![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
Wartime sexual violence exhibits significant variation both across and within civil conflicts. When explaining this variation, existing scholarly work primarily focuses on rebel groups and overlooks state actors engaged in sexualized civilian victimization. This paper extends the literature by concentrating on states as perpetrators of wartime sexual violence, linking the variation in state-perpetrated sexual violence to battlefield outcomes. It argues that when states suffer from battlefield losses, sexual violence by state security forces is likely to become more prevalent due to changes in motivations at both the command level and the rank-and-file level. Moreover, it suggests that the effect of battle losses on state-perpetrated sexual violence is greater if the states are particularly strong in terms of their military capabilities. Time-series cross-sectional analyses of all armed conflicts between the years 1989 and 2020 provide support for these theoretical expectations, indicating that the effect of battle losses on sexual violence is conditional on states’ military strength. The findings have important implications for the prevention of sexual violence in conflict zones.
La violencia sexual en tiempos de guerra muestra una variación significativa, tanto entre diferentes conflictos civiles como dentro de cada uno de ellos. Al explicar esta variación, el trabajo académico existente se centra principalmente en los grupos rebeldes y tiende a pasar por alto a los agentes estatales involucrados en la victimización civil sexualizada. Este artículo amplía la literatura existente, concentrándose en los Estados como perpetradores de violencia sexual en tiempos de guerra y vinculando la variación existente en la violencia sexual perpetrada por el Estado con los resultados en el campo de batalla. El artículo argumenta que, cuando los Estados sufren pérdidas en el campo de batalla, resulta más probable que la violencia sexual por parte de las fuerzas de seguridad del Estado se vuelva más frecuente, debido a los cambios en las motivaciones tanto a nivel de mando como a nivel de base. Además, el artículo sugiere que el efecto que ejercen las derrotas en batalla sobre la violencia sexual perpetrada por el Estado es mayor si los Estados son particularmente fuertes en términos de capacidades militares. Usamos análisis transversales de series temporales de todos los conflictos armados entre los años 1989 y 2020 para respaldar estas expectativas teóricas, indicando que el efecto de las derrotas en batalla sobre la violencia sexual está condicionado por la fuerza militar de los Estados. Nuestras conclusiones tienen implicaciones importantes para la prevención de la violencia sexual en zonas de conflicto Les violences sexuelles en temps de guerre présentent d’importantes variations, d’un conflit à l’autre, mais aussi en leur sein.
Pour expliquer ces variations, les travaux de recherche existants se concentrent d’abord sur les groupes rebelles et omettent les acteurs étatiques qui s’adonnent à la victimisation civile sexualisée. Le présent article élargit les horizons de cette littérature en se concentrant sur l’État en tant qu’auteur d’actes de violence sexuelle par temps de guerre, et met en lien les actes de violence perpétrés par l’État avec les issues sur le champ de bataille. Il affirme que lorsque les États y subissent des pertes, les actes de violence commis par les forces de sécurité étatiques ont des chances de se renforcer, à cause de l’évolution des motivations au niveau des donneurs d’ordre comme à celui des subalternes. De plus, il indique que l’effet des pertes sur le champ de bataille sur les actes de violence sexuelle commis par l’État s’intensifie quand les dépenses militaires de l’État en question sont particulièrement importantes. À l’aide d’une analyse des données chronologiques transversales de tous les conflits armés entre 1989 et 2020, je trouve des éléments pour venir étayer ces attentes théoriques; ils indiquent que l’effet des pertes sur le champ de bataille sur les violences sexuelles est fonction de la force militaire des États. Les conclusions s’accompagnent d’implications importantes pour la prévention des violences sexuelles dans les zones de conflit.
Acknowledgements
The author extends gratitude to Burcu Savun, Michael Colaresi, Merve Keskin, Hyunjung Park, as well as the two anonymous Reviewers for their valuable comments on an earlier draft of the paper.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Correction Statement
This article has been corrected with minor changes. These changes do not impact the academic content of the article.
Notes
1 As Benson and Gizelis (Citation2020) show, wartime sexual violence is a distinct form of civilian victimization. That is, it does not necessarily correlate with other types of political violence against civilians. Problematizing the focus on the lethal forms of violence in the civilian victimization literature, Cohen (Citation2013) also argues that “it is both uncertain and unlikely that theories developed to explain the incidence of homicide can be readily applied to another type of violence that affects a distinct population (462)”.
2 Dawson and Prewitt (Citation1968) define socialization as “a process of inducting actors into the norms and rules of a given community, the endpoint of which is internalization” (cited in Checkel Citation2017, 594).
3 Baaz and Stern (Citation2013), among other critiques of weapon of war approach, warn that intentions of commanders should not be directly inferred from the consequences of sexual violence. For example, when the widespread prevalence of sexual violence leads to social disruption in society, one should avoid labeling this consequence as the intention of the perpetrator group without supporting evidence (Aranburu Citation2010).
