Why Protectors Become Predators? Violence against Civilians by AMISOM Peacekeepers in Somalia

Abstract

This study investigates the phenomenon of peacekeepers resorting to lethal violence against civilians, despite their mandate to protect them. While existing literature acknowledges that some peacekeepers perpetrate human rights abuses, including sexual exploitation and excessive force, there has been a notable neglect of studies on lethal violence within the broader spectrum of peacekeeping misconduct. I propose and evaluate two explanations: the revengehypothesis, suggesting emotional responses to attacks on comrades as drivers of violence, and the show of force hypothesis, where violence is an unintentional byproduct of deterring perceived threats. Using original survey data and an event dataset from AMISOM in Somalia, I leverage the random exposure to Improvised Explosive Devices (IEDs). Results suggest that while IED attacks on peacekeepers are linked to subsequent violence against civilians, lethal force is primarily used following the loss of a fellow peacekeeper due to an IED attack, supporting the revenge mechanism. These insights reveal the intricate interplay of emotions and peacekeeping dynamics, providing insights into the origins of abusive behavior.

Este artículo investiga el fenómeno referente al personal de mantenimiento de la paz que recurre a la violencia letal contra los civiles, a pesar de tener la obligación de protegerlos. Si bien la bibliografía existente reconoce que algunos miembros del personal de mantenimiento de la paz cometen abusos contra los derechos humanos, entre los que se incluyen la explotación sexual y el uso excesivo de la fuerza, existe una falta notable de estudios relativos al uso de la violencia letal dentro del espectro más amplio de las faltas de conducta en el mantenimiento de la paz. En este contexto, proponemos y evaluamos dos explicaciones: la hipótesis de la venganza, que sugiere que las respuestas emocionales a los ataques a los camaradas son los impulsores de la violencia, y la hipótesis de la demostración de fuerza, donde la violencia es un subproducto no intencional de la disuasión de las amenazas percibidas. Utilizamos datos de encuestas originales, así como un conjunto de datos de eventos de AMISOM en Somalia, que nos permiten utilizar los datos relativos a la exposición aleatoria a Dispositivos Explosivos Improvisados (IED, por sus siglas en inglés). Los resultados sugieren que, si bien los ataques con IED contra el personal de mantenimiento de la paz están vinculados a la violencia posterior contra la población civil, la fuerza letal se utiliza principalmente después de la pérdida de un compañero del personal de mantenimiento de la paz debido a un ataque con un IED, lo que respalda el mecanismo de venganza. Estas observaciones revelan la intrincada interacción existente entre las emociones y las dinámicas de mantenimiento de la paz, y proporcionan información sobre los orígenes del comportamiento abusivo.

Cette étude se penche sur le phénomène des officiers du maintien de la paix qui ont recours à la violence mortelle à l'encontre de civils, malgré leur mandat de protection de ceux-ci. Bien que la littérature existante reconnaisse que certains officiers du maintien de la paix commettent des violations des droits de l'homme, y compris l'exploitation sexuelle et la violence excessive, l'on note que les études omettent la violence mortelle dans le spectre des inconduites du maintien de la paix au sens large. Je propose et évalue deux explications : l'hypothèse de la revanche, qui suggère qu'une réponse émotionnelle aux attaques sur des camarades motive la violence, et l'hypothèse de la démonstration de force, dans laquelle la violence serait un produit dérivé involontaire de la dissuasion des menaces perçues. À l'aide de données de sondage inédites et d'un ensemble de données d’événement d'ANISOM en Somalie, j'exploite l'exposition aléatoire aux engins explosifs improvisés (EEI). D'après les résultats, bien qu'il existe un lien entre les attaques aux EEI sur les officiers du maintien de la paix et les violences ultérieures à l'encontre de civils, la violence mortelle est principalement utilisée après la perte d'officiers de maintien de la paix lors d'une attaque à l'EEI, ce qui vient étayer le mécanisme de la revanche. Ces renseignements mettent au jour l'enchevêtrement de liens étroits entre les émotions et les dynamiques de maintien de la paix, ce qui nous renseigne sur les origines des comportements agressifs.

“We fight according to certain principles, but if something shoots at us, they ‘[explicit]’ die.”

Opening remarks from US Army Capt. McClean in the Netflix documentary on Afghanistan “The Fighting Season.”

Introduction

On December 19, 2017, a Hawiye traditional clan elder openly criticized AMISOM on a local FM radio station in Mogadishu in response to the reported deaths of six civilians by the African Union Mission in Somalia (AMISOM) in Qoryooley district, Lower Shabelle.¹ The elder held AMISOM responsible for the civilian deaths and cautioned that “if such harm continued, Somalis would take up arms to fight against them.” Modern UN peacekeeping operations have evolved beyond traditional roles to address twenty-first-century challenges, incorporating counter-terrorism and counter-insurgency aspects (Karlsrud 2018, 2019; Rosas Duarte and Souza 2024). However, these shifts face criticism for their limited capacity to effectively handle counter-insurgency demands and protect civilians in conflict (dos Santos, William, and Salvator 2017). To address these issues, the UN has collaborated with regional organizations like the African Union (AU), which have shown effectiveness in containing violence (Bara and Hultman 2020; Schumann and Bara 2023). These partnerships, while promising in addressing UN mission underperformance (UNSC Res 2436 2018), also face allegations of human rights violations, such as those by AMISOM in Somalia (Williams 2019a; OHCHR 2017). Events such as the one in Wardinle town, on July 17, 2016, where AMISOM inflicted one of the most severe civilian casualties in a single incident,² resulting in fourteen civilian deaths and three injuries (OHCHR 2017), highlight concerns about the legitimacy of any partnerships involving the UN. In light of such events, ensuring stringent oversight is crucial to balance security and human rights, as the Protection of Civilians (POC) remains a core peacekeeping tenet requiring adherence to International Humanitarian Law (IHL) (Fjelde, Hultman, and Nilsson 2019).

Extensive scholarly work has explored the motives behind intentional violence against civilians by state and nonstate armed groups (Balcells 2017; Humphreys and Weinstein 2006; Kalyvas 2006; Valentino 2014; Wood 2010; Weinstein 2006; Wood and Kathman 2014). However, the peacekeeping literature has largely neglected the systematic examination of lethal violence perpetrated by peacekeepers against civilians.³ While empirical research on violence by government soldiers against civilians is growing (Bell 2022; Bell, Gift, and Monten 2022; Van Wie and Walden 2023), exploration of human rights violations by peacekeepers has primarily focused on sexual exploitation and abuse (Beber et al. 2017; Moncrief 2017), with limited studies examining broader human rights abuses by peacekeepers (Rodriguez and Kinne 2019; Horne, Lloyd, and Pieper 2022). This research gap, particularly on lethal violence by peacekeepers, is significant given the prevalence of contemporary peacekeeping operations under Chapter VII mandates similar to counter-terrorism and counter-insurgency activities (Cruz and Cusimano 2017; Friis 2010; Karlsrud 2019). As peacekeeping and counterinsurgency (COIN) operations increasingly intersect, understanding how violence against civilians shapes civilian sentiments is crucial for maintaining support for peacekeeping efforts (Ruggeri, Gizelis, and Dorussen 2013).

