;2009;Gilligan, Pasquale, and Samii 2014). But despite the fact that protest represents one of the most paradigmatic types of collective action and a centrally important form of politics, rarely has this scholarship considered how legacies of violence might condition protest participation. Moreover, the few studies that do examine protest as a dependent variable rely on self-reported measures in surveys, rather than the incidence of actual protest events. 4 FIGURE 1. Nonviolent Uprisings in the Wake of Intrastate Conflict: 1940Conflict: -2010 Source: NAVCO campaign dataset and COW intrastate war dataset.
2 Here we are referring to intra-state conflicts that weaken state capacity and produce broken or fragmented regimes. We might expect protest dynamics to work somewhat differently following wars that produce strong and durable revolutionary dictatorships, which are often quite effective at repressing protest. 3 Though in this paper we adopt a sub-national research design, one could also imagine studying the relationship between armed conflict and unarmed uprisings in a cross-national framework-we leave this worthy task to future research. 4 To the best of our knowledge, only two quantitative studies have examined how violence shapes participation in protest, both using survey data. In the first, Grosjean (2014) studies the legacy of violent conflict in Europe on collective action, including joining a demonstration or strike or signing a petition. In the second, Rapp, Kijewski, and Freitag (2019) compare the effect of war victimization in Sri Lanka on institutionalized political participation (i.e., voting) versus noninstitutionalized participation (i.e., protesting and signing petitions). Both studies find a positive effect of wartime violence on protest.
Another shortcoming of this literature is that, while many studies pay careful attention to identifying causal effects, they often have less to say about the mechanisms linking experiences of violence with later pro-social outcomes. As Bauer and co-authors sum up in their review of this field: "the research to date has done a far better job of establishing the effect of war violence on later cooperation than of explaining it" (Bauer et al. 2016, 266).
In the next section, we build on the findings in this literature, positing that legacies of wartime violence are likely to enhance capacities for mass mobilization. Moreover, in order to theorize how and why such a relationship might exist, we draw on insights from the sociological literatures on conflict and social networks. In addition to explaining our argument below, we also summarize it graphically in Figure 2.

