Civil War Mediation and Rebel Use of Violence Against Civilians

Abstract

Violence against civilians is portrayed as an antecedent of civil war, a cause, or both. Civil war creates opportune environments for planning and carrying out these acts that in turn can have detrimental effects on peace processes. Since not all civil war factions will see peace as beneficial, some actors may use violence to undermine the peace talks. The rebels may use indiscriminate violence to demonstrate their ability to exact costs on the government thus forcing the latter to negotiate. This article focuses upon acts of violence committed by rebel groups during mediated peace process. The central hypothesis is that violence against civilians increases the probability of mediation that in turn increases the prospects for violence. Using all civil war episodes from 1970 to 2008 as observations results from bivariate probit analysis endogenizing the choice of mediation bear out this theoretical argument.

Keywords

rebel violence mediation violence against civilians civil wars spoiler violence

Violence against civilians can be understood as an antecedent of civil war, a cause, or both (Boulden, 2009; Krueger & Laitin, 2008). Civil wars create opportune environments for planning and carrying out acts of violence against civilians. Since not all civil war factions will see peace as beneficial, some actors may use violence as a strategy to undermine the process. Actors engaged within the peace process, as well as outside parties who are excluded from a peace process, might perceive the peace as a threat to their power, status, and interests (Stedman, 2000). Therefore, these groups may use violence to undermine attempts to achieve peace (Stedman, 2000). This article focuses upon acts of violence against civilians carried out by rebel actors during a specific form of peace processes.

The relationship between third-party mediation and violence against civilians in civil wars is likely nonrecursive. Mediators are more likely to be involved in costly wars involving high casualties (Bercovitch & Gartner, 2006), such as the type involving violence against civilians. In turn, mediation is associated with peace processes and thus could be linked to the use of violence against civilians. It might decrease the scale of violence if it successfully offers alternatives to waging war. However, mediation could be associated with violence in civil wars if the rebels groups, or groups that splinter from the main organization, are dissatisfied with mediated peace process and resort to violence as a means of undermining the peace process. Stedman (1997) refers to such groups as inside spoilers, who are involved in peace talks but over displeasure with the peace process choose to disrupt it. Jonas Savimbi or Radovan Karadzic were such actors who used violence to derail peace process, nevertheless were engaged by United Nations (UN) mediators when searching for negotiated solution of the conflict (Zahar, 2010, p. 265). As for another example may serve the Real Irish Republican Army broke from the main Provisional Irish Republican Army in 1997 in the wake of the ceasefire with the United Kingdom and began carrying out criminal and violent actions during the mediated Northern Ireland peace process. Their most notable attack came in 1998 in Omagh and resulted in 29 deaths.

We argue the use of violence against civilians attracts mediation, but mediation in turn may trigger spoiler violence that occurs in the wake of mediated talks between governments and rebels. The study will further the civil war mediation (CWM), spoiler, and rebel violence knowledge bases and provide practical information for counterterror policy makers and mediators.

This potential two-way relationship is explored using the bivariate probit modeling method. Bivariate probit is specified with two dependent variables that are expected to impact each other in the same time period (Wiggins & Poi, 2010). In this case, the two dependent variables are rebel violence against civilians and mediation. Bivariate probit allows for the “endogenous choice of mediation” (Beardsley, 2008, p. 731).

This article is situated in the CWM, logic of violence, and spoiler literatures. We provide empirical evidence of negative side effects of mediation that shows mediation can lead to increased violence against civilians. To our knowledge, this is an original contribution to the ever-expanding mediation literature. There are empirical studies that consider the effect of mediation on resolution of the conflict, but here we ask different questions whose answers can help shed light on why mediation might not always be the answer or might need to be redesigned in some cases.

The study proceeds as follows. First, we report findings from studies on the use of violence against civilians in internal conflict and a brief review of the literatures on CWM. This is followed by research design, findings, and conclusion.

Theoretical Background: Mediation and Violence Against Civilians Within the Context of Civil War

Civil wars often place the civilian population directly in the midst of conflict: Insurgents need civilian support to survive and thrive, while the government seeks to deny rebels access to the population. Both sides use violence against civilians as a resource to compel collaboration and deter defection. This violence may occur before or after civil war, but research shows that it is more likely to occur during the civil war (Findley & Young, 2012). For example, in Mozambique, there were more acts of violence against civilians during the civil war (1977–1992), than in prewar or postconflict periods.

There is a considerable literature explaining the causes and effectiveness of targeting civilians in civil war (e.g., Balcells, 2010; Downes, 2007; Hoffman, 1998; Piazza, 2008). Governments can use violence against civilians as a substitute for fighting if traditional combat is not possible. Since rebels often rely on the population for support, the government can use violence to weaken the insurgency by killing civilians (Azam & Hoeffler, 2002). However, in the context of civil war, violence against civilians is more likely to be used by the insurgents (Eck & Hultman, 2007). This article concentrates on violence against civilians committed by rebels. Rebel violence, or threat of use of violence against civilians, is used for political ends—for example, to achieve national independence, influence government policy, or overthrow the government (Lake, 2002; Sandler, 2003). However, there are several, sometimes contradictory, explanations for the use of violence against civilians by rebels. One line of reasoning suggests rebel groups are less likely to target civilians because they are dependent on the civilian population for resources, logistical support, and/or funding (Zahar, 2001, p. 112). Another line asserts rebels are more likely to kill civilians, because by punishing the broad population, extremists seek to provoke a response from the government, to secure compliance, to demonstrate how determination (Kalyvas, 1999, 2004, 2006), and to mobilize additional recruits and financial support (Humphreys & Weinstein, 2006; Lake, 2002).

Another position argues weak insurgent groups have an incentive to target civilians because they lack the capacity to provide sufficient benefits to entice loyalty (Wood, 2010). Rebel violence against civilians can also be used to improve rebels’ bargaining position with the government. Hultman (2007) finds rebel groups adopt extreme methods under time pressure or when they are losing battles to compensate for political weakness and to impose extra political and military costs on the government. Finally, the rebel groups may use terrorism¹ strategically when an organization’s goal is to spoil the peace (Findley & Young, 2012; Kydd & Walter, 2006), or if they believe it will get them to the bargaining table (Thomas, 2014).

A use of violence against civilians affords rebels opportunities to fulfill other functions related to the war. Pape (2003) reports suicide terrorism during civil wars follows a strategic logic designed to achieve specific political purposes. This purpose is typically the goal of forcing a democratic occupying force to withdraw. Suicide attacks aimed at forcing Israel to withdraw from the West Bank and Gaza, Rwanda’s Committee for the Defense of the Revolution use of violence to prevent the implementation of the Arusha Accords, and al Qaeda actions on 9/11 (and after) pressuring the United States to withdraw from the Arabian Peninsula are three recent examples. According to Pape (2003), the aim of suicide terrorists is to mobilize the public in the democratic occupying state to weigh in domestically against the occupation.

