Abstract
Combatants used sexual violence in approximately half of all civil conflicts since 1989. We expect that when groups resort to sexual violence they are organizationally vulnerable, unlikely to win, and as such they are inclined to salvage something from the conflict by way of a settlement. Using quantitative analysis of data on civil conflicts in the post-Cold War period, we find that a higher prevalence of sexual violence perpetrated by government forces precipitates negotiated outcomes. This is particularly true in contexts where both government and rebel forces utilize comparable levels of wartime rape and other forms of sexual abuse.
Roughly 52% of civil conflicts between 1989 and 2008 featured combatants engaging in sexual violence against non-combatants. 1 In this period of time, wartime sexual abuse occurred in conflicts across Africa, Latin America, Eastern Europe, Asia, and Oceania. Sexual violence is perpetrated during identity conflicts and ideological ones, in struggles over territory and over control of the central government. Given the wide-ranging nature of conflicts affected by sexual violence, we ask: does the adoption of sexual abuse as a strategy in warfare impact the ability of the government and rebels to achieve a negotiated outcome to their incompatibility?
The Burundian civil war in the 1990s and early 2000s provides some insight into this question. Particularly during the second half of the conflict, both the Tutsi government’s armed forces and Hutu rebel groups utilized sexual violence against non-combatants as a regular part of their armed campaigns, often at comparable levels of severity. Previous efforts at negotiating peace between the warring parties had struggled to take hold, but as rebels from the National Council for the Defense of Democracy-Forces for the Defense of Democracy (CNDD-FDD) began to match the Burundian armed forces both in terms of the need to engage in forced recruitment as well as in the perpetration of moderate to high levels of rape and other forms of sexual violence, peace talks began in earnest, with a negotiated settlement between these two combatants being reached in 2003. However, the conflict did not completely terminate because another mainly ethnically Hutu rebel group, Palipehutu-National Liberation Front (FNL), escalated its military offensive and began engaging in higher levels of sexual abuse of civilians—mirroring such abuse strategies maintained by the Burundian military. Shortly thereafter, the Burundian government and Palipehutu-FNL rebels agreed to a settlement, ending their conflict in 2008.
This case highlights a number of potentially important dynamics regarding the effect of wartime sexual abuse on the prospects for terminating civil conflicts generally. In particular, it is worth examining whether the perpetration of sexual violence as a strategy in combat helps facilitate compromise or victory, and whether the balanced employment of sexual abuse during conflict creates a hurting stalemate of sorts that ultimately compels combatants to pursue peace talks in earnest.
Some argue that the utilization of sexual violence is strategic during armed conflict, especially with respect to recruitment of fighters (e.g. Bloom, 2011; Cohen, 2013). While a growing area of research addresses how and when civilians are targeted by combatants, scholars have only recently begun to assess how civilian abuse systematically shapes outcomes of these conflicts. Wood and Kathman (2014) highlight a curvilinear relationship between non-combatant casualties and the likelihood of negotiated settlements in African conflicts, while Thomas (2014) and Fortna (2015) find conflicting results with respect to terrorism’s impact on concessionary behavior of governments. We extend this line of inquiry by focusing on a typically non-lethal, yet potentially equally impactful, form of civilian targeting in civil conflicts: sexual violence. We find that negotiated outcomes are somewhat more likely when governments engage in higher levels of wartime sexual abuse, and that this form of termination is particularly common when both sides are engaging in comparable levels of sexual violence.
Why do combatants target civilians?
Before delving into a discussion linking wartime sexual violence to the likelihood of successful negotiations, it is important to understand why combatants are motivated to perpetrate crimes against non-combatants in the first place. Existing literature provides a variety of explanations for why civilians are targeted during conflicts.
