UN Peacekeeping and Protection from Sexual Violence

Abstract

Recent years have seen an increased emphasis on both protection of civilians and the problem of sexual violence. We explore the impact of United Nations (UN) peacekeeping on the occurrence of wartime sexual violence. Acknowledging the difficulty in reducing sexual violence, we propose two conditions under which peacekeepers are more likely to be successful: when the mission has a protection mandate and when the conflict actors exercise a high level of control over their forces. We find that the ability of peacekeepers to reduce sexual violence in general is weak. Only police within protection missions reduces the risk of sexual violence by rebels. However, when the actors exercise control, the number of peacekeepers is associated with a lower risk of sexual violence by both governments and rebels. We conclude that dealing with sexual violence by weak and fragmented actors is a challenging task beyond the current capacity of UN peacekeeping.

Keywords

civil wars conflict management international peacekeeping wartime sexual violence

There is an increasing awareness that protection against sexual violence is an important component in achieving international peace and security. This is reflected in the United Nations Security Council’s (UNSC) deployment of peacekeepers for which mission mandates increasingly include clauses on sexual and gender-based violence. However, increased attention does not necessarily translate into effective protection. The United Nations (UN) itself admits that combating sexual violence is one of its most challenging tasks and that the tactics by peacekeepers often is ad hoc rather than streamlined across various missions (UNSC 2012; UN Women 2012a). Even more troubling are the frequent reports of sexual abuse by peacekeepers themselves and the initial unwillingness of the UN to investigate such allegations. A crude look at the changes in sexual violence after the deployment of UN peacekeepers implies that the effect varies greatly across missions. There is indeed a significant number of missions, including those in the Democratic Republic of Congo (DRC) and Chad, where sexual violence appears more prevalent after years of deployment than at the outset. The UN track record however also includes relative successes; in cases such as Sierra Leone and Haiti, there are reports of less sexual violence when peacekeepers leave than when they enter (Cohen and Nordås 2014).

In this article, we address the question under what conditions UN peacekeeping can serve to protect civilians from sexual violence by the warring actors. Previous research has demonstrated that peacekeeping is a successful instrument for protecting civilians from lethal violence (Hultman, Kathman, and Shannon 2013; Bove and Ruggeri 2015; Kathman and Wood 2016) and preventing genocidal violence (Melander 2009). However, armed actors’ repertoires of violence vary greatly, and wartime sexual violence does not necessarily follow the same logic as other forms of lethal and nonlethal violence (Gutiérrez-Sanín and Wood 2017). Moreover, sexual violence encompasses both public events, such as seemingly strategic mass rapes, to more semiprivate events, such as opportunistic rape as well as sexual violence encouraged for the purpose of enhancing cohesion is smaller military units. While this makes peacekeeping even more challenging, it also means that we cannot automatically use studies of peacekeeping and lethal violence to draw inference about the effectiveness of peacekeepers in protecting civilians from sexual violence. In fact, it is not uncommon that missions report a decrease in other types of violence while sexual violence consistently remains a security issue (UN Women 2012a).

In addressing the question under what conditions peacekeeping is effective in protecting civilians from sexual violence, we focus on conditions pertaining to characteristics of both the mission and the perpetrating armed actors. First, we argue that peacekeeping should be more effective when the mission has a mandate to protect civilians, since those mission mandates put more emphasis on the interaction between the armed actors and the civilian population. Second, we argue that peacekeeping should be more effective when the mission faces an actor that exercises a high level of internal control. This is particularly important, given the shifting nature of sexual violence, from strategic to opportunistic. The level of internal control determines whether an actor is capable of responding to peacekeepers as a unitary actor and restraining its fighters from engaging in sexual violence.

The hypotheses we pose require a focus on the armed actors as the unit of analysis. Our sample has global coverage and includes all organized armed actors—governments and rebel groups—involved in an intrastate, armed conflict and the five years following its termination for the period 1989–2009 (Gleditsch et al. 2002). We use data on reported levels of sexual violence provided by the Sexual Violence in Armed Conflict (SVAC) dataset (Cohen and Nordås 2014). Our findings indicate that, on average, peacekeeping is not an efficient tool for managing sexual violence—even with a protection mandate. However, larger police deployments within protection missions are associated with a lower risk of sexual violence perpetrated by rebel groups. Moreover, we find that actors with a high level of internal control are less likely to perpetrate sexual violence as the number of peacekeepers increase. This applies to both states and rebel groups. The findings suggest that peacekeeping can be a useful tool for protection also against sexual violence, although only under particular circumstances.

To position our study, we begin this article with a review of contemporary research on sexual violence in conflict. Thereafter, we outline our theoretical foundation from which we draw testable hypotheses about the conditions under which peacekeeping successfully reduce sexual violence. We explain how we test these hypotheses in the Research Design section, after which we present our results.

Conflict-related Sexual Violence

Policy makers tend to describe conflict-related sexual violence as a weapon of war used to advance military goals and punish the population (Eriksson Baaz and Stern 2013).¹ Several scholars have also emphasized these strategic aims of systematic sexual violence against women used to destroy the cohesion of the enemy’s community (Seifert 1996) or to devastate their morale (Sharlach 2000). In some conflicts, sexual violence appears to shift according to the power dynamics between the warring parties, indicating that sexual violence by states somehow compensates for military weakness in relation to its opponent (Leiby 2009). Complementing, and to some extent in contrast to, this view stands a growing literature on the causes of sexual violence that indicates that it can serve several purposes beyond purely strategic. Explanations are instead sought in the constitution and internal structures of armed groups. It is argued that sexual violence is more common in military units with low social cohesion where gang rape and other sexual offenses are used for socialization and hazing purposes (Cohen 2013; Cohen and Nordås 2015). From the perspective of restraint, Wood (2009) and Hoover Green (2011) find the prevalence of sexual violence to be associated with the ability and will of the leadership to constrain the extent to which a conflict environment enables members of armed groups to commit sexual violence. This is also consistent with Muvumba Sellström’s (2015) research in which she claims that sexual violence is more likely in rebel groups lacking explicit and consistently enforced prohibition against sexual violence.

Common for these strands of research is the explicit accountability placed on the political and/or military leadership. There is no evidence suggesting that rape and sexual violence is an inevitable consequence of war or that it happens at random. Instead, there is a wide variation across conflicts and armed groups, which demonstrate the feasibility to curb it (Cohen, Hoover Green, and Wood 2013). Among all conflicts between 1989 and 2009, there is evidence of documented sexual violence in 57 percent. The variation in prevalence between these cases is large, ranging from isolated reports of sexual violence such as in the Philippines in the early 1990s to accounts of massive sexual violence as in Burundi between 2003 and 2005 (Cohen and Nordås 2014).

The responsibility to take action against sexual violence does however not end with the warring parties. As a consequence of the parallel, normative developments of the Responsibility to Protect (R2P) on the one hand and the Women, Peace, and Security (WPS) agenda on the other, wartime sexual violence has become a matter of increasing international concern. Since the UN resolution 1325 on WPS in 2000, four resolutions have been dedicated to prevention and protection from sexual and/or gender-based violence (Tryggestad 2015; Kirby and Shepherd 2016). Conflict-related sexual violence commonly also takes a prominent role in the country reports on WPS submitted to the UN Security Council (Kirby and Shepherd 2016). The increased awareness of conflict-related sexual violence is also reflected in UN peacekeeping. Indeed, the Department of Peacekeeping Operations (DPKO) was one of the first UN bodies to implement changes according to 1325 by ensuring gender advisors being part of every mission since the year 2000. Policy directives have been adopted in 2006 and 2010, which include clauses on safety in relation to gender-based violence (UN DPKO 2006; UN DPKO/DFS 2010). In 2009, the UN Security Council mandated the deployment of women’s protection advisors to conflict areas with reported problems of sexual violence (UN DPKO/Department of Field Support [DFS] 2014a). Until recently, however, no study has empirically examined the adequacy, let alone effectiveness, of peacekeeping in actually reducing sexual violence (for an exception, see ongoing work by Sarwari 2016). Before taking on this task, we outline our theoretical basis from which we draw testable implications.