This study investigates the factors that lead peacekeepers to employ lethal violence against civilians despite their mandate to protect them from harm. The argument applies to Chapter VII peace enforcement missions, where peacekeepers are deployed in high-intensity conflict zones and face deliberate, lethal attacks aimed at provoking violent responses, such as the insurgent use of IEDs in AMISOM Somalia. Although rare in UN missions, incidents such as the near-overrun of a UN base by the Allied Democratic Forces in the Democratic Republic of the Congo (DRC), along with multiple attempts to capture UN bases in Mali, underscore the persistent risks peacekeepers face in volatile enviornments (Williams 2023). These contexts often necessitate militarized and coercive operations, signaling a shift from traditional peacekeeping towards missions with COIN characteristics, where peacekeepers operate with expanded mandates and military capabilities (Friis 2010; Howard 2019; Rosas Duarte and Souza 2024), including the use of artillery and attack helicopters by UN peacekeepers in conflict zones such as the Central African Republic and the DRC (Hunnicutt, Nomikos, and Williams 2021).

Amid growing militarization and overlap with COIN, peacekeepers continue to uphold POC and IHL, aware that their legitimacy depends on balancing military engagement with humanitarian responsibilities, as civilian harm risks eroding local faith in peacekeeping missions (Hultman, Kathman, and Shannon 2013). Peace enforcement missions like AMISOM in Somalia demonstrate how POC strategies are integrated into military operations. By 2012, AMISOM had begun implementing POC measures and reinforcing IHL adherence, years before the UN Security Council's 2019 POC mandate (Williams 2018). Consistent with IHL's distinction between civilians and combatants (Henckaerts and Alvermann 2005), AMISOM introduced specialized training in 2012, enabling peacekeepers to distinguish combatants, avoid indiscriminate attacks, and exercise restraint (AMISOM 2012). This dual emphasis on military capability and civilian protection reflects an evolving peacekeeping model where civilian protection is pivotal to mission success.

I propose and evaluate two hypotheses to explain violence against civilians by peacekeepers: “revenge,” suggesting emotional responses to attacks on fellow peacekeepers, or “show of force,” indicating civilian casualties arise as an unintended byproduct of military actions aimed at deterring perceived threats. I investigate AMISOM in Somalia, an innovative regional-UN peacekeeping model, as a compelling case study. I use two measurement strategies to capture civilian abuse by peacekeepers. First, I conducted surveys with 600 active AMISOM peacekeepers to create an attitudinal human rights index based on IHL. Second, I rely on the United Nations’ daily security incident reports in Somalia, a nonpublic resource, to build an original event dataset capturing every confrontation involving AMISOM peacekeepers from June 2014 to April 2019. For my independent variable (IV), I draw on existing conflict literature highlighting insurgents’ use of Improvised Explosive Devices (IEDs) to disrupt counter-insurgency strategies (Condra et al. 2018; Sexton 2016). This variable encompasses both self-reported and actual IED attacks targeting peacekeepers from their arrival in Somalia up until the time of the survey. I focus on this specific timeframe for identification, as it remains independent of the frequency of IED attacks or AMISOM casualties.

My findings reveal that the likelihood of civilian abuse by AMISOM peacekeepers is closely tied to the occurrence of a fellow peacekeeper's death resulting from indiscriminate attacks. Specifically, an increase in one unit in an IED attack corresponds to a 0.60-unit rise in the probability of peacekeepers resorting to lethal violence against civilians. Furthermore, there is a 0.19 standard deviation increase in my attitudinal standardized index compared to scenarios without IED attacks against peacekeepers. Additional tests reveal a significant increase in the likelihood of peacekeepers using lethal violence against civilians following the loss of a comrade in an IED attack. This increase aligns with the “revenge” mechanism, reflecting emotional responses to indiscriminate and intensified explosions, rather than the “show of force” rationale.

This research makes two key contributions. First, it distinguishes between deliberate and unintended but foreseeable violence against civilians, highlighting how mission-specific factors explain why peacekeepers resort to lethal violence, particularly in response to indiscriminate IED attacks targeting them (Moncrief 2017). The study identifies emotional triggers—such as anger and fear—that lead peacekeepers to act out of revenge, especially in COIN-like operations. While existing literature has explored factors influencing intentional violence by government soldiers (Bell 2022), this study offers micro-level insights into the role of emotions in instigating acts of revenge and violence (Balcells and Stanton 2021; Petersen 2002). Methodologically, the research contributes by focusing on the UN–AU partnership as a case study, an area that has received limited empirical attention (see Schumann and Bara’s (2023) recent work). While qualitative research and journalistic accounts often provide descriptive details of peacekeeper-inflicted harm, this study is one of the first to survey peacekeepers and compile robust data on their use of lethal violence against civilians.

While the findings have limitations in generalizability, this research provides critical insights into insurgent tactics, highlighting their strategic use of IEDs to provoke peacekeeper responses. Although not all IED attacks are intended to elicit such reactions, certain attacks are deliberately designed to incite retaliation, which can inadvertently harm civilians and undermine mission legitimacy. By exploiting this dynamic, insurgents manipulate local perceptions of peacekeepers, further complicating their operational environment and jeopardizing the broader objectives of peacekeeping missions. These findings have significant policy implications for influencing peacekeeper behavior in combat scenarios, particularly given the growing resemblance between current peacekeeping operations and COIN efforts (Cruz and Cusimano 2017). They shed light on the challenge of reconciling the formidable combat capabilities of regional bodies like the AU, designed to confront ongoing security threats, with the necessity to uphold civilian protection and human rights principles inherent in contemporary UN peacekeeping missions.

Lethal Violence against Civilians by Peacekeepers

Scholars have proposed various mechanisms to explain instances of civilian abuse by state and nonstate armed groups. Among these, conventional rationalist explanations emphasize the evaluation of costs and benefits by actors involved in conflicts (Kalyvas 2006; Humphreys and Weinstein 2006; Wood 2010; Valentino 2014). While these discussions provide valuable strategic insights into why government troops might deliberately target civilian populations during civil conflicts, they fall short of comprehensively explaining peacekeepers’ potential involvement in lethal violence against civilians. This limitation stems from two key factors. First, in UN-sponsored peacekeeping missions, civilian harm—whether intentional or not—is strictly prohibited. Unlike combat operations, where collateral damage may be deemed an unfortunate but acceptable risk in certain conditions (Walzer 2015), peacekeepers operate under a “do no harm” principle (United Nations 2023), which mandates preventing civilian harm in all situations. Even without an explicit POC mandate, peacekeepers must prioritize civilian safety and actively mitigate harm (Williams 2018, 262).

Additionally, peacekeepers work to build trust with local communities, using initiatives like quick impact projects to engage and stimulate economic activity (Bakaki and Dorussen 2023; Gizelis and Cao 2021; Kassem 2017; Leloup 2019; Sauter 2022). In Haiti, peacekeepers gained cooperation by participating in relief efforts, improving public opinion, and encouraging information-sharing (Gordon and Young 2017). Even in missions resembling counter-insurgency operations, such as in Somalia, peacekeepers implement public goods initiatives, including health clinics, schools, police posts, and food distribution programs (AMISOM 2024). Despite these efforts, instances of peacekeepers harming civilians persist, highlighting the tension between their protective mandate and the operational realities of conflict zones. Notable examples include the fatal shooting of four civilians by AMISOM troops in November 2018 (Amnesty International 2018) and an incident in February 2023 where eight civilians were killed in a clash with UN peacekeepers in East Congo (Reuters 2023). These cases emphasize the ongoing challenge of balancing the duty to protect civilians with the pressures of complex operational environments.