First, attacks on non-combatants can be used systematically to decrease support for the enemy. Terrorizing non-combatants is employed as a rational strategy to maximize civilian support by deterring defection (Kalyvas, 1999). During the Algerian civil war, civilians who supported the opposition or defected were singled out based on their allegiance and were subsequently killed. Even ambivalence, or “fence-sitting,” was considered an unacceptable position for civilians to take by insurgents and the incumbent, with both actors taking violent measures to punish such behavior. With respect to sexual violence in particular, during the Guatemalan Civil War, almost half of all recorded sexual violence events were part of a generalized terror campaign against the civilian population (Leiby, 2009). Government forces deliberately perpetrated rape publicly to demonstrate to the entire community that the state should not be reckoned with and support for rebels would be severely punished.
While warring parties may rationally undertake a terrorization strategy to instill fear amongst the populace, there is the potential that such attacks will backfire, with civilians resenting these actors in the short and long term. Rather than submitting to their terrorizers, civilians sometimes mobilize in support of the opposition because these actors are not committing crimes against them (Zhukov, 2013). The empirical record demonstrates that civilian support is an integral factor in an armed group’s success in achieving conflict objectives (Wood, 2010, 2014), so combatants make a considerable gamble when terrorizing non-combatants. Nevertheless, the aforementioned examples highlight a calculated logic behind the implementation of civilian targeting, with an objective by the perpetrator to intimidate the population into submission at the hands of the perpetrating group.
Second, combatants have been known to victimize civilians as part of a broader campaign seeking revenge for perceived injustices and deep-seated ideological, ethnic, or religious hatred. Balcells (2010) demonstrates that patterns of victimization during the Spanish Civil War correspond to pre-war areas of political contestation as well as earlier wartime experiences that instilled a desire for revenge. In the case of Bosnia–Herzegovina, Serbian rebels raped and massacred Bosniaks in order to destroy family and community bonds and to force the ethnic group out of desired territory (Card, 1996). The Rwandan genocide also exemplifies how long-running tensions between identity groups can escalate to widespread civilian victimization in the context of civil war violence.
Third, and with respect to violent non-state actors terrorizing non-combatants, some argue that civilian targeting is a method for rebels to punish the government via public opinion; fostering an environment of vulnerability within the broader population can decrease the regime’s desire to fight and force the provision of concessions to the rebels (Wood and Kathman, 2014). In addition to Thomas (2014), who demonstrates that terrorism committed by rebels during civil war increases opportunities to negotiate and concessions from the government, Pape (2003) finds that the use of suicide bombing leads to some policy reform success in liberal democracies.
However, there is considerable debate as to whether civilian targeting by way of terrorism is an effective strategy in helping perpetrators achieve their goals. Fortna (2015) reasons that rebels face a dilemma: while terrorism may help rebels survive through continued fighting, the choice of employing terrorist attacks may come with the (expensive) price of discrediting their larger political ambitions. Thus, terrorist attacks are “cheap” shots that rebels can utilize in order to survive their sustained fight but, in adopting this strategy, they lose their ability to credibly bargain for their primary goal. Abrahms and Gottfried (2014) contend that groups who employ terrorist attacks in their repertoire limit the bargaining space they have with governments, which in turn decreases the likelihood of gaining favorable concessions. They demonstrate that attacks targeting civilians significantly lower a group’s bargaining success. This suggests that rebels utilizing such strategies impair their ability to extract concessions from the government, since terrorist tactics damage their reputation.
Finally, regarding the perpetration of sexual violence in particular, scholars have reasoned that combatants employ sexual abuse during conflicts for a number of reasons. One notion holds that individuals perpetrating wartime rape do so when their leaders are unable or unwilling to punish this behavior. In this sense, wartime sexual violence occurs because these groups lack strong leadership to act as strict disciplinarians (Humphreys and Weinstein, 2006; Butler et al., 2007; Wood, 2009). This has been witnessed in conflicts around the world, including the Sierra Leonean and Bosnian civil wars. Sexual violence in Sierra Leone increased considerably after the breakdown of the Lomé Peace Accord. One group, the Kamajors, was responsible for multiple acts of sexual violence against civilians, particularly when forces were moved away from their native areas and traditional chiefs. Once rebels were no longer geographically proximate to their leaders, the Kamajors became undisciplined and the level of sexual violence increased (Human Rights Watch, 2001). Additionally, in interviews with soldiers who were convicted of raping women in Bosnia, members of the government’s armed forces claimed they did so because superiors rewarded them with women to rape as acknowledgment of their good behavior. Soldiers who refused this “token of appreciation” were purportedly assigned to worse positions (Stiglmayer, 1994).