Peacekeeping and Sexual Violence

A year after the adoption of Resolution 1325, Skjelsbaek (2001) concluded that sexual violence constitutes a particular challenge to peace operations. Nevertheless, we still know little about the impact of peacekeeping for reducing the levels of sexual violence and under what circumstances peacekeepers are more likely to successfully achieve such results. The mass rape in Walikale in the DRC demonstrates the difficulties of protecting civilians from sexual violence. In this event, armed actors raped around 300 civilians over the course of just a few days in 2011. Close to this village, there was a UN base, with peacekeepers that also patrolled the area. Despite this, they did not manage to detect and stop the ongoing mass rape. Arguably, when sexual violence is nonlethal, it may be particularly difficult to detect since it does not leave any bodies to be found. On the other hand, one could expect more reports of nonlethal violence as not only witnesses but also survivors potentially could raise awareness about the situation (Chu and Braithwaite 2018, 6). However, victims may be reluctant to report such abuse to peacekeepers in time for them to effectively stop it—either because of feared stigmatization or because civilians are afraid the warring actors will punish them. This indicates that peacekeeping, in general, may not be an effective instrument for addressing sexual violence. In this study, we are interested in identifying the circumstances under which peacekeeping may in fact work to reduce the risk of sexual violence.

The focus in this article is on conflict-related sexual violence by organized conflict actors. When theorizing about the effect of peacekeeping on sexual violence, we rely on the judicial definition stipulated by the Rome Statute of the International Criminal Court (ICC 2011).² This is a broad definition of sexual violence that is widely used in research interested in the entire spectrum of conflict-related sexual offenses. According to this definition, sexual violence is not limited to rape but includes acts of sexual slavery, enforced prostitution, forced pregnancy, and enforced sterilization (ICC 2011, Article 7 (1) (g)-1–6). Conceptually, we follow the example by Wood (2009, 133) which also includes sexual torture and mutilation.

We acknowledge that sexual violence may follow different logics—including both opportunistic violence, sexual violence as a practice within military units, and the strategic use of sexual violence. Since these theoretical perspectives and suggested mechanisms of sexual violence cannot be empirically observed in our study, we theorize about the general impact of peacekeeping, assuming that sexual violence may be carried out for a range of reasons. We are interested in the potentially curbing effect of peacekeeping, regardless of motivation, and suggest that peacekeepers can protect civilians from sexual violence by (1) increasing the expected risk for the individual perpetrators and/or (2) increasing the expected costs for the organized group/government. For the individual-level mechanism to work, missions need to be large and have the mandate to prioritize protection. Therefore, we would expect the efficiency of peacekeeping to be contingent on the size of the deployment and whether or not it is explicitly mandated to protect civilians. For the actor-level mechanism to work, perpetrators of sexual violence must act as unitary actors. Therefore, we would expect peacekeeping deployments to be more successful when facing actors that exercise a high level of internal control over their own forces. Below we develop these arguments.

Prioritizing Protection—The Role of Size and Mandate

Deployment size matters for any peacekeeping mission—not the least as a signal of political commitment. Large missions signal credible commitment to the task on behalf of the UN and its troop-contributing countries. Higher degree of involvement increases the costs of potential failure, making it all the more essential for the mission to succeed in ending ongoing atrocities (Hultman, Kathman, and Shannon 2013). Larger commitments from third parties such as the UN also alleviate warring parties’ security dilemma to a larger extent. This may be particularly true for the military forces within peacekeeping operations. By physically separating warring parties, military peacekeepers can facilitate the demobilization process whereby conflict actors may feel safer to loosen the grip over specific populations. To the extent that rape and other sexual abuses is an expression of power demonstrations between the warring parties or threats to deter side switching of civilians, we should thus observe a decline in sexual violence the larger mission size we observe. On the argument’s flip side, less ambitious or poorly fulfilled mandates instead indicate ambiguous engagement with questionable prospects of increased involvement, should the need arise. This gives missions less leverage to impose strategic costs on the warring parties in case of ceasefire violations and bolsters a sense of uncertainty for the warring parties. At worst, this may trigger more sexual violence as it could heighten the security dilemma and make actors more inclined to forcefully tie civilians to their group.

Deployment size also matters for more fundamental reasons. While large deployments always have a numeric advantage to smaller deployments, the difference in impact may be more significant for sexual violence than other types of violence. When armed actors rape individuals in order to spread fear among larger masses of people, it may very well be carried out in public spaces in order for as many civilians as possible to observe. Often, however, combatants commit sexual violence in private spaces such as the home or on the field. To address sexual violence of this kind requires peacekeepers to go beyond the conventional “battlefield” into private spaces which usually fall outside of peacekeepers’ reach (UN Women 2012a, 10). This is resource demanding and necessitates large mission sizes in order to reach out across territories and populations. In particular, it may require larger military and police units that are able to patrol among both armed actors and civilians.

The functions performed by peacekeepers discussed above are partly dictated by the mandate authorized by the UNSC. Therefore, we also consider the possibility that peacekeeping primarily is an effective way of reducing sexual violence when the mission has a strong mandate to focus on the protection of the civilian population. In general, third-party interventions risk stirring up the power balance between warring parties, which triggers behavior resulting in more exploitation of civilians (Wood, Kathman, and Gent 2012; Johansson and Sarwari 2017). Sexual violence may be particularly sensitive to this, given that rape and other sexual offenses often are unregulated in ceasefire and peace agreements (UN Women 2012b). This should make the cost of continued sexual violence lower compared to the potential costs of continued killing, thereby lowering the incentives for armed groups to take action against sexual offenses by their combatants. Yet while we acknowledge that sexual violence for these reasons pose particular challenges to any mission, missions with a protection of civilians (PoC) mandate should have a greater focus on the interaction between the warring actors and the civilian population. These missions should prioritize tasks relating to abuse of the civilian population and their personnel are more likely to have received training focusing on protection. The UN police units that operate under a protection mandate are, as an illustration, supposed to pay particular attention to the needs of vulnerable groups. This has been increasingly emphasized over the years and is specifically clear in a central policy document from 2014 where it is stated that UN police “shall prioritize supporting the host State in the prevention, investigation and prosecution of sexual and gender-based violence” (DPKO/DFS 2014b, 8, emphasis added).

In addition, PoC mandates almost always come with the authorization of robust use of force. This means that peacekeepers in these missions have the ability to use force at the tactical level for the purpose of protecting civilians. That should enhance the peacekeepers’ deterrent effect. For any armed actor in the conflict environment, the threat of peacekeepers responding with the use of force should increase the expected costs of engaging in sexual violence. Even smaller military units should perceive the peacekeepers as a credible threat when missions prioritize protection and operate in areas where sexual violence is perpetrated by one or several warring parties. While a mandate is not a guarantee for such operations, a protection mandate in combination with a sizable force should at least be more likely to succeed in deterring sexual violence. We thus expect that an increase in the number of peacekeepers with a protection mandate should be associated with a reduction in sexual violence by the armed actors. This is expressed in our first hypothesis:

Hypothesis 1: The more peacekeepers with a mandate to protect civilians, the lower the probability of sexual violence by warring parties.

Ability to Restrain: The Role of Internal Control within Armed Groups

Until now, we have discussed the effect of peacekeeping as uniformly affecting all conflict actors equally. However, we know that warring parties have different levels of organization and control, which influence the way they respond to peacekeeping.

Recent scholarship on sexual violence emphasizes nonstrategic explanations, suggesting that sexual violence is a practice that may serve other functions, such as socialization in heterogeneous fighting units (Cohen 2013) or simply be tolerated within military organizations for other reasons not directly related to the war aims (Muvumba Sellström 2015). Furthermore, it is argued that norms within military groups shape the likelihood of sexual violence (Wood 2009) and that armed actors that uphold control through political education are less likely to engage in rape (Hoover Green 2016). In line with this, Eriksson Baaz and Stern (2009) emphasize poor organization as a driver of looting and sexual violence. That said, control over military forces could result in both high and low levels of violence, depending on the strategic goals of the leadership (cf. DeMeritt 2015). As a general principle, control and internal functioning institutions are keys to manage commanders’ dilemma of having to promote effective fighters willing to use violence and at the same time maintaining the ability to exercise restraint over unwanted abuse (Hoover Green 2016). For understanding the violence following this dilemma, one must consider both incentives and constraints (Wood 2014). What is common for actors with a high level of control, however, is the ability of the principal to convince the agents to follow orders—regardless of whether those are orders of violence or restraint.