CIVILIAN VICTIMIZATION, LOCAL NETWORKS, AND PROTEST
We argue that the experience of wartime violence against civilians may bolster the ability of local communities to stage collective protests long after the end of armed hostilities. More specifically, we posit that the violence inflicted on civilians during wartime strengthens local social networks, as individuals turn to their friends, neighbors, communal groups, and social organizations for support. These networks endure beyond the end of the conflict and may later be "reactivated" to organize and support protests when new political opportunities or grievances emerge (Della Porta 2017; Parkinson 2021). Further, we argue that the identity of the perpetrator determines the extent of network strengthening and, therefore, the possibilities for later mobilization. When violence is inflicted by a well-defined out-group, social ties will be strengthened through in-group solidarity and a sense of collective victimization. But when violence is inflicted by a socially "closer" or unknown actor, networks may be disrupted because of suspicion or mistrust, leading to lower mobilizational capacity in the future.
Per Diani and Mische 2015, we understand "social networks" to be the product of patterned, relatively stable interactions. As in prior research on conflict and social networks, we emphasize relational mechanisms that operate at the meso-level-i.e., at the level of small cities, towns, or urban neighborhoods (Gould 1995;1999;Krause 2018;Majed 2020b;Mazur 2020;. Though these effects may certainly "scale up" to produce nationwide uprisings, the types of repeated interactions constitutive of network formation, transformation, and sustenance take place primarily within these geographic locales. As we explain below, our subnational research design aligns with this meso-level theorization. For the first link in our causal sequence, we argue that violence inflicted upon civilians contributes to strengthening local social networks. As Wood points out, civil wars generally have the effect of "creating new networks, dissolving some, and changing the structure of others" (Wood 2008, 540). While these transformations can take many forms, in this paper, we focus on two key transformations: the strengthening of informal, interpersonal ties through collective coping, and the creation and strengthening of ties to various social and communal organizations. Armed conflicts generate sudden and acute grievances, ranging from casualties to kidnappings to shortages in basic goods or destruction of infrastructure. To cope with the effects of this violence, people turn to their local communities for support. The experience of giving and receiving support among friends, neighbors, coworkers, coreligionists, and extended families generates feelings of gratitude, trust, and reciprocity, imbuing these social ties with strong emotive meaning. Civilians may furthermore come to depend on communal organizations, such as civil society organizations, unions, neighborhood councils, religious organizations, or tribes, especially when such organizations come to serve as focal points for mutual aid and protection. These social organizations not only create "bridges" between otherwise isolated neighborhood-level networks, but in collaborating with each other in providing support they also form interorganizational ties and networks that may knit together an entire city or district.
The idea that wartime violence strengthens local networks through processes of collective coping is supported by some existing research in the legacies of violence literature. For example, in explaining why war-afflicted communities in Nepal demonstrate higher capacities for collective action, Gilligan, Pasquale, and Samii (2014) find evidence to support two mechanisms: (1) individuals who are less socially invested in communities are more likely to flee during the conflict and (2) the individuals who remain must band together to cope with the hardships of the war. In another study, Dorff (2017) finds that a link between crime victimization and pro-social behavior is conditional on an individual being embedded in strong kinship networks.
Importantly, we posit that the strengthening of local networks may depend on the social identity of the perpetrator. We argue that violence inflicted by a clearly defined out-group-like a foreign invader or a non-coethnic militia-will generate particularly strong social ties. In these instances, the ties binding individuals to each other and to communal organizations will be based not just on gratitude and reciprocity, but also a sense of collective, identity-based victimization. Conversely, when violence emerges from an in-group or from an unknown perpetrator, it is likely to breed distrust, fear, and suspicion, which weakens preexisting social ties or undermines network formation. 5 The sociologist Georg Simmel was perhaps one of the first to argue that shared enemies can breed social cohesion (Simmel 1955). This argument also appears in the literature on ethnic boundary-making, with scholars arguing that violence occurring across clear ethnic lines can generate stronger bonds within ethnic groups (Brubaker 2006;Wimmer 2013). Wood (2003; similarly argues that civil wars can make "master cleavages," at the level of the nation-state, suddenly resonant to local communities. Likewise, Blaydes (2018) finds that strategies of collective punishment under Saddam Hussein's Ba'ath regime in Iraq heightened and politicized intergroup boundaries and identity divisions. Similarly, legacies of violence studies have found that attitudes of altruism and trust among individuals who experience violence may prove "parochial" in nature, that is, strengthening bonds with members of a socially defined in-group, but hardening attitudes toward out-groups or society as a whole (Beber, Roessler, and Scacco 2014;Calvo et al. 2020;Cecchi, Leuveld, and Voors 2016;Grosjean 2014;Grossman, Manekin, and Miodownik 2015;Hadzic, Carlson, and Tavits 2020;Lupu and Peisakhin 2017;Rapp, Kijewski, and Freitag 2019).
In the second step of our causal chain, networks forged in the face of violence endure-potentially for months or years following the armed conflict. Previous studies have found the social structures, networks, and organizations formed during wartime to be quite durable, with the potential to pattern violence and other political behaviors well into the postwar period (Daly 2012;Grandi 2013). Finkel's (2015; 2019) research on legacies of repression highlights a similar dynamic, with intense repression in earlier periods creating "resister's toolkits" that activists may redeploy during later repressive episodes. Social movements scholars have likewise emphasized that networks may "persist in a state of latency over a long period of time," only later becoming activated when an opportunity for mobilization arises (Diani and Mische 2015, 309).
This brings us to the final link in the causal sequence: strong networks forged in the context of earlier conflict provide the mobilizing infrastructure for protest during the postwar period. Gould (1995) highlights two ways in which strong networks facilitate mobilization. First, networks provide the organizational infrastructure for protests, channeling resources and providing movement leaders with the means to recruit participants and sanction free riders (see also Marwell and Oliver 1993;Olson 1965;Parkinson 2013;Petersen 2001). Second, networks help to foster collective identities, creating affective links between potential participants. As Mazur (2020) argues, these affective links help to "conduct" grievances that might not otherwise be shared by all network members, heightening their resonance and translating them to a local context. For both of these reasons, networks that may have been forged under particular circumstances may later be repurposed to respond to new scenarios. Gould further specifies that two types of ties within a given community are important in fostering mobilization: highly localized ties based on interpersonal and informal interactions, and more organizational ties that bridge these dense localized clusters and create the potential for more wide-scale collective action. As we have argued above, we believe that both types of ties are likely to result from protracted experiences of violence-particularly when violence is perpetrated by an out-group.
Together, these arguments generate two relatively clear expectations. First, we expect that communities subjected to intense civilian violence are more likely to become sites of protest in later periods. Second, we expect communities that experienced violence specifically at the hands of well-defined out-groups-like foreign invaders or non-coethnics-to be especially prone to protest.
An important caveat is that we do not claim our causal sequence to be the only possible pathway by which wartime violence may pattern subsequent protests. As Wood (2008) notes, there are myriad ways in which wars transform society and reconfigure relations between civilians, organizations, and other powerful actors. For example, in line with Gurr's (1970) theory of relative deprivation, we might expect that the destruction of infrastructure during war creates social inequalities that may spur later protests. More broadly, war may leave behind troubling political practices and socioeconomic conditions, like entrenched corruption and lack of political accountability, which may form the basis of future protest grievances. Districts where conflict is especially acute may also suffer deterioration in state capacity, including police capacity, making protest easier during later periods. Wars also empower rebel groups, some of whom may take on the role of nonstate governors (Arjona 2016;Mampilly 2011;Stewart 2021), and if these groups survive after the end of the war, they may take the lead in organizing anti-state protests. As we explain in the following sections, our quantitative analyses do control for these alternative arguments, and the Fallujah case study helps to substantiate the presence of our network-based mechanisms, in some cases alongside these alternative pathways. But theorizing and testing these arguments in full remains beyond the scope of this paper-we leave this exercise to future researchers.

