Fueling Factionalism? The Impact of Peace Processes on Rebel Group Fragmentation in Civil Wars

Abstract

While peace processes increase the likelihood that a civil war is resolved, they can also complicate peace by increasing the risk of rebel fragmentation. In this article, we argue that negotiations exacerbate pre-existing structural and substantial divisions within rebel organizations, therefore increasing the likelihood of a rebel split. More specifically, we put forward a theoretical framework that specifies why factions within a rebel group may disagree with the onset of negotiations, the conclusion of a peace agreement, or the implementation of an agreement—and thus break away during the peace process. We empirically assess the merit of this framework by systematically comparing the impact of these phases in a peace process on the fragmentation of rebel organizations. Using data that more accurately reflect the moment a rebel split takes place than earlier studies, we find that peace processes have a greater substantial impact on rebel fragmentation than previously assumed.

Keywords

civil war fragmentation rebel splintering negotiation mediation peace process

Introduction

Peace processes are intended to bring conflict to an end, but often fail to do so. In 1976, for example, negotiations between the Philippine government and the Moro National Liberation Movement (MNLF) resulted in the conclusion of the Tripoli Agreement, but the civil war in Mindanao resumed after less than five months. Not only did the MNLF itself return to arms, dissatisfaction about the provisions of the agreement also led to the emergence of multiple breakaway factions, including the Moro Islamic Liberation Front (MILF) (Plank 2015). In the following years, the international community undertook several renewed attempts at resolving the conflict between the Philippine government and the MNLF, eventually culminating in the signing of the Jakarta Accord in 1996. Many MNLF members were reportedly dissatisfied with the implementation of the agreement, however, because the rebel group’s leaders monopolized the benefits of the accord. This discontent resulted in large-scale defection to the MILF and Abu Sayyaf (another rebel organization that had split from the MNLF), as well as the creation of two new splinter groups (Plank 2015). The MILF was not immune to splintering either: in 2011, the Bangsamoro Islamic Freedom Movement (BIFM) broke away from the group because it did not agree with the MILF leadership negotiating on autonomy rather than full independence (Chalk 2013).

The case of Mindanao is by no means unique: peace processes often lead to the emergence of splinter factions that subsequently spoil the peace (Reiter 2016). As such, negotiations can have the paradoxical consequence of prolonging the conflict they are intended to resolve: the more rebel groups are involved in a conflict, the longer they tend to last (D. E. Cunningham 2006). What is it about peace processes that makes rebel splintering more likely? And in which phase of a peace process are splits most likely to occur? In this article, we argue that peace negotiations exacerbate preexisting structural and substantial divisions within rebel organizations,¹ therefore increasing the likelihood of a rebel split. More specifically, we put forward a theoretical framework that specifies why factions within a rebel group may disagree with the onset of negotiations, the conclusion of a peace agreement, or the implementation of an agreement—and thus break away during the peace process.

The link between peace processes and rebel fragmentation has not gone unobserved in the negotiations literature—quite the contrary. In a seminal study, Stedman argues that the greatest risk to peace comes from spoilers that break away from a rebel group and continue to fight because the “peace emerging from negotiations threatens their power, worldview, and interests” (Stedman 1997, 5). Moreover, Zartman notes how rebel fragmentation might occur because a subgroup within a rebel movement may try to abandon the collective negotiation effort in order to reach a better deal with the government side than this subgroup would have obtained as a result of the collective negotiation effort (Zartman 1995). In addition, Cronin points out that “Despite the successful negotiated outcomes that can result between major parties, a common effect of political processes is the splintering of groups into factions that support the negotiations (or their outcome) and those that do not” (Cronin 2006, 25).

In short, existing literature seems to suggest that peace processes increase the likelihood of rebel fragmentation—but there are few quantitative studies that empirically examine this link. One important exception is a previous study by Olson Lounsbery and Cook, which finds that international mediation in civil wars indeed seems to make rebel fragmentation more likely (Olson Lounsbery and Cook 2011).² We build on this study in this article, but make important theoretical and methodological innovations.

