Conflict-Related Sexual Violence

This article reviews the remarkable growth in empirical literature in political science on wartime sexual violence against civilians, including rape, sexual slavery, forced marriage, and other forms. Early work, motivated by ongoing conflicts in the former Yugoslavia and Rwanda, tended to portray these forms of violence as inevitable, ubiquitous, and either opportunistic or strategic. However, recent literature and new data sources have documented substantial variation in sexual violence across countries, conflicts, perpetrators, and victims and survivors. Building on this observed variation, scholars have developed and tested a wealth of theories about when, where, why, and under what conditions sexual violence occurs as well as its consequences. We highlight the core findings from the literature, explain the key debates among experts, and explore several avenues for future research. We conclude by detailing what the study of wartime sexual violence—both the findings and the research process—offers to a broader set of political science scholars.


INTRODUCTION
While wartime rape has long been a subject of scholarship for feminist scholars, legal thinkers, and journalists (e.g., Brownmiller 1975, MacKinnon 1994, it is only within the last 15 years that political scientists have begun to engage with conflict-related sexual violence as a topic of focused study. Over this time, there has been a marked increase in the volume of studies. Research on conflict-related sexual violence has appeared in top disciplinary and subfield journals in political science, featuring nuanced theoretical and empirical approaches to analyzing its causes, dynamics, and consequences. The increasing attention to research on conflict-related sexual violence echoes the call for a more comprehensive perspective on political violence (Davenport et al. 2019). It also forms part of a broader trend in conflict studies toward expanding our analyses to include more forms of violence, as well as a wider range of armed actors, and toward understanding the complexities of wartime dynamics for individuals, societies, and states in the short and the long term.
The introduction of sexual violence as a phenomenon worthy of serious theoretical and empirical study in political science has contributed to overcoming the problematic historical neglect of gender issues in conflict studies (McDermott 2020). Sexual violence is a key component of the burgeoning literature on civilian victimization in conflict (Balcells & Stanton 2021) and of both positivist and interpretivist/post-positivist studies of gender and international relations (e.g., Reiter 2015, Sjoberg et al. 2018.
In this review, our main focus is systematic empirical research on conflict-related sexual violence that relies on quantitative and qualitative data and methods. We begin by discussing how scholars define and conceptualize conflict-related sexual violence and trace the development of the field of study in political science. We then describe major findings that emerge from the literature before turning to a set of open questions regarding methodology and substantive themes, highlighting where more research is needed. We posit that this literature has important implications for a range of related topics on conflict and political violence, such as the internal workings of armed organizations, peacekeeping and transitional justice, and postconflict social cohesion. In short, the study of conflict-related sexual violence contributes valuable insights into core questions of war and peace, the dynamics of warfare, and the research process, including issues of ethics and data.

WHAT IS CONFLICT-RELATED SEXUAL VIOLENCE? DEFINITIONS AND COMPLEXITIES
There is no consensus definition of conflict-related sexual violence in the literature. Existing definitions tend to vary along two dimensions: which forms of violations should be included and what violence should be considered conflict related. Most definitions include rape and consider sexual violence to include a wide range of violations. Wood (2006, p. 308), for example, defines sexual violence as "a broader category that includes rape, coerced undressing, and nonpenetrating sexual assault." Some scholars also include nonviolent forms, such as humiliation and "improper sexual comments" (e.g., Hynes et al. 2004, p. 301).
Recent quantitative cross-national studies often draw on the Sexual Violence in Armed Conflict (SVAC) data set (Cohen & Nordås 2014) and adopt its definition. Drawing on the definition given by the International Criminal Court (ICC), the SVAC definition includes direct force or physical violence and/or the threat of force or coercion. This definition covers seven distinct forms of violence: (a) rape, (b) sexual slavery, (c) forced prostitution, (d) forced pregnancy, (e) forced sterilization/abortion, ( f ) sexual mutilation, and (g) sexual torture (Cohen & Nordås 2014). Each of these forms, in turn, has its own definition; for example, following Wood (2006, p. 308), rape is defined in the SVAC data set as "the coerced (under physical force or threat of physical force against the victim or a third person) penetration of the anus or vagina by the penis or another