FROM INVASION AND ARMED CONFLICT TO UPRISING IN TWENTY-FIRST-CENTURY IRAQ
In this section, we provide details on the historical context of Iraq's armed conflict post-2003 and protest dynamics. Iraq is an ideal case in which to study mobilization in the aftermath of conflict. First, the country has witnessed several waves of protest in the decade since the end of the war, allowing us to examine the legacies of violence across multiple temporal horizons. Second, Iraq's vibrant postwar media landscape allows us to assemble fine-grained protest data from multiple sources, a technique known to enhance the validity of event catalogs (Earl et al. 2004). Third, Iraq exemplifies the multidimensional conflict dynamics common to contemporary intrastate wars, with foreign troops fighting alongside domestic militias and insurgent groups. The case, therefore, allows us to investigate how casualties inflicted by different types of actors within the same conflict generate divergent social legacies.
Much of the literature on Iraq has focused on questions of sectarianism, military conflict, and power sharing (e.g., Berman, Shapiro, and Felter 2011;Condra and Shapiro 2012;Haddad 2014;McGarry and O'Leary 2007), with few studies looking at dynamics of protest or mobilization. 6 But, in fact, Iraq has witnessed several waves of mass protest over the last decade, including in 2011, 2013, 2015, and 2018, along with more persistent, low-scale protests staged by various labor and social sectors. More recently, the country was engulfed in a major wave of revolutionary mobilization from October 2019 to March 2020 (Berman, Clarke, and Majed 2020).
Of course, Iraqi society did not always exhibit such high levels of mobilization. Before 2003, during the rule of the Ba'ath under Saddam Hussein, a violent authoritarian regime considerably curtailed opportunities for mobilization (Blaydes 2018). Political opposition under Ba'ath rule mainly occurred through secret organizations that could not protest in public (Ismael 2008). Similarly, under Saddam Hussein's regime, most social and economic organizations such as labor unions or women's organizations were coopted by the regime (Ali 2018;Majed 2020a). However, the strong grip of the Ba'ath regime did not totally foreclose the emergence of social struggles, labor disputes, or political rebellion as shown in several historical accounts of dissent and mobilization (Blaydes 2018; Bet-Shlimon 2019; Rizk Khoury 2013).
The period after the U.S. invasion witnessed two profound transformations in Iraq's social and political landscape. The first involved citizens taking advantage of the political opening brought about by the end of the Ba'ath to mobilize and form grassroots associations. Indeed, as Isakhan (2011, 192) has put it, there is "an entirely 'secret' history to democracy in post-2003 Iraq" that is still to be told. This period witnessed the return of many activists from abroad and the reestablishment of various parties that were previously banned (Majed 2020a). Similarly, the labor movement began to rebuild itself (Isakhan 2021), women's organizations flourished (Ali 2018), and a vibrant, yet heterogeneous, civil society started to take shape (Alshamary 2022). The period also witnessed the transformation of more traditional types of social organizations, such as tribes 7 and religious institutions, which came to play a major role in Iraq's political and social life (Alexander 2018). These political parties, labor syndicates, tribal networks, religious institutions, and civil society organizations constituted a loose network of actors that initiated, organized, and sustained the protests that began to emerge in 2010.
The second and perhaps better-known transformation that Iraq experienced after 2003 was the descent into militarized conflict (Haddad 2013). This period, in which U.S.-led Coalition forces battled sectarian militias, who also fought each other, wracked vast regions of the country with violence-though it did so unevenly. For example, while Iraqi Kurdistan was barely affected by the fighting, regions in Federal Iraq that were considered to be militarily or economically strategic or that had been strongholds of the Ba'ath regime experienced intense fighting-e.g., Fallujah, Kirkuk, Tikrit, Nasiriya, Basra, Karbala, Najaf, and Baghdad. As the war continued, myriad resistance and insurgency groups emerged and guerilla warfare intensified. This armed violence caused immense social and economic harm to the Iraqi population, with huge numbers of dead, injured, and disappeared people, a massive destruction of infrastructure and livelihoods, waves of displacement and migration, the sectarianization of society and politics, and a political class steeped in corruption and cronyism.
Iraqis endured the travails of this war in part by relying on informal social networks, religious organizations, long-standing tribal and familial ties, and newly emerging civil society organizations. For example, the anthropologist Kali Rubaii found that after Mosul fell under ISIS control residents relied on informal trust networks to smuggle medication to hospitals in the besieged city (Rubaii 2020). In addition to falling back on such informal networks, Iraqis also turned to the new grassroots initiatives and local organizations that had been formed after 2003, which helped to forestall full-scale social collapse.
By the end of 2007, armed violence started to decrease and a process of stabilization began. 8 Prime Minister Nouri al-Maliki managed to consolidate his power while simultaneously cracking down on insurgent violence and civil liberties. In addition, the years between 2007 and 2010 witnessed a deterioration in basic services, an increase in youth unemployment, and a stark rise in corruption (Al-Ali 2014). These features of Iraq's emerging political system-particularly the sectarianized corruption and unaccountability of the political class-motivated much of the protest activity over the subsequent decade. Indeed, the armed conflict had only just subsided, when in summer 2010 residents in Basra, which had been suffering from a lack of electricity in the sweltering days of June, took to the streets to denounce the collapse of basic infrastructure. These protests were heavily repressed, but they also set the stage for the major protest wave that emerged a few months later inspired by the uprisings in Tunisia and Egypt.
The protest wave in Iraq in the wake of the Arab Uprisings of 2011 initially started in the capital city of Baghdad in January 2011, and quickly expanded into a nationwide movement. Demands largely coalesced around social, economic, and anti-corruption issues, including electricity, clean water, jobs, and an end to the sectarian-based quota system (known as muhassassa). The mobilization culminated on February 25, 2011 with a "Day of Rage" in which tens of thousands of Iraqis protested across the country (El Helfi 2017, 224). The movement remained active for more than 10 months despite confronting serious state repression. In Figure 3, we plot the weekly count of protests across this full mobilization wave, using our original event dataset (which we introduce more fully in the next section). As the figure makes clear, protest levels spiked during the month of February, but remained elevated through the summer and fall of 2011. Protests also reached many of Iraq's 104 districts, as shown in Figure 4.
Though protest levels declined in October 2011, Iraqi society remained mobilized. In 2012 and 2013, there were numerous protests in the predominantly Sunni western region motivated by high unemployment rates, corruption, the prolonged economic crisis, and the marginalization of Sunnis under the al-Maliki government (Yehya 2017, 21). Next, in summer 2015, large protests erupted in the southern city of Basra and quickly spread to the other parts of the country, including Baghdad. Again, these protests focused on issues of socioeconomic rights, corruption, and redistribution (Jabbar 2018). In 2018, Basra exploded in protest once again, sparking another nationwide wave of contention. Finally, in late 2019, Iraqi society launched the most intense wave of mobilization in the entire postwar period-a massive uprising dubbed the "October Revolution" that raised many of the same demands and grievances as previous waves (Ali 2019b;Berman, Clarke, and Majed 2020). Moreover, while the protest spikes in 2015, 2018, and 2019 garnered major headlines, even in the periods between these uprisings, Iraqis continued to protest in smaller numbers over myriad social, economic, and sectoral issues.
In this overview of Iraq's recent history, we have emphasized two crucial, inter-related processes. The first is the violent conflict that emerged following the 2003 U.S. invasion, which wracked certain Iraqi towns and cities with intense violence. The second is the remarkably persistent mobilization in the post-2003 period, when Iraq experienced multiple waves of mass protest and continuous everyday, small-scale protests. In the next section, we begin our analysis of how these two dynamics may, in fact, have been linked.