Theoretically, we go beyond just considering mediation efforts, also looking at the impact of bilateral peace process in which no third parties are involved. A second theoretical contribution is that we examine when peace processes are most likely to lead to the fragmentation of rebel organizations. In this article, we put forward a theoretical framework that specifies why rebel splintering is more likely in three distinct phases of a peace process that are commonly recognized: the start of a peace process, the signing of a peace agreement, and the implementation of an agreement (Duursma 2014). We empirically assess the merit of this framework by systematically comparing the impact of these phases in a peace process on the fragmentation of rebel organizations. This is not only important from a theoretical point of view, but also from a practitioner’s perspective: knowing during which stages of a peace process rebel groups are most at risk of fragmenting and why might provide insights regarding how to prevent rebel fragmentation.

Methodologically, we depart from the study by Olson Lounsbery and Cook by operationalizing the fragmentation of rebel organizations in a more precise temporal manner. Olson Lounsbery and Cook focus on the moment splinter groups first engage in armed conflict with the government (Olson Lounsbery and Cook 2011), but there is often a significant lag between the moment a splinter group breaks away and the moment it engages in armed conflict for the first time. We therefore focus on the moment a rebel split actually occurs. This is not a trivial difference: only 31% of the splinter factions emerged in the UCDP dataset in the same year they broke away. The research design section of this article further addresses this issue.³

In short, this article revisits the apparent association between peace processes and rebel fragmentation. A robust assessment of this association is important since it sheds light on possible negative consequences of peace efforts. We find that peace processes have a greater substantial impact on rebel fragmentation than previously assumed, significantly increasing the number of rebel splits at the onset of a peace process, while a peace process is ongoing, and when a peace agreement is concluded. We do not find a significant increase in rebel splits during the implementation phase of a peace process.

The Causes and Consequences of Rebel Fragmentation

In recent years, the literature on civil wars has increasingly recognized the importance of rebel fragmentation. Starting from the observation that warring parties are frequently marred by internal divisions, one strand of this literature sheds light on the negative consequences of rebel fragmentation, demonstrating that it makes civil wars longer, more violent, and more difficult to resolve (Lidow 2016). Indeed, Kathleen Gallagher Cunningham and colleagues demonstrate that internally divided rebels use more violence against the state, civilians, and co-ethnic factions (K. G. Cunningham et al. 2012),⁴ while David Cunningham shows that a higher number of actors involved in a conflict is associated with an increase in the duration of civil wars, because it makes the achievement of a negotiated settlement more difficult (D. E. Cunningham 2006). Rudloff and Findley further show that even after a settlement is reached, rebel splintering tends to affect the sustainability of the post-civil war peace, as it hastens the recurrence of civil wars (Rudloff and Findley 2016). Rebel splits thus greatly exacerbate the instability caused by civil conflict.

Given the detrimental effects of rebel fragmentation, another strand of the civil wars literature investigates what causes the emergence of internal divisions in warring groups. One of the prevailing arguments in this literature is that the cohesion of rebel organizations, as well as broader rebel movements, depends on the strength of the social bases underpinning them. More specifically, Staniland demonstrates that organizations with fragile processes of central control will have more trouble retaining the unity and loyalty of key leaders within the movement (Staniland 2012, 2014).⁵ This finding is echoed in a study by Asal et al., which finds that ethnopolitical groups with competing leadership structures are more at risk of fragmentation (Asal et al. 2012).

In addition to leadership structure, leadership qualities are also likely to affect organizational cohesion. Indeed, Doctor (2020) finds that rebel groups headed by leaders with combat experience are more likely to remain cohesive, while those run by leaders with political experience are more likely to fragment. Nagel and Doctor (2020) further demonstrate that rebel leaders’ socialization strategies also matter for organizational cohesion: while top-down socialization processes (such as political education) may generate cohesion, bottom-up processes like the perpetration of conflict-related sexual violence are found to increase the likelihood of organizational splits. Although the perpetration of conflict-related sexual violence increases cohesion at the battalion level, this paradoxically makes sub-commanders more likely to split from the parent group because they are more confident that their subordinates will follow them.