Building on this tension, recent research has started to examine factors influencing human rights abuses by peacekeepers, identifying the institutional characteristics of troop-contributing countries as a significant determinant. Findings point to peacekeeping forces from countries with extractive institutional frameworks or low compliance with international norms as more likely to engage in civilian abuse (Horne, Lloyd, and Pieper 2022; Rodriguez and Kinne 2019). While these studies provide critical insights into the role of institutional origins in shaping peacekeeper behavior, they also highlight the necessity of broadening our understanding to encompass a range of factors contributing to civilian harm. Rather than attributing violence to unintentional “collateral damage” due to rising attacks or operational constraints, it is crucial to recognize the broader spectrum of intentional and structural human rights violations in peacekeeping missions. In the following section, I present two micro-level explanations for lethal violence perpetrated by civilians against civilians, which delve deeper into the motivations and circumstances surrounding these acts.

Emotional Drivers of Revenge

Understanding peacekeepers’ violence against civilians requires considering emotional responses and situational factors that shape military conduct in conflict settings (Balcells 2017; Balcells and Stanton 2021; Petersen 2002). While micro-level emotion-based theories are insufficient alone (Balcells 2017), scholars acknowledge that armed groups engaging in civilian victimization share typical backgrounds (Dutton, Boyanowsky, and Bond, 2005). Situational factors, such as those elaborated in social-psychological research (Zimbardo 2011) and historical analyses of mass violence (Christopher et al. 1992), can influence the inclination to harm civilians. This suggests that revenge, fueled by emotions like anger, could explain peacekeeper violence, particularly in response to attacks on them.

To understand the dynamics of revenge in peacekeeping operations, I propose a theory that centers on the role of emotions—such as shame, disgust, resentment, and anger—in driving acts of revenge killings (Balcells 2017; Petersen 2011). Building on this, I conceptualize a two-step framework to explore how these emotional drivers unfold in conflict settings. The first step integrates Wood’s (2010) concept of “relative capability,” which emphasizes how the interaction between the capacities of peacekeepers and insurgents profoundly shapes conflict outcomes. Scholars have highlighted how armed groups engage in strategic calculations that target military personnel and aid workers as part of a broader competition for civilian support (Kalyvas 2006; Narang and Stanton, 2017; Sexton 2016).

In regions with a large peacekeeper presence, rebel attacks on civilians typically diminish as a direct consequence of the peacekeepers’ presence (Hultman, Kathman, and Shannon 2013). As rebels contest government troops for civilian support, rebels use violence against peacekeepers to limit peacekeeper interactions with the population and undermine the mission's ability to enforce peace (Hunnicutt, Nomikos, and Williams 2021; Fjelde, Hultman, and Lindberg Bromley 2016; Ruggeri, Gizelis, and Dorussen 2013). Rebels may therefore strategically target peacekeepers—a tactic that exploits the asymmetry in military capabilities, where rebels—often constrained by limited resources—seek to weaken peacekeepers’ operational effectiveness through direct assaults (Fjelde, Hultman, and Nilsson 2019). This strategic targeting undermines peacekeepers’ mobility and their ability to perform protection duties, as evidenced by armed groups obstructing peacekeepers in Darfur, which significantly hindered peacekeeping efforts to safeguard civilians (Duursma 2019; Fjelde, Hultman, and Nilsson 2019). Moreover, deliberate violations against civilians can also hamper the gathering of local intelligence on insurgent activities, thereby obstructing broader international stabilization efforts (Hultman 2009). In contexts where rebel groups possess military superiority over government forces prior to peacekeeper deployment, rebels have been observed to resort to even more extreme tactics, such as terrorism, to “overcome the physical barrier” and enhance their bargaining power relative to the government (Di Salvatore, Polo, and Ruggeri 2022).

If the initial phase involves a resource disparity that leads insurgents to strategically target peacekeepers, the subsequent stage of the “revenge” mechanism explores how peacekeepers’ retaliatory violence is fueled by intense emotional responses, particularly anger and fear, following insurgent attacks. This process, as described by Petersen (2002, 22–3), operates through “belief formation” in a sequential pattern, where emotional responses are shaped by an individual's perception and interpretation of the situation. According to Petersen (2002, 17–8), emotion acts as a “mechanism that triggers action to satisfy a pressing concern.” In contemporary peacekeeping missions, which often resemble COIN operations, peacekeepers are frequently subjected to lethal IED attacks (Williams 2019b), setting the stage for an emotional cascade. Driven by the instinct for self-preservation, the aftermath of such attacks triggers heightened anger and fear, compelling peacekeepers to seek revenge and retaliate against those they perceive as threats. The emotional residues left by these intense experiences, such as anger, humiliation, and resentment, continue to influence peacekeepers’ motivations for future retaliation. As Petersen (2002) emphasizes, these emotions are not simply reactions to immediate threats; they serve as ongoing catalysts, shaping future actions and reinforcing cycles of violence. Balcells (2017) illustrates similar dynamics in the Iraq and Afghanistan wars, showing how emotions rooted in loss and the pursuit of revenge distort perceptions and justify harming perceived wrongdoers as a form of repayment for transgressions.

Emotions can compel individuals in conflict to pursue revenge, leading to violence that surpasses strategic objectives and often lacks tangible gains (Frank, 1988, 53). In the aftermath of an IED attack, the emotional upheaval of battle may blur the line between civilians and combatants, causing peacekeepers to violate operational boundaries and engage in violence against noncombatants. As Balcells (2017, 38) notes, vengeful motives can overshadow rational decision-making, driving individuals, fueled by anger from victimization, to attempt to correct perceived injustices. This shift from rationality to emotion explains why actors may act against their strategic interests. Elizur and Yishay-Krien (2009, 261) illustrates this phenomenon in a qualitative study of Israeli soldiers in Gaza, showing how situational pressures can push individuals beyond self-defense into excessive violence. Petersen (2011, 56) offers a conceptual framework to further understand these behaviors, distinguishing between “discriminate violence,” which incites anger, and “indiscriminate violence,” which instills fear. The latter form of violence, driven by uncertainty and heightened threat perception, can lead to “opportunistic violence” fueled by personal gain or status (Manekin 2020, 7). This concept aligns with the idea that emotional intensification distorts threat perception, causing actors to engage in violence that deviates from military goals. By connecting opportunistic violence to personal incentives and the broader conceptual distinction between discriminate and indiscriminate violence, we gain a deeper understanding of the motives underlying violent behavior. Grossman (2014) provides an empirical perspective by linking civilian harm by US soldiers in Vietnam to the psychological toll of their actions, emphasizing the profound emotional and psychological costs borne by perpetrators which underscores how emotions, personal motives, and situational pressures interact to drive revenge-driven behaviors. Collectively, these studies offer crucial insights into the complex dynamics that lead to civilian abuse by peacekeepers, highlighting how emotions and incentives operate within conflict environments.