LEGACIES OF VIOLENCE AND THE 2011 PROTEST WAVE
In this section, we study patterns of conflict, civilian casualties, and protest at the subnational level to establish whether communities in Iraq that suffered more wartime violence later participated in protests at higher levels. These analyses rely on two types of district-level data: (1) an original database of Iraqi protest events spanning the period from July 2010 to June 2012, coded from three local Arabic-language news sources and (2) data on conflict events and civilian casualties during the period of 2004 to 2009, collected by the ESOC project. 9 The first dataset represents a particularly valuable empirical contribution, as it captures protest events using local Arabiclanguage sources, which are not typically captured in off-the-shelf event datasets that rely primarily on English-language sources (Clarke 2023). These data, therefore, provide a strong measure of subnational protest dynamics in Iraq before, during, and after a major wave of mobilization. In the first appendix of the Supplementary Material, we describe both datasets at length and explain their appropriateness for evaluating our argument; we also include maps laying out the spatial distribution of our main independent and dependent variables.

Empirical Strategy
We use a series of regression models to evaluate the effect of local civilian casualties during the Iraqi war on patterns of protest during the 2011 protest wave (Figure 4). We use a time series approach with a district-month unit of analysis. As the spatial unit of analysis, we use Iraq's 104 administrative districts. Our protest event catalog captures the period from July 2010 to June 2012. Thus, our full sample includes 2,496 district-month observations. Our time series approach allows us to account for the temporal dynamics of protest waves, both in our control strategy (see below) and in our modeling of month-to-month continuity in protest activity. (As a robustness check, we present cross-sectional models using Iraq's 104 districts in Tables A6 and A7 in Appendix 5 of the Supplementary Material.) The dependent variable is a count of protest events in a district-month. To model this outcome, we follow standard conventions in quantitative research on protest and use negative binomial regressions. Standard errors are clustered at the district level. 10 The main independent variable is the per capita number of civilian casualties in a district, sourced from ESOC. In subsequent models, leverage the fact that these casualties data are broken down according to four different types of perpetrators: the U.S.-led Coalition; insurgents in conflict with Coalition forces; sectarian militias battling other local Iraqi forces; and "unknown" perpetrators, where claims of responsibility were not made and/or a clear perpetrator could not be identified. Importantly, these casualty data capture only individuals killed through collateral damage, rather than those deliberately targeted by combatants. For our purposes, civilian casualty counts are a more theoretically relevant measure than aggregate wartime deaths, since we are trying to understand effects of violence on noncombatant citizens and their communities.
Importantly, we control for the level of underlying conflict in a district using ESOC's SIGACT variable, which counts the number of conflict events that occurred in a district. In an influential paper using the ESOC data, Condra and Shapiro (2012) argue that once they control for the number of conflict events in a district, the number of civilian casualties may plausibly be considered exogenous. They reason that any given attack may vary in its casualty rate due to stochastic factors, such as the number of civilians passing through a given blast radius and the effects of wind on shrapnel. They also show empirically that the civilian casualty and SIGACT conflict measures are uncorrelated, both geographically and temporally. Though we do not claim strict causal identification based on this plausible exogeneity, we do believe that by controlling for conflict events in a district, we are able to address major issues of endogeneity (i.e., the fact that the locations where conflicts occur are not randomly determined) and to discount several alternate causal pathways. For example, we are able to rule out the possibility that more protest-prone communities in Iraq were more likely to suffer civilian casualties, perhaps because their members were more likely to join the insurgency. We are also able to dismiss other channels through which the armed conflict might have shaped protest patterns (e.g., by affecting policing capacity or damaging infrastructure). These types of issues of endogeneity, bias, and alternate causal pathways have plagued many past studies examining violence legacies and our ability to address them using a conflict control variable is among the strengths of this research design. 11 Finally, numerous studies have established that violence against civilians during wartime may itself be geographically patterned, by factors ranging from demographic composition (Balcells 2017;Hägerdal 2019) to the presence of resources that governments or rebels wish to extract (Stewart and Liou 2016).
Further, protest levels may be alternatively (or jointly) driven by a range of local social factors, such as poverty levels or the proximity of major urban centers. Our models, therefore, include a range of district-level covariates motivated by these concerns. Model 1 includes only our measures of per capita civilians casualties and per capita conflict events. Model 2 introduces demographic and geographic controls, including logged district population, the district's sectarian population breakdown, the percentage of urban households, a logged measure of distance to Iraq's three largest urban centers (capturing potential coreperipheral dynamics), and the volume of oil pipelines passing through the district (capturing the availability of resource rents). Model 3 introduces temporal covariates, including a lagged protest count variable, and two dummy variables representing potential catalysts for high, nationwide protests. 12 Model 4 introduces social and economic covariates, including a district's rate of illiteracy, the percentage of households belonging to lowest and highest income quintiles, the district's unemployment rate (capturing economic activity), and Source: Original data collected by the authors. 11 In Tables A13 and A14 in Appendix 10 of the Supplementary Material, we run models with an alternative measure of underlying conflict events: the number of U.S. airstrikes in a district during the war (sourced from Papadogeorgou et al. 2022). Results are similar when including this alternative conflict control. 12 Ideally, we would also control for levels of protest in districts during the period before the U.S. invasion (i.e., the pretreatment period). However, we lack high-quality protest data for the final years of the Ba'ath regime. In Appendix 4 of the Supplementary Material, we examine relevant data from the GDELT project to address this shortcoming. the percentage of households reporting continuous availability of electricity (capturing quality of infrastructure). District-level covariates are sourced from the ESOC database and from a unique household survey conducted by the World Food Program in 2007. Descriptive statistics for these and all other variables are included in Appendix 2 of the Supplementary Material. 13

Results
Baseline Model: Effects of Wartime Casualties on 2011 Mobilization Table 1 presents our baseline regression model. We observe a robust and positive relationship between per capita civilian casualties and aggregate protest levels. The relationship persists across all four models, though statistical significance is somewhat attenuated as we increased the number of covariates. Additionally, the relationship between conflict events and protest levels is found to be negative and significant. This divergence suggests that conflict events and civilian casualties operate differently on citizens' protest behaviors in the post-conflict era. Finally, we note several significant covariate effects, including a negative relationship between illiteracy rates and protest levels, a positive relationship with the unemployment rate, and a positive relationship with the percentage of urban households. These associations are in line with prior findings on the social determinants of protest mobilization.
To more clearly represent the relationships captured in these models, Figure 5 shows the marginal effect of wartime civilian casualties on the number of predicted Iraqi protests. The predicted level of protests per district-month during the Iraqi protest wave of 2011 nearly doubles as the independent variable shifts from 50 to 100 civilian casualties per 10,000.