TRACING THE EXPLOSIVE GROWTH OF A FIELD
Once an obscure and marginalized topic of study, conflict-related sexual violence is increasingly studied by a wide range of scholars using multiple epistemologies, methodologies, and data types and sources. The development of the research field from the early to the contemporary phase is characterized by significant shifts in how scholars conceptualize the phenomenon and what methods they use to study it. In terms of the share of peer-reviewed publications on the topic, the disciplinary loci of the research have shifted from being almost exclusively in the fields of women's and gender studies and feminist legal scholarship to the majority of the current research studies being in political science. According to the Web of Science, the first article published in the field of political science or international relations on the topics of "sexual violence" and "war" was in 2001 (Skjelsbaek 2001). From fewer than five peer-reviewed publications per year from 2001 to 2006, the annual number of publications on these topics in political science and international relations has increased tenfold to 58 in 2018. In 2019, publications on wartime rape and/or sexual violence generated 755 citations, a more than 40% increase from the year before (536 citations). Broadening the scope to include all disciplines, there were over a hundred publications on the topic and almost 2,300 citations in 2019 alone. 1 We use the terms victim and perpetrator throughout this article because these are standard within the criminal justice system and in much of the literature on violence. We acknowledge that some people who have been assaulted prefer the term survivor as a signal of their agency and strength, and also that not all victims of sexual violence in fact survive. We therefore refer to "victims and survivors" throughout.
The scholarly study of wartime sexual violence is rooted in global politics, and it is inextricably linked to the processes by which sexual violence became a war crime. In this sense, the study of sexual violence should be situated in the moments that made it historically and politically salient to a global audience. The history of the field is intimately tied to politics and activism, particularly to feminist movements. Arguably, the motivating spark for the field was the Bosnian conflict, which served-in the eyes of Western academics and policy communities-to put a familiar (and therefore shocking) European face on the problem of wartime rape (Crawford 2017).
Scholarly attention to sexual violence can also be attributed to improvements in sex and gender diversity in the academy. The rising tide of women in the discipline has increased the visibility and impact of scholarly work by women, who have introduced different research topics and asked novel questions. At the same time, however, the scholarship on conflict-related sexual violence has been dominated by researchers from the United States, Europe, and Australia; future work might address how this has shaped the study of sexual violence and which voices have been marginalized and overlooked in the development of the research agenda.
The timeline of research on conflict-related sexual violence is divided into two distinct periods of growth. The first occurred after the outbreak of the war in the former Yugoslavia in 1992-1995 and the 1994 genocide in Rwanda, both of which were marked by widespread and horrific episodes of mass rape. Deeply affected by the ongoing conflicts and their harrowing aftermaths, scholars at the time, led by pioneering feminist legal scholars, offered strong normative arguments about why sexual violence should be taken seriously, both in scholarship and in justice processes. This early work often included explicit accounts of violence to demonstrate the realities of the problem and to connect it to ethnic conflict (MacKinnon 1994, Seifert 1996. Inspired by feminist activists and advocates at the time, as well as by the conflicts of the day, scholars argued that rape was a "weapon of war" (Card 1996, Farwell 2004) and advocated for understanding rape (even absent lethal violence) as a crime of genocide (Sharlach 2000).
A common thread throughout the literature of this early period is that conflict-related sexual violence was considered a ubiquitous feature of war perpetrated by men against women. Scholars sought to make sense of the brutality and extent of sexual violence by asking why women were being targeted with sexual violence. This early work laid important conceptual foundations for different ways to understand sexual violence: as a part of gender analysis (Brownmiller 1975, Enloe 2000, as a central element of the politics of identity (Weitsman 2008), and as a form of violence replete with symbolic social and political meanings (Milillo 2006). Different strands of gender analysis emphasized why women in general were targeted (to assert militarized masculinities) and why specific groups of women were selected (to attack the perceived bearers of a group's ethnic, religious, or political identity or to feminize and humiliate an ethnic, religious, or political group) (Skjelsbaek 2001, p. 215). Some scholars viewed sexual violence as a natural and unavoidable side effect of war, due to male sexual urges (e.g., Farwell 2004), while later contributions to conceptualizing wartime sexual violence also considered rape as a violation of "family honor" rather than an individual harm (Mackenzie 2010). In this early phase of scholarship on sexual violence in war, the empirical evidence primarily functioned to illustrate rather than test theories about dynamics, causes, and consequences.
Finally, many scholars in this first period argued that wartime rape exists on a continuum of violations against women carried over from peacetime. The continuum argument originated with feminist scholars, who maintained that wartime sexual violence is closely connected to the political, social, and economic status of women in the prewar period. In this view, the types of violations women faced before the war were driven by patriarchy and are causally related to what women experience during the war (Card 1996, Enloe 2000. As we discuss below, the continuum argument still resonates in contemporary academic debates and policy circles. A second phase of the research on conflict-related sexual violence was spurred by the critical recognition that rather than being ubiquitous, wartime sexual violence exhibits significant variation (Wood 2006) across conflict and actors, forms of violence, identities of the targeted groups (including male victims and survivors), and locations. Wood (2009) argued explicitly that some armed organizations do not perpetrate sexual violence. This observation had profound implications for the trajectory of the scholarship, challenging two fundamental (sometimes implicit) ideas in the earlier literature: (a) that sexual violence is a constant and inevitable aspect of warfare, and (b) that all armed groups/soldiers engage in sexual violence if given the opportunity. Even within the same country or conflict setting, some armed groups commit sexual violence while others do not (Wood 2006). Earlier explanations of sexual violence, such as those focused on ethnic tensions and near-constant factors such as patriarchy, fail to explain these differences. Wood's (2006Wood's ( , 2009 seminal work on variation and restraint forged a path for researchers working on both statistical analyses and micro-level, small-n empirical work. Wood's scholarship also placed the focus squarely on armed organizations, rather than cultures, states, or even conflicts, as the key unit of analysis to understand variation in wartime sexual violence. The second wave of scholarship was marked by a wealth of new theories explaining why sexual violence varies. The most prominent among these developed a third category of motivation for sexual violence as a practice, located somewhere between such violence as a strategy and as an opportunistic act (discussed in more detail in the next section). Following Wood, scholars also began to document the variation, and the field turned more comparative and global in nature. For example, Leiby's (2009b) comparative study of sexual violence in Guatemala and Peru used statistical data coded from a sample of testimonies given to each country's truth commission. Others used the comparative case method to look for commonalities across incidents in terms of prevalence, perpetrators, targets, and locations (e.g., Farr 2009).
Scholars also began to examine "negative" cases, in which sexual violence was not reported or was known to have been limited. The move away from only studying well-documented cases of mass rape (in effect, selecting on the dependent variable) and toward studying the universe of cases allowed for better and more rigorous testing of generalizable explanations for sexual violence.
Two large-n global data sets enabled the rapid growth of the statistical studies of wartime rape and sexual violence (McDermott 2020). The first data set covers rape during civil wars by state armed forces and rebels (Cohen 2013a(Cohen , 2016. The second, the SVAC data set, was developed as a public good and includes numerous forms of sexual violence across all types of armed conflicts and conflict actors (states, rebel groups, progovernment militias) (Cohen & Nordås 2014). Both include estimates of the prevalence of sexual violence (i.e., level of severity) across conflict actors over time. The SVAC data set, in part because it is fully compatible with the widely used Uppsala Conflict Data Program (UCDP) data sets, has become the predominant source of quantitative data on conflict-related sexual violence, and it has made sexual violence a characteristic of conflict that scholars are readily able to take into account in statistical analyses of war and its consequences.