To summarize, insurgents, driven by their inferior capability, strategically employ violence against peacekeepers. Their dual objective is to incite peacekeepers into violent acts against civilians, thereby triggering punitive responses. Consequently, civilians might withdraw support and share intelligence with insurgents, influencing conflict trajectories (Shaver and Shapiro 2021). If successful, this strategy should lead to the portrayal of peacekeepers as predatory aggressors, “alienating citizens from the state” and fostering reliance on nonstate alternatives (Blair 2019, 372). Recognizing the potential of abusive treatment of civilians to reshape internal conflict dynamics by affecting civilian cooperation with insurgents, I argue that insurgents adopt diverse strategies against peacekeepers to indirectly pressure peacekeepers into committing abuses. Based on these discussions, I develop a framework with a hypothesis for why peacekeepers commit civilian abuse within peacekeeping operations resembling counter-insurgency operations: 

Alternative Logic—Show of Force Resulting in Collateral Damage

An alternative argument posits that civilian abuse by peacekeepers may emerge from institutional pressures that require a delicate balance between force protection and the duty to safeguard civilians (Bell 2022). In this mechanism, the imperative to project strength and maintain operational control inadvertently leads to harm. The “show of force”—characterized by aggression, dominance, and performative displays of strength—becomes central to decision-making processes, driven by the need to assert authority and manage perceived threats (Enloe 2014). Unlike violence motivated by revenge, which is rooted in personal emotions such as anger or fear and manifests as deliberate retaliation, violence stemming from a “show of force” arises from operational pressures. These pressures, though not malicious in intent, often result in unintended harm to civilians. This type of harm is commonly categorized as “collateral damage”—unintended injury to noncombatant targets resulting from otherwise lawful actions (Crawford 2013, 38). In contrast, deliberate or indiscriminate attacks on civilians are characterized by intentional targeting of noncombatants (Cronin 2013, 176).

Contemporary peacekeeping operations, which prioritize civilian protection, increasingly assign peacekeepers to roles that require empathy and compassion (Carreiras 2010; Karim and Beardsley 2016). However, the deeply ingrained military identity—shaped by norms emphasizing strength, toughness, and aggression—poses challenges in adapting to these new, more human-centered responsibilities. This conflict between combat-oriented military norms and the contrasting expectations of peacekeeping missions, which demand a nuanced approach to conflict resolution, often results in internal friction. Peacekeepers must reconcile their military training, which prioritizes control and dominance, with the humanitarian goals of de-escalation and civilian protection (Duncanson 2009). Narratives that portray peacekeepers as “saviors of the war-torn citizenry” further reinforce these militarized ideals, embedding the notion of strength as a necessary component of peacekeeping (Higate and Henry 2017, 493). This internal struggle to balance competing demands exacerbates the tension between military norms and the expectations of peacekeeping operations.

The tension between military identity and peacekeeping expectations becomes particularly pronounced when “warrior identities” intersect with the pressures of maintaining peace. Military norms, which emphasize strength, resilience, and toughness, may lead peacekeepers to resort to a forceful yet ostensibly benign “show of force.” While such displays are intended to assert authority and protect the peacekeepers’ unit, they are frequently directed at managing perceived threats, often resulting in collateral harm to civilians. This dynamic is deeply ingrained in peacekeepers’ decision-making, rooted in peacekeepers’ military training. Thus, violence, though not intended, escalates as a consequence of this mechanism, which aims to project authority and maintain control over the operational environment.

Caught between their roles as “warriors versus humanitarians,” peacekeepers may begin to view local populations through a lens of suspicion and negative stereotypes. This view, coupled with an emphasis on military strength, frequently overshadows the humanitarian objectives of their mission. Consequently, while violence against civilians is rarely deliberate, it remains a foreseeable consequence of peacekeepers’ actions. This dynamic is reflected in autobiographical accounts of peacekeepers’ experiences in Bosnia during the 1990s, where aggression towards civilians often stemmed from the clash between combat-trained identities and the demands of peacebuilding (Duncanson 2009).

As the tension between warrior identities and peacekeeping roles persists, the inclination toward an aggressive “show of force” continues to escalate, with unintended harm to civilians. This pattern mirrors the civilian casualties caused by US forces in Iraq and Afghanistan, where efforts to demonstrate military dominance and maintain operational momentum led to significant violence against noncombatants (Crawford 2013; Duncanson 2009). In Chapter VII peacekeeping missions, peacekeepers may inadvertently cause harm when operating within a structural framework designed to project authority and deter threats. Despite these unintended consequences, peacekeeping operations are governed by the fundamental principle of “do no harm,” which strictly prohibits violence against civilians (United Nations 2023). This ethical principle serves to differentiate between accidental harm, such as collateral damage resulting from a “show of force,” and intentional violence or punitive actions against civilians. While peacekeepers may sometimes cause harm unintentionally, it remains distinct from deliberate violence, emphasizing the legal and ethical boundaries within which peacekeeping forces must operate (Williams 2013). Building on this understanding, the alternative hypothesis posits: 

Case Study––The African Union Peacekeeping Mission in Somalia (AMISOM)

The utilization of Regional-UN peacekeeping models, such as the AMISOM in complex security contexts like Somalia, signifies an evolving strategy to tackle entrenched conflicts (Schumann and Bara 2023). Such peacekeeping operations are closely intertwined with COIN strategies, involving peacekeepers facing deliberate targeting, necessitating the development of enhanced coercive capabilities and broader mandates compared to previous missions (Karlsrud 2019). In this landscape, AMISOM peacekeeping in Somalia stands out from UN peacekeeping missions for its distinct emphasis on COIN operations and its active involvement in combatting nonstate armed groups (Williams 2023). Despite facing significant internal and external obstacles, AMISOM has demonstrated notable achievements in stabilizing Somalia. These include thwarting Al-Shabaab's attempts to overthrow the Transitional Federal Government following the Ethiopian withdrawal in January 2009, successfully reclaiming major urban areas in South Central Somalia from Al-Shabaab control by 2014, and facilitating the electoral processes that led to the establishment of new Somali Federal Governments in 2012 and 2017 (Williams 2018).

AMISOM's efforts to enhance security have been, however, marred by widespread reports of human rights abuses by its troops, casting doubt on the credibility of UN-supported missions (Williams 2018). UN reports frequently cite instances of AMISOM troops conducting indiscriminate shelling of villages and committing acts of violence, including murder, assault, and rape against civilians, often seen as retaliatory responses to attacks targeting AMISOM forces (OHCHR 2017). For example, reports indicate that between March 2007 and December 2018, approximately 1,884 AMISOM peacekeepers were killed, with 383 of these casualties attributed to IED attacks (Williams 2019b). This suggests that Al-Shabaab's extensive use of IEDs against AMISOM has been a significant factor in triggering forceful reactions against civilians. While the explicit inclusion of “protection of civilians” within the AMISOM mandate was formalized after 2019 (UNSC Res 2472 2019, see para 15), both the UN Security Council and the AU Peace and Security Council have consistently urged AMISOM to adhere to international humanitarian and human rights law since 2013 (Williams 2013). Moreover, recent recommendations call for AMISOM to prioritize activities aimed at monitoring, mitigating civilian harm, and avoiding unnecessary harm resulting from operational activities (UNSC Res 2372 2017, see para 16).