We argue that the level of control an actor exerts over its armed forces should affect the degree to which peacekeepers can contribute to the reduction of sexual violence. While parties that have strong central command and act as unitary actors may sometimes engage in sexual violence for strategic or instrumental purposes, they are also able to restrict such behavior within their forces when faced with peacekeepers. With these groups, peacekeepers should be able to impose costs at the strategic level and deter the use of sexual violence. In contrast, actors with weak control are more difficult to influence as it requires much more resources to engage with the behavior of individual fighters. The actors are thus not deterred at the group level, but peacekeepers need to engage with the rank and file in order to protect civilians. Hence, the mechanisms through which we hypothesize that peacekeeping works are less effective against weak state armies or rebel groups with weak internal discipline. We therefore expect peacekeeping to be primarily effective against actors with a high level of control. The observable implication is expressed in our second hypothesis.

Hypothesis 2: Given that actors have control over their armed forces: the more peacekeepers that are deployed to a peacekeeping mission, the lower the probability of sexual violence.

Research Design

We examine the relationship between peacekeeping and sexual violence perpetrated by all governments and rebel groups involved in internal armed conflicts³ between 1989 and 2009. We observe these conflict actors annually for the duration of the active conflict plus up to five interim years if the conflict is temporarily inactive. We also include five postconflict years. In total, the dataset covers 120 conflicts, involving 72 governments and 274 rebel groups. Our unit of analysis is conflict-actor-year, allowing for time series cross-section analyses of 1,429 state observations and 2,303 rebel observations. This setup allows us to evaluate the impact of peacekeepers with no peacekeepers as the point of reference. However, as a robustness check, we also restrict the sample to only cases of ongoing peacekeeping operations.

For the outcome of interest, we use data on the prevalence of sexual violence from the SVAC dataset (Cohen and Nordås 2014). The dataset contains information on the levels of sexual violence per conflict-actor-year as reported in three separate sources: the US State Department Reports, Human Rights Watch, and Amnesty International. We collapse the three sources and use the maximum value for each observation. The original variable ranges from 0 to 3. A value of 0 represents “no reports” of sexual violence, 1 represents “isolated reports” of sexual violence, 2 represents reports of “widespread” sexual violence, and 3 represents reports of sexual violence as “massive.” Our dependent variable, sexual violence, is a dummy variable measuring whether there were reports of high levels of sexual violence (levels 2 and 3 in the SVAC dataset) compared to no or only isolated reports of sexual violence. By coarsening the data into a dichotomous measure, we reduce the problem with underreporting bias. Our positive values thus only reflect instances with widespread or massive levels of sexual violence, which are at least less likely to be unnoticed. The zeros should thus not be interpreted as no instances of sexual violence but rather comparatively lower levels of reported sexual violence.

For our main independent variable, we use peacekeeping data from Kathman (2013). The original dataset contains monthly counts of the number of UN personnel deployed in each mission. We use the maximum number of peacekeepers deployed at any point in time during the year. In cases where there are multiple peacekeeping operations, such as in Sudan, the total number of personnel for the missions is taken. While theoretically we only refer to peacekeepers as one category, we choose to empirically discriminate between military forces and police, considering their different tasks that could influence their ability to address sexual violence. Our main independent variables are thus UN troops and UN police—both counting the number of each type deployed to a country.

To assess Hypothesis 1 about the impact of peacekeepers in protection missions specifically, we include a variable whether missions have an explicit mandate to protect civilians from the UN Security Council. This is coded based on the language in the resolutions providing the mandate (typically formulated as a task to protect civilians under imminent threat of physical violence), and we account for the fact that these mandates may change over the course of a mission. This variable is then interacted with the number of peacekeepers (troops or police), which enables us to evaluate the interaction effect suggested by the hypothesis. For ease of interpretation of the interaction effect, we interact the number of peacekeepers with a dummy that is coded 0 when there is a protection mandate and 1 for no protection mandate.

To assess Hypothesis 2 about the degree of control exercised by the warring party, we use two different proxies for states and rebels. For states, we code a dummy for whether a state exercises control over its territory, using the variable state authority over territory taken from V-Dem version 6.1 (Coppedge et al. 2016). This provides us with a proxy for actors that exercise control over their territory, which indicates the capacity and discipline of the armed forces (since states with a weak army more easily lose control over territory during civil wars). To generate the dummy, we use the median value in the sample, which is 86.2.⁴ For rebels, we identify groups with a strong central command based on the Non-state Actor Dataset (Cunningham, Gleditsch, and Salehyan 2009). This provides us with measure for rebel actors who have the organization to control and discipline the rank and file. Both of these dummy variables are coded as 0 for control and 1 for no control, which in turn interacted with the number of peacekeepers (troops or police) as a way of evaluating the interaction effect suggested by the hypothesis.

We include a number of control variables that may influence the deployment of peacekeepers or the occurrence of sexual violence. Starting with the overall conflict environment, we account for battle deaths as this has been found to correlate with both peacekeeping interventions (Gilligan and Stedman 2003) and sexual violence (Cohen 2013, 471-72).⁵ Using data from the UCDP Battle-related Deaths Dataset (2015), we include a measure of the annual number of battle deaths per conflict, starting at twenty-five deaths. Since the number of civilian casualties may also be correlated with both peacekeeping deployment (Hultman 2013) and sexual violence, we control separately for one-sided violence, which is a measure of the number of civilians killed in direct and deliberate targeting provided by the UCDP One-sided Violence Dataset (Eck and Hultman 2007). Based on evidence that sexual violence at times appears to compensate for military weakness (Johansson and Sarwari 2017; Leiby 2009) and because the UN tend to be more inclined to intervene in conflicts where the rebel groups are strong (Fortna 2008), we account for the relative strength of each rebel group compared to the government.⁶ This variable is an ordinal variable ranging from 1 (much weaker) to 5 (much stronger), taken from the Non-state Actor Dataset (Cunningham, Gleditsch, and Salehyan 2009). For years without information, we use the previous year’s estimate. To account for whether the actor is already perpetrating sexual violence in the conflict, we include one measure indicating the level of sexual violence in the previous year. This variable is thus not a strict lagged dependent variable but reflects the ordinal scale of the original SVAC data.

Next, we include a few peacekeeping-related control variables. First, when testing Hypothesis 2, we control for whether the mission has a PoC mandate. Expecting that UN peacekeeping operations also improve their knowledge and effectiveness over time through learning (Howard 2008), we add a mission duration variable counting the number of years since its initial deployment. We also include a dummy variable denoting the period post the United Nations Security Council Resolution 1325 (post UNSCR 1325), since this resolution partly aimed to highlight problems with sexual violence and put them on the UN agenda. This resolution could influence the way in which the UN addresses sexual violence, thereby changing the relationship between peacekeepers and sexual violence.

Proceeding to characteristics of the country in conflict, it has been argued that democracies are less inclined to abuse their citizens due to, for example, institutional constraints and audience costs (e.g., Butler, Gluch, and Mitchell 2007; Harff 2003; Valentino, Huth, and Balch-Lindsay 2004). This can have a reversed impact on violence by rebel groups (Stanton 2013). We account for democracy level by including the electoral democracy index from V-Dem version 6.1 (Coppedge et al. 2016). The index ranges from 0 to 1 and corresponds to the question: “To what extent is the ideal of electoral democracy in its fullest sense achieved?” (Coppedge et al. 2016, 44). Another characteristic that may influence the dynamics of sexual violence is the general level of gender equality in the country (Davies and True 2015). As a proxy for this, we control for fertility rate, using data from V-Dem. The peacekeeping literature has shown that peacekeeping missions are less likely to be deployed in large countries (Gilligan and Stedman 2003). Large-scale sexual violence, like other forms of political violence, may also be more likely the larger the population. Therefore, we include the natural log of the size of the population, with data provided by the UN (2015). Lastly, we try to account for the problem of reporting bias, and the effect that might have both for the likelihood of a conflict getting enough attention to attract a UN peacekeeping mission and the numbers of sexual violence reported in international media. For this purpose, we include free media, which is an ordinal variable from V-Dem, ranging from 0 (not free) to 2 (free). Summary statistics for all variables are reported in the Online Appendix (Tables A1 and A2).

