Regulating Militias

Abstract

In nearly two-thirds of civil wars since 1989, governments have received support in their counterinsurgency operations from militias. Many scholars predict higher levels of violence in conflicts involving pro-government militias because governments are either unable or unwilling to control militias. This article challenges this view, arguing that governments can and do often control militia behavior in civil war. Governments make strategic decisions about whether to use violence against civilians, encouraging both regular military forces and militia forces to target civilians or restraining regular military forces and militia forces from attacking civilians. In some cases, however, government and militia behavior differs. When a militia recruits its members from the same constituency as the insurgents, the militia is less likely to target civilians, as doing so would mean attacking their own community. Statistical analyses, using new data on pro-government militia violence in civil wars from 1989 to 2010, support these arguments.

Keywords

civil wars civilian casualties internal armed conflict human rights conflict

In late December 2003, Sudanese military forces working with Janjaweed militias attacked several villages near Murnei in the Darfur region of Sudan. Sudanese armed forces first fired on the villages from helicopters; the air attacks were then followed by ground attacks, in which military and Janjaweed forces acted together, burning the villages and killing more than 100 civilians (Amnesty International 2004a; Human Rights Watch 2004a). This incident was not an isolated event; Janjaweed militia forces, in collaboration with Sudanese military forces, committed severe atrocities against civilians in the context of counterinsurgency operations throughout the conflict in Darfur. Cases such as this raise questions about the consequences of militia involvement in civil wars. Do pro-government militias commonly engage in violence against civilians? How much does militia behavior vary across cases of civil war? Why might some militias, such as the Janjaweed in Sudan, engage in severe forms of violence against civilians, while other militias do not?

A growing literature has begun to examine the dynamics of civil war violence, looking in particular at violence against civilians. This literature has focused primarily on the behavior of governments and rebel groups. Yet civil wars are often more complex, involving more than just one single government fighting against one single rebel group (Jentzsch, Kalyvas, and Schubiger 2015). This article explores one aspect of this complexity, that is, the role that pro-government militias play in civil war violence.¹ Throughout the article, I use the term militia to refer to armed groups operating in support of the government but outside of the government’s regular police or military forces.²

As the introduction to this special issue points out, existing research on militias has primarily taken the form of case studies of militia involvement in specific civil wars (Kowalewski 1992; Campbell and Brenner 2000; Romero 2003; Staniland 2012), and missing from existing research is a systematic examination of patterns of militia violence across conflicts (Jentzsch, Kalyvas, and Schubiger 2015).³ This article seeks to fill this gap by examining pro-government militia violence against civilians in all civil wars from 1989 to 2010, looking, in particular, at the relationship between militia violence against civilians and government violence against civilians. The focus of the article is thus on the consequences of militia involvement in civil war, rather than its causes.

One of the commonly cited negative consequences of using militias is the potential loss of government control over the use of violence within its territory (Mitchell, Carey, and Butler 2014). Because militias operate outside of the regular command structure of government security forces, governments may be unable to control militia violence. Or, as some have argued, governments may intentionally avoid exerting control over militias, allowing militias to use any tactics necessary to combat the insurgency and thus outsourcing the most brutal tasks of counterinsurgency. Governments can therefore reap the benefits of a harsh counterinsurgency campaign, while denying responsibility for this violence (Campbell 2000; Campbell and Brenner 2000; Mitchell, Carey, and Butler 2014; Carey, Colaresi, and Mitchell 2015).

Yet, as the data presented in this article show, in many cases, governments do control militia violence. If governments choose to target civilians, they enlist both their regular military forces and militia forces in carrying out this violence. If governments choose not to target civilians, they limit violence among both their regular forces and militia forces. The fear that militias will act outside of the control of the government, targeting civilians even when the government does not, is not borne out in the evidence. In addition, the findings in this article demonstrate that deliberate government outsourcing of violence to militias is not as common in civil wars as many scholars have posited. Outsourcing arguments would predict patterns of violence in which government forces refrain from violence against civilians, while militias carry out atrocities against civilians, but the evidence shows that militias rarely target civilians when government forces refrain from violence.

This does not mean, however, that government and militia forces always behave similarly. In some conflicts, government forces target civilians, while militia forces do not. When a militia recruits its members from the same ethnic or religious group as the insurgents, violence against civilians is likely to be low; militia members are unlikely to be willing to direct severe violence against their own communities. Militias who draw members from an ethnic or religious group outside of the rebel group’s base of civilian support, in contrast, are more likely to be willing to engage in extreme violence against the rebel group’s civilian constituents.

I begin by discussing these hypotheses in greater detail. I then introduce an original data set on pro-government militia violence against civilians in all civil wars from 1989 to 2010. Although space constraints preclude detailed case studies, the last section of the article compares militia violence in the following two illustrative cases: Sudan and Uganda.

The Outsourcing Hypothesis

Governments create or support militias for a variety of reasons (Jentzsch, Kalyvas, and Schubiger 2015). Governments may use militias as a way of minimizing the costs of counterinsurgency; increasing their access to local knowledge; extending their reach into peripheral regions; or even bolstering the government’s legitimacy, making it seem as if the government possesses a broad base of public support (Kalyvas 1999; Kalyvas 2006; Jentzsch, Kalyvas, and Schubiger 2015; Carey, Colaresi, and Mitchell 2015). In other cases, governments may enlist militias to carry out violence against civilians suspected of sympathizing with the rebels (Campbell 2000; Campbell and Brenner 2000; Mitchell, Carey, and Butler 2014; Ahram 2014).

The ways in which governments collaborate with militias are similarly varied.⁴ Some militias are initiated by the state, with the government recruiting, training, and financing the militia from its inception; while in other cases, militias are initiated by the community, with the government providing material or operational support to sustain existing local groups (Jentzsch, Kalyvas, and Schubiger 2015). In some cases, regular military or police forces work closely with militias, commanding them to carry out specific tasks or incorporating them in joint counterinsurgency operations. In other cases, the links between regular forces and militias are looser, with the government giving general directives to militias, without providing guidance on specific operations.

Because militias are, by definition, groups that operate outside of the regular command structure of government military and police forces, many have warned that governments may have difficulty controlling militias, leading to higher levels of violence against civilians (Mitchell, Carey, and Butler 2014).⁵ Others have argued that governments may even use this loose command structure to their benefit, outsourcing to militias the “dirty” tasks of counterinsurgency while allowing the regular security forces to maintain their distance from human rights abuses. In some cases, governments may explicitly delegate violence to militias, commanding groups to burn villages or crops in regions of rebel activity, to kill groups of civilians as an example to potential rebel collaborators, or to expel civilians from contested territory. In other cases, the delegation of violence may be less explicit; the government may not order militias to engage in atrocities but may continue to arm or finance militias despite reports of abuse. Or the government may be lenient in its policing, refusing to investigate or file criminal charges against militia members responsible for violence. Governments can thus reap the benefits of a harsh counterinsurgency campaign, while denying responsibility for violence (Campbell 2000; Campbell and Brenner 2000; Mitchell, Carey, and Butler 2014; Carey, Colaresi, and Mitchell 2015).