With mission closures in Mali and growing uncertainties about the future of operations in the DRC, the UN's peacekeeping presence in Africa is diminishing. As funding mechanisms for AU-led operations materialize, this trend may accelerate the decline of traditional UN peacekeeping missions, potentially facilitating African-led initiatives with UN support (Gowan and Forti 2023). Collaborations between regional organizations and the UN have been shown to enhance operational effectiveness, surpassing unilateral interventions (Schumann and Bara 2023). Nonetheless, any departure by such partnerships from the UN's human rights and “protection of civilians doctrine” could jeopardize the UN's credibility (Williams 2018). Given the intricate interplay between AMISOM's accomplishments and obstacles, as well as its capacity to impact the UN's legitimacy, the AMISOM case, therefore, serves as a valuable testing ground for investigating the dynamics of peacekeepers’ involvement in civilian abuse.

Empirical Strategy

My research design incorporates both survey and observational data. The first dataset consists of individual surveys with active duty AMISOM peacekeepers, capturing attitudinal outcomes on human rights and civilian abuse. The second dataset utilizes the UN SAGE event dataset (Duursma 2017) and a separate UN dataset on IED attacks to construct an original human rights dataset. See Section B in the Online Appendix for a detailed discussion about the UN SAGE dataset.

Dependent Variable

I employ two different empirical strategies to encompass both attitudinal and behavioral measures to capture my dependent variable (DV)—violence against civilians by peacekeepers.

Attitudinal Measures

For my attitudinal outcomes, I built a “Human Rights” index derived from a set of six questions from an original survey with 600 serving AMISOM peacekeepers randomly selected from twenty-nine bases scattered across South Central Somalia (see figure 1 and Online Appendix Section A.1 for my sampling strategy). Survey work with active peacekeepers in conflict settings faces two main challenges. First, discussing civilian abuse, especially by peacekeepers, is sensitive, and peacekeepers may be reluctant to talk about it. As noted by Wood and Kathman (2014) and Balcells (2017), the lack of reliable data significantly complicates the measurement of civilian abuse. Second, due to the sensitivity of human rights abuse topics, peacekeepers may exhibit Hawthorne effects, altering their behavior because they know they are being observed, which can distort responses. To address these challenges, we sought permission to conduct surveys asking peacekeepers about their field experiences without explicitly focusing on civilian abuse. This approach did not involve deception, as the questionnaire was reviewed by higher authorities, and all six questions were derived from UN and AMISOM training materials.

Given the lack of established survey instruments for civilian abuse by state armed groups, I adapted Bell, Gift, and Monten’s (2022) approach with military cadets to assess peacekeepers’ normative attitudes towards conduct during battle and civilian protection. AMISOM troops receive regular training on the Law of Armed Conflict (LOAC) upon deployment.⁴ Using these training materials, I developed six questions addressing two main themes: Questions 1–3 focused on attitudes towards civilians during peace, while questions 4–6 addressed civilian casualties during military operations (see Table 1). These questions were integrated into a broader survey on attitudes towards peacekeeping to ensure candid responses. Responses were recorded on a four-point Likert scale and combined into a “Human Rights” index using inverse covariance weighting to enhance statistical power and address multiple comparison concerns (Anderson 2008).

Behavioral Measures

I separately construct an original dataset on civilian abuse using UN daily security situation reports, capturing all AMISOM-related security events in Somalia from June 2014 to April 2019. This approach addresses potential biases in self-reported survey outcomes from AMISOM peacekeepers, who might underreport sensitive issues like exposure to IED attacks due to fear of disclosure to authorities (Blair Coppock, and Moor 2020). The human rights event data, sourced from the UN's Department of Peace Operations’ Situational Awareness Geo-spatial Enterprise (SAGE) database (Duursma 2021, 10), enhances the reliability of information by providing real-time event acquisition. UN reports detail three key variables: whether clashes were initiated by AMISOM or were attacks against them, the use of small arms fire or IEDs, and any resulting AMISOM casualties. Crucially, these reports also indicate if AMISOM peacekeepers harmed civilians immediately following clashes (see Section B in the Online Appendix for coding details). I used this dataset to operationalize the behavioral measure for my DV, which is the actual number of civilian casualties caused by AMISOM troops from January 2014 to April 2019. Out of 490 clashes recorded over 58 months, 105 incidents involved AMISOM's use of lethal violence, resulting in 307 civilian injuries and 170 deaths. Notably, 55 of these 105 incidents occurred immediately after an IED attack. My main DV is the number of civilian casualties, both killed and injured by AMISOM troops, ranging from 0 to 34, with an average of 0.97 casualties per incident.

Independent Variable

Building on previous conflict literature, I use direct peacekeeper exposure to IED attacks as my IV (Condra et al. 2018; Sexton 2016). My analysis incorporates three sources of IED attacks against AMISOM peacekeepers. First, my survey analysis captures whether peacekeepers have personally witnessed their military convoy being hit by an IED, coded as “1” for yes and “0” for no. Second, I utilize geo-coded data on IED attacks against each individual's specific Forward Operating Base (FOB) from their arrival until the survey date. This data, sourced from United Nations records, includes the exact location, intensity, and consequences of each IED attack on AMISOM, with attacks ranging from 0 to 26 and averaging 3.92 IEDs per peacekeeper. Finally, in my event dataset, the IV is coded as “1” or “0” based on whether the encounter between AMISOM and Al-Shabaab involved a deliberate IED attack. During the study period, there were 490 clashes between AMISOM and Al-Shabaab, with 212 incidents (43%) involving deliberate IED attacks against AMISOM peacekeepers.

Identification Strategy

I utilize three sources of exogenous variation in this study. First, individual peacekeepers are deployed to various FOBs. Although assignment is not entirely random, units of about 150 peacekeepers are allocated to each FOB for a 1-year tour. Rotation occurs in smaller batches of 30–40 soldiers at unpredictable intervals for security reasons, rendering it exogenous to IED attacks. The causal argument regarding my primary survey outcomes hinges on the timing of a peacekeeper's arrival at an FOB similar to recent literature (Bol et al. 2021). Figure 2 illustrates how each randomly selected individual peacekeeper p_i’s exposure to the number of IED attacks is dependent only after the peacekeeper's arrival in each base FOB_i in time t_a to the time of the incident captured by my survey on the survey date t_a + s. Hence, although the IED attacks are expected, the probability of taking the survey before or after peacekeeper p_i’s direct exposure to an IED attack seems to be “as good as random.”

Second, the stochastic nature of IED attacks introduces exogenous variation (Condra et al. 2018; Sexton 2016). Both Al-Shabaab and peacekeepers cannot predict specific outcomes of IED attacks, such as which vehicle or peacekeeper will be affected, making the effects essentially random. Although AMISOM patrols are not always randomized, there is plausible exogenous variation in peacekeepers’ exposure to IED attacks. Encounters with Al-Shabaab are nonrandom, but the likelihood of these encounters resulting in small arms fire or an IED attack remains uncertain to the peacekeeper. In my survey sample, 57% reported their convoy had never been directly hit by an IED attack, 31% reported being hit 1–3 times, 5% 3–4 times, and 7% more than 5 times. In my event dataset of 490 incidents involving AMISOM, 43% included IED usage by Al-Shabaab. Figure 2, focusing on Sector 1, shows no discernible pattern in IED attacks against AMISOM convoys, confirming their unpredictability. Online Appendix Figure D11 in the annex further illustrates the random nature of IED attacks across all sectors, supporting my use of IED incidents as the IV. Despite sector assignments, cross-sectional variation exists in the likelihood of individual peacekeepers encountering IED attacks.