To conduct our analysis, we use logistic regression analyses. We cluster the standard errors on the conflict. While selection bias is always a concern when estimating the impact of peacekeeping, we do not find any strong evidence that nonrandom deployment is a problem for this study. In the Online Appendix (Table A7), we report results from a recursive bivariate probit, where we use UNSC membership as an instrument for peacekeeping deployment (Vivalt 2017).⁷ While the instrument is significant as expected, ρ is not significant, indicating that selection bias is not necessarily a problem. We also note that peacekeeping is more likely in cases with high levels of sexual violence, which would bias against finding support for our hypotheses (Tables A3 and A4 in the Online Appendix).

Empirical Analyses

Despite the general picture often conveyed by international media outlets, a large share of all conflicts do not generate reports of sexual violence, and to the extent they do, the majority of these reports concern isolated events. In our dataset, there are 558 observations (15 percent of all observations in the dataset) with reports of sexual violence. One hundred and eighty-five of these refer to accounts of widespread or massive sexual violence, which we use as our dependent variable in this study. In Tables 1 and 2, we show cross tabulations of UN peacekeeping presence and high levels of sexual violence by state forces and rebel groups, respectively. When there is no peacekeeping mission present in the country, sexual violence by states occurs in 5 percent compared to 6 percent where there is a peacekeeping mission. We note reports of sexual violence perpetrated by rebel groups in 3 percent of the cases without any peacekeepers and in 5 percent of the cases where peacekeepers were present. Taken at face value, these numbers could indicate that peacekeeping operations on average are unable to reduce sexual violence by both state forces and rebel groups. We can however not know to what extent this indeed is the case or if the numbers rather tell us that sexual violence attracts peacekeepers. After all, there is some, though ambiguous, empirical evidence that reports of sexual violence increase the likelihood of peacekeeping deployment (Hultman and Johansson 2017). Moreover, we do not know if the impact of peacekeeping operations depends on their size and mandate or rather on the type of actors they face. These are questions, in addition to the question of temporal order, we deal with in the ensuing analysis.

Table 1.

UN Peacekeeping and Wartime Sexual Violence by States.

Sexual violence prevalence	No UN PK	UN PK	Total
No sexual violence	960 (95 percent)	395 (94 percent)	1,355
Sexual violence	46 (5 percent)	26 (6 percent)	72
Total	1,006 (100 percent)	421 (100 percent)	1,427
Pearson χ²(1) = 1.5923, Pr = .207

Note: UN PK = United Nations Peacekeeping.

Table 2.

UN Peacekeeping and Wartime Sexual Violence by Rebels.

Sexual violence prevalence	No UN PK	UN PK	Total
No sexual violence	1,514 (97 percent)	712 (95 percent)	2,226
Sexual violence	41 (3 percent)	34 (5 percent)	75
Total	1,555 (100 percent)	746 (100 percent)	2,301
Pearson χ²(1) = 5.8999, Pr = .015

Note: UN PK = United Nations Peacekeeping.

We begin to evaluate the first hypothesis suggesting that the number of peacekeepers should be associated with a lower risk of sexual violence but conditioned by a PoC mandate. To allow for potentially varying influence on state-perpetrated violence compared to violence perpetrated by rebel groups, we study the two actor types separately. Table 3 reports the findings looking at state actors and Table 4 reports the findings for rebels. We stratify the number of peacekeepers according to whether they belong to the military or the police contingency. In so doing, we are able to tease out potentially different effects of their respective presence. For each of these tables that focus on states and rebels, respectively, we report three models: (1) the independent effect of PoC mandate, (2) the number of troops interacted with PoC mandate, and (3) the number of police interacted with PoC mandate. Starting with state actors, these findings provide little support for the first hypothesis. The number of peacekeepers does not reduce the risk of sexual violence—even if they have a protection mandate. Turning to rebel actors in Table 4, we find similarly discouraging results in the first two models. Peacekeepers do not appear to have an influence on the likelihood of sexual violence by rebel groups. Nevertheless, when we specifically study the effect of protection-mandated police (Model 3), we find a negative and significant effect based on both the standalone and the interaction term. This suggests that the UN police manage to address sexual violence, by insurgents, when adequately mandated by the UNSC.

Table 3.

The Impact of UN Peacekeeping Size and Mandate on Sexual Violence by State Forces.

	Model 1: Effect of Mandate	Model 2: Interacting Mandate and Troops	Model 3: Interacting Mandate and Police
No PoC mandate (0 = PoC)	−0.789	0.0469	−0.549
No PoC mandate (0 = PoC)	(0.642)	(1.006)	(0.690)
UN troops, t − 1	0.0139	0.0874	0.0161
UN troops, t − 1	(0.0925)	(0.131)	(0.0929)
UN police, t − 1	0.108	0.0447	0.377
UN police, t − 1	(0.389)	(0.504)	(0.440)
UN troops × No PoC mandate		−0.477
UN troops × No PoC mandate		(0.412)
UN police × No PoC mandate			−3.684
UN police × No PoC mandate			(5.566)
Free media	−0.210	−0.245	−0.224
Free media	(0.374)	(0.374)	(0.371)
Gender inequality	0.245*	0.250*	0.246*
Gender inequality	(0.0890)	(0.0896)	(0.0884)
Post–UNSCR 1325	0.636*	0.643*	0.604*
Post–UNSCR 1325	(0.273)	(0.287)	(0.273)
Mission duration	−0.0232	0.00511	−0.0131
Mission duration	(0.0414)	(0.0512)	(0.0452)
Battle deaths, t − 1	0.206*	0.191*	0.203*
Battle deaths, t − 1	(0.0632)	(0.0654)	(0.0639)
One-sided violence, t − 1	0.243**	0.254**	0.240**
One-sided violence, t − 1	(0.0488)	(0.0456)	(0.0484)
Population	0.0894	0.0836	0.0744
Population	(0.140)	(0.138)	(0.140)
Democracy	−0.499	−0.699	−0.527
Democracy	(1.105)	(1.164)	(1.123)
Rebel strength	−0.0390	0.00865	−0.0356
Rebel strength	(0.219)	(0.258)	(0.222)
Sexual violence, t − 1	0.836**	0.848**	0.843**
Sexual violence, t − 1	(0.183)	(0.192)	(0.183)
Constant	−6.932*	−7.650*	−6.871*
Constant	(2.971)	(3.137)	(2.950)
Observations	1,263	1,263	1,263

Note: Standard errors are in parentheses, clustered on conflict. UN = United Nations; PoC = protection of civilians.

^† p < .10.

*p < .05.

**p < .001.

Table 4.

The Impact of UN Peacekeeping Size and Mandate on Sexual Violence by Rebel Groups.