Other authors have demonstrated that the cohesion of rebel organizations is also affected by rebel leaders’ capacity to attract support from either external sponsors or the networks on which they rely. Lidow (2016), for example, demonstrates that the risk of factionalization decreases when rebel leaders have access to external financing (but increases when they have access to lootable resources). Tamm shows instead that external support can lead to a split in a rebel organization if resources are allocated to both the leader of a rebel organization and his rival, allowing the rival to defy the leader but not to overthrow him—thus making an organizational split the most likely outcome (Tamm 2016, 2019). Mosinger (2019) advances a similar argument, but based on the degree to which rival rebel leaders can cultivate the loyalties of their recruitment and operational networks: if the power bases of the incumbent leader and his challenger are relatively equal, “neither side has the strength to win, and the rebel group is likely to split into two splinter groups” (Mosinger 2019, 941).

Looking at the policies that states employ vis-à-vis rebel groups on their territory, existing scholarship further finds that both repression (McLauchlin and Pearlman 2012; Seymour et al. 2016) and accommodation (K. G. Cunningham 2014; Seymour et al. 2016; Fjelde and Nilsson 2018) increase the chance of fragmentation in rebel movements.⁶ Military action is also associated with rebel fragmentation in several ways: Seymour and colleagues, for example, find that the onset of a civil war is associated with an increase in movement fragmentation (Seymour et al. 2016), while other authors have pointed out the importance of battlefield gains and losses as a trigger for splits in rebel groups (Christia 2012; Woldemariam 2016, 2018).

Lastly, and most relevant for the purposes of this article, attempts to end civil wars have also been found to foster fragmentation. Fjelde and Nilsson (2018) and Lidow (2016) both find that the onset of peace talks increases the risk of rebel fragmentation, while Olson Lounsbery and Cook (2011) further find that mediation, too, can lead to rebel group splintering. Our theoretical framework builds on these findings. We recognize that different conflict dynamics can (interact to) cause rebel fragmentation; a peace process is just one of these dynamics, but its effect has thus far been underestimated and undertheorized. This article seeks to fill this gap.

Peace Processes and Rebel Splits

Peace processes are intended to bring an end to a conflict between different parties, but often reveal the divisions within them. Indeed, peace processes tend to exacerbate pre-existing fault lines in rebel organizations, which can be of either a structural or a substantive nature. Firstly, structural divisions reflect the fact that rebel organizations often do not have a unitary organizational structure. Especially in groups that operate in large swathes of territory, rebel organizations instead tend to consist of multiple subunits that operate with relative autonomy. In such cases, rather than having a centralized command-and-control structure, the leadership of a rebel organization authorizes regional subcommanders to administer violence on its behalf (Johnston 2008). While such delegation of authority may increase organizational effectiveness in wartime, structural divisions can also act as fault lines at the basis of organizational splits when a peace process is taking place. If rebel leaders cannot convince regional subcommanders that a negotiated settlement is in their best interest, and do not exercise sufficient control to force them to lay down their arms, factions of the organization may decide to break away to continue the armed struggle.

Substantial divisions, on the other hand, relate to divergences in the aims or priorities of different rebels. Like in any other political organization, differences of opinion are always present in rebel groups. As explained by Woldemariam, “rebel organizations are amalgamations of identifiable groups that possess distinct interests. To a greater or lesser degree, they are coalitions united by the pursuit of a common goal—the violent contestation of state power—rather than an identical set of preferences” (Woldemariam 2018, 38). Aware of the risk of this internal heterogeneity, some rebel organizations actively attempt to defer discussions over matters that could lead to internal disagreement for as long as they can. For example, in its struggle for an independent Western Sahara, the Polisario Front prefers to postpone deliberations about ideological issues until after independence is achieved, because it “does not want to split the unity of the nationalist movement while the struggle for self-determination is still taking place” (Hacene-Djaballah 1985, 89). During a peace process, however, such discussions can often no longer be avoided, because negotiations force rebel leaders to make concrete their demands vis-à-vis the government—which others in the organization might not agree with. If they cannot convince the rebel leadership to change its position, certain factions may decide to split from the rebel organization in the hope that they can secure their own place at the negotiating table (Zartman 1995).