CORE THEORIES AND FINDINGS
The result of the rapid growth in literature is a core set of established findings about conflictrelated sexual violence. We highlight four important findings with wide-ranging implications for theory, empirics, and policy. These include findings about (a) variation in perpetrators, victims and survivors, conflicts, and countries, and the implications of this variation for the earlier continuum argument; (b) motivations for sexual violence, including arguments about opportunity, strategy, and sexual violence as a practice; (c) rebel groups and their internal organizational processes regarding sexual violence; and (d) the long-term consequences of sexual violence for individuals, communities, and states.

Variation (and Its Implications for the Continuum Argument)
A misconception from the first period of research is that sexual violence is ubiquitous in war. Extensive research has now clearly documented that this is not the case. Both theoretical and empirical work now focus on exploring this variation and studying perpetrators to understand the motivations for conflict-related sexual violence. The study of armed groups illuminates patterns of perpetration and restraint-and offers greater insight into the causes of sexual violence than analyses that extrapolate perpetrators' motivations from the experiences of survivors and witnesses (Eriksson Baaz & Stern 2009, Haer et al. 2015, Cohen 2016. We now know that there is significant variation in sexual violence across conflicts, between different conflict actors, and over time (e.g., Cohen 2013aCohen , 2016, even within the same conflict (Wood 2009, Cohen 2016). In addition, whereas public discourse has often emphasized that sexual violence is perpetrated by unruly rebel groups, sexual violence is far more often reported to be perpetrated by states (Cohen & Nordås 2014). Nonetheless, a substantial proportion of armed groups are not reported as perpetrators. SVAC data show that the highest share of rebel groups reported as perpetrators in the period 1989-2015 was 40% (in 2002), whereas the highest observed share for states was about 60% (in 2002 and 2012). Although the fact of variation has been largely accepted by scholars, it has been less accepted by some policy and practitioner communities. For example, it is a guiding principle of the International Committee of the Red Cross to use what it calls a "reversed burden of proof," whereby staff is required to operate under the assumption that unless proven otherwise, rape is endemic in any given context (ICRC 2020). Although this assumption is a well-intentioned attempt to address underreporting, an unfortunate consequence may be that scarce resources are stretched far too thin, limiting access by those most affected by sexual violence.
Although there are no reliable cross-national data on victim counts, there is clearly important variation in who the victims and survivors of sexual violence are. Evidence shows that sexual violence is perpetrated not only by men against women. Men are victims and survivors of many types of sexual violence. The exact prevalence remains unknown; however, male victimization seems to be far more frequent than previously assumed, and silence about male victimization is now less common (Carpenter 2006, Sivakumaran 2007, Edström & Dolan 2019, Traunmüller et al. 2019). In addition, there is growing evidence regarding female perpetrators in particular cases ( Johnson et al. 2008, Cohen 2013b, Sjoberg 2016 as well as recognition of the targeting of gender and sexual minorities, although data remain sparse (Tschantret 2018).
Another problematic, often implicit, assumption of earlier studies was the notion that sexual violence is closely correlated with other forms of wartime violence, such as killing and looting. If sexual violence can be modeled as a function of other wartime violence, this diminishes the need for studies documenting and theorizing sexual violence as a meaningful and distinct category. However, although sexual violence may sometimes be correlated with other forms of violence, micro-level evidence from case studies has shown that the correlation is not universal; for example, the years with the highest levels of reported rape are not necessarily the years with the highest levels of reported killing (Cohen 2016). There is therefore increasing awareness that "sexual violence is distinct from other aspects of civilian victimization in civil wars" (Benson & Gizelis 2020, p. 167). There is also significant variation in terms of the forms of sexual violence, to which we return in the section discussing open questions and future directions.
The fact of variation has important implications for the continuum argument-i.e., the idea that the prewar status of women in society is predictive of wartime sexual violence. Rooted in older debates, the continuum argument is still very much a part of the literature on conflict-related sexual violence. Whether it is useful to view wartime sexual violence as a continuation of peacetime violations is one of the most contested topics among scholars of sexual violence. For example, in explaining the sexual violence in the Guatemalan conflict, Boesten (2017, p. 507) argues that "much of the scale and cruelty of these experiences was certainly exceptional and strongly conflict related, [but] the script for these acts-immersed in racism and sexism. . .-predated the conflict." Some scholars are also critical of what they see as a tendency to treat sexual violence as an "exceptional-if not aberrant-phenomenon in war" (Meger 2016, p. 149), opposing the proposition that wartime sexual violence can or should be studied separately from more common peacetime violations, such as intimate partner violence. It is less contested, but perhaps also less central to the continuum argument, whether and how wartime patterns of sexual violence shape postwar patterns of gendered violence; these connections are explored in recent work in Peru (Østby et al. 2019) and the Democratic Republic of Congo (DRC) (Lindsey 2019).
Others argue that the continuum argument fails to offer much analytical leverage. While sexual violations occur during peacetime, war, and the postconflict period, the forms, severity, and perpetrators of sexual violence vary dramatically across these periods. Some forms are reportedly common across all periods; for example, intimate partner rape is the most common form of sexual violence, even in areas with the highest levels of wartime rape (Peterman et al. 2011). However, other forms-such as forced pregnancies in so-called rape camps (Bosnia), forced marriages to combatants (Liberia), public acts of gang rape (Sierra Leone), and sexual mutilation of prisoners (El Salvador)-are generally rare outside the context of war and are both more brutal and more organized than peacetime violence (Wood 2014, Cohen 2016. Nonetheless, the choice of rape and other forms of sexual violence is most certainly gendered and is shaped by structural forms of sex and gender inequality, including patriarchy. Scholars have argued persuasively that rape can have a role in "feminizing" an enemy (e.g., Sivakumaran 2007). Patriarchy in its myriad manifestations is necessary for wartime rape and sexual violence; however, patriarchy is not sufficient, and it fails to explain variation in where, when, how, by whom, and against whom wartime sexual violence occurs (Cohen et al. 2013). In other words, structural gender inequality is likely far too common across space and time to explain the incidence of conflictrelated sexual violence. It may be the case that variation in structural gender inequality is not substantial enough to explain differences in conflict-related sexual violence, or the best current measures are not fine-grained enough to capture it. Either way, there is little evidence for structural gender inequality as a major "causal explanation" for wartime sexual violence (Davies & True 2015, p. 495).