	Model 1: Effect of Mandate	Model 2: Interacting Mandate and Troops	Model 3: Interacting Mandate and Police
No PoC mandate (0 = PoC)	−0.812	−1.154	−3.025**
No PoC mandate (0 = PoC)	(1.122)	(1.274)	(0.610)
UN troops, t – 1	−0.0767	−0.117	−0.0125
UN troops, t – 1	(0.111)	(0.150)	(0.0397)
UN police, t – 1	0.542	0.444	−33.59**
UN police, t – 1	(0.685)	(0.518)	(9.977)
UN troops × No PoC mandate		0.0975
UN troops × No PoC mandate		(0.152)
UN police × No PoC mandate			34.33**
UN police × No PoC mandate			(9.989)
Free media	−0.0905	−0.0910	−0.147
Free media	(0.363)	(0.363)	(0.350)
Gender inequality	0.353*	0.350*	0.351*
Gender inequality	(0.154)	(0.153)	(0.157)
Post–UNSCR 1325	0.0585	0.0731	0.0330
Post–UNSCR 1325	(0.344)	(0.342)	(0.334)
Mission duration	0.0207	0.0155	0.00118
Mission duration	(0.0698)	(0.0668)	(0.0678)
Battle deaths, t − 1	0.0803	0.0789	0.0802
Battle deaths, t − 1	(0.0610)	(0.0601)	(0.0649)
One-sided violence, t − 1	0.294**	0.289**	0.307**
One-sided violence, t − 1	(0.0578)	(0.0579)	(0.0637)
Population	−0.481*	−0.478*	−0.499*
Population	(0.155)	(0.156)	(0.161)
Democracy	−0.276	−0.177	0.0498
Democracy	(0.971)	(0.968)	(1.058)
Rebel strength	0.127	0.131	0.0000637
Rebel strength	(0.196)	(0.196)	(0.212)
Sexual violence, t − 1	1.277**	1.300**	1.321**
Sexual violence, t − 1	(0.155)	(0.158)	(0.179)
Constant	1.591	1.845	4.255
Constant	(3.294)	(3.271)	(3.394)
Observations	1,981	1,981	1,981

Note: Standard errors are in parentheses, clustered on conflict. UN = United Nations; PoC = protection of civilians.

^† p < .10.

*p < .05.

**p < .001.

The substantive impact of police is also significant. In a high-risk scenario with a PoC mandate, post–UNSCR 1325, with a high level of sexual violence in the preceding year, increasing the number of police from 0 to 1,000 reduces the risk of sexual violence from 83 percent to close to 0 percent.⁸ This can be compared to a situation where the mission lacks a protection mandate: then the risk of sexual violence in fact increases from 23 percent to 38 percent in the same scenario of increasing the number of police.

Next, we take a closer look at the empirics behind the negative association between protection-mandated police forces and sexual violence. Part of the story are conflicts such as the civil war in Ivory Coast to which more than 1,000 police forces were sent but only isolated reports of sexual violence were issued throughout the conflict. While it is difficult to assess the counterfactual in this case, that is, to what extent the presence of police forces had a preventative effect on the actual prevalence of sexual violence, there are a number of other cases where initially high prevalence levels of sexual violence are reduced as the number of deployed police increases. One example is the conflict in Haiti. There were isolated but persistent reports of sexual violence that plagued the Haitian conflict every year since 1991, primarily perpetrated by nonstate forces under Raol Cedras (Nordås and Cohen 2013). From the second year onward of the UN police deployment in 1994, this type of violence appears to have stopped.⁹ While this mission did not get its protection mandate until 2004, “gender training” of domestic police forces has been part of the mission mandate since its inception and joint UN/Haitian camp patrols have been mentioned as useful tools to curb and prevent further sexual abuse (UN Women 2012a, 29).

Another case that stands out in our data is the conflict in Darfur. Starting with an initial deployment of slightly less than 300 police forces, the UN eventually deployed more than 4,500 police officers to Sudan. During the same time, the prevalence of reported sexual violence dropped for a number of domestic armed groups, such as JEM, Janjaweed, SLM/A, and SLM/A MM. Notably, sexual violence by the state forces remained a severe problem (Nordås and Cohen 2013). While this is consistent with our quantitative results for state actors, this calls for in-depth analyses of why the UN police efforts appear to have left the behavior of the Sudanese state forces unchanged.

The analyses thus far have focused on the role of the mandate, implicitly assuming that the effect of these missions is the same across different types of actors, regardless of their organization or level of control. Nevertheless, as hypothesized, the degree to which an actor exerts internal control may determine the effectiveness of peacekeeping. In particular, we expect that parties who exert high levels of control are more capable of acting as unitary actors and enforce compliance within the rank and file in response to peacekeepers. Therefore, we now move on to evaluate the second hypothesis. Tables 5 and 6 focus on states and rebels, respectively, and for each of these we report three models: (1) the independent effect of control, (2) the number of troops interacted with control, and (3) the number of police interacted with control. States with a low level of control are less likely to perpetrate sexual violence, indicating that this is not just an artifact of ill-disciplined forces. When interacting control with peacekeepers, we find that for actors with a high level of control, increasing the number of peacekeepers does reduce the likelihood of sexual violence. The effect is significant at 90 percent in Model 2 and at 95 percent in Model 3. The substantive effect further supports the argument. Increasing the troop presence from 0 to 10,000 peacekeepers in a high-risk scenario with high level of control reduces the risk of sexual violence from 45 percent to 5 percent. The impact of peacekeepers on the risk of sexual violence is even stronger for police: when increasing the number of police in the same scenario from 0 to 1,000, the risk is reduced from 55 percent to below 1 percent. The UN police forces are commonly mandated to assist in law enforcement and crime prevention, and in many missions, they engage in active training of the local police forces and even collaborate by conducting joint patrols as in Haiti. Perhaps such close interaction with the government works to reduce sexual violence, both through monitoring and through training, in particular for government actors with a high level of control who are then able to restrain the behavior of the rank and file in response to high political costs that come with close monitoring by UN peacekeepers.

Table 5.

The Impact of Control on Sexual Violence by State Forces.

	Model 1: Effect of Control	Model 2: Interacting Control and Troops	Model 3: Interacting Control and Police
Weak control (0 = control)	−0.293	−0.555^†	−0.531
Weak control (0 = control)	(0.317)	(0.324)	(0.345)
UN troops, t − 1	0.0144	−0.399^†	0.0166
UN troops, t − 1	(0.0944)	(0.218)	(0.0892)
UN police, t − 1	0.142	−0.0927	−36.96*
UN police, t − 1	(0.412)	(0.356)	(12.97)
UN troops × Weak control		0.435*
UN troops × Weak control		(0.218)
UN police × Weak control			36.91*
UN police × Weak control			(12.84)
No PoC mandate, t − 1	−0.887	−1.035	−1.379*
No PoC mandate, t − 1	(0.656)	(0.631)	(0.663)
Free media	−0.241	−0.326	−0.297
Free media	(0.389)	(0.410)	(0.395)
Gender inequality	0.244*	0.260*	0.247*
Gender inequality	(0.0870)	(0.0867)	(0.0857)
Post–UNSCR 1325	0.618*	0.636*	0.551*
Post–UNSCR 1325	(0.273)	(0.289)	(0.274)
Mission duration	−0.0258	−0.0105	−0.0236
Mission duration	(0.0421)	(0.0420)	(0.0439)
Battle deaths, t – 1	0.216*	0.215*	0.224*
Battle deaths, t – 1	(0.0706)	(0.0702)	(0.0712)
One-sided violence, t − 1	0.245**	0.265**	0.262**
One-sided violence, t − 1	(0.0486)	(0.0428)	(0.0427)
Population	0.0743	0.0505	0.0211
Population	(0.144)	(0.141)	(0.143)
Democracy	−0.628	−0.641	−0.466
Democracy	(1.138)	(1.148)	(1.134)
Rebel strength	−0.0452	−0.0622	−0.0638
Rebel strength	(0.221)	(0.223)	(0.224)
Sexual violence, t − 1	0.819**	0.849**	0.886**
Sexual violence, t − 1	(0.187)	(0.198)	(0.192)
Constant	−6.370*	−5.748^†	−4.949
Constant	(3.162)	(3.176)	(3.255)
Observations	1,263	1,263	1,263

Note: Standard errors are in parentheses, clustered on conflict. UN = United Nations; PoC = protection of civilians.

^† p < .10.

*p < .05.

**p < .001.

Table 6.

The Impact of Control on Sexual Violence by Rebel Groups.