If this outsourcing argument is correct, in conflicts involving pro-government militias, one would expect to see regular government forces limit their use of violence against civilians—to maintain the government’s ability to deny responsibility for abuses—while militia forces engage in higher levels of violence against civilians.⁶ Carey, Colaresi, and Mitchell (2015) argue that governments outsourcing violence to militias do not necessarily limit the use of violence by regular forces, positing that even when regular government forces target civilians, the presence of militias may create some ambiguity about responsibility for violence, benefiting governments. This may be true in cases of moderate government violence against civilians, but is less applicable to the cases addressed in this article, in which government violence is high. In cases of high government violence, shifting responsibility to militias is likely to be difficult, if not impossible. This is particularly true of conflicts in recent decades, as advances in communication technology have increased the availability of information regarding wartime atrocities.

Hypothesis 1: Among conflicts involving pro-government militias, high levels of militia violence against civilians are likely to be associated with low levels of government violence against civilians.

An Alternative Understanding of Militia Violence in Civil Wars

While governments may outsource violence to militias in some small number of conflicts, outsourcing is unlikely to be able to explain patterns of government and militia violence across cases of civil war because the outsourcing argument does not provide a complete picture of how governments weigh the costs and benefits of targeting civilians. First, as the growing literature on civil war violence against civilians has pointed out, governments often have incentives to target civilians, incentives that are strong enough for the government to use regular military forces to commit violence. Governments may use violence as a means of deterring civilians from providing material support or intelligence to insurgents. By gaining the cooperation of the civilian population, governments can thereby extend their control over territory (Kalyvas 2006). In some cases, governments may adopt a more extreme counterinsurgency strategy, seeking to separate insurgents entirely from their civilian base of support by cleansing civilians from contested territory (B. Valentino 2004; Downes 2008). Although civilians are often essential sources of material support and intelligence for insurgents, they are also political constituents. Governments, thus, may attack an insurgent group’s civilian constituents as a means of increasing the costs of the conflict for insurgents beyond the military costs associated with losses of soldiers or military equipment.⁷ If a government allows or instructs its regular military forces to carry out violence against civilians as part of its counterinsurgency campaign, the government is likely to instruct or permit pro-government militia forces to target civilians as well.

Second, for some governments, the costs of engaging in violence outweigh the potential gains from targeting an insurgent group’s civilian base of support (Stanton 2009). Violence may bring domestic costs. If key domestic constituencies oppose the government’s attacks on civilians, this may threaten a government’s source of domestic authority and political power. Violence may also bring international costs. Government abuses of civilians may prompt international criticism, leading to diplomatic or economic sanctions, a decline in foreign aid, a loss of diplomatic support during peace negotiations, and in some cases, even foreign military intervention. Governments that are particularly sensitive to these domestic and international costs of violence are likely to be concerned about maintaining close ties to militias engaging in atrocities against civilians. Even if the government’s regular military forces refrain from violence against civilians, the government is unlikely to be able to absolve itself of all responsibility for abuses committed by militias. Although a government can deny responsibility for violence perpetrated by a militia and may even be able to hide any ties to the militia, the government is still likely to face some criticism for not doing a better job of policing such extreme violence on its territory. And for governments that face high costs to engaging in violence against civilians, even a loose association with a pro-government militia that engages in violence against civilians would be too risky. These governments, therefore, are likely to limit the use of violence by both regular military forces and militia forces.

Hypothesis 2: In conflicts involving pro-government militias, the level of militia violence against civilians should be similar to the level of government violence against civilians.

Yet even if governments want to restrain militia violence, they may not always have the capacity to do so. Although most existing work in this vein has focused on the ability of governments or rebel groups to control their own forces (Weinstein 2005, 2007; Humphreys and Weinstein 2006; Wood 2006, 2009; Hoover Green 2011; Manekin 2013), it is possible to extend this argument to apply to militias (Mitchell, Carey, and Butler 2014). Militias tend to have little professional training and may have poorly developed or decentralized command structures. Furthermore, because militia forces often operate outside of the regular military or police command structure, governments may have difficulty monitoring and controlling the activities of militias. In this view, governments may not want militia soldiers to attack civilians but may be unable to reign in these organizations once they have been recruited and armed (Dasgupta 2003). When a militia operates under the direct command of government or military officials, the government is more likely to be able to exert control over the militia, reigning in violence if the government so desires. In contrast, when a militia operates with a high level of autonomy and is not under the direct command of government officials, the government is likely to have greater difficulty controlling the use of violence among militia soldiers. Problems of control may be particularly severe for governments with low levels of state capacity, with few resources available to police militia behavior.

Hypothesis 3: When government violence against civilians is low, militia violence against civilians is likely to be higher when the militia has a high degree of autonomy from the government than when the militia has limited autonomy from the government.

Drawing on the logic of the outsourcing hypothesis, some might argue that governments intentionally distance themselves from militias that they know will engage in high levels of violence against civilians, allowing these militias to remain autonomous, in order to be able to deny any association with militia violence. The evidence presented in the next section shows, however, that very few cases exist in which government violence is low, but militia violence is high. In addition, in these few cases, little evidence exists that these governments intentionally manipulated the nature of their ties with militias because they were aware that the militia would use violence.

All three of the hypotheses laid out thus far focus on the ways in which governments encourage or restrict the use of violence by militias. Yet militia preferences matter as well.⁸ Different types of militias may face different incentives or costs to civilian targeting, leading to different patterns of behavior. Militias can be divided into two broad types, based on their membership—groups that recruit individuals from within the same ethnic or religious constituency supporting the rebellion and groups that recruit individuals from outside of the ethnic or religious constituency supporting the rebellion. Even when a rebel group draws support from a particular ethnic or religious group, rarely do all members of this group enthusiastically back insurgents. Individual civilians may prefer to side with the government from the outset of war; others may become disillusioned with the insurgency over the course of a long, costly conflict. Some civilians may come to oppose the rebellion following a rebel attack on a friend or family member, while others may be willing to accept payment to fight on behalf of the government. For these and many other reasons, individual civilians may choose to join a pro-government militia, even if other individuals from their ethnic or religious group are backing the rebels.⁹ Examples of militias that recruit from within the rebel group’s civilian constituency include the Groupes de Légitime Défense in Algeria, pro-government Tamil militias fighting against the Tamil Tigers in Sri Lanka, and the Bow and Arrow militias that formed in opposition to the Lord’s Resistance Army (LRA) in Uganda.