The goal of this article is to provide empirical evidence of the relationship between mediation and use of violence against civilians. Specifically, we explore how rebel violence in civil wars might attract third-party mediation. Mediation is a peaceful means of noncoercive engagement that does not entail direct military physical force (Bercovitch, Anagnoson, & Willie, 1991, p. 8). In mediation, the third party acts as intermediary using strategies ranging from a simple communicative function entailing the transfer of information to a more proactive directive style involving inducements such as aid or threats of punitive actions such as sanctions or withholding aid. Mediation can entail mundane tasks such as timing, agenda setting, and other logistical functions (Bercovitch & DeRouen, 2004). It can also provide information, security guarantees, and promises to facilitate the implementation of peace agreements. A mediator can be an individual, private organization, state, a nongovernmental organization, or an intergovernmental organization. Often, there are multiple mediators in the same dispute.

Mediation is a common conflict resolution mechanism both in interstate and intrastate conflicts and has led to the successful termination of many conflicts (Bercovitch & DeRouen, 2004). Long-term success includes clear evidence of a sustainable including the absence of recurrences of violence and the implementation of substantive agreement provisions. Successful mediators derive certain benefits such as international prestige, boosts in domestic public approval, and influence in the conflict state. Civil wars with indiscriminate rebel violence make reaching a negotiated settlement more difficult (Cronin, 2008). Nevertheless, mediators are active in such conflicts. Examples include El Salvador (1979–1992), Peru (1981–1999), and Mozambique (1977–1992; Crocker, Hampson, & Aall, 2004; Herz & Nogueira, 2002; Negroponte, 2012; Wilkerson, 2008).

If a civil war is intractable and no negotiated settlement is in sight, then it becomes a likelier candidate for a third party offer of mediation (Bercovitch & Gartner, 2006; Gartner, 2013). Mediators are also more likely to be drawn to civil wars with civilian casualties (Kreutz & Brosche, 2013). Negotiated settlements do not necessarily entail mediation as disputants can negotiate between themselves. There is an underlying temporal causation. The indiscriminate rebel violence makes negotiated settlement between the parties less likely, which in turn, attracts mediation. Whether or not mediation occurs is the selection aspect of our modeling effort below where indiscriminate violence is coded as 1 only if it is prior to the mediation.

Mediation requires the consent of each disputant; therefore, the question is why would governments agree to negotiate with rebels that use terror. For their part, rebels might perceive participation in a mediation process positively, as it provides an opportunity for formal recognition (Crenshaw, 1981) and gives the rebels the opportunity to have their demands heard (Zartman, 1993). However, the rebels might also resort to mediation if they are not successful in achieving their goal by targeting civilians and instilling fear. This latter scenario becomes more likely when rebel violence against civilians becomes counterproductive, the popular support for rebels is waning, or the rebels are losing ground in the conflict (Cronin, 2010). Rebels might also want mediation in the event it could entail an amnesty against violence charges.

Governments, however, should be reluctant to negotiate with groups that take up terror, as this grants legitimacy and/or priority to such groups and deprioritizes those rebels that do not use terror and may, in fact, prefer to gain concessions peacefully (Neumann, 2007; Toros, 2008; Wilkinson, 2006). If so, why despite a government’s reluctance, does mediation with these violent rebel groups often occur? Especially, when the government appears to be winning?

Explaining how acts of terrorism convince a state to negotiate with rebels, scholars find that increased costs or a hurting stalemate (Zartman, 1993) could be why the government’s willingness to negotiate is subject to change. This means that negotiations can begin when terrorism offers rebels the power to hurt the government. Violence can be used to demonstrate the state’s illegitimacy when it fails to provide adequate protection (Hultman, 2007, 2009; Kydd & Walter, 2006; Wood, 2010). Hultman (2009, p. 823) suggests that by demonstrating the state is unable to provide security, a rebel group can force the government to withdraw and offer them concessions. The civilians may cooperate with rebels, because the government failed to take actions to protect them from rebel violence, and in order to protect themselves, the groups search for other security arrangements. On the other hand, a harsh government’s response to terror attacks may also cause civilians to support rebels (Goodwin & Skocpol, 1989, p. 493). If civilians cooperate with rebels and the government loses control, the state should be more willing to search for a peaceful resolution to the conflict. In other words, as terror tactics used by the rebels undermine the government’s position, the government should be expected to engage in mediated talks with the rebel groups in civil wars. Finally, even though governments may vastly overpower rebels, it does not always mean a military victory is imminent. Recent examples of this include civil wars in East Timor, Bougainville, Aceh, and numerous low-intensity conflicts in the far east of India. Although it was not a civil war, the United States never lost a major battle to the Vietcong and maintained an extreme preponderance of power, yet it entered peace talks. Israel agreeing to talks with Palestinian factions might provide another paradoxical example. The reader is referred to Thomas (2014) for a closer look at why governments reward exceptionally bad behavior by rebels (e.g., terrorism). In fact, she demonstrates that less extreme groups are relatively lower probability of having the opportunity to enter mediation and explains that governments make negotiation offers to groups “that have the power to hurt them.”

The disputants are likely to accept (or request) mediation when it offers a mutually beneficial solution that is more attractive than a “mutually hurting stalemate” exemplified by the realization by both sides that continuing the war is more costly than peace (Zartman, 2000). A situation in which neither party thinks it can win a given conflict without significant costs, and in which both are suffering from a continuation of fighting may be sufficient to foster mediation. A hurting stalemate, which often follows rebel violence against civilians, is a propitious moment for third-party mediation. This leads to our first hypothesis.

Hypothesis 1: The probability of CWM increases in the presence of rebel violence against civilians.

As mentioned, Thomas (2014) reports that rebel groups are both more likely to be granted the opportunity to participate in negotiations and offered more concessions when they execute a greater number of terror attacks during civil wars in Africa. In our study, we not only examine whether rebel violence against civilians may increase the likelihood of mediation but also whether subsequent rebel violence is reduced as the result of mediation.

Despite evidence that mediation helps resolve civil war (Gartner, 2013; Gartner & Bercovitch, 2006; Regan & Stam, 2000; Wilkenfeld, Young, Quinn, & Asal, 2005), the process can have the unintended effect of triggering the use of violence against civilians by rebels. Civil wars that entail extreme acts of violence against civilians tend to be complex, and to involve high stakes. These features make mediation more difficult. Scholars find mediation tends to work best once the conflict reaches a level of “ripeness” in terms of casualties (Bercovitch et al., 1991; Zartman, 1993). Characteristics of the rebel group, the conflict history, and the war type may also undermine mediation efforts.