It has also been argued that rape and other forms of sexual abuse are employed strategically during civil conflicts, instead of or in addition to its seemingly stochastic occurrence. The use of sexual violence by a combatant signals to the opposition and to non-combatants that the perpetrators are willing and able to inflict pain on others in pursuit of their aims, as suggested by the aforementioned Guatemalan example. However, the utilization of sexual violence may also signal a weak actor reliant on a system of rewards or shared experiences to compel individuals to fight for and remain loyal to the armed group. Perpetrating organizations tend to be less cohesive owing to forced recruitment and conscription (“pressganging”), leading to their adoption of systematic sexual violence as a perverse way of rewarding and socializing recruits (Cohen, 2013).
This is in comparison to participants in conflicts mobilized along ideological or ethnic similarities, which have been theorized to exert more credible commitment to the cause (Gates, 2002). In order to entice recruits to continue fighting in lieu of ethnic or ideological motivations, the leadership must offer selective benefits to members in order to maintain their allegiance (Kalyvas and Kocher, 2007). These benefits encourage other individuals to join as well, promoting active and productive participation in wartime activities. The downside to offering rewards for service is that it is likely to result in a flood of opportunistic individuals who join the group, rather than those committed to the cause in the long term. This ultimately produces a fighting force that is weak organizationally, as the diversity of backgrounds among recruits impedes social cohesion; the group lacks the ability to filter low- and high-committed individuals, and the leadership must continually offer rewards to keep participants motivated to fight (Weinstein, 2005). Unlike material benefits, such as lootable gemstones, illicit drugs, or money, which can prove difficult to maintain in regular supply, providing opportunities to engage in sexual activity is a “reward” that group leaders can offer to their members on a more routine basis if selective benefits are required to facilitate and maintain organizational cohesion.
The effect of sexual violence on the likelihood of negotiated outcomes
Groups reliant on civilian targeting—and sexual violence in particular—suffer in terms of their broader legitimacy across the population. It is reasonable to expect that decisive military victory becomes more difficult as groups lose this support, especially with respect to rebel groups that are largely reliant on domestic support and assistance from civilians. Combatants able to recognize that their victory is unlikely should be more willing to pursue a negotiated outcome to the conflict. In this way, they are able to achieve some gains (rebels) or maintain some power (government), while not having to suffer the costs of continued fighting or risk outright defeat at the hands of their enemy.
Cohen (2013) highlights that both rebels and government forces engage in sexual violence during civil conflicts; she attributes this to an inherent lack of cohesion amongst those forces when brought together through conscription or abduction. Forced recruitment necessitates some sort of mechanism by which individual combatants can be joined together through shared experiences or private goods. Weinstein (2005) alludes to challenges facing forces cobbled together through forced or materially incentivized recruitment; these rebel organizations struggle to achieve their goals and are often defeated (e.g. the Revolutionary United Front in Sierra Leone), or have to agree to some compromise with the government (e.g. Renamo in Mozambique), in large part because of issues with defection and fighting effectiveness. Weinstein’s logic reasonably extends to the government as well; when the government relies on troops that are not participating voluntarily, but are brought into the movement through conscription and indoctrinated through sexual violence, the organization as a whole suffers—especially when fighting rebels who organize without the need to strong-arm participation. “Pressganging”—forced conscription into the government’s armed forces—was prevalent in conflicts where the government regularly engaged in sexual violence, including civil wars in Sudan, Liberia, and Peru.