Militias that recruit their members from within the constituency supporting the insurgency are unlikely to engage in extreme violence against civilians. These groups often live and operate in their own local communities, where they have access to good sources of information to aid in identifying suspected rebel collaborators. This availability of information means that militia punishments are likely to be targeted at specific individuals believed to be aiding the insurgency, rather than at whole families or whole villages (Lyall 2010). Although punishments may be severe, with suspected collaborators tortured or killed, this type of selective violence is likely to remain limited. And because these militias live and operate in their own communities, they are unlikely to be willing to use more extreme forms of violence such as massacres of groups of suspected rebel sympathizers, large-scale burning of villages or crops, or expulsions of villages and towns.

Pro-government militias of the second type—militias that recruit members from outside of the ethnic or religious constituency supporting the rebellion—are likely to behave much differently. In some cases, these militias may develop on their own; for example, a town or region neighboring an area of insurgent activity may create a militia to defend their region against rebel attacks or may volunteer to aid the government in making inroads into insurgent territory, in a “community-driven” process of mobilization (Jentzsch, Kalyvas, and Schubiger 2015). This town or region may be made up primarily of individuals from a different ethnic or religious group than the one backing the insurgents. In other cases, the government may deliberately target a particular ethnic or religious group when recruiting members for local militias, in an effort to build on existing divisions within the civilian population. Examples include Serb militias that fought alongside the Yugoslav government against Croatian independence, Interahamwe militias in Rwanda, and Janjaweed militias fighting against Darfur insurgents in Sudan.

When militias are composed of members from groups “outside” of the insurgent base, violence against civilians is more likely. These militias often do not live and operate within the communities they target for violence; they may live nearby, in a neighboring village or a neighboring region, but they have some degree of distance from the civilians they attack. Although these militias may have some local knowledge—of terrain, for example—they are unlikely to have access to detailed information about specific individuals aiding the insurgency. For this reason, these groups are likely to have greater difficulty in targeting violence at specific rebel collaborators and may direct punishments at groups of civilians or even whole villages, in an attempt to deter civilians from supporting the insurgency (Lyall 2010). Because these militias do not operate in their own communities, they are also more likely to be willing to impose severe costs on the rebel group’s civilian base as part of a strategy intended to cleanse territory or to punish the rebel group. Massacres, the burning of villages, and expulsions are all within the realm of possibility for these militias.

Hypothesis 4: In conflicts involving pro-government militias, violence against civilians is likely to be higher when the militia recruits members from outside of the rebel group’s civilian constituency than when the militia recruits members from within the rebel group’s civilian constituency.

While the above-mentioned hypotheses posit that government and militia incentives for violence and restraint are key to explaining militia violence during civil war, these are not the only possible factors that might drive behavior. A number of scholars have argued, for example, that other group characteristics, such as the relative strength of the belligerents, may influence the likelihood of civilian targeting (Hultman 2007; B. Valentino, Huth, and Balch-Lindsay 2004). Governments that are weak militarily may have difficulty confronting insurgents in direct military engagements; instead, these governments may attack the insurgents’ civilian base of support as a means of imposing costs on insurgents while minimizing their own losses. When the government is weak, pro-government militias may also try to limit military losses by focusing their attacks on soft, civilian targets. Characteristics of the conflict, too, may create incentives for both government forces and militia forces to target civilians. Some posit that the costs of the conflict may influence belligerent behavior; as the costs of a conflict rise, belligerents may attack soft civilian targets in an effort to conserve their waning military resources.¹⁰ Others contend that because rebel groups using guerrilla warfare rely more heavily on civilians for support, governments have stronger incentives to attack civilians in conflicts involving guerrilla warfare than in conflicts involving conventional warfare (B. Valentino, Huth, and Balch-Lindsay 2004); this may extend to pro-government militia groups as well. Like Hypothesis 2, these alternative arguments would predict similar levels of violence among regular government forces and militia forces during civil war; thus, controlling for group- and conflict-level characteristics that might drive militia behavior is essential.

Methodology

I test these arguments using a new data set on pro-government militia violence against civilians in civil wars from 1989 to 2010. The list of civil wars is drawn from the Uppsala Conflict Data Program (UCDP)/Peace Research Institute Oslo (PRIO) Armed Conflict Dataset, which defines internal armed conflict as “a contested incompatibility that concerns government or territory or both where the use of armed force between two parties results in at least 25 battle-related deaths” and “occurs between the government of a state and internal opposition groups” (N. P. Gleditsch et al. 2002, 618-19). Following standard definitions of civil war, I focus on conflicts with at least 1,000 battle-related deaths (Doyle and Sambanis 2006; Fearon and Laitin 2003).

For each conflict, I identified whether the government collaborated with militias in carrying out its counterinsurgency campaign by using data from the Pro-Government Militias Database (PGMD; Carey, Mitchell, and Lowe 2013). The PGMD defines a pro-government militia as an armed group with some level of organization, that is, “identified…as pro-government or supported by the government” and is “identified as not part of the regular security forces” (Carey and Mitchell 2013, 5). The PGMD includes all pro-government militias active during wartime and peacetime, but the data set for this article includes only those pro-government militias that were involved in counterinsurgency operations during a civil war. Thirty-five of the 102 civil wars occurring from 1989 to 2010 do not involve pro-government militias. These civil wars do not enter into the data set. Among conflicts involving pro-government militias, the number of active militias varies across cases. In about two-thirds of the conflicts involving militias, multiple pro-government militias were active. The data set thus includes 186 militias, active in sixty-seven different civil wars.

Measuring Government and Militia Violence

Government Civilian Targeting and Militia Civilian Targeting are dichotomous variables measuring whether each government and each militia engaged in deliberate attacks on civilians. Differentiating between violence perpetrated by regular government forces and violence perpetrated by pro-government militias is challenging, as regular government forces often collaborate closely with militias. For this reason, it is not possible to compare the number of civilians killed by government forces to the number of civilians killed by militia forces. Instead, to measure the extent of government and militia violence against civilians, this data set focuses on forms of violence against civilians (Stanton 2009).

A government or militia group is coded as having targeted civilians if it engaged in any of the following four forms of violence against civilians: massacres; scorched earth campaigns; cleansing of a particular ethnic or religious group from territory; or deliberate bombing and shelling of civilian targets. I define massacres as the killing of more than five civilians in the same place and at the same time, through the direct contact of armed forces with civilians. Scorched earth strategies involve the intentional burning or destruction of villages and/or agricultural land, while cleansing is the forced, permanent removal of a particular civilian population from a particular territory. Coding the deliberate bombing and shelling of civilian areas is complicated, as belligerents often claim that aerial bombardment or artillery shelling is aimed at military targets, only mistakenly hitting nearby civilian targets. For this reason, I include only sustained bombing or shelling that was directed intentionally at civilian targets such as residential areas of towns or cities.

I focus on violence against civilians that is intentional, excluding incidents in which civilians are killed or civilian targets are destroyed accidentally during a military engagement with the opponent. In addition, I am interested in capturing patterns of behavior; thus, I do not include cases in which one or two incidents of violence against civilians are reported and there is no evidence that these incidents constituted a pattern of behavior. I use multiple sources in my coding, including the US Department of State Country Reports on Human Rights Practices; annual reports and in-depth reports published by Amnesty International, Human Rights Watch, International Crisis Group, and the United Nations; as well as secondary historical sources and newspaper reports.