Understanding Motivations: Opportunism, Strategic Violence, and Violence as a Practice
The early literature often understood sexual violence as opportunistic, driven primarily by private motives and individual urges (Gottschall 2004, Wood 2009). This view was countered by arguments that wartime sexual violence is a weapon or a tool of war, perpetrated by commanders pursuing military goals (e.g., Skjelsbaek 2001). Describing sexual violence as a weapon or tool was a successful framing choice by advocates and activists, which enabled sexual violence to become accepted as a critical security issue and as a policy priority for the highest levels of politics (Crawford 2017). Scholars and activists also commonly highlighted that rape was a cheap weapon, highly effective and readily available. The strategic benefits of sexual violence, according to these arguments, include sowing fear in the civilian population to induce collaboration or compliance, humiliating and breaking the morale of the enemy, and expelling populations from contested territory that an armed group seeks to control. Others argued that commanders adopt sexual violence as a tool of war by using rape to coerce women to become suicide bombers in order to mitigate the shame of being a rape victim (Bloom 2011). Sexual violence may also be strategic when it functions as an institutional reward or compensation for combatants, part of the so-called spoils of war (Wood 2014). Among others, Eriksson Baaz & Stern (2013) are critical of the weapon-of-war discourse, cautioning against conflating the consequences of widespread sexual violence with the intentions of the commanders.
A theoretical lens that scholars have used to structure discussions has been to consider sexual violence a result of principal-agent dynamics, and several studies emphasize the role of oversight, discipline, and ongoing education as a prerequisite for restraint (Butler et al. 2007, Hoover Green 2018, Whitaker et al. 2019). The underlying assumption in these studies reflects an understanding of sexual violence as opportunistic; the argument suggests that agents have a preference for committing sexual violence and will show restraint only if they are sufficiently controlled by the principal or materially incentivized to refrain from rape. Principal-agent dynamics also inform discussions of organizational characteristics, which we address in the next section.
Between these extremes of opportunity and strategy, Wood (2018) develops a more nuanced understanding of sexual violence as a practice. Sexual violence as a practice is not directly ordered or authorized from the top as part of a clearly defined military strategy; rather, leaders tolerate it or simply fail to punish it. Sexual violence as a practice suggests that when commanders are permissive, sexual violence can occur even at very high frequencies due to the horizontal socialization between peers and combatants' private preferences for rape (Wood 2018, p. 522). An important implication is that a high frequency of atrocities is not necessarily indicative of a strategic logic: Sexual violence can occur, even on a massive scale, without direct orders.
The different modalities of sexual violence as opportunistic, strategic, and a practice can all be possible, even within the same conflict setting, and are not mutually exclusive. For example, sexual violence as a practice may also occur because of opportunism, and commanders' tolerance for sexual violence can be strategically beneficial for the organization even if it is not ordered or orchestrated.
In line with the notion that sexual violence emerges as a practice, Cohen (2013aCohen ( ,b, 2016 explores why, where, and by whom rape-and particularly gang rape, the most commonly reported form of wartime rape-occurs during war. Cohen focuses on the dynamics that make gang rape useful for building intragroup cohesion among fighters within armed groups; the central finding is that armed groups that recruit fighters by force are more likely to commit rape. Cohen argues that rape is neither opportunistic nor strategic; the argument does not assume that perpetrators independently desire to commit rape for private reasons, nor that rape is the result of direct orders by commanders. Combining large-n statistical work on global patterns and fieldwork in three post-conflict countries, Cohen (2013Cohen ( , 2016 finds that combatant socialization helps explain gang rape by both state forces and rebel groups; building on this work, Cohen & Nordås (2015) find the argument also extends to progovernment militias.
Besides forced recruitment, scholars have highlighted several other conditions under which an armed organization will embrace some form of sexual violence as a strategy or commanders will tolerate rape as a practice. These include the types of resources on which the armed group relies (Whitaker et al. 2019) and the group's ideology and training (points we address below). Although other factors are part of the strategic logic of certain cases-for example, the role of ethnicity as a driver of rape in the Bosnian conflict-few of these drivers have been shown to be broadly generalizable when tested cross-nationally (e.g., Cohen 2016).
Studies now challenge the widespread perception that sexual violence is a cheap weapon of war, a cornerstone of the opportunism and strategy arguments. If sexual violence (rape, in particular) were as costless, easy, and effective as the discourse suggests, many more armed groups would be expected to use it; the notion of costless violations overpredicts sexual violence relative to the empirical evidence. Instead, sexual violence often comes with significant costs to the armed group and individual perpetrators (Wood 2009), including the risk of turning the civilian population against the group and undermining its political goals as well as the emotional toll on perpetrators, the time it takes to perpetrate, and the risks of disease (particularly sexually transmitted infections) that can harm a group's ability to fight (Cohen 2016).