While effective mediation is often assumed to require neutrality and a lack of bias, Svensson (2007) argues mediators may never be truly unbiased. He posits a mediator biased toward the government can actually be beneficial as rebels who agree to such a mediator are signaling a credible commitment to the process. Rebels often stand to gain ground during peace processes in terms of access to political power, and if the rebels are willing to accept a mediator who is more likely to check unbridled power grabs in the postconflict phase, the government concerns will be assuaged.

The central task of the mediator is to integrate and understand the elements of the dispute: people, dynamics, issues, and interests. Communication problems and misinterpretation may occur throughout the process. When the parties cannot reach agreement, a negotiation deadlocks occurs. If mediation fails to facilitate agreement, the disputants may turn to direct talks without mediation, arbitration, adjudication, legislation, political action (e.g., election), or renewed fighting.

Because of the risk of new violence, peace process should be designed to thwart potential spoilers and change a culture of violence. Darby (2001, pp. 116–123) writes peace processes should include groups that have the capacity for violent obstruction. The Liberian experience suggests that it was beneficial for the peace process to involve political parties, civil society actors, and other interest groups in the negotiations (Nilsson, 2010). Zahar (2010) in her critique of spoiler theory argues that mediators can limit the capabilities of all actors to use violence by shaping the opportunity structure to make violence a costly option to them.

When mediation does not meet the expectations of one or more of the actors, a use of violence against civilians may be taken up to change the direction of the peace process or to possibly end the dialogue (Findley & Young, 2012). The rebels, or a splinter group within the rebels, may have an interest in maintaining conflict and provoking violence from the state, and use violence against civilians to “spoil” the peace process (Stedman, 1996). Kydd and Walter (2006) report that attacks of violence during negotiations have a negative effect on the number of treaties put into effect. Stedman (1996, pp. 369–371) postulates these spoilers are threatened by a peace that could potentially undermine their status and will turn to violence to make peace a less likely outcome. Cronin (2010) discusses the difficulties mediation faces in light of rebel attacks against civilians. She observes negotiations facilitated by third parties can be promising when further use of violence by rebels is counterproductive, the popular support for rebels is waning, terror groups are losing ground in the conflict, and there are strong leaders on both sides. She notes spoilers make success less likely.

Mediators are drawn to the more intractable and costly wars, and extreme violence is often a hallmark of these long wars (e.g., in Sri Lanka, Israel and the Territories, Northern Ireland). On the surface, we may expect that mediation would appease the rebels. However, the spoiler literature tells us that once the peace talks begin or appear imminent, mediation could have the opposite effect on rebel behavior. Informed by Stedman’s (1996, 1997, 2000) arguments, we posit that mediated peace processes increase the probability carried out by rebels. Stedman (2000) suggests that spoilers can be inside or outside a peace process. Our focus is on inside spoilers, that is, elements within the rebel faction who are involved in peace talks but choose to undermine the process because it represents a threat. Spoilers may feel the mediation process is eliciting too many concessions such as foregoing foundational principles of the organization. If mediators show favoritism, rebel factions can feel especially threatened by the peace process and the probability of spoilers emerging increases (Newman & Richmond, 2006).

Finally, Thomas (2014) also explains why mediated settlements can yield more violence against civilians if rebels are willing to target civilians. Governments will avoid agreements with insurgents who appear much weaker. However, the targeting civilians inflict costs upon the government and can push the government to pursue a negotiated settlement. As it becomes apparent that the rebels can strike civilians at will, the government has a relatively weaker bargaining position. According to Thomas, the government will offer overly ambitious concessions making credibly commitment difficult. In turn, extremist rebels will likely renege on the agreement, and we are left with a situation with neither side being able to commit to an agreement and the likelihood of more violence against civilians.

From this discussion, we derive our second hypothesis.

Hypothesis 2: Mediation increases the probability of rebel violence against civilians.

If these two hypotheses are empirically supported, we will have provided an explanation for why mediation sometimes fails to stem violence in civil wars.

Research Design

Our observations are civil war episodes from 1970 to 2008. As mentioned, the bivariate probit method is applicable in this case, as we wish to model the endogenous choice of mediation (see Beardsley, 2008) and the effect of mediation on the probability of rebel violence. Although we do have the number of violent acts, to our knowledge, there is no modeling technique that allows us to endogenize mediation in a count setting.

Bivariate probit modeling is specified with two equations with dichotomous outcomes (Beardsley, 2008; Greene, 2011). The first equation’s dependent variable is the presence or absence of mediation. The second equation’s dependent variable is the presence or absence of rebel violence against civilians. Greene (2011) observes that it is acceptable for the dependent variable in the first equation to be specified as an independent variable in the second equation thus affording the opportunity to explore the determinants of mediation and in turn to assess the impact of mediation on rebel use of violence against civilians. The equations are solved simultaneously while taking into account the correlation of the error terms across the two equations (Beardsley, 2008). If supported empirically, this would represent the first statistical study of which we are aware to demonstrate the spoiler effect and would add to recent work of Beardsley (2008, 2011) questioning the sometimes overly optimistic mediation literature.

We test our hypotheses with seven model specifications. The specifications differ on the basis of the rebel strength variables in the violence outcome half of the model. A seventh specification uses the Heckman selection (Heckman, 1979) method for robustness reasons. There is a possibility that a selection effect exists in that only certain cases are mediated. If that is the case, it would mean the sample in the rebel violence against civilians equation is not random.

A host of independent variables are specified: war duration, war type, third-party intervention, battle deaths, number of recurrences, rebel strength, and regime polity score.

The data on rebel violence are from the University of Maryland’s Global Terrorism Database (GTD), 1970–2008 (La Free, 2011). For mediated civil wars episodes, GTD data were collected for two periods: (1) total number of violence acts by rebels between conflict start date and mediation start date and (2) total number of acts of violence committed by rebels between mediation start date and conflict end date. Because there are many overlapping, simultaneous mediations, we collect data between the date of the first mediation effort in a given episode and the end date of the conflict episode. For nonmediated conflict episodes, we collected number of violence acts during the entire conflict period.

We use the information included in the GTD variable specifying target/victim type (“targtype1”) to capture acts of violence against civilians. If the GTD indicated acts were targeting tourists, transportation, private and citizens’ properties, schools, and airports, we coded them as acts against civilians. We collected data on these violent attacks committed by rebels against civilians by carefully matching the names of “sideB” (rebels) with names of terror groups in the GTD to ensure these were the same groups committing acts of violence against civilians. Since the rebels are engaged in the peace processes, they are, by definition, internal to the peace process. This study does not consider groups that were external to the peace process.