	Model 1: Effect of Control	Model 2: Interacting Control and Troops	Model 3: Interacting Control and Police
Weak control (0 = control)	−1.158**	−1.477**	−1.316**
Weak control (0 = control)	(0.349)	(0.378)	(0.378)
UN troops, t − 1	−0.0697	−0.302*	−0.0560
UN troops, t − 1	(0.101)	(0.105)	(0.0810)
UN police, t − 1	0.511	0.394	−2.052
UN police, t − 1	(0.840)	(0.463)	(2.123)
UN troops × Weak control		0.320**
UN troops × Weak control		(0.0867)
UN police × Weak control			2.811
UN police × Weak control			(2.006)
PoC mandate, t − 1	−1.060	−0.933	−1.070
PoC mandate, t − 1	(1.024)	(1.179)	(0.926)
Free media	−0.272	−0.407^†	−0.347
Free media	(0.286)	(0.240)	(0.257)
Gender inequality	0.322*	0.328*	0.325*
Gender inequality	(0.125)	(0.118)	(0.121)
Post–UNSCR 1325	0.0505	−0.0534	0.00864
Post–UNSCR 1325	(0.408)	(0.468)	(0.432)
Mission duration	0.00328	0.0328	0.0250
Mission duration	(0.0660)	(0.0647)	(0.0598)
Battle deaths, t − 1	0.0939	0.0570	0.0839
Battle deaths, t − 1	(0.0676)	(0.0741)	(0.0702)
One-sided violence, t − 1	0.321**	0.330**	0.319**
One-sided violence, t − 1	(0.0581)	(0.0572)	(0.0560)
Population	−0.697**	−0.555**	−0.667**
Population	(0.166)	(0.165)	(0.158)
Democracy	−0.417	−0.523	−0.238
Democracy	(1.364)	(1.277)	(1.360)
Rebel strength	−0.0743	−0.0340	−0.0872
Rebel strength	(0.198)	(0.175)	(0.202)
Sexual violence, t − 1	1.193**	1.211**	1.206**
Sexual violence, t − 1	(0.149)	(0.156)	(0.149)
Constant	6.648*	4.521	6.294^†
Constant	(3.358)	(3.119)	(3.348)
Observations	1,820	1,820	1,820

Note: Standard errors are in parentheses, clustered on conflict. UN = United Nations; PoC = protection of civilians.

^† p < .10.

*p < .05.

**p < .001.

In Table 6, we report the same models but for rebel groups. Model 1 shows that, similar to state forces, rebel groups with weak control are less likely to perpetrate sexual violence. When interacting control with the number of peacekeepers, we find that troops reduce the risk of sexual violence by rebels when rebels have a functioning control structure. Police, on the other hand, do not have a significant effect. Hence, rebels may require stronger enforcement mechanisms through military means in order to change their behavior, especially since these groups with high level of control are in fact more likely to perpetrate sexual violence to begin with. It is also plausible that the results evince that the UN police are less likely to engage in joint training and patrols with rebel groups. Looking at the substantive effect, the risk of sexual violence in a high-risk scenario reduces from 54 percent to 7 percent when increasing the number of military troops from 0 to 10,000 (compared to a null effect for groups with a low degree of control).

Figure 1 displays the marginal effect of rebel control on the risk of high levels of sexual violence when increasing the number of peacekeeping troops. It shows that the impact of troops is significantly different when facing groups with control compared to groups without control at least up to almost 15,000 troops. If deployments are large enough, that difference is no longer as apparent. Since most peacekeeping missions deploy well below 15,000 troops, this means that in reality, the internal control of rebel groups is an important factor conditioning the violence-reducing effect of peacekeepers.

Figure 1.

The marginal effect of rebel control when increasing the number of troops.

The findings are supportive of the argument and suggest that peacekeeping indeed works to reduce sexual violence when facing actors with high levels of control. According to our findings, rebels who have a high level of control are in fact also more likely to commit sexual violence in the first place. There are a significant number of cases of high control and high levels of sexual violence in our sample. Among governments, widespread or massive sexual violence occurs is 3.2 percent of cases with high control compared to 7.2 percent in cases with low control. Among rebels, we see widespread or massive sexual violence in 6.2 percent of cases with high control compared to 2.4 percent for actors with weak control. Hence, the fact that peacekeeping may work to reduce sexual violence in these instances is of practical relevance. Control could potentially also influence where peacekeepers are deployed. Descriptive statistics show that peacekeeping is more common where governments are in control but where rebels have weak control (Online Appendix Tables A5 and A6). However, the results from our bivariate probit (Online Appendix Table A7) suggest that control does not significantly influence the likelihood of peacekeeping deployment.

Robustness Checks

Our analyses have shown that peacekeeping works to reduce the risk of sexual violence under certain circumstances. Two main findings stand out. First, while peacekeepers on average are not effective at reducing sexual violence by governments, the UN police with a protection mandate does seem to reduce the risk of rebel-perpetrated sexual violence. Second, peacekeepers are on average effective in reducing the risk of sexual violence when facing actors that exercise internal control; for government actors, the UN police is particularly effective, while for rebel actors, military troops are instead the personnel type associated with a lower risk of sexual violence. But how robust are these results?

We begin by addressing one aspect of peacekeeping, which commonly receives more attention in media outlets than in academic journals (for scholarly exceptions, see Beardsley and Karim 2016; Nordås and Rustad 2013; Beber et al. 2017): sexual abuse by peacekeepers themselves. These scandals have not only damaged the credibility of UN protection efforts, they may also be taken as an indication of a dysfunctional military practice that renders these missions unfit to effectively protect civilians from sexual violence also by the warring parties. Nordås and Rustad (2013) find that sexual violence by warring parties increases the probability of sexual abuse by peacekeepers. To account for the plausible, yet unexamined, potential that peacekeeping missions accused of perpetrating sexual violence should be worse in curbing sexual violence than other missions, we rescrutinize our significant findings adding a variable indicating whether or not there were allegations of sexual violence by peacekeepers (Online Appendix Tables A8 and A9). Due to data availability, this test reduces our sample to the period 1999–2009. Our findings for the first hypothesis are consistent when we include sexual abuse by peacekeepers. The findings for the second hypothesis change slightly but still in line with our argument: for actors with strong control, military troops have a negative effect on sexual violence by states, while police has a negative effect on sexual violence committed by rebels.

Second, we reestimate all our models using only years of peacekeeping presence. Since we rely on the SVAC dataset, our results reflect potential changes in sexual violence as reported by the US State Department, Amnesty International, and Human Rights Watch. There are good reasons to believe that sexual violence is underreported in these sources. By restricting the sample to cases only with an international presence through the UN, we hope to reduce the problem of underreporting of sexual violence since these are cases that attract international attention in the form of both journalists and international agencies. Our findings remain largely the same (Online Appendix Tables A10 and A11).

Lastly, we account for the intuition that armed groups’ conduct in terms of sexual violence is influenced by other groups’ conduct. While this relationship is largely unexplored empirically (though mentioned by Johansson and Sarwari [2017]), it has been noted that most sexual violence in the SVAC dataset takes place in conflicts where there are reports of sexual violence by both the government and the rebel side (Marks 2016, 10). Our analyses do not indicate a significant correlation between different parties’ behavior and our findings remain largely unchanged with this variable included (Online Appendix Tables A12 and A13).

Conclusions

Peacekeeping was not primarily designed to deal with the challenge of sexual violence by organized actors. However, with an increased emphasis on women, peace, and security, the expectations that the UN should be able to address such violence has intensified. Against this background, this article has examined under what conditions peacekeepers could potentially contribute to a decreased risk of sexual violence during and in the aftermath of conflict. We suggest two conditions under which peacekeeping should be more likely to aid in the PoC from sexual violence. First, when the mission has the mandate to protect civilians, an increase in the number of peacekeepers should be associated with a decline in sexual violence. Second, peacekeepers should be more successful in preventing and reacting to sexual violence by armed actors that have a high level of internal control. We find partial support for these conjectures. Protection mandates are not sufficient for reducing sexual violence by government actors. For rebel groups, we do however observe a decreasing risk of sexual violence with the larger deployment of protection-mandated police. Our findings are stronger for the second argument that peacekeeping primarily is successful in reducing sexual violence by actors with a higher level of internal control. These actors can be deterred and they can control the behavior of their own fighters. If sexual violence is driven by opportunistic motivations or carried out by fragmented groups, peacekeepers are not well equipped to address such behavior.