When the government’s armed forces must resort to sexual violence to maintain organizational unity and membership, they have issues not only with group capacity and cohesion but also with broader civilian support. This should be particularly concerning for regimes that rely on some degree of public legitimacy to stay in office. Thus, when the government resorts to sanctioning, or at least turning a blind eye to, sexual violence to reward or unify its fighters, it suffers on multiple levels and should seek to end the conflict expediently. This can be best achieved through some sort of negotiation process rather than trying to pursue a military victory that may not succeed given the tenuous connections uniting the military. Thus, we expect that governments forced or choosing to resort to sexual violence will not want to maintain these activities or the campaign for a long period of time, and will seek to end the conflict as rapidly as possible.
Drawing again upon Weinstein (2005) and Cohen (2013) to suggest that rebel groups employing sexual violence are likely to have tenuous organization structures, we also expect that these groups will struggle to sustain a long-running and successful campaign once forced to resort to civilian abuse to maintain some semblance of cohesion. One might argue that combatants receiving this private good for participating in a group that allows or encourages sexual violence will want to prolong the conflict, as it can be more difficult to obtain these “rewards” once the conflict terminates and some semblance of law and order is restored (e.g. “Rebellion-as-Business” according to Collier et al., 2004). However, unlike conflicts facilitating access to lootable resources that provide material profit, sexual violence does not translate in the same way and is not something that can be used to purchase basic necessities. Thus, we assume it does not have the same conflict-prolonging effect as other (material) forms of private benefits used to recruit and retain members in an armed group.
Beyond being an organizational mechanism, rebels can use sexual violence to compel the government to come to the negotiating table. In a sense, sexual abuse used systematically by armed groups in the context of an anti-government campaign could be considered alongside the strategic use of terrorism. Pape (2003) and Thomas (2014) demonstrate that groups using terrorism have been able to successfully extract concessions from their governments; it is reasonable to extend this logic to groups using sexual violence as another form of civilian targeting as well. When governments face a backlash from citizens targeted for sexual abuse by violent non-state actors, they may feel pressured to offer concessions and even seek a negotiated end to the conflict in order to stop rebels from abusing non-combatants. Hultman (2007), as well as Wood and Kathman (2014), argue this with respect to the killing of civilians, and this should apply in the context of non-lethal attacks as well, especially since victims of sexual abuse are often still alive to voice their discontent at being unprotected. Thus, as Thomas (2014) argues with respect to rebel groups using terrorism in African civil wars, we expect that rebels using sexual violence are more likely to achieve negotiated outcomes in civil wars as compared with groups abstaining from this form of civilian targeting.
Beyond organizational signals and incentives for combatants to negotiate an end to conflict in the midst of perpetrating sexual violence, we would be remiss in ignoring the critical role civilians can play in ensuring that combatants not only attempt talks but also ultimately agree to end the fighting. Victims of sexual abuse have a vested interest in ensuring that negotiations succeed, as failure and return to the battlefield probably means a return to civilian abuse as well. Thus, we should expect that when victims of sexual abuse are able to leverage their experiences of abuse to push for negotiations between combatants, they do so in pursuit of long-term solutions rather than talks that are mere window dressing. For example, the Women of Liberia Mass Action for Peace pressured rebels and the Liberian government not only to come to the negotiating table, but to actually achieve a settlement. Many of its members were women who had been sexually abused over the course of the Liberian conflict, and they used their experiences as victims to mobilize a movement that rallied public support and met with combatants on all sides until a peace agreement was reached (Gbowee and Mithers, 2011).