Dependent Variables

Rebel violence after mediation

This variable refers to acts of rebel violence committed against civilians between the start date of the first mediation event in a given dispute and the end date of the conflict episode. This was coded as 1 if there were rebel attacks committed between the mediation start date and the conflict end date and 0 otherwise.

Mediation

This variable is coded as 1 if a third party mediated a given civil war episode as it was active or in the subsequent peace spell and 0 otherwise. Mediation data are from the CWM data set of DeRouen and Bercovitch (DeRouen, Bercovitch, & Pospieszna, 2011). The CWM data set uses the Uppsala Conflict Data Program (UCDP) Armed Conflict Termination (ACT) data set assembled by Kreutz (2010) as a framework. By definition, these armed conflicts between government and organized rebel group involve at least 25 battle-related deaths in a year. Mediation efforts tied to each war episode were identified by coders through searches of online sources such as Keesing’s World News Archive, as it provides consistent coverage over time and from a wide variety of foreign newspapers. After Keesing was exhausted, coders turned to other online sources such as the New York Times. We then used Regan, Frank, and Aydin’s (2009) diplomatic actions data set as cues to locate missed cases and then searched for these in Keesing’s and other online sources.

Independent Variables

Rebel violence prior to mediation

This relates to the first hypothesis and is tested in the mediation equations. Recall that Hypothesis 1 postulates that rebel violence against civilians increases the probability of mediation. We collected data on acts of rebel violence against civilians using the GTD data set. The variable is coded as 1 if there were acts of violence against civilians committed by rebels between the start date of conflict and the start date of the first mediation and 0 otherwise.

Incompatibility

This is tested in both equations and is coded as 1 if the war is fought over autonomy or secessionist claims (territorial wars) and 0 otherwise. The data are from Kreutz (2010). The literature on civil wars makes an important distinction between territorial and governmental incompatibilities. In general, scholars find that territorial wars last longer and are more likely to recur than wars over control of government (Bercovitch & DeRouen, 2004; Fearon, 2004). The higher recurrence rate is likely because territorial civil wars are less likely to end with military victory than are government wars. Therefore, because of the relatively more intractable nature of territorial wars, it is reasonable to expect that territorial wars are more likely to be mediated than governmental wars (DeRouen et al., 2011). Therefore, we use this variable as a possible factor affecting mediation. We also contend that territorial conflicts will have a positive effect on the probability of rebel violence against civilians in the outcome equation (e.g., Sri Lanka, Kashmir, Israel/West Bank, Turkey, Russia/Chechnya). Eck and Hultman (2007) report rebels in territorial wars kill almost 6 times more civilians than do governments.

War duration

This is tested in both equations. Using the episode start and end date from the UCDP Conflict Termination data set, we have calculated the number of days the war lasted. This measure is logged, as there are outliers in the data. For ongoing wars, we calculated duration of conflicts based on end of the data set, that is, December 31, 2008. Long conflicts can cause devastating financial costs, moral, and physical harm and can motivate violent activity.² Therefore, it is appropriate to control for war duration in the rebel violence equations. We also include war duration as a possible factor affecting mediation. Some scholars have reported mediation is more likely in longer wars (e.g., Bercovitch & Jackson, 2001; Greig, 2005; Greig & Diehl, 2006; Svensson, 2007). The duration of a conflict affects the acceptability of mediation from warring parties, as combatants may realize that they are unable to win by military means and thus are more willing to move on to the negotiating table to possibly find an agreement (Pillar, 1983; Schwarzer, 1998). This ties in to the mutually hurting stalemate argument of Zartman (1989, 1993).

Battle deaths

This is tested in the mediation equation. The literature generally indicates that costly and lengthy conflicts increase the probability of mediation (Bercovitch & Jackson, 2001; Greig, 2005; Greig & Diehl, 2006; Svensson, 2007). The number of casualties is a viable indicator of this cost. If the bloodshed is great, the warring parties might be more willing to agree on mediation. However, another view on dispute intensity implies exactly the opposite. This line of research states that the greater the intensity of a conflict, and the higher the level of costs borne by each party, the more polarized and entrenched will be their positions, and the more reluctant will they be to engage in conflict management and mediation (Burton, 1969). However, we side with scholars who argue that warring parties are more likely to request and accept mediation when a conflict becomes costly. We measure battle deaths as the yearly average using data from Lacina and Gleditsch (2005).

Internationalized war

This is tested in both equations and is coded as 1 in the presence of third-party military intervention and 0 otherwise. The variable is from the ACT data set compiled by Kreutz (2010). This variable could impact both mediation and the use of violence against civilians by the rebels. Pape (2003) reports suicide terrorist campaigns have been used to compel foreign military interventions to abandon a country. On the other hand, intervention could decrease overall violence because third-party military involvement could derail the use of rebel violence against civilians. It is difficult to predict the effect of intervention on the probability of mediation. Mediation may be expected to be less likely in internationalized internal conflicts as presumably the intervention is on behalf of one of the disputants. This disputant would be in a better position and would ostensibly prefer victory to a mediated settlement. However, third-party military involvement could increase the chances of successful mediation by signaling that the settlement will be enforced and implemented. In this line of reasoning, internationalized war would increase the probability of mediation. Internationalized civil wars may be more violent with the presence of a third party(ies) and this heightened intensity could also attract mediation

Rebel strength (access to lootable natural resources)

In order to address alternative explanations of the use of violence by rebels, we focus on Weinstein’s (2007) rebel organizational structure argument. He theorizes rebel groups emerging in environments rich in natural resources or with the external support of an outside patron tend to commit higher levels of indiscriminate violence in civil wars. However, movements that arise in resource-poor contexts and are formed on the basis of social endowments (ethnicity, ideology) perpetrate far fewer abuses and employ violence selectively and strategically. The reason rebel groups in resource-rich environments are more likely to abuse civilian population is that their structures, which are organized around economic endowments, are populated by opportunists, lack mechanisms for discipline, and to force cooperation (Weinstein, 2007, pp. 14–15). In contrast, groups organized around ethnicity or ideology draw committed recruits and maintain institutional structures that facilitate cooperation and discipline in the rebel ranks. Weinstein (2007) also argues that rebel groups with the external support of an outside patron tend to commit high levels of violence, because outside actors may offer financial resources that permit rebel groups to equip combatants with armament thus changing a character and structure of these groups. Moreover, resources generated from external patrons require little if any civilian labor, so groups with such support remain largely independent of civilian populations for their survival (Weinstein, 2007, p. 173). This line of reasoning is tested in the rebel violence equations.