While these sources provide information on the behavior of many of the pro-government militias involved in civil wars, some militias are small groups about which little is known. I code Militia Civilian Targeting as missing in these cases; coding these as cases of no violence might skew the results, making it seem as if these groups were not engaging in violence when, in fact, the absence of reports of violence is simply an indication of the general lack of information about these groups. The data set thus codes Militia Civilian Targeting for 130 militia groups, active in sixty different civil wars. In these sixty civil wars, 60.0 percent of governments (36 out of 60) targeted civilians, while forty-six of the 130 militia groups, or 35.4 percent, targeted civilians.¹¹

Measuring Militia Characteristics

While the first two hypotheses focus on the relationship between government and militia violence—the first predicting an inverse relationship between government and militia violence and the second predicting a positive relationship between government and militia violence—the third hypothesis focuses on the degree to which militia forces operate with autonomy from the government. An ideal test of this hypothesis would look at the organizational command structure linking government and militia forces. Comprehensive data on the nature of the organizational ties linking government and militia forces are not available, but it is possible to examine organizational ties in a smaller subset of cases, as I discuss in greater detail subsequently. As an additional test of this hypothesis, I include gross domestic product (GDP) per capita as a measure of state capacity.¹² Even if governments have incentives to limit the use of violence among regular and militia forces, when state capacity is low, governments may have difficulty controlling violence.

The fourth hypothesis posits that militia violence is likely to be higher when the militia recruits its members from outside of the rebel group’s constituency. To measure militia recruitment patterns, I begin with data on militia recruitment from the PGMD, which identifies whether a militia recruited its members from a particular ethnic or religious group (Carey, Mitchell, and Lowe 2013). The PGMD does not specify the name of the ethnic or religious group from which the militia recruited members; but using a variety of secondary sources, I was able to obtain this information. Comparing this to the name of the main ethnic or religious group from which the rebel group recruited members, I constructed a dichotomous variable, Recruits Rebel Constituents, coded as 1 if the militia recruits members from the same ethnic or religious group as the rebel group and coded as 0 if the militia recruits members from a different ethnic or religious group than the rebel group or does not recruit members from any particular ethnic or religious group. Only 14.6 percent of militias recruit members from within the same ethnic or religious group as the insurgents.

In addition, I include several other measures of militia characteristics, drawn from the PGMD. Government Trained captures whether the militia received training from the government,¹³ while Semi-Official Militia measures whether the militia “has a formally and/or legally acknowledged status” (Carey and Mitchell 2013, 12). About half of the militias received training from the government and 43.1 percent are formally or legally acknowledged.

Control Variables

Scholars have suggested a number of other factors—mainly characteristics of the conflict—that might influence patterns of wartime violence. Further, some of these conflict characteristics might shape government and militia behavior in similar ways, leading to similar patterns of government and militia violence toward civilians. It is possible, therefore, that any positive correlation between government and militia violence is a result of the conflict context and not, as the second hypothesis contends, government incentives for violence. Controlling for conflict characteristics helps to address this concern.

To measure the strength of belligerents, Relative Strength is a ratio of average government troop strength to average rebel group troop strength, constructed using annual data from The Military Balance and the SIPRI Yearbooks.¹⁴ While it would be useful to control for militia troop strength as well, reliable data on the size and strength of militia forces are not available. To measure the costs of the conflict, I include two variables. The first—Conflict Intensity—captures the costs to the government’s military and measures the average annual battle-related deaths in the conflict.¹⁵ The second—Rebel Group Civilian Targeting—captures the costs to the government’s civilian constituents and measures whether the rebel group engaged in violence against civilians; this variable is measured in the same way as Government Civilian Targeting and Militia Civilian Targeting. To address the possibility that incentives for government and militia violence are higher in conflicts involving guerrilla warfare, Guerrilla Conflict measures whether the war was fought primarily using guerrilla warfare tactics as opposed to conventional tactics.¹⁶ Within the literature on terrorism, some have argued that violence is higher in conflicts involving multiple armed groups, as these groups use violence in an effort to outbid one another for popular support (e.g., Chenoweth 2010). Multiple Militias thus indicates whether the conflict involved multiple pro-government militias. I also control for the type of conflict, as measured by whether the conflict is a Separatist Conflict, in which the rebel group seeks autonomy or independence for a particular region of the country.¹⁷

Data Analysis

Table 1 shows the distribution of militia and government violence against civilians for conflicts involving pro-government militias. Several patterns evident in Table 1 are relevant to assessing the hypotheses put forth in this article. First, when the government does not target civilians, it is rare for a militia group to target civilians. Among conflicts in which the government did not target civilians, the majority of pro-government militias—82.9 percent—did not attack civilians; only seven militias targeted civilians when the government did not. This piece of evidence runs counter to the predictions of the outsourcing hypothesis. If governments were outsourcing the dirty tasks of counterinsurgency to militias in an effort to maintain distance from these abuses, then one would expect to see a significant number of cases in which government violence is low—indicating that the government is restraining regular security forces in an effort to keep its own hands clean—but pro-government militia violence is high.

Table 1.

Government and Pro-government Militia Violence Against Civilians.

	Militia Targets Civilians	Militia Does Not Target Civilians	Total
Government targets civilians	39 (43.8%)	50 (56.2%)	89 (100%)
Government does not target civilians	7 (17.1%)	34 (82.9%)	41 (100%)

A second pattern evident in Table 1 is the high degree of correspondence between government and militia behavior. In more than half of the conflicts involving pro-government militias, government and militia forces behaved similarly—either both avoiding violence against civilians or both targeting civilians. This high degree of correspondence supports the strategic interpretation of government and militia violence posited in the second hypothesis in which governments make decisions about whether to target civilians and then either enlist both regular and militia forces to carry out these violent counterinsurgency tasks or restrict the use of violence by both regular and militia forces.

Perhaps the most surprising pattern is the large number of cases in which the government engaged in high levels of violence against civilians, but the militia group did not use violence against civilians. This pattern indicates that, in some cases, the preferences of the government and the preferences of the militia differ; this divergence in preferences is consistent with the fourth hypothesis, positing that militia incentives for violence vary across conflicts.

To examine further the variation in pro-government militia violence across cases of civil war, I run binary logit regressions with Militia Civilian Targeting as the dependent variable. Table 2 presents the results of these analyses; I report robust standard errors, with errors clustered by country, to account for country-specific factors that may lead to correlation among conflicts occurring in the same country.

Table 2.

Binary Logit Results: Pro-government Militia Violence Against Civilians.