Organizational Factors: Cohesion, Ideology, Governance, and Training
Following the emphasis on armed groups as the key unit of analysis, important studies of the inner workings of nonstate groups have illuminated the role of sexual violence for rebel organizations in recent wars (e.g., Baines 2014, Cohen 2016, Muvumba Sellström 2019, Whitaker et al. 2019, Sawyer et al. 2021. Drawing on interviews with current and former fighters, scholars have investigated how sexual violence is perceived by fighters inside armed organizations, filling a gap in the literature (e.g., Eriksson Baaz & Stern 2009, Cohen 2016. For example, Cohen's (2013aCohen's ( ,b, 2016) assessment of the Sierra Leone conflict focuses on the role of sexual violence for group cohesion, demonstrating how forced recruitment practices (and the resulting poor intragroup cohesion) help explain why the Revolutionary United Front (RUF) was responsible for far more rape than other groups. The RUF inculcated that rape was a serious offense, but due to selective enforcement, such rules were lost in practice (Marks 2013). Focus groups of senior commanders' wives in the Lord's Resistance Army in Uganda reveal how the group institutionalized forced marriage within the organization, while severely punishing sexual violence outside regulated sexual relations; these joint efforts served as a political project to "build a 'new Acholi' nation" and to enforce hierarchies within the group (Baines 2014, p. 405).
Investigating the causes of restraint by armed organizations follows from the emphasis on variation in the sexual violence literature. Rather than only trying to explain why sexual violence is perpetrated, it is equally important from an academic and policy perspective to understand why some actors rarely commit acts of sexual violence. Scholars have pointed to ideology as a critical factor in the explanation of restraint. For some organizations, such as leftist armed groups, the use of sexual violence conflicts with ideals for which the organization is claiming to fight, such as a gender-equal social order. Some insurgent organizations may recruit members already committed to the organization's ideology. Alternatively, organizations known not to commit sexual violence may be better able to recruit women-for example, as a means for women to protect themselves against sexual violence by state armed forces, as in El Salvador. In these cases, sexual violence would be detrimental to the organization's internal and external legitimacy and would erode efforts to build trust with the civilian population. Organizations known to suppress effectively the use of sexual violence (despite the use of other forms of violence) by their troops include the Farabundo Martí National Liberation Front (FMLN) in El Salvador, the Liberation Tigers of Tamil Eelam (LTTE) in Sri Lanka, and Sendero Luminoso in Peru (Wood 2006(Wood , 2018Leiby 2009b;Cohen 2016;Hoover Green 2018).
In contrast to restraint, the role of ideology as a motivation for sexual violence has also been highlighted, for example, with regard to the use of systematic sexual violence by extremist Islamist armed groups. For the Islamic State, ideology led to organizational policies that ordered or authorized sexual violence against specific social groups, and it served to create institutions that regulated the conditions under which violence occurred (Revkin & Wood 2021).
This scholarship also has links to the rebel governance literature, which associates rebel leaders' authority, recruitment, and resource endowments with their wartime treatment of civilians (e.g., Weinstein 2007, Mampilly 2012. For sexual violence, the types of resource endowments are consequential: In a cross-national statistical analysis, Whitaker et al. (2019) show that rebel groups www.annualreviews.org • Conflict-Related Sexual Violence that rely on civilian assistance in smuggling networks are incentivized to enforce restraint in the use of sexual violence. Leadership and principal-agent dynamics are often highlighted as important for understanding variation in sexual violence by different organizations. Leaders' degree of investment in political education for rank-and-file fighters can help explain groups' constraint on sexual violence (Hoover Green 2018). Conversely, short time horizons and pressures on a group's survival could increase sexual violence by heightening commanders' tolerance for this violence despite presumed long-term costs ( Johansson & Sarwari 2019). Finally, rebel organizations that elect their leaders are less likely to perpetrate sexual violence, possibly because leaders in these organizations do not want to alienate supporters or abuse potential recruits and because elections mitigate principal-agent problems (Sawyer et al. 2021).