Fearon (2004) finds that conflicts with rebel groups that have access to contraband goods tend to last significantly longer than other conflicts. Zahar (2010) observes that spoilers are more likely if they can sustain themselves with contraband financing. Therefore, we believe that contraband goods facilitate the capacity of rebels since they provide a lucrative source of finance to sustain a long-running war (Ross, 2006). We added Fearon’s drugs variable³ (referred to as contraband financing). This variable is coded 1 if evidence of significant contraband financing of rebels and 0 otherwise. Lujala, Gleditsch, and Gilmore (2005) focus on the presence of diamonds. They argue primary diamonds (those present in underground rock formations and mined by large companies) are less accessible to rebel forces during a conflict and therefore less likely to lead to war. By contrast, secondary diamonds are easier to find and can be exploited by rebels and thus increases the risk of conflict. We employ secondary diamonds variable from this source and code it as 1 if there is secondary diamond production during the conflict and 0 otherwise.

The remaining three variables that capture access to lootable natural resources are from Buhaug, Gates, and Lujala (2009) who argue easily extractable resources increase the rebel funds and the prospect of personal enrichment makes rebel recruitment easier. The dichotomous variables are drugs, coded 1 if there are drugs in production in the conflict zone; hydrocarbon, coded as 1 if there are hydrocarbons in the conflict zone; and gemstones, coded as 1 if there are gemstones (including secondary diamonds) in the conflict zone (each coded as 0 otherwise). Based on Weinstein (2007), we also control for third-party support for rebels. The variable intervener support is coded as 1 if the intervener supports rebel or opposition groups and 0 otherwise. These data are from the International Military Interventions data set (Kisangani & Pickering, 2008).

Regime type

This is tested in the mediation equations. We use the average score of the polity2 variable of a country at the beginning and in the end of the conflict as an indicator of the country’s regime type. We converted this variable to a 1–21 scale. The Polity IV data (the source of the polity2 variable) are from Marshall, Jaggers, and Gurr (2011). The impact of regime type on mediation is not entirely clear. Scholars argue that democracy should increase chances of mediation because democracies are more amenable to peaceful resolution of disputes (DeRouen & Goldfinch, 2005; Dixon, 1994, pp. 14–32; Greig, 2005). Wallensteen, DeRouen, Bercovitch, and Möller (2009), for example, report that some level of democracy is required before mediation can take place. However, Savun (2009) finds that democracy has a negative impact on the probability of mediation in interstate war. Democracies may consider themselves capable of reaching agreements using internal democratic processes.

Stratum

This variable is tested in the mediation equations and reflects whether the civil war is in its first iteration, second iteration, third iteration, and so on. It is likely that successive wars are not independent and memories of previous events could shape the outcome of subsequent wars (DeRouen, Bercovitch, & Wei, 2009). DeRouen, Bercovitch, and Pospieszna (2011) find a negative effect of previous war on the likelihood of mediation, which may suggest that mediators may be more reluctant to intervene in a civil war that recurred. We control for this variable in the mediation equations only. This variable is from Kreutz (2010).

Findings

The results are reported in Table 1. Starting with the mediation equations we find, as expected, the use of violence against civilians by rebels has a positive effect on the probability of mediation thus confirming the first hypothesis. Recall that these acts of violence are measured prior to the beginning of mediation. This finding is robust across all seven specifications. Oddly, longer wars are less likely to be mediated. This may be because the international community has given up on these wars or because some very long wars are stalemated and relatively low cost in terms of battle-related deaths. For example, some ethnic civil wars fought in the periphery in Myanmar and India have lasted decades with little to no mediation. This finding could also be a result of an artifact in the measure of war duration. Recall that many wars are ongoing at the expiration of the data set thus the true duration of some of the conflicts is potentially underestimated.

Table 1.

Bivariate Probit Models: Mediation and Rebel Violence Against Civilians, 1970–2008.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
DV2 = Rebel violence against civilians after mediation
Territorial war	0.099 (0.165)	0.098 (0.162)	0.101 (0.164)	0.114 (0.173)	0.117 (0.170)	0.182 (0.175)	0.017 (0.269)
Internationalized war	–0.755* (0.289)	–0.811* (0.291)	–0.778* (0.293)	–0.835* (0.306)	–0.856* (0.306)	–0.821* (0.307)	–0.846* (0.417)
War duration (logged)	0.068* (0.029)	0.069* (0.028)	0.067* (0.029)	0.050^† (0.028)	0.048^† (0.028)	0.046 (0.028)	0.054 (0.044)
Mediation	1.558** (0.110)	1.555** (0.110)	1.542** (0.113)	1.601** (0.117)	1.617** (0.120)	1.591** (0.118)
Contraband financing		0.565* (0.266)
Secondary diamonds			0.133 (0.134)
Drugs				0.037 (0.056)
Hydrocarbon					0.128 (0.091)
Gemstones						0.213^† (0.111)
Intervener support							0.078 (0.140)
Constant	–0.871** (0.204)	–0.898** (0.199)	–0.886** (0.201)	–0.777** (0.197)	–0.823** (0.202)	–0.819** (0.192)	1.025* (0.384)
DV1 = Mediation
War duration (logged)	–0.064* (0.032)	–0.069* (0.028)	–0.066* (0.032)	–0.048^† (0.028)	–0.048^† (0.028)	–0.051^† (0.030)	–0.066^† (0.038)
Average battle death	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
Territorial war	–0.062 (0.174)	–0.026 (0.187)	–0.033 (0.188)	–0.044 (0.192)	0.011 (0.191)	0.032 (0.187)	0.261 (0.209)
Internationalized war	1.048** (0.321)	1.093** (0.331)	1.037** (0.327)	1.066** (0.341)	1.078** (0.344)	1.078** (0.342)	1.353** (0.330)
Polity2 average	0.002 (0.009)	0.000 (0.000)	0.000 (0.002)	0.003 (0.005)	0.000 (0.000)	0.000 (0.000)	–0.003 (0.018)
Stratum	–0.020 (0.092)	–0.036 (0.133)	–0.069 (0.159)	–0.034 (0.055)	–0.128 (0.091)	–0.012 (0.040)	–0.110 (0.158)
Violence prior Mediation	1.348** (0.121)	1.344** (0.133)	1.345** (0.139)	1.352** (0.135)	1.344** (0.124)	1.311** (0.130)	0.733** (0.179)
Constant	–0.544* (0.225)	–0.519* (0.204)	–0.528* (0.214)	–0.684** (0.213)	–0.649** (0.205)	–0.674** (0.209)	–0.665* (0.290)
N	248	248	248	231	231	231	238
Wald test	1,617.47**	1.042	63.264**	29.269**	10.476**	.908	5.150*
Censored observation							162
Uncensored observation							76

Note. Robust standard error clustered on dyad in parentheses; column 7 is a Heckman probit model. Two-tailed. DV = dependent variable.

^†p < .1. *p < .05. **p < .005.