Civilian Casualties Disaggregated by Perpetrator
Table 2 present models disaggregated by the perpetrator of casualties, using negative binomial regressions and the same set of covariates. Here, as with the previous model, we also present marginal effects plots for each of these four casualty variables. These models make clear that who perpetrates violence against civilians determines whether that violence builds subsequent protest capacity. Casualties inflicted by coalition forces have the strongest relationship with later protest levels, accounting for most of the general association between civilian casualties and protest ( Figure 6). The coefficient on the variable capturing casualties inflicted by sectarian militias is also positive and significant, though this relationship is not statistically significant in Model 4. Further, we find a significant and negative effect of unknown violence on subsequent protest activity.
Finally, there appears to be no robust relationship between casualties and protest when these casualties were inflicted by local militias in the context of fighting Coalition forces. Overall, then, we find that in places with high levels of violence perpetrated by a clear out-group-e.g., the army of a foreign occupier, or to a lesser extent, a non-coethnic militialevels of protest in subsequent periods are higher. But when violence occurred at the hands of an unknown culprit, there was actually a small demobilizing effect.

How Protests become Sustained
In a final event data analysis, we explore whether the social ties born of civil conflict may allow protests not just to erupt, but to become sustained within a particular community. We model this relationship as an interaction between the measure of wartime civilian casualties and the lagged protest count variable, which measures protest levels in the district-month prior to the observed district-month. A strong, positive interaction effect suggests that the effect of protests on the likelihood of subsequent protests may, in fact, be conditional on these conflict legacies.
Predicted probabilities plotted in Figure 7 help to interpret this interaction. We plot predicted protest levels as a function of the lagged protest count, with three curves representing divergent levels of wartime civilian casualties. Universally, the relationship between lagged protests and current protests is positive, but the effect size increases as levels of civilian casualty increase. At the mean level of 18 civilian casualties per 10,000 population, an increase from 0 to 10 protests in the prior month corresponds to an increase of roughly 0.3 predicted protest events. At a rate of 1 per 10,000 wartime casualties, this predicted increase is roughly 0.1. At a rate of 30 per 10,000 wartime casualties, this predicted increase is greater than 0.5. These results indicate that legacies of violence may not only shape protest onset or frequency, but also the ability of communities to sustain mobilization over time-a dynamic effect that is crucial for building protest waves.

LEGACIES OF VIOLENCE AND INDIVIDUAL PROTEST PARTICIPATION
In the previous section, we established that districts that suffered from higher numbers of civilian casualties during Iraq's conflict were more likely to be sites of contention during the 2011 protest wave. Moreover, we saw that the relationship comes primarily from casualties inflicted by Coalition forces and, to a lesser extent, sectarian militias. To further substantiate these findings 13 Further, in our appendices of the Supplementary Material, we include several additional models which demonstrate the robustness of our results. For example, our results hold even when removing the outlier districts of Nassriya, Basrah, and Al Resafa, which have especially high protest levels. They also hold when removing all districts in the capital city of Baghdad and when removing Kurdish districts. and to probe the longevity of these potential effects, in this section, we examine the relationship between civilian casualties and individuals' past protest participation, as recorded in a nationally representative survey. We use Wave V of the Arab Barometer, a regular series of nationally representative surveys fielded across the Arab World, which asks about a variety of individual attitudes and behaviors, including protest participation. The survey was also fielded in 2018 and 2019, several years after the first wave of Arab Uprisings and nearly a decade after the end of the war, allowing us to assess how long the legacies uncovered in the previous analyses might endure. We include more extensive description of this survey in the first appendix of the Supplementary Material. Because here the passage of time since the end of the war is longer than it was in the previous section's analyses, we do have to account for dynamics of internal migration. As explained in the first appendix of the Supplementary Material, we only run analyses using respondents who claimed to be living in their current location at the end of the war.

Empirical Strategy
We use binomial logistic regressions to evaluate the likelihood of a respondent having participated in protest. The question in the survey is phrased as follows: "During the past three years, did you participate in a protest, march or sit-in?" We constructed a binary variable from answers to this question, with those claiming to have participated in protest "once" or "more than once" given the value of 1 and those who said they had never participated in a protest given a value of 0. In total, 341 respondents, or 14% of the sample, claimed to have protested over the previous 3 years. As our independent variable, we once again use the civilian casualties data from ESOC, with SIGACT attacks as a control. We, therefore, rely on the same identifying assumption that we used in the previous section, that is, that conditional on a given level of conflict the number of casualties in a district is plausibly exogenous. In addition to controls for the number of attacks in a district, we include a battery of individual-level controls that other studies have shown to affect protest participation. These include the gender of the respondent, whether therespondent attended college, the respondent's age, a squared age term (to capture non-monotonic effects of age), whether the respondent's income was below the national median, whether the respondent was unemployed, whether the responded reported being religious, whether the respondent lived in an urban location, whether the respondent was a student, and the religious identity of the respondent (Shia or Sunni). We also include the same district-level controls that we did in the previous section. And, we cluster robust standard errors at the district level. 14