It is often the case that both rebels and government forces employ sexual abuse during civil conflicts. While the aforementioned discussions speak to the motivations of specific actors for engaging in this behavior and why that would compel them to seek compromise, contexts in which both actors engage in sexual violence suggest a unique environment “ripe” for negotiation (Zartman, 1989). In particular, when both sides in the conflict resort to civilian abuse of this kind, it provides information about the organizational structures and capabilities of these combatants in a more open, visible manner. If the rebels recognize that the government needs to engage in sexual violence (and vice versa), but it too is in a position of contrived organization, then a compromise might be optimal rather than continuing the conflict and risking the enemy gaining access to more resources such that the balance of capabilities tips in their favor. When both sides engage in comparable levels of sexual abuse, suggesting some degree of equality in terms of strategy and capacity, we expect that combatants will seize upon the opportunity to negotiate a settlement. Revisiting the example in the introduction, negotiations in the Burundian civil war were attempted across several years of the conflict, but an agreement seemed to become truly possible at a time when rebels and government forces were engaging in comparable levels of sexual violence. Thus, we expect the following addendum to the previous hypotheses, taking into account actions by both adversaries involved in a conflict:
Research design
The unit of analysis in our study is the government-rebel group dyad-year, from 1989 to 2008. We consider violent intrastate conflicts included in the UCDP Armed Conflict Database resulting in at least 25 battle deaths per year, involving the government and at least one organized armed group (Pettersson and Wallensteen, 2015).
Our dependent variable is derived from the UCDP Conflict Termination Dataset (Kreutz, 2010). We construct a five-point variable indicating the conflict status at the end of the year in question. Outcome is coded 0 if the conflict is still ongoing, 1 if it ended in a military victory for the government, 2 if rebels achieved military victory, 3 if a negotiated outcome was reached (ceasefire or settlement), and 4 if the conflict terminated in low activity by dropping below the 25 battle death threshold. Across the sample used in our analyses, we observe 797 dyad-years in 286 dyadic conflict episodes. Of these, 23 episodes terminated in government military victory, 13 in rebel victory, 75 in a negotiated outcome, and 115 in low activity. The remaining episodes had not ended as of December 2008.
We employ a variety of independent variables capturing different aspects of sexual violence during civil conflicts. These measures are drawn from the Sexual Violence in Armed Conflict Database, which relies on accounts of sexual abuse included in reports from the US State Department, Amnesty International, and Human Rights Watch (Cohen and Nordås, 2014). First, we include two variables, Government SV Prevalence and Rebel SV Prevalence, denoting the level of sexual violence perpetrated by the government and by the rebel group, respectively. These variables are measured each year of the conflict along a four point index, where 0 indicates no reported sexual violence, 1 indicates that the combatant group committed some sexual violence related to the conflict, 2 represents an actor that engaged in common or widespread abuse, and 3 is attributed to groups that used sexual violence on a “massive” scale.
We then utilize these prevalence scores to construct a binary variable indicating whether the combatants engaged in a comparable level of sexual abuse in a given year. Equal Abuse is coded 1 if, in the year of observation, both government and rebels engaged in sexual violence at the same prevalence level (in other words, both Government SV and Rebel SV are coded 1, or both 2, or both 3). It is coded 0 if the actors engaged in different levels of sexual violence, or if there was no record of one or both combatant sides perpetrating this type of abuse. Our sample includes 52 dyad-years in which both the government and rebel group engaged in the same level of sexual violence.
We include a number of control variables important in considering the likelihood of negotiated outcomes in civil conflicts. When assessing conflict termination, it is important to take into account characteristics of the conflict itself. We include Weak Rebels, an indicator of whether a rebel group is weaker (coded 1) or stronger than/at parity with (coded 0) the government in terms of fighting capacity. We also control for foreign intervention and the provision of non-military assistance to the government and rebels; Government Support and Rebel Support are coded 1 if that side received foreign sponsorship and 0 if it did not. These variables are derived from the Non-State Actor Dataset (Cunningham et al., 2009). We also consider a few measures of conflict severity. First, Battle Deaths is the natural log of a yearly measure of the best estimate of total battle-related deaths, from the UCDP Battle-Related Deaths Dataset v.5-2015 (UCDP, 2015). Second, we include data on non-combatant fatalities in conflicts; Government OSV is the natural log of civilian killings perpetrated by the government, and Rebel OSV is the complimentary measure for the rebel group in a given dyad. These data come from the UCDP One-sided Violence Dataset v.1.4-2015 (Eck and Hultman, 2007). We also construct a country-year indicator of whether UN Peacekeepers were present in the previous year, using data from Kathman (2013).