4 For a discussion of principal-agent problems in the context of sexual violence, see Butler, Gluch, and Mitchell (Citation2007) and Lee and Tomashevskiy (Citation2023).
5 For studies on rebel-perpetrated sexual violence, see, for example, Asal and Nagel (Citation2021); Faulkner and Welsh (Citation2022); Sarwari (Citation2021); Sawyer, Bond, and Cunningham (Citation2021); Whitaker, Walsh, and Conrad (Citation2019); Wieselgren (Citation2022).
6 There are few cases in which the rebel groups are held accountable by international organizations for engaging in war crimes and crimes against humanity. For example, in 2012, the United Nations Security Council imposed sanctions on members of the rebel group, M23, for their acts of civilian victimization in the Democratic Republic of Congo (BBC 2012). Moreover, recent research finds that some rebel groups express their commitment to the “law of war” which would make it more costly for them to engage in sexual violence (Fazal Citation2018). I leave the question of, why states are more likely to engage in sexual violence than rebel groups?, for future research, and focus on the variation in sexual violence committed by state actors in this study.
7 Hoover Green (Citation2016) makes this argument for both states and rebel groups, yet only tests its empirical validity for rebel groups.
8 While the identity of civilian victims is undeniably a critical aspect, it falls outside the central scope of this analysis. The mechanisms explored in this paper suggest that victims are often perceived as supporters of enemy forces, underscoring potential significance in the geographical and ethnic dimensions of these atrocities. However, it is crucial to note the limitations in existing datasets, which do not adequately capture the ethnicities of victims or provide comprehensive information to assess the prevalence of sexual violence across different locations and ethnic groups. Consequently, this study does not establish a direct link between the ethnic identities of state security forces and civilian victims concerning sexual violence perpetration. Wieselgren (Citation2022) offers insights into the ethnic dimensions of sexual violence, but his research specifically focuses on rebel groups in Africa, and thus does not extend to the behavior of state security forces. This acknowledgment of the study’s limitations in identifying the victims’ identities underlines the necessity for more comprehensive research in this area, particularly large-N studies that explore the ethnic dimensions of the perpetrator-victim dyads in sexual violence committed by state actors.
9 This explanation for sexual violence resonates with what Baaz and Stern (Citation2009) call “evil rape”. Baaz and Stern (Citation2008) make a distinction between “lust rape” and “evil rape” and argue that while the former reflects the sexual needs of perpetrators, “evil rapes” arise from anger and rage generated by soldiers’ inabilities to inhabit idealized notions of heterosexual masculinity.
10 In this context, sexual violence functions as a form of communication between hegemonic and subordinate masculinities (Goldstein Citation2003; Seifert Citation1993). Hegemonic masculinity has been conceptualized as norms and institutions that seek to maintain men’s authority over women and over subordinate masculinities (Tosh Citation2004). Tosh (Citation2004) points out that the term “implies that control (even oppression) is in some way integral to masculinity, providing a framework for placing men in relation to women and those males whose manhood is for some reason denied” (51).
11 Acknowledging the empirical limitations in distinctly identifying these mechanisms is crucial, given the overlap in their manifestations and the secrecy typically surrounding military operations. This study utilizes available data and contextual analysis to infer the most likely motivations behind wartime sexual violence while recognizing the inherent challenges in this approach. The complexity of definitively attributing specific acts of sexual violence to one mechanism over another underscores the need for further research in this area, particularly research that involves access to classified military records or detailed perpetrator testimonies, which could provide more definitive insights into the specific conditions activating each mechanism.
12 The histogram illustrating the variation in sexual violence can be found in the appendix.
13 The histogram illustrating the variation in battle losses sustained by states can be found in the appendix.
14 When the polity score is employed instead of the electoral democracy index, the results remain unchanged.
15 Caprioli et al. (Citation2009) assert that the fertility rate represents the most suitable measure of gender inequality. This is due to its ability to encompass not only cultural aspects, such as individual preferences and the desire for children, but also the presence of discrimination against women and structural disparities, such as reduced access to education, employment opportunities, and political representation.
16 Multiple imputation is utilized to create five completed datasets, a convention well-established on theoretical grounds (Horton and Lipsitz Citation2001). To synthesize these multiple datasets into a single-point estimate for analysis, Rubin’s Rule is employed (Rubin Citation2018).
17 The ordered logit model assumes that the effects of the coefficients are the same across outcomes. The likelihood ratio and Wald tests confirm that the models do not violate this proportional odds assumption.
18 To generate predicted probabilities, I use model 4 and hold the control variables at their observed sample values. I consider states to be militarily weak if they fall within the 10th percentile of the observations. Conversely, I consider states to be militarily strong if they fall within the 90th percentile. This cutoff does not significantly impact the results. If the 95th and 5th, or 75th and 25th percentile cutoffs are chosen, the results remain similar.
19 For discussions on sexual violence and reporting biases, see Cohen (Citation2013); Leiby (Citation2009).
20 Aid dependency is calculated using the natural logarithm of the total aid expressed as a proportion of the recipient’s GDP per capita.