Consequences of Sexual Violence
One of the most recent developments in sexual violence research is a growing number of studies, many drawing on the SVAC data, that consider the consequences of wartime sexual violence on conflict processes and outcomes. At the international level, reports of sexual violence are associated with a higher number and level of diplomatic actions required by UN Security Council resolutions (Benson & Gizelis 2020), reinforcing the recognition of wartime sexual violence as a critical international security problem. Nagel & Doctor (2020) connect sexual violence to rebel group fragmentation; commanders who have built group cohesion through sexual violence are more likely to split from the main organization as they are confident that their soldiers will follow them. In addition, based on a cross-national statistical analysis, the likelihood of conflict management mediation efforts increases when sexual violence by rebel groups is public knowledge, defined as being reported in NGO and US State Department reports (Nagel 2019). Shifting to state actors, the higher the prevalence of sexual violence by state forces is, the more likely states are to reach negotiated outcomes; Chu & Braithwaite (2018, p. 233) argue that sexual violence is an indicator of weakness, and actors that perpetrate it "are inclined to salvage something from the conflict by way of a settlement." Studies have also documented increased mobilization by women as a response to or in the aftermath of conflict-related sexual violence (Berry 2018, Kreft 2019, connecting with the broader literature on posttraumatic growth resulting from conflict-related violence (Bauer et al. 2016). Kreft (2019, p. 220) argues that affected women understand "sexual violence as a violent manifestation of a patriarchal culture and gender inequalities," which inspires mobilization around a broad range of women's issues "with the goal of transforming sociopolitical conditions." Among sexual violence survivors and their families, studies have found inspiring evidence of resilience and growth. In Sierra Leone, survivors and their families took active measures to counteract stigma and the negative social consequences associated with being a victim (Koos 2018). While stigma is still present, Koos's study suggests that it is neither insurmountable nor inevitable, a hopeful corrective to a dominant view in previous scholarship that suggests indelible stigma and long-term social exclusion for survivors. Empowerment programming for sexual violence survivors can also counteract some of the negative consequences of sexual violence (Amisi et al. 2018).
Community-based stigma nonetheless remains a major concern. In contrast to the mobilization documented by some studies, some victims and survivors of sexual violence suffer from shame, fear, ostracism, and distrust, and they may retreat from social interactions (Wood 2008). Negative perceptions of and objections to the social inclusion of sexual violence survivors remain strong even in the eastern DRC, the site of scores of programs to shift the stigma (e.g., Finnbakk & Nordås 2019). The negative long-term effects of conflict-related sexual violence are also apparent in a statistical study of how wartime violence affects postwar violence in the private sphere. In Peru, areas with higher rates of wartime sexual violence became hotspots for intimate partner abuse in the years after the end of the war (Østby et al. 2019). These consequences have significant implications for efforts to mitigate the effects of war and recovery efforts, and they are linked to debates about whether wars increase or decrease sex and gender equality (Lindsey 2019, Webster et al. 2019).

OPEN QUESTIONS AND FUTURE DIRECTIONS IN METHODS AND TOPICS
In this section, we address paths for future research, both methodological and substantive. From a methodological perspective, we address documentation challenges and issues of aggregation and disaggregation. Substantively, we highlight two promising areas for future research: a focus on state armed actors as perpetrators and an evaluation of the evidence for closing the impunity gap, which has emerged as the primary solution for policy actors seeking to deter future acts of sexual violence.

Documentation Challenges and Opportunities
Data collection on sexual violence involves significant methodological challenges and closely connected and complex ethical considerations. In addition to concerns over the risks of retraumatization and privacy, the research program is fraught with problems of social desirability bias, shame, and fear of reprisal for survivors and their families. Survivors navigate difficult social environments during and after war, choosing to report or conceal their experience for strategic reasons (Utas 2005). Scholars of sexual violence, particularly those who engage in fieldwork and have direct interactions with victims and survivors, have been at the forefront of the writing about the ethical issues at stake in research on sensitive topics and with vulnerable populations (e.g., Aroussi 2020).
One common contention is that sexual violence is chronically underreported. This is true in official sources, including reports to police and justice authorities, even in peacetime, and it is likely that sexual violence is severely underreported during wartime. Creative research methods can mitigate underreporting and reveal previously hidden reporting biases. For example, a list experiment in a survey in Sri Lanka uncovered a prevalence of sexual violence 10 times higher than found when direct questions were asked, as well as a previously hidden population of male survivors (Traunmüller et al. 2019). Latent variable modeling has also been used to analyze the problems of observability, which can lead to patterns of some instances of sexual violence being reported and others overlooked (Krüger & Nordås 2020).
Underreporting can arise because the language used to describe sexual violence is opaque compared to the language used to describe killings and other types of violations (Leiby 2009a). Victims, survivors, and witnesses may engage in self-censorship, using hard-to-decipher euphemisms and indirect descriptions. This makes data extraction through automation and machine learning, which holds great promise in the field of political violence, far more complex when studying sexual violence. Scholars have also argued that sexual violence may be overreported if benefits and aid are directly tied to sexual violence survivor status (Utas 2005) or if advocacy groups selectively highlight sexual violence to garner public attention to a crisis (Cohen & Hoover Green 2012).
There are also gendered biases in reporting. Men and boys as victims and survivors of sexual violence have been historically overlooked (Carpenter 2006). Issues of shame and stigma may be intensified with male victims and survivors (Sivakumaran 2007). Although the literature on male survivors is growing (e.g., Schulz 2018, Edström & Dolan 2019, Traunmüller et al. 2019, the accumulation of knowledge is hampered by the stark lack of reliable data. This is also the case with research on sexual violence against sexual and gender minorities, although there are detailed case studies of, for example, sexual violence against LGBT combatants in the www.annualreviews.org • Conflict-Related Sexual Violence Revolutionary Armed Forces of Colombia (FARC) (Thylin 2020). Finally, there are insufficient data on combatant victims and survivors who face sexual violence from within armed organizations; research on sexual assault within the US armed forces (e.g., Wood & Toppelberg 2017) is an exception.
The rapid expansion of quantitative studies of sexual violence has brought concerns about data bias and limitations to the fore (e.g., Davies & True 2015, Krüger & Nordås 2020. Although these scholars surface important critiques, qualitative data are affected by many of the same biases, such as underreporting and stigma. The most successful studies of conflict-related sexual violence combine multiple methods, using statistical analyses of large-n data sets and intensive field-based interviews, surveys, and focus groups. Future research should systematically analyze variation in observability under different conditions and invest in more sophisticated assessments of data uncertainty to better calibrate analyses and conclusions (Krüger & Nordås 2020). This could be coupled with comparative case studies or new methods of data collection, including tablet use and audio computer-assisted self-interviewing (ACASI), that eliminate the need for participants to verbally report sensitive information to enumerators. Such investment in innovation in data collection methods and practices at multiple levels of analysis ought to be a priority moving forward, and it could vastly improve the evidence base and our ability to predict and prevent sexual violence in the future.