Concluding that an armed actor’s response to peacekeeping has to do with its organizational strength, we ought to also acknowledge the variability of strength over time. Organization, at least within rebel groups, tends to be rather path dependent (Weinstein 2007). However, the ability to uphold control can vary, for instance, as a function of military performance. When a party loses territorial control or military strength, it is likely to experience internal challenges, which may result in loss of control and, in the most severe cases, group fragmentation (Christia 2012). The presence of peacekeepers could exacerbate these challenges as the conflicting parties want to secure as strong positions as possible, should the conflict come to an end and/or a settlement (cf. Fjelde, Hultman, and Lindberg Bromley 2016). This might spur more violence and a more complex situation for peacekeepers to manage. However, it has equally been argued that actors in these situations are more open to negotiated settlements (Chu and Braithwaite 2018). These more complex dynamics are thus an interesting issue for future research to explore further.

Our findings are interesting in light of recent research that shows that peacekeeping indeed is effective in protecting civilians from lethal violence (Hultman, Kathman, and Shannon 2013; Bove and Ruggeri 2015; Kathman and Wood 2016). In comparison, it appears that sexual violence poses a particular challenge that peacekeepers are not always adequately trained and equipped to deal with. The fact that it works under certain circumstances, nevertheless, is encouraging and should serve to promote the development of peacekeeping strategy in cases with widespread sexual violence.

Supplemental Material

Supplemental Material, JCR-17-0164.R1 - UN Peacekeeping and Protection from Sexual Violence

Supplemental Material, JCR-17-0164.R1 for UN Peacekeeping and Protection from Sexual Violence by Karin Johansson and Lisa Hultman in Journal of Conflict Resolution

Footnotes

Authors’ Note

Reversed alphabetic order; equal authorship. Earlier versions of this article have been presented at the ISA Annual Convention in Atlanta, 2016; the Jan Tinbergen European Peace Science Conference in Milan, 2016; and the Research Paper Seminar at the Department of Peace and Conflict Research at Uppsala University, September 15, 2016.

Acknowledgments

We received very helpful comments at each of the occasions we presented this paper and we thank everyone who offered feedback. In particular, we acknowledge helpful input from Kyle Beardsley, Erika Forsberg, and Erik Melander. Lastly, we thank the Journal of Conflict Resolution editor and anonymous reviewers whose comments have helped to improve this study considerably.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is funded by Knut and Alice Wallenberg (Grant Number KAW2014.0162).

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Beardsley

Kyle

Karim

Sabrina

. 2016. “Explaining Sexual Exploitation and Abuse in Peacekeeping Missions: The Role of Female Peacekeepers and Gender Equality in Contributing Countries.” Journal of Peace Research 53 (1): 100–15.

Beber

Berndt

Gilligan

Michael

Guardado

Jenny

Karim

Sabrina

. 2017. “Peacekeeping, Compliance with International Norms, and Transactional Sex in Monrovia, Liberia.” International Organization 71 (1): 1–30.

Bove

Vincenzo

Ruggeri

Andrea

. 2015. “Kinds of Blue. Diversity in U.N. Peacekeeping Missions and Civilian Protection.” British Journal of Political Science 46 (3): 681–700.

Butler

C. K.

Gluch

Mitchell

N. J.

. 2007. “Security Forces and Sexual Violence: A Cross-National Analysis of a Principal Agent Argument.” Journal of Peace Research 44 (6): 669–87.

Christia

Fotini

. 2012. Alliance Formation in Civil Wars. Cambridge, UK: Cambridge University Press.

Chu

S. Tiffany

Braithwaite

Jessica Maves

. 2018. “The Effect of Sexual Violence on Negotiated Outcomes in Civil Conflicts.” Conflict Management and Peace Science 35(3): 233–47.

Cohen

Dara Kay

. 2013. “Explaining Rape during Civil War: Cross-national Evidence (1980–2009).” American Political Science Review 107 (3): 461–77.

Cohen

Dara Kay

Green

Amelia Hoover

Wood

Elisabeth

. 2013. “Wartime Sexual Violence: Misconceptions, Implications, and Ways Forward.” United States Institute of Peace. Accessed June 21, 2018. http://www.usip.org/publications/wartime-sexual-violence-misconceptions-implications-and-ways-forward.

Cohen

Dara Kay

Nordås

Ragnhild

. 2014. “Sexual Violence in Armed Conflict: Introducing the SVAC Dataset, 1989–2009.” Journal of Peace Research 51 (3): 418–28.

10.

Cohen

Dara Kay

Nordås

Ragnhild

. 2015. “Do States Delegate Shameful Violence to Militias? Patterns of Sexual Violence in Recent Armed Conflicts.” Journal of Conflict Resolution 59 (5): 877–98.

11.

Coppedge

Michael

Gerring

John

Lindberg

Staffan I.

Skaaning

Svend-Erik

Teorell

Jan

Altman

David

Bernhard

Michael

. 2016. “V-Dem Country-year Dataset v6.” Varieties of Democracy (V-Dem) Project. Accessed November 15, 2016. https://www.v-dem.net/en/data/data-version-6/.

12.

Cunningham

David E.

Gleditsch

Kristian Skrede

Salehyan

Idean

. 2009. “It Takes Two: A Dyadic Analysis of Civil War Duration and Outcome.” Journal of Conflict Resolution 53 (4): 570–97.

13.

Davies

Sara

True

Jacqui

. 2015. “Reframing Conflict-related Sexual and Gender-based Violence: Bringing Gender Analysis Back In.” Security Dialogue 46 (6): 495–512.

14.

DeMeritt

Jacqueline

. 2015. “Delegating Death: Military Intervention and Government Killing.” Journal of Conflict Resolution 59 (3): 428–54.

15.

Eck

Kristine

Hultman

Lisa

. 2007 “One-Sided Violence against Civilians in War: Insights from New Fatality Data.” Journal of Peace Research 44 (2): 233–46.

16.

Eriksson Baaz

Maria

Stern

Maria

. 2009. “Why Do Soldiers Rape? Masculinity, Violence, and Sexuality in the Armed Forces in the Congo (DRC).” International Studies Quarterly 53 (2): 495–18.

17.

Eriksson Baaz

Maria

Stern

Maria

. 2013. Sexual Violence as a Weapon of War? Perceptions, Prescriptions, Problems in the Congo and Beyond. London, UK: Zed Book.

18.

Fjelde

Hanne

Hultman

Lisa

Bromley

Sara Lindberg

. 2016. “Offsetting Losses: Bargaining Power and Rebel Attacks on Peacekeepers.” International Studies Quarterly 60 (4): 611–23.

19.

Fortna

Virginia Page

. 2008. Does Peacekeeping Work? Shaping Belligerents’ Choices after Civil War. Princeton, NJ: Princeton University Press.

20.

Gilligan

Michael

Stedman

Stephen John

. 2003. “Where Do the Peacekeepers Go?” International Studies Review 5 (4): 37–54.

21.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Håvard

. 2002. “Armed Conflict 1946-2001: A New Dataset.” Journal of Peace Research 39 (5): 615–37.

22.

Gutiérrez-Sanín

Fransisco

Wood

Elisabeth Jean

. 2017. “What Should We Mean by ‘Pattern of Political Violence’? Repertoire, Targeting, Frequency, and Technique.” Perspectives on Politics 15 (1): 20–41.

23.

Harff

Barbara

. 2003. “No Lessons Learned from the Holocaust? Assessing the Risk of Genocide and Political Mass Murder since 1955.” American Political Science Review 97 (1): 57–73.

24.

Hoover Green

Amelia

. 2011. “Repertoires of Violence: The Role of Armed Group Institutions and Ideologies.” Doctoral dissertation,

Yale University,

New Haven, CT.

25.

Hoover Green

Amelia

. 2016. “The Commander’s Dilemma: Creating and Controlling Armed Group Violence.” Journal of Peace Research 53 (5): 619–32.

26.

Howard

Lise Morjé

. 2008. UN Peacekeeping in Civil Wars. Cambridge, UK: Cambridge University Press.

27.