We also control for country-specific characteristics associated with conflict outcomes, including the natural logs of (government) Military Size in terms of personnel, Population, and GDP Per Capita, as well as the level of Democracy, all in the previous year. Data on military size comes from the Correlates of War National Material Capabilities dataset (Singer et al., 1972). Population and economic data are drawn from the Maddison Project (Bolt and van Zanden, 2014), and we employ the Polity IV index to capture regime type dynamics (Marshall et al., 2014). Summary statistics for all variables are reported in Table A1 of the Online Appendix.
Given our particular focus on negotiated settlements, we are interested in determining whether sexual violence hastens conflict termination by way of compromise as compared with the four other possible outcomes (continued conflict, military victory for government or for rebels, and low activity). We employ Fine and Gray’s (1999) competing risks model to test our hypotheses. This model is appropriate for testing our expectations for several reasons. First, it does not require that we make an assumption about the functional form of the baseline hazard. Other models (such as the Weibull and Gompertz) are parametric and assume that we know the distribution and shape of the hazard rate. Since we cannot make that assumption, a semi-parametric functional form is more appropriate.
Second, we prefer the competing risks setup to a semi-parametric Cox regression model. The latter modeling strategy assumes that there is only one way to reach a certain outcome. However, we are interested in multiple forms of conflict termination: government victory, rebel victory, low activity, and negotiated outcomes in particular. Hence, a competing risks model accounting for the varying ways a conflict can end is ideal.
Third, running a competing risks model gives the subdistribution hazard, or the instantaneous risk of reaching a specific outcome from a particular covariate given the observation has not already experienced the said outcome. The subdistribution hazard accounts for dyads (observations) that experience termination other than a negotiated outcome by keeping it in the risk set. This means that the subhazard calculated accounts for dyads (observations) that could have potentially ended with successful negotiations, but reached another type of conflict termination first. Other models decrease the risk set by removing an observation if it achieves conflict termination other than a negotiated outcome. Keeping an observation that has already experienced termination in the risk set helps calculate hazard rates as if it had not been removed. For the purposes of our hypotheses and research question, it is essential that we retain observations in the risk set since we expect that the utilization of sexual violence changes the conflict termination type. This also allows us to estimate the cumulative incidence function (CIF) for a specific event outcome, for a direct calculation of the likelihood of the a negotiated settlement when competing with other alternative outcomes (see, e.g. Wood and Kathman, 2014).
Results
Table 1 presents findings from five competing risks regressions examining the relationship between the perpetration of sexual violence and the likelihood of negotiated outcomes in civil conflicts. All models compare the risk of observing this particular result as compared with other forms of conflict termination across the duration of the conflict episode. We report coefficient estimates rather than hazard ratios, and as such, a positively signed coefficient means that the given covariate shortens the time to a negotiated outcome as compared with termination of a conflict via decisive military victory or stalemate. Conversely, a negative coefficient means that it will take longer to achieve a successful settlement in the conflict relative to the time it would take to defeat the enemy militarily or for a state of low activity to prevail. Models 1–3 provide the bivariate relationship between our main independent variables measuring wartime sexual violence and the likelihood of negotiated settlements as compared with other outcomes. Models 4 and 5 offer multivariate analyses controlling for several potential confounding factors. While we only report coefficients for observing a negotiated outcome given the focus of our hypotheses, coefficients for alternative forms of conflict termination based on Models 4 and 5 are presented in Tables A2 and A3 of the Online Appendix, respectively.
Effect of sexual violence on the likelihood of negotiated outcomes
Conflict episode dyad-years, 1989–2008. N = 797, N Conflicts = 286, N Failures = 75.