(Dis)aggregation and Context
There are competing pressures of aggregation and disaggregation within the broader study of conflict-related violence that also are consequential for the study of sexual violence. Scholars are beginning to disaggregate the concept of sexual violence into its specific forms, arguing that the logics behind different forms likely differ. Donnelly (2019), for example, focuses on forced marriage to uncover its distinct logic and dynamic, a necessary step toward understanding the particular conditions under which this form of sexual violence occurs. Spatiotemporal data, such as data from the Armed Conflict Location and Event Data Project (ACLED), are another means of disaggregating information to understand variation in event-based conflict data. ACLED has also recently started to collect events-based information on violence against women, including sexual violence (Kishi et al. 2019), a useful addition to the cross-national data.
Despite its benefits, disaggregation risks introducing false precision (Hoover Green & Cohen 2021), specificity that the underlying information cannot credibly support. The problems of observability are not necessarily eliminated, and might even be exacerbated, by more disaggregated data structures, as more information risks being left unusable due to not being specific enough. For instance, a description of sexual violence being perpetrated by "some rebel groups" cannot be coded in a data structure that requires information about specific forms of violence and named groups of perpetrators. Disaggregation also carries the risk of violating respondents' privacy through deductive disclosure-inadvertently revealing respondents' identities-a particularly acute risk with stigmatized violence such as rape. In addition, disaggregation may miss larger contextual dynamics such as whether sexual violence was a complement or a substitute for other forms of violence, or whether, for example, looting and rape-violent acts that are often assumed to move together-are indeed closely correlated. Studying sexual violence in tandem with other violations sheds light on both conflict processes and political history, as shown in Inal's (2013) study on the puzzle of why pillaging was prohibited by treaty law, the Hague Conventions of 1899 and 1907, a century before rape was prohibited the Rome Statute of 1998.
An alternative to a singular focus on sexual violence is to focus on broader repertoires of violence. By studying multiple forms together, scholars seek to understand the types of violence that armed groups use and the degree to which they are employed. Hoover Green (2018) explicitly theorizes the notion of repertoire and studies sexual violence, looting, and killing in the Salvadoran civil war. Similarly, Gutiérrez-Sanín & Wood (2017) advocate for greater conceptual clarity in understanding patterns of violence, including repertoire, targeting, frequency, and technique. The causes of sexual violence are likely to vary with its specific form (e.g., sexual torture, gang rape, and sexual slavery might be driven by different processes). Data that document the repertoire of sexual violations-i.e., what combinations of specific forms are perpetrated at what level of prevalence by particular organizations-can facilitate more analyses of such patterns. Furthermore, embedding sexual violence in its broader context can mitigate the critique from feminist scholars (e.g., Meger 2016) regarding the tendency to elevate sexual violence as the worst possible harm (for women, especially), reifying patriarchal notions of women's sexual purity.

State Actors as Perpetrators
A powerful myth is that wartime sexual violence is mainly perpetrated by undisciplined rebels (Cohen et al. 2013, Wood 2013. However, the evidence clearly shows that states are far more likely to be reported as perpetrators of sexual violence (Cohen & Nordås 2014). Even in cases where rebels show restraint, states may perpetrate sexual violence against their citizens. For example, the state was responsible for the overwhelming majority of sexual violence in the civil wars in both Peru (Leiby 2009b) andEl Salvador (Hoover Green 2018).
Given the frequency with which states and state-affiliated actors are reported as perpetrators of sexual violence against civilians, the causes, dynamics, and consequences of such violations by states ought to be studied more systematically. Based on anecdotal evidence, sexual violations are common in states' repressive repertoires, in situations of mass killings, and in authoritarian states where the coercive apparatus is engaged in other forms of repression. Systematic data on state violence by a range of actors, both during conflict and during peacetime, are needed to rigorously analyze patterns of such violations across time and space.
To date, existing arguments explaining why state actors perpetrate sexual violations center on principal-agent dynamics and the assumption that sexual violence is more likely to be associated with "overwork" by agents because of private motives or personal desire (e.g., Mitchell 2004). A global statistical analysis supports this argument, but it is limited to data from a single year (Butler et al. 2007). Because the argument emphasizes the lack of control and discipline of security forces as the main driver of sexual violence, it contrasts sharply with other work in the human rights literature emphasizing that state repression stems from the strategic motives of, and coordination by, leaders (Davenport 2007). Different theoretical models should therefore be considered in future work.
While it is possible that agents might be perpetrating sexual violence of which the principal is unaware, ignorance seems implausible in many cases. For instance, Leiby (2009b, p. 459) finds that in Guatemala and Peru, one-third and about half, respectively, of reported incidents "occurred under circumstances that negate the possibility that state leaders had no knowledge of the violence." This is echoed in research on rape in civil war that found that the vast majority of reported state-perpetrated sexual violence took place in detention (Cohen 2016). In addition to the finding that states do not seem to delegate sexual violence to progovernment militias to achieve plausible deniability (Cohen & Nordås 2015), these results together cast significant doubt on the idea that state leaders are unaware of and unable to stop sexual violations by state representatives.
Beyond state armed forces and progovernment militias, future work ought to build on research that considers the wider set of actors that are part of the state's coercive apparatus both at home and abroad. This includes research on sexual violations by UN peacekeepers (Nordås & Rustad 2013, Karim & Beardsley 2017, Westendorf 2020, and on patterns of sexual violence by other state actors linked to conflict, such as private military contractors (Snell 2011). Both in peacetime and during periods of armed conflict, sexual violence by police may be central to understanding the state's use of sexual violations, but it is not yet well documented. Both the strategic and opportunistic elements of the states' coercive apparatuses and their use of sexual violence, as well as how it compares to other violence and repressive repertoires perpetrated by states, are critical topics for future research. A renewed study of state actors who are active across the spectrum from peacetime to war makes state-actor-centric research particularly valuable for addressing debates about the continuum of sexual violence in future research.