Internationalized wars are more likely to be mediated. One line of reasoning above pointed out intervention could heighten the intensity of the war and attract mediation. None of the other variables are significant.

Turning attention to the second equation, we find rebel violence against civilians determined by mediation (thus confirming the second hypothesis), whether the war is internationalized, war duration and two of the natural resource measures. Internationalized wars have less acts of violence against civilians used by the rebels. This is somewhat unexpected as Pape (2003) depicts suicide terror as a common response to foreign intervention/occupation. Recent events in Iraq and Afghanistan in the 21st century indicate a positive relationship between intervention and rebel violence. Perhaps third-party military involvement may deter the use of violence by rebels, as it becomes more difficult to organize and carry out attacks of violence against civilians. Longer wars are more likely to experience rebel violence against civilians. This could be an indication that violence against civilians is used after a war has endured without yielding a military victory or concessions.

The presence of drugs and gems each has a positive effect on rebel violence against civilians after mediation. As Weinstein (2007) argues, resource-rich environments create an opportunity for the emergence of rebel groups that will be organized around economic endowments. It is difficult for rebel leaders to maintain discipline in this type of organizational structure. The final column in Table 1 includes findings on the effect of intervention on behalf of the rebels. The variable is positive and significant as predicted by Weinstein’s theory of rebel violence.

The findings shed light on mediation, rebel violence, and spoiler literatures. Hypothesis 1 is supported as the results in Table 1 demonstrate that rebel violence against civilians increases the probability of mediation. Hypothesis 2, derived from spoiler theory, is supported as mediation has a positive significant effect on the probability of rebel violence against civilians. Moreover, we report that two forms of natural resources—drugs and gems—increase the likelihood of violence against civilians, which supports Weinstein’s theory of rebel violence.

The underlying theoretical explanation of this relationship is that rebel violence is indicative of a costly war for both sides that increases the willingness to accept mediation. Consistent with spoiler theory, the diplomatic efforts of mediators appear to mobilize groups whose status is threatened. This is a probabilistic argument and we do not claim mediation will inevitably lead to violence against civilians committed by the rebels. Rather we demonstrate that mediated talks between governments and rebels shape the potential for spoiler violence.

Finally, it should be noted that the results for the Heckman model in the final column are virtually identical to the bivariate probit specifications with the caveat, of course, that mediation does not appear as an independent variable in the outcome equation because only those cases that are mediated (uncensored) appear in the outcome equation.

The Wald test of independent equations is significant in four of the seven models indicating the null hypothesis of independent equations can be rejected in these cases. Robustness checks were also performed. We also ran the first six models with average battle deaths, polity score, and each of the resource variables in both equations rather than only one. The results (available from the authors) are remarkably similar to those presented in Table 1. The only notable differences pertain to the resource variables. In the robustness checks, the presence of gemstones in the conflict zone is not significant in the rebel violence equation; however, the impact of the presence of drugs and hydrocarbon is positive and significant. Fearon’s (2004) variable for contraband financing as well as drugs and hydrocarbon variables are each negative and significant in the mediation equations in the robustness checks. Notably, the key results related to mediation and rebel violence against civilians are unchanged in the robustness models.

Conclusion

This article has expanded the literature on an important yet understudied relationship between rebel violence against civilians, civil war, and mediation. The statistical analysis reveals the following:

Mediation is more likely after acts of rebel violence against civilians. Previous research (e.g., Bercovitch & Jackson, 2001; DeRouen et al., 2011; Greig, 2005; Svensson, 2007) demonstrated mediation is attracted to violent, costly conflicts. Rebel violence against civilian certainly qualifies as costly. Mediation is prescribed in order to prevent the killings of civilians or to stem the spread of violence beyond the borders.

Long wars are less likely to be mediated but increase the probability of rebel use of violence against civilians. The former could be because some of the longest civil wars are low in casualties and have not been mediated. We contend the latter is indicative of rebels who have been unable to secure a timely victory and have turned to violence against civilians out of desperation. Colombia provides an example of a long civil war in which rebels eventually turned to violence against civilians.

Mediation is more likely if the civil war is internationalized. These wars are typically more violent and as such more likely to attract mediation.

Rebel violence against civilians is less likely in internationalized wars. Third-party military involvement may make rebel violence logistically difficult to carry out.

Mediation is linked to the probability of rebel violence attacks on civilians. This finding empirically corroborates Stedman’s (1996, 1997) spoiler theory. Elements within the rebel constituency who sense that progress toward peace may undermine their position can turn to violence against civilians in an attempt to derail the peace process.

Weinstein’s (2007) rebel violence findings are partially borne out in terms of all gems, drugs, and foreign intervention on behalf of rebels. While we highlight the role of mediation on rebel violence against civilians in this article, we are well aware this type of rebel violence has multiple determinants.

In an era when civil war is the dominant form armed violence and mediation and the use of violence against civilians are on the rise, the empirical study of spoiler theory demands close attention. One obvious question for further scrutiny is whether there are mediation strategies that are relatively less likely to lead to spoiler activity. For example, if the mediator exhibits bias toward the government and/or the mediator employs a directive (forceful) style of mediation, potential spoilers could begin to question the process and become concerned any negotiated settlement will not be flexible and accommodating. Potential spoilers could feel forced out of the process. Relatedly, the findings here tell us the dangerous potential of lootable resources. Mediation strategies tied specifically to complicated resource issues should be considered in terror-prone civil wars (e.g., Colombia).

The results also suggest that peace agreement design in the presence of mediation and rebel violence against civilians also merits further analysis. Some agreements might be more successful than others in overcoming spoilers. Power-sharing provisions could set potential spoilers at ease. Special regional concessions (e.g., autonomy, language rights, being allowed to form a regional party, religious and cultural allowances) could similarly defuse the spoiler problem. The issue of natural resource revenue distribution must be carefully considered in peace agreements (e.g., the final agreement in the Aceh civil war).

A final word on the impact of the long-term impact of spoilers is warranted. Newman and Richmond (2006) remind us that an uptick in violence by spoilers during a peace process does not necessarily mean the process is doomed. On the contrary, they assert that violence could be emblematic of peace processes moving in the right direction and that this progress is threatening “marginalized groups [who] see the processes undermining their position” (Newman & Richmond, 2006, p. 7).

In other words, if potential spoilers fear a loss of status, they may change the underlying conflict dynamics. Thus, a further line of inquiry could entail the long-term effect of mediation in the presence of spoilers.

Our study sends a message to mediators that their efforts may trigger insurgent violence against civilians. We are not suggesting that this threat should stand in the way of mediation efforts. On the contrary, by highlighting this possibility, mediation strategies can prepare for this eventuality. Mediation design should head off this form of violence against civilians by designing especially inclusive peace processes. Insurgent factions whose status is threatened by a peace process might not resort to violence if the third-party mediator can get the government to credibly commit that the interests of those who might potentially use or have used violence against civilians will have a seat at the table in the postagreement era.