Results
Results across four models are reported in Table 4. Models 1 and 2 report results using the aggregate version of the civilian casualties measure and Models 3 and 4 report the results when this measure is broken down according to the perpetrator. As with the models in Table 1, we first show only the bivariate relationship between civilian casualties and protest, controlling for SIGACT, and then we add additional controls in Models 2 and 4. To simplify presentation of results, control variables other than SIGACT are not included in the tables; however, full results appear in Table A3 in Appendix 3 of the Supplementary Material.
We see that the per capita number of civilian casualties in a district is positively associated with the likelihood of a respondent in that district having protested and that this relationship is statistically significant when control variables are included. Moreover, when we break down civilian casualties by the assailant, we see results that align with those in the previous section. For instance, we see that civilian casualties inflicted by Coalition forces are positively associated with the likelihood of protest. And, we see that there is also a positive coefficient on the variable capturing casualties inflicted by sectarian militias. As for the other two types of violence, the relationship between insurgent-inflicted casualties and protest likelihood is somewhat ambiguous; it is negative in the simple model and positive but not statistically significant in the model with full controls. However, in line with the models in the  Table 1. 14 Descriptive statistics for all variables can be found in Appendix 2 of the Supplementary Material. Full regression results corresponding to Models 1-4 in Table 4 can be found in Table A3 in Appendix 3 of the Supplementary Material. previous section, the relationship between casualties of unknown origin and protest likelihood is negative. 15 As with the protest models above, we also include marginal effects plots to show the substantive effect of each of these casualty variables on the likelihood of protesting. We see in Figure 8 that the pooled civilian casualty variable has a positive relationship with the Note: Standard errors are clustered at the district level. * p < 0.1; ** p < 0.05; *** p < 0.01.
15 As a further probe of the hypothesized causal mechanisms behind these relationships, in Table A15 in Appendix 12 of the Extended Supplementary Material (Berman, Clarke, and Majed 2023), we find that these casualties variables are also correlated with higher levels of generalized trust among survey respondents-a rough proxy for the strength of local social networks. likelihood of protesting. In districts where there were very few civilian casualties during the war, the probability of a respondent reporting having protested is less than 10%. However, in districts that were racked by severe violence, with civilian casualties per capita of 100 or more, the probability of protesting is about 20%. Then, in Figure 9, we see clearly the striking differences in size and direction of effects across the four types of civilian casualties: those who suffered violence at the hands of the Coalition or sectarian militias are more likely to have protested, whereas individuals who experienced violence by unknown perpetrators are less likely to have protested.

UNPACKING MECHANISMS: PROTEST IN FALLUJAH
Our quantitative results point to a positive relationship between violence legacies and protest, especially when the perpetrator of violence is from an out-group like a foreign invader or an ethnic militia. To further unpack the mechanisms behind this relationship, in this section, we conduct a brief case study of Fallujah, which witnessed some of the fiercest battles during the war and was a site of significant protest activity during 2011. We selected Fallujah to analyze for several reasons. First, it is a district whose protest activity was relatively high and is "well predicted" by our regression models (see Appendix 11 of the Extended Supplementary Material [Berman, Clarke, and Majed 2023] for further details). Second, it is a district where our models suggest that civilian casualties had a considerable impact on protest; when we remove civilian casualties from the model, the predicted protest levels in Fallujah fall well below observed values. Third, there are abundant historical and qualitative data on Fallujah, since it was a major site of conflict during the war.  Table 2.
For evidence, we draw on YouTube videos, newspaper articles, leaflets and banners, as well as a number of in-depth interviews. The interviewees included an Iraqi journalist who covered the protests, two heads of local human rights organizations, a social researcher from Anbar who participated in the protests, an ex-military commander in the Ba'ath regime army, and a psychologist who has been working with an international organization in the city since 2003 (for further information on our interview strategy, see Appendix 13 of the Extended Supplementary Material [Berman, Clarke, and Majed 2023]).
Fallujah was the first Iraqi city to see serious armed conflict in the aftermath of the 2003 U.S.-led invasion. The popular insurgency in Fallujah started as early as April 2003, when U.S. soldiers fired at Iraqi crowds who were protesting inside schools, killing 13 civilians and wounding more than 91 (Malkasian 2006). At that time, the city was heavily affected by the de-Ba'athification campaign and the dissolution of the army, which left between 40% and 60% of its residents without a job or sufficient income (Malkasian 2006, 429). Ahmad, 16 an ex-military commander during the Ba'ath regime, explained how this process of de-Ba'athification created resentment among many of the residents of Fallujah and contributed to a process of sectarianization. He said: "You know, Fallujah is a Sunni-only area, let's not hide behind our finger… The Sunnis of this region during the era of Saddam Hussein were mostly in highly-ranked position in the army: commanders, lieutenants, secret services, etc. They were employees in the highest ranks … so when Bremer came and dissolved the army, he effectively dissolved thousands of families… the income of these families ended… you know, the ex-soldiers are given a lump sum of around $300 monthly. I was a commander, I had servants and assistants and bodyguards, then I suddenly got transformed to someone who would have to wait at the bank's door to get $300… well, I won't accept it for my dignity!" Later, Ahmad explained that throughout this post-2003 period, Sunnis in the Anbar region, including Fallujah, were made to feel like "second-class citizens." He also explained how Fallujah became isolated due to the checkpoints that sealed off the city from other regions. This claim aligns with the work of Martínez and Sirri (2020), who have shown how routine encounters at checkpoints deepen grievances and create a spatial dimension to feelings of marginalization.
These feelings of marginalization, sectarianization, and victimization pushed many of Fallujah's residents to organize and join insurgency groups. During the initial phase of the occupation, the Islamist movement gained popular support as a resistance front against the United States. However, it was not too long before the local population in Fallujah started to turn against what became known as Al Qaeda in Iraq. Starting in 2005, the tribal sheikhs and religious leaders played a crucial role in curtailing the power of Al Qaeda and in organizing resistance against them in what became known as the Sahwa or the Awakening (Green 2010;Lynch 2011;Newton 2017).
However, Fallujans did not simply respond to war by joining militias and organizing armed resistance. The  Table 3. 16 All names mentioned in the text are pseudonyms in order to protect the anonymity of our interlocutors. city also witnessed a flourishing of grassroots networks and organizing to help citizens manage the challenges of the war. Khaled, who heads a human rights organization, explained that Fallujah was known as the "City of Mosques," and is famous for its strong tribes. He noted that following 2003, tribal sheikhs and religious leaders assumed a major role in the city, not only helping to organize armed resistance but also providing civilians with support and protection. Similarly, a variety of grassroots organizations and networks emerged to help residents cope with the conditions of war. Residents organized in professional associations and unions, in pressure groups such as families of detainees and victims of war, as well as through religious initiatives and tribal councils, which provided support and services to those most affected by the violence.
Mazen, a human rights activist, explained that residents who remained in the city turned to each other for help and formed tightly knit networks to survive the violence: "After 2003, there was a security void… the state disappeared, so people started to have recourse to their own capacities, to tribalism… the sons of each region had to protect themselves." Local tribes were especially active in supporting citizens, through ties that dated back to the 1990s (Haddad 2014, 95). Various tribal, religious, and civil society groups also collaborated to advocate on behalf of the city's residents. For example, on October 14, 2004, a group of Fallujan organizations sent a letter to the United Nations Secretary-General Kofi Anan calling for an end to U.S. crimes in the city. 17 The letter was sent on behalf of the "people of Fallujah" and it was signed by five organizations: Al-Fallujah Shura Council, the Bar Association, the Teacher Union, the Council of Tribes Leaders, and the House of Fatwa and Religious Education. Not only does this initiative show that diverse local organizations banded together in the face of conflict, but it also demonstrates how confronting a well-defined out-group can enhance this cohesion.
Sami, a psychologist who works in the Anbar region, explained to us that at the same time that the war fragmented Iraqis at the national level, at the local level, social ties were strengthened. He also emphasized the importance of a common enemy: "social divisions got entrenched… people unified at the local level against an outside enemy." Together, this evidence supports the first part of our suggested causal mechanism that wartime violence inflicted by a welldefined out-group creates a feeling of shared victimization and precipitates collective coping, which together strengthen local networks.
These same local ties and social organizations played an important role in organizing and sustaining protests in Fallujah after the war ended, when new opportunities and grievances emerged. Tribal sheikhs and local religious leaders were especially crucial in spurring protest, as they were connected to extensive networks of followers. Most of our interviewees explained the important role played by religious leaders and mosques in announcing protests through minarets and holding Friday sermons on the Ramadi Highway, where protest tents were set up. Ahmad, the ex-army commander, told us that he used to see pick-up trucks coming from the mosques with megaphones calling on people to join Note: Standard errors are clustered at the district level. *p < 0.1; ** p < 0.05; *** p < 0.01.