In some cases, rebel splintering may also be a rational decision on the part of a rebel leader, rather than an unintended consequence of a peace process. As Lidow explains, “In certain situations there may be an incentive for the group to split apart to increase its bargaining leverage during peace talks. More factions, for example, could provide more seats at the bargaining table and a better total payout” (Lidow 2016, 48-49). If the leadership of a rebel organization believes it can exercise effective control over its factions throughout a peace process (while giving its negotiating partner the impression that it does not), rebel leaders may choose to break up the organization to improve their overall bargaining position. In such a calculated manner, too, peace processes can fuel rebel factionalism.

Lastly, peace processes can also alter the cost-benefit analysis of subcommanders. Rebel splits usually generate substantial risks for splinter groups, as they affect the survival and status of the respective factions that emerge. Indeed, Mahoney (2017) finds that splinter groups often die young: while larger core groups tend to survive the breakup of their organization and often gain relatively prominent positions within the insurgency movement, smaller splinter groups are much more likely to succumb under the pressure of government counterinsurgency operations as well as military attacks by their larger rivals. Even if they survive these attacks, they tend to struggle recruiting new members and therefore remain peripheral actors in the larger movement (Mahoney 2017).⁷ Breaking away from an existing rebel organization may thus come at a high cost for the splinter faction. This cost is mitigated during peace talks, however, because they typically coincide with a pause in the fighting (Svensson 2007). During a peace process, breakaway factions are thus less at risk of attack by either government or rebel forces, increasing their chances of survival and decreasing their chances of marginalization. Subcommanders can therefore see peace processes as a window of opportunity to break away and form their own organization. This leads us to the following hypothesis:

Hypothesis 1: The occurrence of a peace process in a given conflict year increases the likelihood of a rebel split.

The Onset of Peace Processes and Rebel Splits

The onset of a peace process is a pivotal moment for rebel organizations, firstly because rebels often disagree over whether peace should be pursued in the first place. As Stedman explains, “it is rare in civil wars for all leaders and factions to see peace as beneficial. Even if all parties come to value peace, they rarely do so simultaneously, and they often strongly disagree over the terms of an acceptable peace” (Stedman 1997, 7). Indeed, it is common for rebel organizations to experience substantial divisions between moderates and hardliners over whether the goals of the group can best be achieved through political or military means. When the leadership of a rebel organization decides to engage in a peace process, this signals a willingness to make concessions, which can motivate more radical factions that are opposed to any sort of compromise to split from the organization (Cronin 2006; Fjelde and Nilsson 2018; Olson Lounsbery and Cook 2011; Stedman 1997; Zartman 1995).⁸

Even when all factions in a rebel organization do agree on the desirability or necessity of a peace process, they may not share the same opinion on who should represent the group in the forthcoming negotiations. This is particularly likely in organizations that lack a robust central leadership, in which “no single voice is capable of negotiating or speaking for the group” (Staniland 2014, 5). For instance, the Conseil National pour la Défense de la Démocratie-Forces pour la Défense de la Démocratie (CNDD-FDD) broke away from the CNDD in May 1998, just before the Arusha peace negotiations mediated by former Tanzanian President Julius Nyerere were about to start, because it did not agree with Nyerere’s endorsement of Léonard Nyangoma as the leader of CNDD instead of CNDD-FDD leader Jean-Bosco Ndayikengurukiye (Khadiagala 2007, 139).

As Pearlman (2009) demonstrates, peace processes can also generate competition between elites and aspirants over the leadership of an organization because they expect to derive personal advantages from being at the negotiating table. If aspirants fail to obtain a leadership role when negotiations start, they may instead break away to form a new organization, hoping that it will allow them to gain their own seat at the table—as well as the payoffs that come with it. Rather than themselves being interested in private benefits, rebel factions may also split off because they do not trust their leadership to be capable of, or interested in, getting the best possible deal for the group as a whole, for example when they suspect the person who enters the peace process on their behalf to be more interested in the accumulation of individual benefits than the pursuit of group benefits (cf. K. G. Cunningham et al. 2012). It can also be a deliberate strategy of the government to invite certain subcommanders while excluding others as part of a divide-and-rule strategy (Fjelde and Nilsson 2018), as such fueling rebel factionalism. Together, these different mechanisms bring us to our second hypothesis:

Hypothesis 2: The start of a peace process in a given conflict year increases the likelihood of a rebel split.