Finally, the variability in peacekeepers’ direct exposure to IED attacks serves as its own identification strategy. Not only is the selection of a specific vehicle from a convoy for impact stochastic but so is the precise point of impact on that vehicle. Online Appendix Figure D10 shows that weaker convoys do not experience a higher rate of IED attacks, contradicting expectations if these attacks were planned. Thus, whether a peacekeeper emerges unharmed, injured, or killed depends on chance, contributing to their potential experience of physical harm or witnessing harm inflicted upon colleagues (figure 3). Among 590 surveyed peacekeepers, 69% reported never witnessing a fellow peacekeeper injured or killed due to an IED attack, 22% reported between 1 and 3 injured or killed, and 9% reported between 4 and 6 peacekeepers injured or killed. Considering these observations, I employ ordinary least squares regression to estimate the following equations using both survey and observational data:

where the point estimate β₁ is the quantity of interest. In the first specification, the Civilian Abuse Index consolidates six questions into a single index, as outlined by Anderson (2008). In the second equation, it refers to real injuries and fatalities caused by individual AMISOM peacekeepers stationed at FOB b in district d during time t. Controls for the survey data include pretreatment variables such as age, sex, marital status, education, rank, previous AU/UN experience, previous operational experience, predeployment training, and current deployment duration to mitigate any deployment duration effects. µ denotes FOB location or district fixed effects. Standard errors are clustered at the individual FOB level. For the observational data, controls include rainfall in millimeters,⁵ and whether the incident occurred during Ramadan, along with location and year-fixed effects. Despite Hodler, Raschky, and Strittmatter’s (2018) findings of a negative impact of Ramadan on terrorist activities, AMISOM forces are placed on high alert during Ramadan due to increased military activities by Al-Shabaab. This caution is warranted, given that Al-Shabaab has historically launched major offensives against AMISOM during Ramadan (Anzalone 2012).

Results

Figure 4 illustrates the main findings regarding the attitudes and behaviors of peacekeepers perpetrating indiscriminate violence that harms civilians in response to IED attacks (Online Appendix Tables A5, A6, and B9). The top left plot represents coefficient plots with the IV) as self-reported IED attacks. The top right plot, the IV is the cumulative IED attacks around the soldier's FOB. For both measures, the DV is the “Human Rights” or civilian abuse index using the six questions in Table 1). The bottom plots are from left to right behavioral outcomes using the event dataset on civilian casualties attributed to AMISOM regressed against real-time IED attacks, and AMISOM offensive operations (refer Online Appendix Tables A5, A6, and B9).

The survey results reveal a statistically significant but negative trend in my civilian abuse index, represented in pooled outcome standard deviations (Cohen's d statistic). These results support my core argument that when peacekeepers self-report having personally witnessed an IED attack (Attitudinal 1) or encounter a higher number of IED incidents around their FOB (Attitudinal 2) after their arrival, they experience a reduction in their commitment to upholding human rights standards by 0.19 (95% CI [–0.37, –0.01]) and 0.03 standard deviations (95% CI [–0.05, –0.01]), respectively.

Furthermore, results from my event dataset in figure 4 (Behavior 1) are also in line with my survey findings, that an IED attack targeting peacekeepers corresponds to a 0.60 unit increase in civilian casualties, suggesting a 14% probability of civilian harm with each IED attack against AMISOM peacekeepers. For robustness, I also use the principal factor analysis to construct the human rights violation/civilian abuse index. This result aligns with the results using the Anderson (2008) civilian abuse index, demonstrating a consistent coefficient of -0.35 standard deviations with p = 0.006 (refer to Online Appendix Table A5, column (8)). Accounts from the ground validate my findings. A UN Human Rights report from December 2017 reveals that during the period from January 1, 2016, to October 15, 2017, AMISOM was associated with a total of 220 civilian casualties, averaging 10 per month. This notably high frequency of casualties is noteworthy within the context of a peacekeeping mission (OHCHR 2017).

My central findings, encompassing both attitudes and behavior show that there is a statistically significant association between intentional violence directed at peacekeepers and instances of indiscriminate violence that harms civilians. My theoretical framework offered two possible explanations: how emotional reactions tied to feelings of “revenge” (Balcells and Stanton 2021) might prompt involvement in abusive behaviors, particularly in situations where peacekeepers suffer casualties due to indiscriminate attacks. Alternatively, the indiscriminate violence causing harm to civilians can be attributed to entrenched military norms that prioritize projecting strength —in the form of “show of force” (Enloe 2014), leading to unintended harm to civilians.

To better understand whether revenge or a show of force influences these negative attitudes and behaviors toward upholding human rights, I analyzed each of the six survey outcomes separately. Individual survey outcomes reveal a significant pattern: both self-reported IED attacks and real-time geo-coded IED attacks are correlated with heightened acceptance of civilian casualties (0.27 and 0.03 units; p < 0.05; Online Appendix Tables A5 and A6, col. (5) in the Online Appendix). Furthermore, self-reported IED attacks also show a significant and negative relationship with the need for POC (0.27 unit; p < 0.05; col. (1)). These findings provide stronger support for H1, suggesting that peacekeepers’ use of lethal violence against civilians stems from a revenge mechanism, rather than H2, which posits that such violence is unintentional and related to a show of force.

Mechanism: Revenge or Show of Force

Although results from figure 4 provide suggestive evidence in support of H1 rather than H2, a critical question regarding internal validity concerns whether the adverse consequences of peacekeepers’ violence against civilians are primarily attributed to IED-triggered “revenge” or to broader “systemic collateral damage” (Cronin 2018). One plausible hypothesis is that IEDs have the potential to sow chaos and escalate conflict, blurring the line between combatants and civilians and thereby increasing the likelihood of collateral damage. To address this issue comprehensively, I undertake several additional analyses aimed at disproving alternative explanations. First, I regress real events of human rights abuse by AMISOM against another variable operationalized as “strategic objective.” Among 490 events in my event dataset, 79 (16%) involved AMISOM launching offensive operations against Al-Shabaab in areas outside AMISOM's full control. For instance, a reported offensive operation in the UN SAGE dataset reads as follows:

On 24 March, 2018, AMISOM along with the Somali National Army (SNA) began operations against Al-Shabaab in parts of Hiraan region. Six AS fighters were killed in Hees, 60 km from Beletweyne and two SNA soldiers were injured. The objective of the operations is to clear the road between Mahas and Beletweyne and will continue for a number of days to build on the recent developments where some communities resisted the payment of zakat to AS.

If AMISOM's use of lethal violence against civilians during offensive operations is driven by unintended collateral damage or strategic aims like shaping civilian behavior or establishing control (Lichtenheld 2020; Kalyvas 2006), we would anticipate a correlation between such violence and offensive operations in areas beyond its full control. However, contrary to this expectation, figure 4 (Behavior 2) reveals a decrease in violence against civilians by 0.98 units (95% CI) with each instance of AMISOM initiating offensive operations. This finding contrasts H2, suggesting that the observed aggressive behavior and civilian casualties are more likely driven by revenge than collateral damage.