	Model 1	Model 2	Model 3
	Basic Model	Control for Additional Conflict Characteristics	Control for Additional Militia Characteristics
Government civilian targeting	1.717** (0.790)	2.437** (1.124)	2.122** (1.049)
Recruits rebel constituents	−1.889* (1.016)	−1.835* (1.088)	−3.042*** (1.045)
Rebel group civilian targeting	−0.082 (0.516)	−0.094 (0.535)	0.799 (0.591)
Relative strength—Ratio of government to rebel troops, logged	0.018 (0.138)	0.070 (0.166)	0.306** (0.149)
Conflict intensity—Average annual battle deaths, logged	0.514** (0.216)	0.723** (0.287)	0.939*** (0.308)
Per capita GDP, logged	0.049 (0.455)	0.131 (0.447)	0.323 (0.422)
Separatist conflict	—	−0.673 (0.913)	—
Guerrilla conflict	—	0.166 (0.586)	—
Multiple militias	—	−0.503 (0.970)	—
Government trained	—	—	3.425*** (0.949)
Semiofficial militia	—	—	−5.143*** (1.018)
Constant	−5.998 (3.887)	−8.316** (3.975)	−12.937*** (4.166)
Wald χ²	14.20**	12.25	34.78***
Pseudo R ²	.1637	.1963	.4920
N	130	124	130

Note: GDP = gross domestic product. Robust standard errors in parentheses.

***p < .01. **p < .05. *p < .10.

The results do not support the outsourcing hypothesis. If governments were delegating their dirty work to militias, then one would expect to see an inverse relationship between government and militia violence. Regular government forces would engage in little violence against civilians, in an effort to maintain the government’s ability to deny responsibility for abuses, while militia forces would engage in high levels of violence against civilians, as they carry out the most distasteful tasks of the counterinsurgency campaign. As the results in Table 2 show, however, the relationship between Government Civilian Targeting and Militia Civilian Targeting is positive—the opposite of what the outsourcing hypothesis would predict. These results are consistent with the findings in Cohen and Nordås (2015), which challenge the outsourcing hypothesis with regard to sexual violence in civil war.

This positive and statistically significant relationship between Government Civilian Targeting and Militia Civilian Targeting provides strong support for the second hypothesis, which predicts correspondence between the behavior of governments and their militia collaborators. When the government’s regular military forces target civilians, militia forces are also more likely to target civilians. This finding suggests that governments, indeed, may be able to influence the behavior of their collaborators; governments whose counterinsurgency strategies involve civilian targeting use both their regular and militia forces to carry out violence, while governments whose counterinsurgency operations do not involve civilian targeting limit the use of violence by both regular and militia forces.

The findings regarding military training lend further support to the second hypothesis. The coefficient for Government Trained is positive and statistically significant, indicating that militias receiving training from the government are more likely to engage in violence against civilians. Without information on the content of military training, it is difficult to assess why government training might be associated with higher levels of militia violence; but this finding does bolster the claim that the positive relationship between government and militia violence is not simply a by-product of the conflict context and rather is tied to government incentives for violence.

The third hypothesis posits that when violence among regular government forces is low, militias that have a greater degree of autonomy from the government are more likely to target civilians. While data on organizational control do not exist for all cases in the data set, it is possible to examine the small number of cases in which government violence is low and militia violence is high, to assess the plausibility of this hypothesis. A closer look at these seven cases shows that, indeed, these are primarily cases in which the militia retained significant autonomy from the government, limiting the government’s ability to exert effective command and control over militia violence. Four of the seven cases involve militias operating in the Democratic Republic of the Congo, which were not under the command of the Congolese government or military. Three of these militias originated in Rwanda, only later becoming embroiled in the Congolese conflict. Another of the seven cases involves vigilante groups who fought against Maoist insurgents in Nepal, but were poorly organized and operated outside any formal command structure. Only one militia of the seven—the Civilian Armed Forces Geographical Unit (CAFGU) militia in the Philippines—was under the direct command of the armed forces, but monitoring of CAFGU activities was limited, making it difficult for the government to control violence. In addition, little evidence of deliberate outsourcing of violence exists in these cases, as most of these militias formed on their own without government involvement. It is possible, however, that the Congolese and Nepalese governments may have intentionally maintained their distance from these militias, having observed militia violence during the early stages of conflict.

Another piece of evidence in support of the third hypothesis comes from the findings regarding Per Capita GDP. When looking at the full universe of cases, Per Capita GDP is not strongly linked to Militia Civilian Targeting. But if the sample is split to compare cases of high government violence with cases of low government violence, a different picture emerges (results shown in appendix). When government civilian targeting is high, the relationship between Per Capita GDP and Militia Civilian Targeting is positive and statistically significant at the .10 level. However, when government civilian targeting is low, this relationship is reversed, with Per Capita GDP showing a negative relationship with Militia Civilian Targeting. The findings suggest that among governments whose regular forces do not target civilians, governments with higher levels of state capacity are better able to control militia behavior. In contrast, among governments whose regular forces engage in civilian targeting, governments with higher levels of state capacity are able to encourage militias to engage in similar forms of violence.

Finally, the fourth hypothesis predicts that militias that recruit their members from the same constituency as the insurgents will be less likely to target civilians. The results of the statistical analyses lend support to this argument. The coefficient for Recruits Rebel Constituents is negative and statistically significant, indicating that militias that recruit members from the same ethnic or religious constituency as the insurgents are less likely to target civilians than militias that recruit members from a different constituency. In addition, this relationship between militia recruitment and militia violence is not simply a by-product of differences between conflicts involving disputes over territory and conflicts involving disputes over ideology. Militias that recruit from within the same ethnic group as insurgents are split fairly evenly between separatist and revolutionary conflicts; the same is true of militias that recruit from outside of the insurgent base of support.

Notably, the findings for the three main variables of interest hold even when controlling for other characteristics of the conflict, including the intensity of the conflict, the relative strength of government and insurgent forces, the type of conflict, the use of guerrilla warfare, and characteristics of the militia.¹⁸ While some scholars have posited a relationship between the relative strength of belligerents, the use of guerrilla warfare tactics, or the number of belligerents and belligerent behavior toward civilians, the results shown here indicate that none of these factors influences significantly the likelihood of militia violence against civilians. The positive and statistically significant relationship between conflict intensity and militia violence, however, does indicate that pro-government militias are more likely to target civilians in conflicts with high military costs, as measured by average annual battle-related deaths. However, this finding is not robust to alternate measures of conflict intensity; total battle-related deaths is not strongly associated with the likelihood of militia civilian targeting nor is conflict duration. Substituting these alternate measures of conflict intensity does not affect the results for the main variables of interest, which remain robust. Finally, the results show that militias that are formally recognized by the government are less likely to target civilians, as indicated by the negative coefficient for Semiofficial Militia. This finding is consistent with the findings on sexual violence in Cohen and Nordås (2015) and suggests some support for the idea that governments may seek to distance themselves from violent militias (Carey, Colaresi, and Mitchell 2015).