The Conclusion of a Peace Agreement and Rebel Splits

Disagreements over whether peace is desirable can also drive factions to break away when a peace agreement is being concluded. More specifically, the (potential) conclusion of a peace process can induce fragmentation along structural fault-lines when subcommanders have a vested interest in sustaining the war because of the power and status it gives them, or because of the economic benefits they derive from the war economy (Stedman 1997; De Waal 2015). When the rebel organization’s leadership cannot deliver sufficient rewards from the peace process to their commanders to convince them to lay down their arms, and does not exercise sufficient control to force them to do so, they might instead form their own organization to continue the war effort (Lidow 2016).⁹ According to a report of the United Nations Security Council, “This is particularly likely when talks are making progress or when agreement is near, since internal divisions within rebel movements or Governments become more pronounced and lead to hard-line break-away factions that are opposed to the process” (Coppedge et al. 2020).

In addition, the conclusion of a peace process is a particularly critical juncture because peace agreements rarely meet all the demands of different rebels. Indeed, the demands of a given rebel group are probably as diverse as the number of rebels within it (Kydd and Walter 2002). In addition, the goals rebels pursue are not necessarily fixed: they can expand or contract based on rational calculations of the costs and risks associated with certain demands (Abrahms 2008; Stedman 1997). Especially during a peace process, such calculations may be subject to change. Hence, the possible conclusion of an agreement can incite divisions within a rebel organization over whether to accept a peace deal or not (K. G. Cunningham 2011). This happened, for example, in the wake of the Darfur Peace Agreement (DPA) in 2006: just hours after the agreement was signed, several factions broke away from the Sudan Liberation Movement/Army because they did not approve of the organization signing the DPA (Reiter 2016; Duursma 2017b).

In short, when the rebel leadership decides to accept a peace agreement that some of its factions are not in favor of—or vice versa, when the rebel leadership rejects a peace agreement that some of its factions did support—this is likely to generate fragmentation.¹⁰ Our third hypothesis thus reads as follows:

Hypothesis 3: The conclusion of a peace agreement in a given conflict year increases the likelihood of a rebel split.

The Implementation of a Peace Agreement and Rebel Splits

Lastly, the implementation of a peace agreement can also generate rebel splits. This type of discontent is distinct from the signing of a peace agreement in that the breakaway faction initially seems on board with the agreement, but becomes frustrated over the way it is implemented. Indeed, if the government does not implement the peace agreement quickly enough (or worse, not at all), part of a rebel organization may lose faith in the government’s ability or willingness to uphold its side of the deal. While the rebel leaders who signed a peace agreement are perhaps more reluctant to abandon it, factions of the organization may decide to break away and resume armed conflict if they are dissatisfied with the way in which a peace agreement is put into practice. As Plank (2017) demonstrates, this is particularly likely when the government fails to implement provisions that address the primary source of the conflict, e.g. autonomy provisions in the case of a territorial conflict, or government posts when the conflict concerns central power.

Alternatively, rebel leaders themselves may also fail to implement an agreement because they fear internal resistance. As noted by the UN Security Council, “Where there is considerable intra-party disagreement between moderates and hardliners [in a rebel organization], the leader may sign an agreement but be afraid to implement it for fear of backlash from hardline constituents” (Coppedge et al. 2020). This lack of implementation may in turn alienate the moderates who are in favor of a peace agreement, generating potential for organizational splintering.

We also expect rebel splits to emerge when discrepancies exist between the expected benefits of a peace agreement and the actual payoffs rebel factions receive. As previous research has demonstrated, rebel leaders often recruit members and inspire cooperation through promises of future rewards (Lidow 2016; Weinstein 2005), but a failure to deliver on these promises can drive factions to abandon a rebel organization. This is particularly likely to happen when certain rebel leaders monopolize the benefits associated with a peace agreement, such as government posts, military appointments, or economic payoffs (Plank 2017).¹¹ For example, The Popular Movement for the Liberation of Azawad (MPLA) in Mali splintered into several factions because the group’s leader monopolized settlement benefits to which others felt entitled (Plank 2015; Schlichte 2009). Our fourth and final hypothesis therefore reads:

Hypothesis 4: The implementation of a peace agreement in a given conflict year increases the likelihood of a rebel split.