Hultman

Lisa

. 2013. “UN Peace Operations and Protection of Civilians: Cheap Talk or Norm Implementation?” Journal of Peace Research 50 (1): 59–73.

28.

Hultman

Lisa

Johansson

Karin

. 2017. “Responding to Sexual Violence: UN Peacekeeping and the Protection Agenda.” Global Responsibility to Protect 9 (2): 129–46.

29.

Hultman

Lisa

Kathman

Jacob

Shannon

Megan

. 2013. “United Nations Peacekeeping and Civilian Protection in Civil War.” American Journal of Political Science 57 (4): 875–91.

30.

International Criminal Court (ICC) 2011. “Elements of Crimes.” Accessed June 21, 2018. https://www.icc-cpi.int/NR/rdonlyres/336923D8-A6AD-40EC-AD7B-45BF9DE73D56/0/ElementsOfCrimesEng.pdf.

31.

Johansson

Karin

Sarwari

Mehwish

. 2017. “Sexual Violence and Biased Military Interventions in Civil Conflict.” Conflict Management and Peace Science. doi:10.1177/0738894216689814.

32.

Kathman

Jacob D.

Wood

Reed M.

. 2016. “Stopping the Killing during the ‘Peace’: Peacekeeping and the Severity of Postconflict Civilian Victimization.” Foreign Policy Analysis 12 (2): 149–69.

33.

Kathman

Jake

. 2013. “United Nations Peacekeeping Personnel Commitments, 1990-2011.” Conflict Management and Peace Science 30 (5): 532–49.

34.

Kirby

Paul

Shepherd

Laura J

. 2016. “The Futures Past of the Women, Peace and Security Agenda.” International Affairs 92 (2): 373–92.

35.

Leiby

Michele

. 2009. “Wartime Sexual Violence in Guatemala and Peru.” International Studies Quarterly 53 (2): 445–68.

36.

Marks

Zoe

. 2016. “Group-level Explanations of Wartime Sexual Violence: Toward an Analytical Framework.” Unpublished manuscript prepared for International Studies Association Working Group on Conflict Trends, Atlanta, GA.

37.

Melander

Erik

. 2009. “Selected To Go Where Murderers Lurk? The Preventive Effect of Peacekeeping on Mass Killings of Civilians.” Conflict Management and Peace Science 26 (4): 389–406.

38.

Muvumba Sellström

Angela

. 2015. “Stronger than Justice: Armed Group Impunity for Sexual Violence.” Report 104, Department of Peace and Conflict Research, Uppsala University, Uppsala, Sweden.

39.

Nordås

Ragnhild

. 2011. “Sexual Violence in African Conflicts.” Policy Brief 01, Center for the Study of Civil War/Peace Research Institute, Oslo, Norway.

40.

Nordås

Ragnhild

Rustad

Siri Aas

. 2013. “Sexual Exploitation and Abuse by Peacekeepers: Understanding Variation.” International Interactions 39 (4): 511–34.

41.

Sarwari

Mehwish

. 2016. “The Effects of United Nations Peacekeeping on Wartime Sexual Violence.” Paper presented at the ISA Annual Meeting, March 16–19, 2016, Atlanta, GA.

42.

Seifert

Ruth

. 1996. “The Second Front: The Logic of Sexual Violence in Wars.” Women’s Studies International Forum 19 (1-2): 35–43.

43.

Sharlach

Lisa

. 2000. “Rape as Genocide: Bangladesh, the Former Yugoslavia, and Rwanda.” New Political Science 21 (1): 89–102.

44.

Skjelsbaek

Inger

. 2001. “Sexual Violence in Times of War: A New Challenge for Peace Operations?” International Peacekeeping 8 (2): 69–84.

45.

Stanton

Jessica

. 2013. “Terrorism in the Context of Civil War.” Journal of Politics 75 (4): 1009–22.

46.

Tryggestad

T. L.

2015. “FN som normspreder: Kvinner, fred og sikkerhet.” Internasjonal Politikk 2015 (2): 275–84.

47.

UCDP Battle-related Deaths Dataset v.5. 2015. Uppsala Conflict Data Program, Uppsala University. Accessed November 15, 2016. www.ucdp.uu.se.

48.

UN (United Nations). 2010. “Report of the Secretary-general on the Implementation of Security Council Resolutions 1820 (2008) and 1888 (2009)”. A/65/592-S/2010/604. Accessed June 21, 2018. http://www.securitycouncilreport.org/atf/cf/%7B65BFCF9B-6D27-4E9C-8CD3-CF6E4FF96FF9%7D/WPS%20S%202010%20604.pdf.

49.

UN (United Nations). 2015. “The National Accounts Main Aggregate Database.” Accessed June 29, 2018. http://unstats.un.org/unsd/snaama/Introduction.asp.

50.

UN Women (United Nations Women). 2012a. “Addressing Conflict-related Sexual Violence an Analytical Inventory of Peacekeeping Practice.” Accessed June 21, 2018. http://www.unwomen.org/∼/media/Headquarters/Media/Publications/en/04DAnAnalyticalinventoryofPeacekeepingPracti.pdf.

51.

UN Women (United Nations Women). 2012b. Women’s Participation in Peace Negotiations: Connections between Presence and Influence, 2nd ed. Accessed June 21, 2018. http://www.peacewomen.org/node/92120

52.

UN DPKO (United Nations Department of Peacekeeping Operations). 2006. “DPKO Policy Directive: Gender Equality in UN Peacekeeping Operations.” Accessed June 21, 2018. http://www.un.org/en/peacekeeping/documents/gender_directive2006.pdf.

53.

UN DPKO/DFS (United Nations Department of Peacekeeping Operations and Department of Field Support). 2010. “Policy: Gender Equality in UN Peacekeeping Operations.” Ref. 2010.25. Accessed June 21, 2018. http://www.un.org/en/peacekeeping/documents/gender_directive_2010.pdf.

54.

UN DPKO/DFS (United Nations Department of Peacekeeping Operations and Department of Field Support). 2014a. “Gender Forward Looking Strategy 2014-2018.” Accessed June 21, 2018. http://www.un.org/en/peacekeeping/documents/DPKO-DFS-Gender-Strategy.pdf.

55.

UN DPKO/DFS (United Nations Department of Peacekeeping Operations and Department of Field Support). 2014b. “Policy: United Nations Police in Peacekeeping Operations and Special Political Missions.” Ref. 2014.01. Accessed June 21, 2018. http://www.un.org/en/peacekeeping/sites/police/documents/Policy.pdf.

56.

UNSC (United Nations Security Council). 2012. “Security Council Presidential Statement Condemns Sexual Violence in Conflict, Post-conflict Situations, Urges Complete, Immediate Cessation of Such Acts | Meetings Coverage and Press Releases.” SC/10555. Accessed June 21, 2018. http://www.un.org/press/en/2012/sc10555.doc.htm.

57.

Valentino

Benjamin

Huth

Paul

Balch-Lindsay

Dylan

. 2004. “Draining the Sea: Mass Killing and Guerrilla Warfare.” International Organization 58 (2): 375–407.

58.

Vivalt

Eva

. 2017. “Peacekeepers Help, Governments Hinder.” Unpublished manuscript. Accessed June 21, 2018. http://evavivalt.com/wp-content/uploads/2014/10/peacekeeping_RR.pdf.

59.

Weinstein

Jeremy

. 2007. Inside Rebellion: The Politics of Insurgent Violence. Cambridge, UK: Cambridge University Press.

60.

Wood

Elisabeth Jean

. 2009. “Armed Groups and Sexual Violence: When Is Sexual Violence Rare?” Politics and Society 37 (101): 131–61.

61.

Wood

Reed

. 2014. “Opportunities to Kill or Incentives for Restraint? Rebel Capabilities, the Origins of Support, and Civilian Victimization in Civil War.” Conflict Management and Peace Science 31 (5): 461–80.

62.

Wood

Reed

Kathman

Jacob

Gent

Stephen

. 2012. “Armed Intervention and Civilian Victimization in Intrastate Conflicts.” Journal of Peace Research 49 (5): 647–60.

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.

0.29 MB

0.00 MB