While the results in figure 4 (Behavior 2) indicate a correlation between AMISOM Offensive Operations and decreased civilian casualties, it does not yet jointly distinguish between the proposed H1 and H2. In my second analysis, I operationalize “revenge” by interacting witnessed IED attack with fellow AMISOM personnel casualties (killed or injured) and “show of force” by interacting witnessed IED attack with engagements with al Shabaab (under fire or offensive operations). The outcomes presented in Table 2 reveal a distinctive pattern. Specifically, the coefficient of the civilian abuse index is statistically significant and negative (col. 2: β = −0.44; p < 0.05) only in cases where fellow AMISOM peacekeepers are killed in an IED attack, supporting H1, while this pattern is absent when the IED attack results in injuries among peacekeepers. These findings align with existing research on conflict dynamics, indicating that government forces tend to retaliate more frequently and swiftly following attacks that result in fatalities compared to those that do not (Kuperman 2001; Hultman 2012). Furthermore, when peacekeepers are either defensively or offensively involved in a firefight with Al-Shabaab alongside an IED attack, yet without resulting peacekeeper casualties, there is no corresponding inclination among peacekeepers to contemplate lethal harm to civilians which is against H2—that violence against civilians by peacekeepers is unintentional as a result of collateral damage.

Finally, to validate the mediating role of the “revenge channel” between my key variables of interest, exposure to IED attacks and civilian abuse, I conduct a formal mediation analysis to establish that the effect of the IV—direct exposure to IED attacks, is at least partially mediated by the mediating variable, fellow soldiers being killed, on the DV “civilian abuse.” Table 3 presents the results from the three separate mediation analyses associated with figure 4. In these analyses, I introduce the mediating variable—self-reported AMISOM casualties, or the number of fellow AMISOM peacekeepers killed by IED attacks since their arrival at the base—into the first-step regression. A key feature of mediation is the reduction in the direct effect of the IV (IED attacks) on the DV (civilian abuse) when the mediator is included in the model. This reduction is evident in all three analyses, with the beta coefficients for the treatment variable decreasing in the second step (from -0.30 to -0.09, from -0.30 to -0.20, and from 0.60 to 0.46). Furthermore, the significance of the coefficients in column (3) decreases in all three analyses, indicating that the effect of IED attacks on civilian abuse is partially explained by the mediator, AMISOM casualties caused by IEDs. The results suggest that a portion of the relationship between IED attacks and civilian abuse is mediated by AMISOM casualties caused by IEDs. Specifically, the mediating variable accounts for approximately 34%, 35%, and 23% of the total effect of IED attacks on civilian abuse in each of the respective analyses. These findings underscore the importance of AMISOM casualties in explaining the relationship between IED attacks and civilian abuse, suggesting that this mediator plays a critical role in understanding why peacekeepers may engage in abusive behavior.

Robustness Test: Survey Experiment

To ensure internal validity, particularly concerning peacekeepers’ understanding of LOAC principles potentially impacting attitude accuracy, I conducted a vignette survey experiment at a remote AMISOM base in Fafadun, involving approximately 100 Kenyan peacekeepers. This survey experiment aimed to address concerns that negative results may have stemmed from peacekeepers who lacked comprehension of LOAC teachings. The two scenarios presented are below in figure 5. The vignettes are identical, except in the treatment condition, soldiers sustained injuries, leading to one fatality, due to an IED attack. Notably, in both scenarios, the fleeing shepherds are unarmed, contrasting with the armed combatants typically involved in IED placement and firing. This underscores the lower perceived threat posed by unarmed civilians compared to armed combatants. Security briefs provided to peacekeepers before convoy departure emphasize the distinction between unarmed civilians and armed combatants, reinforcing the fact that 98% of encountered IEDs are either remote-controlled or triggered through vehicle-operated pressure plates, rather than by nearby individuals using command wires (Interview 2020). After reading the vignette, each peacekeeper responded to the same set of six questions related to Human Rights (see Table 1).

The survey experiment results, depicted in the left coefficient plots of figure 6 (Online Appendix Table C11), show that in the treatment condition, where peacekeepers face IED attacks and the potential loss of fellow peacekeepers, the civilian abuse index increases substantially by 0.8 standard deviations, exceeding the threshold for a “medium” effect size (Cohen 2013). A manipulation test conducted to assess the effectiveness of the experimental design revealed that the intensity of violence in terms of peacekeeper casualties, rather than personal safety concerns, significantly influences violence against civilians by peacekeepers (refer to Online Appendix  Table C14 for details). This finding strongly supports the “revenge” channel, contradicting the expectation that civilian abuse is a consequence of collateral damage. This finding aligns with the argument that collateral damage typically arises in densely populated areas or from artillery/airstrikes, not direct fire (Van Wie and Walden 2023). For example, during Iraq's civil conflict (2004–2008), 38% of civilian casualties resulted from US airstrikes (IBC 2024).

External Validity: Beyond AMISOM

I now attempt to broaden the scope of my investigation beyond AMISOM to ensure external validity by critically evaluating the relevance of these findings to a typical UN-led peacekeeping setting. Two primary issues merit attention in this context. First, there is a concern that peacekeepers predisposed to human rights abuses may exhibit lower baseline attitudes towards human rights (Rodriguez and Kinne 2019; Horne, Lloyd, and Pieper 2022), potentially confounding the attribution of civilian abuse solely to encountered violence intensity. Secondly, the study's scope is subject to consideration, as it is based on a unique case study involving AU peacekeepers engaged in peace enforcement rather than traditional peacekeeping operations. I conduct a comparative survey of Ethiopian peacekeepers deployed in two distinct contexts: those serving in Somalia with AMISOM, experiencing frequent casualties, and those stationed in South Sudan with the UN, where military operation casualties are rare. Although a randomized allocation of Ethiopian peacekeepers between AU and UN missions is unfeasible, I leverage the irregular assignment of Ethiopian peacekeepers to either mission, providing a unique opportunity for comparison.

Results in figure 6 (right plots) demonstrate the human rights index among Ethiopian peacekeepers in Somalia, nearly 0.7 standard deviations lower than those stationed in South Sudan (see Online Appendix Annex Table C13). Interestingly, among Ethiopian peacekeepers in Somalia, outcomes 4–6, which concern military operations, show negative and statistically significant results within the Ethiopian sample in Somalia at conventional levels (with p < 0.05). If the lower civilian abuse observed among AMISOM peacekeepers stemmed from impunity due to weak rule of law in their countries or the ability to evade punishment, we would expect no variation in the civilian abuse index between Ethiopians serving in Somalia and those in South Sudan. Also, human rights abuses by peacekeepers, whether UN or AU, can lead to severe consequences, including repatriation and legal prosecution (Sengupta and Cowell 2014; OHCHR 2017). These results further provide strong evidence supporting the revenge mechanism driven by the intensity of violence.

One challenge to inference that remains is from the possibility that the overall conflict environment in each sector is correlated to the motives behind armed groups targeting them. Recent sub-national studies have found that peacekeepers are deployed often in conflict zones (Ruggeri, Dorussen, and Gizelis 2018; Phayal and Prins 2020). Hence, any observed correlation could have been driven by one Troop Contributing Country (TCC) that is more likely to be deployed to highly violent areas where Al-Shabaab finds it easier to attack peacekeepers. To mitigate concerns that a specific region under the jurisdiction of a particular TCC could be disproportionately targeted by IED attacks, I conduct analyses that systematically exclude each TCC. As depicted in Online Appendix Figure A7, the findings remain consistent for both self-reported IED incidents (at a 90% CI) and for geo-coded IED events (at a 95% CI) across each analysis, which successively omits individual TCCs.