Research Design

To test the hypotheses put forward above, we conduct a large-n quantitative study using an original dataset that combines data from several sources. The unit of analysis in this dataset is a conflict dyad-year in which a rebel group is engaged in armed conflict with a government. Our conflict data are based on data from the Uppsala Conflict Data Program (UCDP). The UCDP defines a civil war, also referred to as an intrastate conflict, as a contested incompatibility that concerns government and/or territory where a government of a state and a rebel group use armed force to fight each other. An intrastate armed conflict is considered to be active when at least 25 battle-related deaths occur in one calendar year (Melander et al. 2016). The use of dyad-years allows us to examine whether a peace process in a given dyad-year led to the fragmentation of a rebel organization engaged in armed conflict with the government.

We rely on data from the Foundation of Rebel Group Emergence (FORGE) Dataset to code the year in which a rebel split occurs. FORGE lists the “birth date” of all rebel groups included in the UCDP (Braithwaite and Cunningham 2020), which allows us to code when a splinter faction breaks away from a rebel group. As already noted in the introduction, this is an important methodological improvement compared to the study by Olson Lounsbery and Cook (2011). We will further address this issue below.

The Problem with Looking at the Emergence of Splinter Groups

In their study on the links between mediation and rebel splits, Olson Lounsbery and Cook code whether the number of dyads in a given conflict has gone up from one year to the next as a proxy for rebel fragmentation. Drawing on a dataset on intrastate armed conflicts occurring between 1945 and 1999 compiled by Regan (2002), they consider “splintering to have occurred [in a given conflict] when group identification from one year to the next involves the previously listed group or groups and a new rebel group” (Olson Lounsbery and Cook 2011, 79). This way of coding splintering assumes that when a new rebel group emerges in given conflict, this group is in fact a breakaway group. As subsequent research has shown, however, this is a flawed assumption. Fjelde and Nilsson demonstrate that most rebel groups that emerge in a given conflict are rebel groups without ties to other rebel groups in a given conflict. Fjelde and Nilsson observe that of the 149 rebel contenders that emerged in the context of an ongoing civil war between 1975–2013, “59 were the result of within-group splits, whereas 90 were joiners, that is, groups without known affiliations to incumbent groups” (Fjelde and Nilsson 2018). Hence, Olson Lounsbery and Cook study the link between mediation and the rise of rebel contenders more broadly, rather than the emergence of splinter groups.

Another fundamental problem with existing quantitative studies on rebel splintering is that they code rebel splits based on the year in which the new rebel group first emerges in the UCDP dataset (Fjelde and Nilsson 2018; Olson Lounsbery 2016; Olson Lounsbery and Cook 2011). Rebel groups are included by UCDP from the moment they engage in armed conflict for the first time, but splinter factions often break away long before that. If the goal is to study what factors influence a rebel organization to splinter, then rebel fragmentation should be coded based on the year in which the splinter group breaks away, rather than the year it first engages in armed conflict.

The distinction between the moment when splinter groups break away and when they engage in armed conflict for the first time is not a trivial difference. Having coded the year in which splinter factions break away, we find that only fourteen out of the fifty-three rebel splits identified in the dataset used in this article occurred in the same calendar year in which the breakaway faction emerged in the UCDP dataset. This is around 26% of the total number of rebel splits. In addition, the average number of calendar years before a breakaway faction enters the UCDP dataset is 3.4 years.

Table 1 above shows the distribution of the number of years a breakaway group appears in the UCDP after the rebel split.¹² The value of the Years variable reflects the difference in years between the splinter group’s birth year and the year it appears in the UCDP for the first time.

Table 1.

The Number of Years between a Rebel Split and the Appearance of the Splinter Group in the UCDP Dataset.