Footnotes

Authors’ Note

An earlier version of this article was presented at the European Consortium for Political Research Joint Sessions Political Institutions and Political Violence Workshop, University of St. Gallen, Switzerland, April 12–17, 2011.

Declaration of Conflicting Interests

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

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We gratefully acknowledge the financial support of the Folke Bernadotte Academy of Sweden for the creation of the mediation dataset and the Fritz Thyssen Foundation of Germany.

Notes

References

Azam

J.-P.

Hoeffler

(2002). Violence against civilians in civil wars: Looting or terror? Journal of Peace Research, 39, 461–485.

Balcells

(2010). Rivalry and revenge: Violence against civilians in conventional civil wars. International Studies Quarterly, 54, 291–313.

Beardsley

(2008). Agreement without peace? International mediation and time inconsistency problems. American Journal of Political Science, 52, 723–740.

Beardsley

(2011). The mediation dilemma. Ithaca, NY: Cornell University Press.

Bercovitch

Anagnoson

J. T.

Wille

D. J.

(1991). Some conceptual issues and empirical trends in the study of successful mediation in international relations. Journal of Peace Research, 28, 7–17.

Bercovitch

DeRouen

(2004). Mediation in internationalized ethnic conflicts: Assessing the determinants of a successful process. Armed Forces and Society, 30, 147–170.

Bercovitch

Gartner

S. S.

(2006). Is there a method in the madness of mediation? Some lessons for mediators from quantitative studies of mediation. International Interactions, 32, 329–354.

Bercovitch

Jackson

(2001). Negotiation or mediation? An exploration of factors affecting the choice of conflict management in international conflict. Negotiation Journal, 17, 59–77.

Boulden

(2009). Terrorism and civil wars. Civil Wars, 11, 5–21.

10.

Buhaug

Gates

Lujala

(2009). Geography, rebel capability, and the duration of civil conflict. Journal of Conflict Resolution, 53, 544–569.

11.

Burton

J. W.

(1969). Conflict and communication: The use of controlled communication in international relations. London: Macmillan and New York: Free Press.

12.

Coleman

P. T.

(2000). Intractable conflict. In Deutsch

Coleman

P. T.

(Eds.), The handbook of conflict resolution: Theory and practice (p. 43, 428, 431, 433). San Francisco, CA: Jossey-Bass.

13.

Crenshaw

(1981). The causes of terrorism. Comparative Politics, 13, 379–399.

14.

Crocker

C. A.

Hampson

F. O.

Aall

P. R.

(2004). Taming intractable conflicts: Mediation in the hardest cases. Washington, DC: U.S. Institute of Peace Press.

15.

Cronin

(2008). How terrorism ends: Lessons from the decline and demise of terrorist groups. Princeton, NJ: Princeton University Press.

16.

Cronin

(2010, 5). When should we talk to terrorists? (U.S. Institute of Peace Special Report 240). Retrieved from http://www.usip.org/files/resources/SR240Cronin.pdf

17.

Darby

(2001). The effects of violence on peace processes. Washington, DC: U.S. Institute of Peace.

18.

DeRouen

Bercovitch

Pospieszna

(2011). Introducing the Civil Wars Mediation (CWM) dataset. Journal of Peace Research, 48, 663–672.

19.

DeRouen

Bercovitch

Wei

(2009). Duration of peace and recurring civil wars in Southeast Asia and the Pacific. Civil Wars, 11, 103–120.

20.

DeRouen

Goldfinch

(2005). Putting the numbers to work: Implications for conflict prevention. Journal of Peace Research, 42, 27–45.

21.

Dixon

(1994). Democracy and the peaceful settlement of international conflict. American Political Science Review, 88, 14–32.

22.

Downes

A. B.

(2007). Draining the sea by filling the graves: Investigating the effectiveness of indiscriminate violence as a counterinsurgency strategy. Civil Wars, 9, 420–444.

23.

Eck

Hultman

(2007). One-sided violence against civilians in war: Insights from new fatality data. Journal of Peace Research, 44, 233.

24.

Fearon

(2004). Why do some civil wars last so much longer than others? Journal of Peace Research, 41, 275–301.

25.

Findley

M. G.

Young

J. K.

(2012). Terrorism and civil war: A spatial and temporal approach to a conceptual problem. Perspectives on Politics, 10, 285–305.

26.

Gartner

S. S.

(2013). Deceptive results: Why mediation appears to fail but actually succeeds. Penn State Journal of International Law and International Affairs, 2, 27–37.

27.

Gartner

S. S.

Bercovitch

(2006). Overcoming obstacles to peace: The contributions of mediation to short-lived conflict settlements. International Studies Quarterly, 50, 319–328.

28.

Goodwin

Skocpol

(1989). Explaining revolutions in the contemporary Third World. Politics and Society, 17, 489–509.

29.

Greig

J. M.

(2005). Stepping into the fray: When do mediators mediate? American Journal of Political Science, 49, 249–266.

30.

Greig

J. M.

Diehl

P. F.

(2006). Softening up: Making conflicts more amenable to diplomacy. International Interactions, 32, 355–384.

31.

Greene

W. H.

(2011). Econometric analysis (7th ed.). Upper Saddle River, NJ: Prentice Hall.

32.

Heckman

(1979). Sample selection bias as a specification error. Econometrica, 47, 153–161.

33.

Herz

Nogueira

J. P.

(2002). Ecuador vs. Peru: Peacemaking amid rivalry. Boulder, CO: Lynne Rienner.

34.

Hoffman

(1998). Inside terrorism. New York, NY: Columbia University Press.

35.

Hultman

(2007). Battle losses and rebel violence: Raising the costs for fighting. Terrorism and Political Violence, 19, 205–222.

36.

Hultman

(2009). The power to hurt in civil war: The strategic aim of RENAMO violence. Journal of Southern African Studies, 35, 821–834.

37.

Humphreys

Weinstein

(2006). Handling and manhandling in civil war. American Political Science Review, 100, 429–447.

38.

Kalyvas

(1999). Wanton and senseless: The logic of massacres in Algeria. Rationality and Society, 11, 243–285.

39.

Kalyvas

(2004). The paradox of terrorism in civil war. The Journal of Ethics, 8, 97–138.

40.

Kalyvas

(2006). The logic of violence in civil wars. Cambridge, England: Cambridge University Press.

41.

Kisangani

E. F.

Pickering

(2008). International military intervention, 1989–2005. Inter-University Consortium for Political and Social Research, Data Collection No. 21282, University of Michigan, Ann Arbor.

42.