Results of a competing risks regression, coefficients with p-values in parentheses.
p < 0.10, **p < 0.05, ***p < 0.01.
Models 1–3 indicate initial support for our three hypotheses. As combatants engage in more wartime sexual abuse, the time it takes to achieve a negotiated outcome is hastened as compared with other forms of termination. This is especially true in cases where the level of sexual violence being perpetrated by combatants is comparable, as shown in Model 3. However, we cannot be certain that these bivariate relationships hold when we account for other factors that might also help promote negotiated outcomes to conflicts. Thus, we present multivariate competing risks analyses using a series of relevant control variables in Models 4 and 5.
Model 4 provides a baseline understanding of how the chances of observing negotiated outcomes are affected by the predominance of sexual violence in a combatant’s toolbox of strategies to inflict costs on the enemy and society writ large. We observe that a negotiated end to civil conflicts is hastened when the government utilizes sexual violence. Furthermore, as the government increases its engagement in this type of abuse, the time it takes to achieve a negotiated outcome (as compared with other forms of termination) decreases. This result offers continued support for our first hypothesis, that negotiated outcomes should be achieved more expediently when the government perpetrates sexual violence during civil conflicts. However, this variable loses statistical significance when we further disaggregate the conflict environment in Online Appendix Tables A5 and A6. While we maintain marginal confidence in the relationship between government-perpetrated sexual violence and the likelihood of negotiated outcomes, there is certainly more work to be done to parse out the particular contexts in which governments are willing and able to make credible concessions that result in successful peace talks.
Figure 1 depicts the CIF for the Government SV variable in Model 4. Essentially, the likelihood of observing a negotiated outcome before time t in the conflict, depicted along the x-axis, is universally higher at any year when the government is utilizing higher levels of sexual abuse as compared with lower or no degrees of this type of civilian targeting. After 11 years of combat (the mean duration of conflicts included in our study), the likelihood of a negotiated outcome when the government is not engaged in sexual violence is less than 20%. Conversely, if the government engages in moderate or widespread sexual abuse during conflict, the likelihood of negotiated outcomes increases to roughly 35 or 45%, respectively.
Model 4 CIF (prevalence of sexual violence committed by government).
While we do have some evidence that increasing prevalence of government-perpetrated sexual violence hastens negotiated outcomes to civil conflicts, we do not see a similar effect for rebel-perpetrated sexual abuse. The level of sexual violence committed by rebel groups does not appear to have a statistically significant effect on the risk of observing successful negotiations as compared with other forms of conflict termination. Thus, our second hypothesis is not supported by the results of Model 4, nor by Model 5. Interestingly, though, results in Table A6 of the Online Appendix suggest that sexual abuse by rebels does hasten negotiated outcomes in places where women have some legal political rights and representation in government. Political empowerment helps enable survivors of abuse to pressure the government to reach an agreement with rebels, as we saw in Liberia, exhibiting an effect in line with Thomas’s (2014) logic of terrorism and concessions in civil wars. This also speaks to work by Haglund and Richards (this issue) and the notion that legal protections for women reduce gender-based violence after conflicts; here we suggest that such conditions also help resolve conflicts when sexual violence occurs. Furthermore, in Table A5 we also observe an increased propensity for rebel sexual violence leading to negotiated outcomes when we distinguish between rebels based on their access to international support. Groups that engage in this form of abuse and are not as reliant on local support (given external assistance and/or bases) are more likely to achieve some negotiated resolution to conflict as compared with groups that lack outside help or that do not perpetrate sexual violence. This indicates a continued need to disaggregate characteristics of rebel groups to better understand how their behavior impacts conflict outcomes. Another possible future line of inquiry would be to leverage data from Wood and Thomas (this issue) to assess whether the proportion of female combatants affects the group’s ability to leverage sexual violence in efforts to force the government to settle.