To provide a clearer sense of the substantive impact of these variables on the likelihood of militia civilian targeting, I calculate the probability that a militia will target civilians, when the government targets civilians and when it does not, and when the militia recruits from within the same constituency as the rebels as well as when it does not.¹⁹ In calculating these probabilities, I use Model 1 from Table 2, holding relative strength, conflict intensity, and per capita GDP at their mean values. I calculate these probabilities for the modal conflict: a conflict in which the rebel group targets civilians. The results are shown in Table 3.

Table 3.

Likelihood of Pro-government Militia Violence.

	Militia Does Not Recruit from Same Constituency as Rebels	Militia Recruits from Same Constituency as Rebels	Difference
Government targets civilians	49.0%	15.3%	33.7 [−0.5, 61.5]
Government does not target civilians	17.3%	4.4%	12.9 [−0.2, 37.9]
Difference	31.7 [2.9, 56.2]	10.9 [0.8, 31.6]

Note: 95 percent confidence interval shown in brackets for the first differences.

As Table 3 shows, the likelihood that a militia will target civilians is substantially greater when the government also targets civilians. Among militias that recruit members from the same constituency as the insurgents, the likelihood of civilian targeting is 15.3 percent when the government also targets civilians, as compared with only a 4.4 percent likelihood of civilian targeting when the government does not target civilians. A similar pattern is evident among militias that do not recruit from the same constituency as the insurgents. For these militias, the likelihood of civilian targeting is 17.3 percent when the government does not target civilians, but increases more than 31 percentage points to 49.0 percent when the government does target civilians. The likelihood of violence is also substantially higher among militias that recruit members from a different ethnic or religious constituency than the insurgents; militias that recruit from a different constituency are three to four times more likely to target civilians than militias that recruit members from the same constituency as the insurgents.

A Closer Look at Variation in Militia Violence across Conflicts

Although the statistical analyses show a strong correspondence between government and militia behavior as well as between militia recruitment patterns and militia behavior, the two brief case studies in this section shed additional light on these patterns in militia violence.

Sudan

In its attempt to defeat the insurgency in Darfur, the Sudanese government collaborated closely with local Janjaweed militias, whose “recruitment was selective and based on ethnicity,” drawing primarily from Arab ethnic groups “with historical grievances against those ethnic groups constituting the rebel movements” (Human Rights Watch 2005, 10). The two rebel movements recruited members from the following three main African ethnic groups: the Fur, Masalit, and Zaghawa. Although the government has denied links to the Janjaweed, numerous reports have provided evidence showing extensive government support for these militias, including the provision of weapons, salaries, training, and supplies (International Crisis Group 2004a; International Crisis Group 2005; Prunier 2005; United Nations 2005; Flint and de Waal 2008). Rather than establishing loose ties with the Janjaweed, as the outsourcing hypothesis would predict, the Sudanese government directly controlled the Janjaweed; within months of the first rebel attack, the government had established bases to train Janjaweed militias and had installed military officials at these bases to command and coordinate counterinsurgency operations carried out jointly by Janjaweed and regular government forces (Human Rights Watch 2005).

The pattern of government and Janjaweed attacks, which began in mid-2003, was similar throughout Darfur.²⁰ While government forces did seek to destroy rebel bases, most of the government’s attacks deliberately targeted Fur, Masalit, and Zaghawa villages, and in most cases, the villages attacked were not located near rebel group bases or areas of activity. Often, attacks began with aerial bombardment and artillery fire on a village. Sudanese military forces and Janjaweed forces, operating in conjunction with one another, or in some cases, operating alone, would then surround the village, firing on civilians using machine guns and rifles, burning homes and crops, looting any remaining property, and frequently raping girls and women. In many cases, villages were attacked repeatedly over the course of several days, weeks, or months, with government or Janjaweed forces returning to the village to kill or displace those who had returned.

Uganda

The LRA rebellion began in the late 1980s, following a series of insurgencies that broke out after Yoweri Museveni took power in 1986. In March 1991, the Ugandan government launched Operation North, a major offensive military operation against the LRA. One of the operation’s central components was the formation of local defense militias, called Arrow Groups, made up of individuals from the Acholi ethnic group living in the districts of Gulu and Kitgum, where the LRA was active at the time (Branch 2005; Gersony 1997).²¹ The LRA’s leader, Joseph Kony, is Acholi, as are most of the LRA soldiers, as well as the soldiers and leaders of the Northern insurgencies that preceded the LRA (Gersony 1997; International Crisis Group 2004b). Arrow Group militia members were unpaid, operated in their own home villages, and remained autonomous. This organizational structure was most likely related to a lack of available government resources to devote to the militias, rather than any deliberate government plan to distance itself from militia activities (Branch 2005).

Information on Operation North is limited, as the government cut off road traffic into and out of Northern Uganda during this period and prohibited access by journalists or aid organizations, but credible sources reported that government forces were responsible for severe human rights abuses. According to one researcher, Operation North “distinguished itself for its brutality and heavy-handedness” (Gersony 1997, 31), with government forces not only carrying out offensive attacks on the LRA but also using widespread arrest, beatings, rapes, and torture of civilians believed to be sympathetic to the insurgency (Amnesty International 1991, 1992). The US State Department reported that Ugandan government forces were responsible for several massacres of civilians, including a series of incidents in which soldiers “herded villagers into huts and burned them to death” killing twenty-two civilians in Kanyoke, fourteen in Angole Awere, and “an unspecified number in other burning incidents in Gulu” (US Department of State 1992, 430). None of these reports, however, indicates any involvement by the Arrow Group militias in violence against civilians.

Although the government did target civilians during the first years of the conflict, the LRA was responsible for the majority of violence against civilians and, in fact, following the creation of the Arrow Group militias, LRA attacks against civilians increased dramatically. Many agree that this increased LRA violence was an attempt to deter civilians from participating in militias (Amnesty International 1991; Branch 2005; Gersony 1997).²² With LRA attacks on the rise and the government’s continued failure to provide adequate weapons and support for militia members, the Arrow Groups demobilized just a few years after they were created (Branch 2005).

Case Discussion

These two cases illustrate several important points. First, as the data analysis shows, governments and militias often both engage in high levels of violence against civilians. Sudan provides an example of this kind of government and militia collaboration. In this case, government officials exerted direct command over Janjaweed operations and did so from the earliest days of the conflict, leading to similar behaviors among regular government forces and militia forces. Second, the Sudan case shows that when levels of government and militia violence are high, government attempts to deny responsibility for violence or to shift blame to militia forces are unlikely to be successful. Third, these two cases illustrate the distinction between militias based on patterns of recruitment. Janjaweed militias did not recruit from the same pool of civilians as insurgents; Janjaweed militias were thus willing to engage in extreme violence against civilians that they did not perceive to be their “own.” The Arrow Group militias in Uganda, in contrast, operated within their own Acholi communities and did not target Acholi civilians, despite the fact that most rebel soldiers were Acholi.