Taken together, these findings consistently support the revenge theory of civilian abuse by peacekeepers (Balcells 2017), rather than civilian abuse resulting from a show of force (Enloe 2014), leading to unintended collateral damage (Cronin 2018). This interpretation is corroborated by various independent tests, including a survey experiment and a comparative analysis across two distinct peacekeeping missions. Furthermore, external sources reinforce this conclusion. The 2017 UN Human Rights report on Somalia highlights that most “civilian casualties occurred during attacks on AMISOM convoys by IEDs, landmines, or grenades” (OHCHR 2017, 40). Additionally, the Kenyan Government's response in the same report states that “it is improbable that the ambushes are mounted by civilians…and the standard anti-ambush procedure is the quick suppression of the covering fire” (OHCHR 2017, 40), emphasizing that peacekeepers’ actions are influenced by situational factors during indiscriminate attacks.

Discussion

What are the broader implications of our findings for understanding the interplay between peacekeepers and civilian abuse, encompassing both AU-led peacekeeping missions and wider contexts? First, I show that insurgent violence against peacekeepers shapes the retaliatory behavior of the latter, a phenomenon consistent with the theoretical framework of relative capability between government and insurgents (Wood 2010). Furthermore, this observation underscores the critical role of casualty figures in shaping retaliatory behavior, particularly highlighting how fatalities can intensify the cycle of violence in conflicts. For example, Israeli retaliations were more frequent and swifter following fatal Arab attacks (Kuperman 2001; Haushofer, Biletzki, and Kanwisher 2010), while significant losses among pro-government forces in Afghanistan correlated with increased civilian deaths (Hultman 2012). Similarly, the Haditha massacre in Iraq shows how an IED attack that killed US Marines led to lethal retaliation against unarmed civilians, illustrating the link between fatalities and revenge (Knickmeyer 2006). As Petersen (2011, 36) posits, “anger is a tool for creating spiraling cycles of violence that can transform an entire conflict,” with high-intensity violence often prompting peacekeepers to prioritize their own safety over that of foreign civilians, leading to heightened moral boundaries and fear-induced group dynamics. Secondly, I provide suggestive evidence that insurgents may deliberately provoke peacekeepers into committing acts of civilian abuse to achieve broader strategic objectives. This strategy aims to alienate civilians from the peacekeeping forces, undermining their legitimacy and support among local populations (Asal et al. 2019). Notably, Al-Shabaab tactics, especially the use of IEDs, aim to amplify civilian casualties caused by peacekeepers, thereby eroding public support for their presence (Williams 2013). Evidence from a UN panel survey conducted among Somalis demonstrates a significant decline in support for AMISOM peacekeepers, from 64% in 2014 (n = 1975) to 28% in 2016 (n = 2306), coinciding with increased civilian casualties (UNSOM 2016, 27). Moreover, insights from a high-level Al-Shabaab commander highlight recruitment success in areas affected by peacekeeper-inflicted civilian casualties (O'Neil and van Broeckhoven 2018), underscoring the strategic implications of such tactics on mission effectiveness and local perceptions.

This research underscores two critical policy implications concerning the fundamental peacekeeping mandate of civilian protection. Firstly, in light of the parallels between contemporary peacekeeping and COIN operations, the propensity of peacekeepers to resort to lethal violence against civilians risks undermining civilian support. Evidence from Afghanistan, Iraq, Peru, and Ukraine indicates that civilian populations, targeted by belligerents, often withdraw support (Pechenkina, Bausch, and Skinner 2019), collaborate with opposition groups (Shaver and Shapiro 2021), or vote for opposition parties in elections (Birnir and Gohdes 2018). Secondly, as fatalities in conflict situations often provoke heightened retaliatory responses, integrating mechanized capabilities and comprehensive training to counter indiscriminate attacks within peacekeeping operations could significantly enhance civilian protection. Recent empirical studies suggest that armored protection improves soldiers’ decision-making regarding lethal force, potentially resulting in reduced civilian casualties, as observed in mechanized units in Iraq (Van Wie and Walden 2023). Scholars emphasize that effective adherence to the law of armed conflict entails both understanding legal principles and analyzing the underlying norms guiding combatant behavior (Stephens 2019).

Conclusion

This article addresses the question of why peacekeepers engage in violence against civilians. I examine two micro-level explanations for such violence: revenge, wherein peacekeepers emotionally respond to attacks on their comrades, and situations where peacekeepers engage in a show of force, with violence being an unintended consequence of collateral damage. Due to the absence of scientifically gathered data on the topic, I constructed an original dataset documenting real-time civilian abuse by AMISOM peacekeepers and separately surveyed 600 AMISOM peacekeepers. My findings underscore the substantial impact of emotional and psychological elements, notably the impulse for revenge following past violent events, which can drive peacekeepers to commit violence against civilians (Petersen 2002; Balcells 2017). This dynamic equips Al-Shabaab with a compelling tool to mobilize the civilian population against both AMISOM and government forces. My results echo recent findings in the peacekeeping literature that rebel forces employ a strategic approach by directing their attacks toward peacekeeping forces in locations deemed to offer the greatest potential for gaining control over the populace vis-à-vis the government (Hunnicutt, Nomikos, and Williams 2021).

Despite the consistency of these findings across multiple tests, they also raise a pivotal question for future research. While my findings establish a connection between peacekeepers and civilian abuse, they do not fully evaluate the extent to which these intentional IED attacks affect peacekeeping activities. An avenue for further investigation could involve delving into the strategic rationale behind insurgents’ choice to subject peacekeepers to high-intensity violence. For instance, such research could explore whether peacekeepers alter their operational behaviors, reduce patrols, or curtail interactions with civilians in response to these attacks. Additionally, my findings challenge conventional notions of moral responsibility within contemporary peacekeeping missions by indicating deliberate, rather than unintended, violence against civilians. The invocation of the “collateral damage” exemption to rationalize civilian casualties incurred during offensive operations raises critical questions about the integrity of IHL and prompts a necessary reassessment of the human costs associated with modern peacekeeping endeavors.

Footnotes

Author Biography

Prabin B. Khadka is a Lecturer at the Department of Government, University of Essex. His research examines the intersection of peacekeeping, counterinsurgency, and efforts to counter violent extremism within the broader economic and political landscapes of conflict-affected regions. His work uses field experiments to assess the mechanisms through which these interventions shape local support, social cohesion, and security dynamics. Prabin was an Explosive Ordnance Disposal Officer with the Nepal Army and served as a Blue Helmet in the Democratic Republic of the Congo in 2003 and 2008. He has also held civilian roles in multiple UN political and peacekeeping missions, including in Chad, Lebanon, Côte d'Ivoire, Gaza, Iraq, and Somalia.

Notes

Author’s Note: This research is funded by grants from New York University (NYU) and the George Downs Data Collection Grant (NYU). This study is approved under the Institutional Review Board (IRB) protocol at New York University, IRB-FY2018-1887.The author would like to thank the editors and the reviewers, M. Gilligan, C. Samii, S. Satyanath, N. Haas, S. Karim, A. Wright, H. Dorussen, F. Ghosn, M. Arslan, B. Welsh, and participants of the 2018 MPSA for their comments on earlier versions of this manuscript. The author is grateful to S. McMenamin from the UN Somalia Office for the coordination with AMISOM Peacekeepers and the AMISOM and UN Peacekeepers who participated in this survey.

References