FIGURE 8. Relationship between Civilian Casualties and Protest Participation, Pooled
Note: Predicted probabilities plot with 95% confident intervals based on Model 2 in Table 4. protests. He also told us about the food that was provided on daily basis by tribal leaders to the protesters in the tents, which facilitated the longevity of the protest campaigns. This central role of tribal and religious leaders is also apparent in our analysis of videos from the Fallujah protests of 2011, where tribal sheikhs are seen leading the protesters and chanting "Allahu Akbar." Other interviewees also emphasized the role played by looser networks of citizens that had formed during the war, including unemployed youth, the council of students, and the mothers and families of detained or disappeared Fallujans. Interestingly, in line with Gould (1995) and Mazur (2020), we see that these local groups contributed to protest not only through their organizational work, but also by connecting the grievances of the broader uprising to the specific concerns of Fallujan residents. For example, in several of the YouTube videos we analyzed, protesters are heard chanting the famous slogan of the Arab Uprisings "The people want to topple the regime," which was being raised in other cities across Iraq and the Arab region. However, this slogan is quickly followed by chants more specific to Fallujah. We see in these protests the same set of demands that were raised throughout the country-over jobs, electricity, infrastructure, and corruption-but in Fallujah we also see these issues being linked to local grievances against the al-Maliki government, the United States, and Iran. In an analysis of a 5-minute-long YouTube video from the March 7, 2011 Day of Regret mobilization (YouTube Video 7), 18 protesters in Fallujah are heard chanting slogans against sectarianism ("Brothers Sunna and Shia, this country we won't sell"), in celebration of Fallujan resistance ("despite the wish of our enemies, Fallujah will remain resistant"), against the United States ("America, your days are counted"), against Iran ("Iran out out, Basra will remain free"), and against al-Maliki ("Oh Maliki, patience patience,  Table 4. 18 Full details for all YouTube videos can be found in Appendix 13 of the extended Supplementary Material. Fallujah will be your graveyard," "Oh Maliki remove your hand, this population doesn't want you"). This medley of chants within a few minutes exemplifies how protest leaders connected local and national grievances.
Here, it is important to remember that these protests began while U.S. troops were still present in Iraq (the United States only withdrew in December 2011). They also occurred following the discriminatory policies adopted by al-Maliki-especially the arbitrary use of Article 4 of the Anti-Terrorism Law to target Sunni leaders (al-Mukhtar 2010). In this context, one of the biggest protests in the city took place on April 9, 2011 on the eighth anniversary of the U.S. invasion of Fallujah. In a video from this long day of protest (YouTube Video 10), a tribal sheikh is heard giving a fiery speech against both the U.S. and the Maliki government. Facing the sheikh stands a man holding a big banner that reads "No to dividing Iraq, no to federalism." In another corner of that protest, another video (YouTube Video 12) shows young men burning both a U.S. flag and a caricature drawing of al-Maliki. The camera focuses on a young man who says: "this is the destiny of traitors, this is the destiny of those who allowed the tanks in… these are the men of Fallujah, these are our free men!" In parallel to these masculinist slogans, we see women in another video of the same protest (YouTube Video 13) gathered around an older lady shouting: "this is a state of traitors, a corrupt state… a corrupt state against the people… a state against the people… a state of muhassassa… … in spirit, in blood, we shall redeem you Iraq." Here again, the celebration of Fallujan resistance and heroism is coupled with a rejection of sectarianism and an enmity toward both the central government of al-Maliki and the U.S. forces. While many slogans addressed national issues like sectarianism and corruption, others focused on a more specific concern to Fallujah: individuals who were imprisoned or disappeared during the war. The salience of this demand can be traced to the important organizational role played by a loose network of families who have lost loved ones during the conflict, and which was formed during the war. In another video from the April 9, 2011 protest (YouTube Video 14), we see a group of women holding photos of young men and banners with slogans asking about the detainees and the disappeared. One woman approaches the camera and says: "Where are human rights from what is happening in Fallujah? Why don't they ask for the punishment of those who have assaulted Fallujah? (… ) Is this a government? Is this a state that rules us? Is this justice? Is this right? We want the punishment of all those who have assaulted Fallujah, and those who sold Iraq for cheap, and we demand the release of our sons from the secret prisons and the Kuwaiti prisons." Here, we see the melding of grievances related to the overall uprising ("Is this a government? Is this a state that rules us?") with demands specific to Fallujah residents. For example, the woman mentions the release of Fallujan prisoners being held in secret prisons and prisons in a Kuwait (a grievance that dates back to the Gulf War). This protest was organized by a network of families formed during the conflict. Moreover, we observed in other protest videos that these demands were also widely upheld by tribal and religious leaders. We, therefore, see that networks formed during the war-between tribes, religious leaders, and families of war victims-later worked together to organize protests, and to translate the grievances of the nationwide uprising into demands and idioms that would resonate in Fallujah's local context.
To sum up, in Fallujah, we see that social networks and grassroots organizations formed and strengthened during the war through collective coping and a sense of shared victimization endured beyond the end of the conflict. Then, when a new political opportunity emerged in 2011, they facilitated mass mobilization. Not only did they directly organize and participate in protests, but they also performed important work in connecting the issues of the broader uprising to those that most affected the city's residents. These efforts help to explain why Fallujah saw such high levels of protest during the 2011 protest wave.