The Impunity Gap: How to End Wartime Sexual Violence
For policy, the fundamental insight that conflict-related sexual violence varies implies that it can be mitigated and even entirely prevented by armed groups. However, it remains unclear what specific interventions can reduce or prevent sexual violence atrocities. Although temporal trends in sexual violence prevalence on the ground are hard to establish due to the "information paradox" (Clark & Sikkink 2013), the reported incidence of sexual violence is not diminishing over time, and in some conflicts it is occurring at very high levels. Moving forward, the scholarly community should continue to theorize and assess which types of interventions can best mitigate conflictrelated sexual violence.
In many practitioner and advocacy circles, the call to end impunity for past perpetrators of sexual violence is heralded as the ultimate solution to the problem of sexual violence (and especially rape) in war. This assumes that prosecution is a powerful deterrent; however, little empirical evidence supports this assumption, at least at the level of the ICC (Cronin-Furman 2013). Under what conditions prosecutions may deter perpetrators of sexual violence, and which types of perpetrators can be deterred, are important questions for future research.
Research in related fields suggests that prosecutions deter some types of human rights violations (Dancy et al. 2019). Future research could consider the impact of efforts to end impunity while weighing the immense costs-both monetary and temporal-of prosecutions against other possible mechanisms to reduce conflict-related sexual violence, drawing from the human rights and state repression literatures. One concern is that states may use prosecutions as a fig leaf covering an explicitly political agenda, in which selective prosecution serves leaders' goals "to garner political legitimacy among key domestic audiences" (Loken et al. 2018, p. 751). Evaluations of whether major policy initiatives have worked and why they have failed are also important to develop a set of best practices for the future (e.g., Kirby 2015). Medie (2020) focuses on the roles of the United Nations and local women's movements in pressing for postconflict justice for survivors of sexual violence, and the key role of women's organizations in institutionalizing and implementing specialized mechanisms to address violence against women. Lake (2018) argues that state fragility creates opportunities for human rights NGOs to initiate legal processes in ways that are impossible in stronger states. Which interventions work to stop or prevent sexual violence is likely to be a function of whether such violence is strategic, is opportunistic, or has developed as a practice at the level of armed groups (Wood 2014). Further, the rare cases where there is clear evidence that rape was ordered or carried out with a strategic intention by leaders-as in Rwanda, Bosnia, and the ongoing Rohingya crisis-may require a different set of tools than situations where the perpetrators were rogue agents.

CONCLUSION
There has been a remarkable expansion in empirical quantitative and qualitative literature in political science on wartime sexual violence against civilians. This growth has been particularly pronounced in the wake of the recognition that the occurrence of such violations varies across countries, conflicts, actors, and time, and as more systematic quantitative data have become available.
While still a relatively small literature, the scholarship has made, and continues to make, important contributions to the broader discipline. The study of sexual violence has led the field in emphasizing the importance of internal organizational and institutional factors of state and nonstate armed groups to understanding patterns of violence. These studies have illuminated how cohesion, ideology, governance, and training are critical for analyses of armed actors' behavior. In addition, the concept of violence as a practice (as distinct from strategy or opportunism) has utility far beyond sexual violence and could be used to explain "all forms of political violence" (Wood 2018, p. 513) as well as other armed group behaviors.
The study of sexual violations holds the potential for contributing important insights into a wider set of political processes, including elections, protest, and repression (e.g., Kreft 2019, Krause 2020). Further, the study of conflict-related sexual violence has brought more attention to sex and gender in conflict studies and international relations writ large and has been made possible by the rising number of women in academia. Calls from scholars of sexual violence, perhaps more in workshops and meetings than in published research, emphasize a more inclusive approach to conflict research, voicing concerns that scholars from conflict-affected regions should be more involved in data collection, analyses, and scholarly debates.
The study of conflict-related sexual violence involves thorny and instructive methodological challenges in terms of how to collect data on hard-to-measure phenomena. More and better data are sorely needed, and the field presents an opportunity to develop new methodologies for data collection that reflect ethical best practices and cutting-edge methods for researching sensitive topics. Given the interdisciplinary nature of the research topic-spanning fields as disparate as public health, medicine, law, women's and gender studies, economics, and computer sciencethis area can and should encourage productive and creative collaborations across a wide range of scholars.

DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.