Years	Number of splinter groups	Names of splinter groups
0	14	UNRF, INPFL, Military faction (forces of Shahnawaz Tanay), USC/SNA, AFRC, CNDD-FDD, MODEL, Al-Shabaab, NRF, WSB, MNLF—NM, MNLF—HM, BMA, and RCSS
1	14	LRA, USC/SSA, UNRF II, AQIM (GSPC), SLM/A-MM, SLM/A-Unity, BIFM, MUJAO, NLFT-B, RIRA, NDFB—RD, Hizb-i Islami-yi Afghanistan-Khalis faction, PBCP-J, and Lebanese Forces-Hobeika faction, PUK
2	3	AMB, ATTF, EPDM, and CRA
3	3	UFM, God’s Army, and UFRA
4	2	ASG, Hizb-i Islami-yi Afghanistan, and FIAA
5	1	FRUD-AD
6	1	Palipehutu-FNL, NLFT
> 6	7	NDA (7 years), FLEC-R (7 years), CDR (8 years), UWSA (8 years), Frolina (8 years), ORPA (9 years), PFNR (9 years), RSO (9 years), MILF (13 years), MNDAA (20 years), and PFLP-GC (21 years)

The Sudan Liberation Movement/Army (SLM/A)—Unity faction breaking away from the SLM/A in 2006 illustrates why the actual rebel split rather than the emergence of a splinter faction in the UCDP dataset is crucial for the purposes of our study. Disagreements about whether to sign the Darfur Peace Agreement contributed to the SLM/A fragmenting in 2006. While the SLM/A—Unity already broke away in 2006, the splinter faction appeared in the UCDP for the first time in 2007. Yet, no peace process took place in the Darfur conflict in 2007 (Duursma 2017b). In other words, operationalizing rebel fragmentation based on when a splinter group emerges in a conflict dataset might not capture whether a split coincided with a peace process.

Furthermore, coding the moment when a splinter group first appears in a conflict dataset leads to the inclusion of six cases in which a splinter group broke away prior to the parent group actually engaging in active conflict. For instance, Frolina broke away from Palipehutu in 1989, though the civil war between the Government of Burundi and Palipehutu is coded by the UCDP as starting in 1991. This split thus preceded the outbreak of civil war, as well as the peace efforts within it. It makes little sense from a theoretical perspective to include cases in which a splinter group emerged before armed conflict started in a dataset used to examine the association between peace processes and rebel fragmentation.

An Alternative Operationalization: Splinter Groups Breaking Away During and After Civil War

It follows from the discussion above that if the goal is to study what factors cause rebel splits rather than studying what factors influence a splinter group to engage in armed conflict, then rebel fragmentation should be coded based on the year the rebel group actually breaks away. This is what we do in the empirical analysis of this article.

However, this way of operationalizing rebel fragmentation is not without problems either. A major disadvantage of coding the time of a split based on conflict data is that splits cannot be coded if the parent rebel group is not engaged in active armed conflict with the government. Of the factions included in our dataset, twenty-nine factions broke away in a year in which the parent rebel group was engaged in an armed conflict with the government that surpasses the twenty-five battle-related deaths threshold, while twenty-seven splinter groups broke away in a year in which the parent rebel group was not engaged in active conflict (see Table A-2 in the appendix).

Not being able to take into account rebel splits that take place in a year in which the parent rebel group is not engaged in active armed conflict is particularly problematic when the goal is to study the impact of peace processes on rebel fragmentation. While conflict parties at times continue fighting the government when engaged in a peace process (Sisk 2009), conflict parties often temporarily stop fighting each other while engaged in negotiations. When we only use data on active conflict years, however, rebel splits that are a trigged by a peace process are not taken into account if the peace effort is accompanied by a lull in fighting. Furthermore, peace efforts aimed at the implementation of agreements often take place in inactive conflict years. This means that solely looking at rebel splits in active conflict years stacks the deck against finding a significant impact of the implementation of peace agreements on rebel splintering.