CONCLUSION
Together, the analyses above provide a wealth of evidence that violence inflicted on civilians during wartime can lead to higher and more sustained protest mobilization after the conflict ends. Further, when we disaggregate these wartime casualties, we see evidence that who inflicts the violence matters a great deal. When the violence comes from clear out-groups-in this case, the U.S.-led Coalition forces or militias from another sectarian group-it is more likely to generate strong and cohesive networks, which later facilitate the generation of protest. But when those casualties are inflicted by unknown actors or members of an in-group, there will be limited network strengthening, resulting in lower protest levels in later periods.
Though this paper is based on in-depth examination of a single country, we do expect our findings to be relevant to comparative knowledge and theorybuilding. As noted in the introduction, there are many historical instances of mass uprisings arriving on the heels of armed conflict. Of course, because our argument is a spatial one, and emphasizes the important of highly localized social networks formed during war, we would only expect it to apply to cases where the armed conflict and the unarmed mobilization occurred in the same location. To demonstrate that other cases with these dynamics do exist, and that our argument has potential utility in explaining them, we include in Appendix 14 of the Supplementary Material three brief shadow cases of Peru, Nepal, and Georgia. Though our analyses of these cases are nowhere near as detailed as those we have conducted above, we offer some preliminary evidence that the mechanisms observed in Iraq may also prove relevant in these cases.
Two theoretical caveats (noted first in our theory section) are worth repeating here. First, we do not claim that the causal pathway we have spelled out in this paper-which links wartime civilian casualties to post-conflict mobilization via the mechanism of strengthened local networks-is the only potential pathway connecting armed conflict to subsequent mass mobilization. Second, we do not claim that protest is the only political or social outcome that is likely to be affected by the network-strengthening effect of violence. We have sought to manage the complexity of the social processes at work by focusing on a singular causal pathway and we marshal evidence to substantiate the specific empirical patterns implied by that argument. We encourage future researchers (in the case of Iraq and elsewhere) to build on our study by investigating alternative pathways connecting conflict and protest or by expanding the dependent variable beyond protest mobilization.
From a methodological standpoint, it is important to be upfront about the limitations of the findings and analyses in the paper. The most obvious empirical limitation is that, in our quantitative sections, we rely on observational analyses of protest and survey data, which means we cannot entirely mitigate concerns about endogeneity, omitted variables, or other threats to causal identification. Nonetheless, we believe that our observational and multimethod approach provides a number of important advantages in addressing a complex research question. Our choice to measure actually occurring protests (rather than self-reported measures of support or participation) greatly enhances the validity of our measurements. In our regression analyses, we included carefully selected and theoretically motivated control measures, reflecting multiple data sources and, most importantly, capturing multiple dimensions of underlying conflict. Our interviews provided micro-level substantiation of our causal mechanisms at work and enabled critical theory-building through close study of one city that exemplifies the strong relationship between wartime violence against civilians and later protest mobilization. We hope that readers will take confidence from the strong consistency in the patterns we observe across different types of data, representing difference sources, scales, and time periods.
We hope that our project may spur future research in several additional areas. First, our study contributes to the nascent literature on Iraq's "hidden democracy." While much social science research on Iraq to date has focused on armed conflict per se, our study joins other newer work that emphasizes the importance of civil society and mass mobilization in shaping Iraq's politics. Building on our study of protest in Fallujah, we hope to see further site-specific research on protest mobilization in Iraq, focused either on major protest waves or on intervening periods of routine mobilization.
Second, the study provides strong evidence that the "prosocial" effects of violence identified by other scholars also apply to protest mobilization. Our paper also finds evidence to support the idea that there may be an unfortunate "dark side" to these pro-social effects (Calvo et al. 2020;Grosjean 2014;Hadzic, Carlson, and Tavits 2020). Local networks strengthened through experiences of collective victimization and in-group solidarity may provide capacities for subsequent mobilization, but they may also contribute to the hardening of sectarian cleavages and the weakening of national affiliations, a pattern that is also borne out by Iraq's recent history. Future research may wish to focus more on the potential connection between sectarian identification and protest mobilization.
Third, our research contributes to debates about the processes underlying protest, mobilization, and mass uprisings. Rarely have scholars considered the relationship between civil conflict and protest. Future research might explore this question in a crossnational setting, examining whether societies that have recently experienced war are more likely to launch civil uprisings. More generally, we hope that this paper will push scholars to integrate studies of armed conflict with those of unarmed protest-two forms of "contentious politics" (McAdam, Tarrow, and Tilly 2001) that may prove more tightly linked than we previously believed.

SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055423000254.

DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available in the American Political Science Review Dataverse at https://doi. org/10.7910/DVN/QQKR3F. Some limitations on data availability are discussed in the README file.