Alternative arguments cannot account for the patterns of militia violence observed in these two cases. Some have argued that when governments are weak or when conflict costs are high, governments (and by extension, pro-government militias) may have incentives to target civilians. Neither government was weak in comparison to the insurgency, with the Sudanese government possessing more than ten times the troop strength of insurgents and the Ugandan government more than forty times the LRA’s troop strength. In both cases, the government engaged in violence against civilian populations from the start of its counterinsurgency operations, before the government had suffered considerable losses. In Sudan, the government in cooperation with Janjaweed militias attacked Fur, Masalit, and Zaghawa villages from the first weeks of the insurgency; the most intense and destructive government and Janjaweed attacks on Darfur villages occurred in the first year of the conflict. And while the conflict in Sudan was a guerrilla conflict, so was the conflict in Uganda; this cannot explain why Janjaweed militias attacked civilians, but Arrow Group militias did not.

Conclusion

In nearly two-thirds of civil wars since 1989, governments have received support in their counterinsurgency operations from militias. Many have warned that militia involvement in civil wars may have severe negative consequences—namely, a loss of government control over the use of violence within its territory. Relying on evidence from a new data set on militia violence against civilians in all civil wars from 1989 to 2010, this article demonstrates that these fears of unchecked militia violence are largely unfounded. Militia forces rarely use higher levels of civilian targeting than their government counterparts. The evidence does not support the claim that governments regularly outsource the most abusive counterinsurgency tasks to militias.

Instead, this article argues that when governments have incentives to target civilians, they use both their regular security forces and militia forces to carry out attacks against the insurgent group’s base of civilian support. For many governments, the costs of engaging in violence against civilians are high and outsourcing violence to militias is not a viable option, as even a loose association with militia atrocities would be too costly. These governments seek to limit the use of violence, both among their regular forces and militia forces. Together, these arguments predict a high degree of correspondence between government and militia behavior. The statistical evidence provides strong support for this hypothesis, showing a positive correlation between government and militia violence, while the Sudanese case provides an example of civilian targeting carried out jointly by government and militia forces. Yet government and militia forces do not always behave similarly. Militias that recruit members from the same constituency as the insurgents are less likely to target civilians than militias who draw members from outside of the insurgent group’s base of civilian support, as the statistical results and the contrast between the Janjaweed militias and the Arrow Group militias demonstrate.

These findings have important implications for the study of civil war violence. While much of the existing research has focused on the behavior of governments and rebel groups, the findings in this article, and throughout this special issue, highlight the value of expanding research to include militia groups. Nearly two-thirds of civil wars fought between 1989 and 2010 involved pro-government militias and more than 35 percent of these militias targeted civilians in the context of counterinsurgency operations. A more complete understanding of the dynamics of wartime violence thus requires an understanding of these militia groups.

Footnotes

Acknowledgments

I would like to thank participants in the Yale University Conference on Paramilitaries, Militias, and Civil Defense Forces; the editors of the special issue; and two anonymous reviewers for their valuable feedback on this article. I would also like to thank Kelsey Baker, Marissa Dwyer, Erodita Herrera, Megha Jain, Laura Resnick, and Tanya Thanawalla for their research assistance.

Declaration of Conflicting Interests

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

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Notes

Supplemental Material

The online appendices are available at .

References

Ahram

Ariel I.

2014. “The Role of State-sponsored Militias in Genocide.” Terrorism and Political Violence 26 (3): 488–503.

Amnesty International. 1991. “Uganda: Human Rights Violations by the National Resistance Army.” Amnesty International Report, December, AI Index: AFR 59/020/1991, Amnesty International, London, UK.

Amnesty International. 1992. “Uganda: The Failure to Safeguard Human Rights.” Amnesty International Report, September, AI Index: AFR 59/005/1992, Amnesty International, London, UK.

Amnesty International. 2004a. “Sudan: At the Mercy of Killers—Destruction of Villages in Darfur.” Amnesty International Report, July, AI Index: AFR 54/072/2004, Amnesty International, London, UK.

Amnesty International. 2004b. “Sudan: Darfur—‘Too Many People Killed for No Reason.’” Amnesty International Report, February, AI Index: AFR 54/008/2004, Amnesty International, London, UK.

Branch

Adam

. 2005. “Neither Peace Nor Justice: Political Violence and the Peasantry in Northern Uganda, 1986-1998.” African Studies Quarterly 8 (2): 1–31.

Campbell

Bruce B.

2000. “Death Squads: Definition, Problems, and Historical Context.” In Death Squads in Global Perspective: Murder with Deniability, edited by Campbell

Bruce B.

Brenner

Arthur D.

, 1–26. New York: St. Martin’s Press.

Campbell

Bruce B.

Brenner

Arthur David

. 2000. Death Squads in Global Perspective: Murder with Deniability. New York: St. Martin’s Press.

Carey

Sabine C.

Colaresi

Michael P.

Mitchell

Neil J.

. 2015. “Governments, Informal Links to Militias, and Accountability.” Journal of Conflict Resolution 59 (5): 850–876.

10.

Carey

Sabine C.

Colaresi

Michael

Mitchell

Neil J.

. 2015. “Risk Mitigation, Regime Security, and Militias: Beyond Coup-proofing.” Working Paper, University of Mannheim, Mannheim, Germany.

11.

Carey

Sabine C.

Mitchell

Neil J.

. 2013. “Pro-government Militias Database (PGMD) Codebook, Version 1.1.” Accessed March 5, 2015. www.sowi.uni-mannheim.de/militias/.

12.

Carey

Sabine C.

Mitchell

Neil J.

Lowe

Will

. 2013. “States, the Security Sector, and the Monopoly of Violence: A New Database on Pro-government Militias.” Journal of Peace Research 50 (2): 249–58.

13.

Chenoweth

Erica

. 2010. “Democratic Competition and Terrorist Activity.” The Journal of Politics 72 (1): 16–30.

14.

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): 877–898.

15.

Dasgupta

Sunil

. 2003. “Understanding Paramilitary Growth: Agency Relations in Military Organization.” Working Paper Presented at Conference on Curbing Human Rights Violations by Non-state Armed Groups, Centre of International Relations, Lui Institute for Global Issues, University of British Columbia, Vancouver, BC.

16.

Downes

Alexander B.

2008. Targeting Civilians in War. Ithaca, NY: Cornell University Press.

17.

Doyle

Michael W.

Sambanis

Nicholas

. 2006. Making War and Building Peace: United Nations Peace Operations. Princeton, NJ: Princeton University Press.

18.

Fearon

James D.

Laitin

David D.

. 2003. “Ethnicity, Insurgency, and Civil War.” American Political Science Review 97 (1): 75–90.

19.

Flint

Julie

Waal

Alex de

. 2008. Darfur: A New History of a Long War. 2nd ed. London, UK: Zed Books.