Kreutz

(2010). How and when armed conflicts end: Introducing the UCDP conflict termination dataset. Journal of Peace Research, 47, 243–250.

43.

Kreutz

Brosche

(2013). A responsibility to talk: Mediation and violence against civilians. Canadian Foreign Policy Journal, 19, 26–38.

44.

Kriesberg

Northrup

T. A.

Thorson

S. J.

(1989). Intractable conflicts and their transformation. New York, NY: Syracuse University Press.

45.

Krueger

Laitin

D. D.

(2008). Kto Kogo? A cross-country study of the origins and targets of terrorism. In Keefer

Loayza

(Eds.), Terrorism, economic development, and political openness (pp. 148–173). New York, NY: Cambridge University Press.

46.

Kydd

Walter

(2006). The strategies of terrorism. International Security, 31, 49–80.

47.

La Free

G. L.

(2011). Generating terrorism event data bases: Results from the global terrorism database, 1970–2008. College Park, MD: University of Maryland.

48.

Lacina

Gleditsch

N. P.

(2005). Monitoring trends in global combat: A new dataset of battle deaths. European Journal of Population, 21, 145–166.

49.

Lake

D. A.

(2002). Rational extremism: Understanding terrorism in the twenty-first century. Dialog-IO, 1, 15–29.

50.

Lujala

Gleditsch

N. P.

Gilmore

(2005). A diamond curse? Civil war and a lootable resource. Journal of Conflict Resolution, 49, 538–563.

51.

Marshall

Jaggers

Gurr

T. R.

(2011). Polity IV project: Dataset users’ manual. Retrieved from http://www.systemicpeace.org/inscr/p4manualv2012.pdf

52.

Negroponte

D. V.

(2012). Seeking peace in El Salvador: The struggle to reconstruct a nation at the end of the cold war. Basingstoke, England: Palgrave Macmillan.

53.

Neumann

(2007). Negotiating with terrorists. Foreign Affairs, 86, 128–139.

54.

Newman

Richmond

(2006). The impact of spoilers on peace processes and peacebuilding (UN University Policy Brief No. 2). Tokyo, Japan: United Nations University Press.

55.

Nilsson

(2010). Turning weakness into strength: Military capabilities, multiple rebel groups and negotiated settlements. Conflict Management and Peace Science, 27, 253–271.

56.

Pape

R. A.

(2003). The strategic logic of suicide terrorism. American Political Science Review, 97, 343–361.

57.

Piazza

J. A.

(2008). A supply-side view of suicide terrorism: A cross-national study. Journal of Politics, 70, 28–39.

58.

Pillar

P. R.

(1983). Negotiating peace: War termination as a bargaining process. Princeton, NJ: Princeton University Press.

59.

Regan

P. M.

Frank

R. W.

Aysegul

(2009). Diplomatic interventions and civil war: A new dataset. Journal of Peace Research, 46, 135–146.

60.

Regan

P. M.

Stam

A. C.

(2000). In the nick of time: Conflict management, mediation timing, and the duration of interstate disputes. International Studies Quarterly, 44, 239–260.

61.

Ross

(2006). A closer look at oil, diamonds, and civil war. Annual Review of Political Science, 9, 265–300.

62.

Sandler

(2003). Collective action and transnational terrorism. World Economy, 26, 779–802.

63.

Savun

(2009). Mediator type and effectiveness of information-provision strategies in the resolution of international conflict. In Bercovitch

Gartner

(Eds.), International conflict mediation: New approaches and findings (pp. 96–114). London, NY: Routledge.

64.

Schwarzer

(1998). The peaceful settlement of interstate conflict: Saar, Austria, and Berlin. Journal of Peace Research, 35, 743–757.

65.

Stedman

S. J.

(1996). Negotiation and mediation in internal conflicts. In Brown

M. E.

(Ed.), The international dimensions of internal conflict (pp. 341–377). Cambridge, MA: MIT Press.

66.

Stedman

S. J.

(1997). Spoiler problems in peace processes. International Security, 22, 5–53.

67.

Stedman

S. J.

(2000). Spoiler problems in peace processes. In Stern

Druckman

(Eds.), International conflict resolution after the cold war (pp. 178–221). National Research Council. Washington, DC: The National Academies Press.

68.

Svensson

(2007). Bargaining, bias and peace brokers: How rebels commit to peace. Journal of Peace Research, 44, 177–194.

69.

Thomas

(2014). Rewarding bad behavior: How governments respond to terrorism in civil war. American Journal of Political Science, 58, 804–818.

70.

Toros

(2008). We don’t negotiate with terrorists: Legitimacy and complexity in terrorist conflicts. Security Dialogue, 38, 407–426.

71.

Wallensteen

DeRouen

Bercovitch

Möller

(2009). Democracy and mediation in territorial civil wars in Southeast Asia and the South Pacific. Asia Europe Journal, 7, 241–264.

72.

Weinstein

J. M.

(2007). Inside rebellion: The politics of insurgent violence. Cambridge, MA: Cambridge University Press.

73.

Wiggins

Poi

(2010, 8). Bivariate probit with partial observability and a single dependent variable. Stata FAQ. Retrieved from http://www.stata.com/support/faqs/statistics/bivariate-probit-with-partial-observability/

74.

Wilkenfeld

Young

K. J.

Quinn

D. M.

Asal

(2005). Mediating international crises. New York, NY: Routledge.

75.

Wilkerson

(2008). Security and democracy in El Salvador: An undeniable connection. Stanford Journal of International Relations, 10, 32–41.

76.

Wilkinson

(2006). Terrorism versus democracy: The liberal state response. London, NY: Routledge.

77.

Wood

R. M.

(2010). Rebel capability and strategic violence against civilians. Journal of Peace Research, 47, 601–614.

78.

Zahar

M.-J.

(2001). Protégés, clients, cannon fodder: Civilians in the calculus of militias. In Adebajo

Sriram

C. L.

(Eds.), Managing armed conflicts in the 21st century (pp. 107–129). London, England: Frank Cass.

79.

Zahar

M.-J.

(2010). SRSG mediation in civil wars: Revisiting the “Spoiler” debate. Global Governance, 16, 265–280.

80.

Zartman

I. W.

(1989). Ripe for resolution: Conflict and intervention in Africa. Oxford, England: Oxford University Press.

81.

Zartman

I. W.

(1993). The unfinished agenda: Negotiating internal conflicts. In Licklider

(Ed.), Stopping the killing: How civil wars end (pp. 20–35). New York: New York University Press.

82.

Zartman

I. W.

(2000). Ripeness: The hurting stalemate and beyond. In Stern

P. C.

Druckman

(Eds.), International conflict resolution after the cold war (pp. 225–250). Washington, DC: National Academy Press.