Model 5 examines the effect that a balanced level of sexual violence perpetrated by combatants has on the likelihood of observing a negotiated outcome in a given year of conflict. In particular, we utilize a binary variable Equal Abuse indicating whether the government and rebel group perpetrated the same level of rape in a given year. Again, with a positive coefficient we observe that when both actors in a dyad employ comparable levels of sexual abuse as a weapon of war, they are more likely to agree to resolve the conflict through negotiations than pursue outright victory or enter into a stalemate. 2
Figure 2 presents the CIF for situations where the rebels and government engaged in differing versus equal levels of sexual violence, as reported with the Equal Abuse variable in Model 5. When both sides of the dyad are engaging in comparable levels of sexual violence in a given year, at any temporal point in the conflict episode, the conflict is more likely to end in a negotiated outcome as compared with situations in which combatants engage in differing levels of abuse. This lends additional support to our third hypothesis.
Model 5 CIF (equal prevalence of sexual violence committed by both sides).
The control variables behave similarly across Models 4 and 5, although few are statistically significant. As the government’s armed forces increase in size, negotiated outcomes take longer to achieve as compared with other forms of civil conflict termination. Conflicts occurring in more democratic countries have a higher likelihood of resolution via negotiated outcomes, perhaps because democratic regimes are more reliable bargaining partners and can credibly promise to allow opposition participation in political institutions. Measures of conflict severity, one-sided violence, and interventions, as well as country-level factors of population size and wealth, are not statistically significant predictors of negotiated outcomes relative to other forms of conflict termination.
Conclusion
Overall, we observe support for some of our expectations. Governments engaging in higher levels of sexual abuse during civil conflicts are somewhat more likely to terminate the violence by offering concessions to the rebels, as compared with those regimes that refrain from wartime sexual violence. Also, when both combatant sides are engaging in comparable degrees of sexual violence against non-combatants, the time it takes to achieve a negotiated outcome is hastened relative to other forms of termination.
Our focus is on final outcomes of conflicts, but it would be compelling to study the effect of sexual violence on initial efforts to negotiate and the propensity of the government to make concessions short of a full settlement. While it is difficult for us to recognize when offers of negotiations are rebuffed, and essentially impossible to determine when governments refrain from extending these offers in the first place, it might be possible to uncover some of these unobserved events through latent variable analysis. It would be useful to assess when we should have seen efforts to negotiate peace but did not observe this in reality. It is also worth extending the analysis from Wood and Thomas (this issue) to investigate whether the prevalence of sexual violence during conflicts, as investigated here, impacts the development of gender-based legal protections and the continued perpetration of abuse after the conflict terminates. This might be especially important for the implementation of settlement terms; if combatants were able to leverage sexual violence to get favorable terms in a settlement, they may be inclined to continue this behavior after the conflict ends in order to receive even more benefits—especially in the absence of enforcable prohibitions on such abuse.
One policy implication that could be derived from these findings speaks to the potential timing of a push for diplomatic interventions as well as support for particular types of social movements. Conflicts appear to be more “ripe for resolution” when combatants engage in comparable levels of sexual violence, and thus would-be mediators might leverage such situations to push for peace talks. Furthermore, these situations of equal abuse by warring parties present opportunities for victims and their supporters to come together and pressure combatants to reach an agreement to end the conflict, as we saw in Liberia in the early 2000s. While we lack definitive evidence as to whether international support helps or hinders such grassroots movements (yet another topic worthy of further study), external actors might also use these conditions of equal abuse to promote a negotiated end to conflict indirectly through assisting civilian-led campaigns for peace.
Footnotes
Authors’ note
An earlier version of this paper was presented at the October 2015 SPGS Kopf Conference: “How Does Gender Shape Violence and Coercion?” at Arizona State University. We thank conference participants, as well as the editors and anonymous reviewers, for their helpful comments and suggestions. The dataset, Online Appendix, and all other supporting materials used to produce this article can be accessed via a supplementary data file hosted on Sage’s CMPS website.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