20.

Gates

Scott

Strand

Havard

. 2006. “Modeling the Duration of Civil Wars: Measurement and Estimation Issues.” Working Paper, PRIO, Centre for the Study of Civil War, London, UK.

21.

Gersony

Robert

. 1997. “The Anguish of Northern Uganda: Results of a Field-based Assessment of the Civil Conflicts in Northern Uganda.” Report Submitted to the US Embassy and USAID Mission, Kampala, August. USAID, Washington, DC. Accessed March 5, 2015. http://pdf.usaid.gov/pdf_docs/pnacc245.pdf.

22.

Gleditsch

Kristian Skrede

. 2002. “Expanded Trade and GDP Data.” Journal of Conflict Resolution 46 (5): 712–24.

23.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Havard

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

24.

Hoover Green

Amelia

. 2011. “Repertoires of Violence Against Noncombatants: The Role of Armed Group Institutions and Ideologies.” PhD Dissertation, Yale University, New Haven, CT.

25.

Hultman

Lisa

. 2007. “Battle Losses and Rebel Violence: Raising the Costs for Fighting.” Terrorism and Political Violence 19 (2): 205–22.

26.

Human Rights Watch. 2004a. “Darfur Destroyed: Ethnic Cleansing by Government and Militia Forces in Western Sudan.” Human Rights Watch Report 16 (6).

27.

Human Rights Watch. 2004b. “Darfur in Flames: Atrocities in Western Sudan.” Human Rights Watch Report 16 (5).

28.

Human Rights Watch. 2005. “Entrenching Impunity: Government Responsibility for International Crimes in Darfur.” Human Rights Watch Report 17 (17).

29.

Humphreys

Macartan

Weinstein

Jeremy M.

. 2006. “Handling and Manhandling Civilians in Civil War.” American Political Science Review 100 (3): 429–47.

30.

International Crisis Group. 2004a. “Darfur Rising: Sudan’s New Crisis.” ICG Africa Report 76, March 25, 2004. International Crisis Group, Nairobi/Brussels, Belgium.

31.

International Crisis Group. 2004b. “Northern Uganda: Understanding and Solving the Conflict.” ICG Africa Report 77, April 14, 2004. International Crisis Group, Nairobi/Brussels, Belgium.

32.

International Crisis Group. 2005. “Darfur: The Failure to Protect.” ICG Africa Report 89, March 8, 2005. International Crisis Group, Nairobi/Brussels, Belgium.

33.

Jentzsch

Corinna

Kalyvas

Stathis

Schubiger

Livia Isabella

. 2015. “Militias in Civil Wars.” Journal of Conflict Resolution 59 (5): 755-769.

34.

Kalyvas

Stathis N.

1999. “Wanton and Senseless? The Logic of Massacres in Algeria.” Rationality and Society 11 (3): 243–85.

35.

Kalyvas

Stathis N.

2006. The Logic of Violence in Civil War. Cambridge, UK: Cambridge University Press.

36.

Kalyvas

Stathis N.

2008. “Ethnic Defection in Civil War.” Comparative Political Studies 41 (8): 1043–68.

37.

Kalyvas

Stathis N.

Balcells

Laia

. 2010. “International System and Technologies of Rebellion: How the End of the Cold War Shaped Internal Conflict.” American Political Science Review 104 (3): 415–29.

38.

King

Gary

Tomz

Michael

Wittenberg

Jason

. 2000. “Making the Most of Statistical Analyses: Improving Interpretation and Presentation.” American Journal of Political Science 44 (2): 347–61.

39.

Kowalewski

David

. 1992. “Counterinsurgent Paramilitarism: A Philippine Case Study.” Journal of Peace Research 29 (1): 71–84.

40.

Lacina

Bethany

Gleditsch

Nils Petter

. 2005. “Monitoring Trends in Global Combat: A New Dataset of Battle Deaths.” European Journal of Population 21 (2): 145–66.

41.

Lyall

Jason

. 2010. “Are Coethnics More Effective Counterinsurgents? Evidence from the Second Chechen War.” American Political Science Review 104 (1): 1–20.

42.

Manekin

Devorah

. 2013. “Violence against Civilians in the Second Intifada: The Moderating Effect of Armed Group Structure on Opportunistic Violence.” Comparative Political Studies 46 (10): 1273–300.

43.

Mitchell

Neil J.

Carey

Sabine C.

Butler

Christopher K.

. 2014. “The Impact of Pro-government Militias on Human Rights Violations.” International Interactions 40 (5): 812–36.

44.

Prunier

Gérard

. 2005. Darfur: The Ambiguous Genocide. Ithaca, NY: Cornell University Press.

45.

Romero

Mauricio

. 2003. “Reform and Reaction: Paramilitary Groups in Contemporary Colombia.” In Irregular Armed Forces and Their Role in Politics and State Formation, edited by Davis

Diane E.

Pereira

Anthony W.

, 178–208. New York: Cambridge University Press.

46.

Staniland

Paul

. 2015. “Militias, Ideology, and the State.” Journal of Conflict Resolution 59 (5): 770–793.

47.

Staniland

Paul

. 2012. “Between a Rock and a Hard Place Insurgent Fratricide, Ethnic Defection, and the Rise of Pro-state Paramilitaries.” Journal of Conflict Resolution 56 (1): 16–40.

48.

Stanton

Jessica A.

2009. “Strategies of Violence and Restraint in Civil War.” PhD diss., Columbia University, New York.

49.

Stanton

Jessica A.

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

50.

United Nations. 2005. Report of the International Commission of Inquiry on Darfur to the United Nations Secretary-General. Geneva, Switzerland: United Nations.

51.

US Department of State. 1992. “Country Reports on Human Rights Practices for 1991.” Report submitted to the Committee on Foreign Affairs, House of Representatives, and the Committee on Foreign Relations, US Senate, Washington, DC.

52.

US Department of State, the Bureau of Democracy, Human Rights, and Labor and the Bureau of Intelligence and Research. 2004. “Documenting Atrocities in Darfur.” State Publication 11182, September 2004, US Department of State, Washington, DC.

53.

Valentino

Benjamin A.

2004. Final Solutions: Mass Killing and Genocide in the Twentieth Century. Ithaca, NY: Cornell University Press.

54.

Valentino

Benjamin

Huth

Paul

Balch-Lindsay

Dylan

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

55.

Weinstein

Jeremy M.

2005. “Resources and the Information Problem in Rebel Recruitment.” Journal of Conflict Resolution 49 (4): 598–624.

56.

Weinstein

Jeremy M.

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

57.

Wood

Elisabeth Jean

. 2006. “Variation in Sexual Violence during War.” Politics & Society 34 (3): 307–42.

58.

Wood

Elisabeth Jean

. 2009. “Armed Groups and Sexual Violence: When Is Wartime Rape Rare?” Politics & Society 37 (1): 131–61.

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.

0.21 MB

0.00 MB