To address this problem, we not only run models on the UCDP data on active conflict years, but also run some of our models with UCDP data on active conflict years supplemented with observations on inactive conflict years. Observations on inactive conflict years are added to the extended dataset on the basis of two sufficient criteria: (1) a conflict dyad-year becomes active again in a later dyad-year. In this case all the previous inactive conflict dyad-years are included; and (2) a peace process takes place in the inactive conflict dyad-year. When a peace process takes place in a later conflict dyad-year, all the previous inactive conflict dyad-years are also included. We are aware that these criteria for inactive years may introduce a bias in the data. The results based on this extended dataset should therefore not be seen as our main findings, but rather treated as a robustness check used to determine whether taking into account peace efforts and rebel splits outside of active conflict years leads to different findings.

It should further be noted that the extended dataset still does not capture all splits. First of all, ten splinter groups that broke away before the parent organization became involved in a conflict that surpassed the twenty five battle-related deaths threshold with the government are not included in the dataset. Another eight splinter factions emerged after the parent group engaged in active conflict for the first time, but did not break away in a conflict dyad-year that meets any of the two criteria for inclusion in the extended dataset. This leaves nine additional splinter rebel groups that are included in the extended dataset. An example of a rebel split included in the extended dataset is the Real Irish Republican Army (RIRA) breaking away from the Irish Republican Army (IRA). Observations on the IRA are included from the first inactive year in 1992 all the way to the Good Friday Agreement in 1998, which was mediated by the United States. This makes it possible to capture the split of the RIRA in 1997, which was a result of the conclusion of a ceasefire.

Explanatory Variables

The main explanatory variable of interest in our study, Peace process, is coded as 1 if a conflict dyad-year experiences peace negotiations between the government and the rebel group. We follow the UCDP definition of negotiations as talks that concern an issue related to the conflict, for example the modalities of a ceasefire or how the conflict parties will resolve their seemingly incompatible goals.¹³ In order to code the peace process variable, we have cross-checked the UCDP Conflict Encyclopedia on negotiations, data on negotiations and mediation from Svensson (Svensson 2014), the Civil Wars Mediation dataset (DeRouen et al. 2011), and data on negotiations and mediation in Africa from Duursma (Duursma 2017a).

Following Duursma and Svensson (Duursma and Svensson 2019), we distinguish between three phases of a peace process: the start of a peace process, the signing of a peace agreement, and the implementation of a peace agreement. We include a dummy variable that is coded as 1 if a peace process starts in a given conflict dyad-year. In order to examine whether it is the peace process aimed at concluding an agreement or the agreement itself that leads to possible rebel group fragmentation, we include a second dummy variable that indicates whether negotiations are ongoing in a given conflict dyad-year (but do not result in an agreement that year) and another dummy variable that indicates the conclusion of an agreement in a given conflict dyad-year. Finally, we include a fourth dummy variable that measures whether efforts to implement an agreement take place in a given conflict dyad-year. More specifically, we look at negotiations aimed at the implementation of a peace agreement as a proxy for implementation efforts.

We control for the impact of several contextual factors that may influence the emergence of splinter groups. First of all, we use UCDP data to code whether multiple rebel groups are active in a given conflict; whether conflict parties fight about a piece of territory rather than control over the government; whether the number of battle-related deaths surpasses 1,000 in a given calendar year; the number of years a conflict episode has been ongoing; whether external states are supporting any of the conflict parties; and whether more than one rebel group is already engaged in an armed conflict with the government (Melander et al. 2016). Moreover, we control for social ties since Fjelde and Nilsson demonstrate that the emergence of new rebel groups is a function of the social ties that exist within rebel movements (Fjelde and Nilsson 2018). However, one important tweak is that the three social network variables—ethnic, religious, and a leftist political orientation—are collapsed into a single “ties” variable to limit the number of variables in the models (see: Achen 2005; Schrodt 2013). We draw on FORGE data to code social ties (Braithwaite and Cunningham 2020). We also control for the country’s level of electoral democracy by including the V-Dem’s electoral democracy index (Coppedge et al. 2020). Since previous research has found that political reform can be a critical juncture in countries and lead to rebel fragmentation (Seymour et al. 2016; Shapiro 2013), we also include a dummy variable that measures whether the Polity-2 score of a country grows by at least two points (Marshall et al. 2016). This measurement of a “democratizing move” is put forward in the work of Fjelde and Nilsson on rebel contenders (Fjelde and Nilsson 2018). Finally, we include a variable indicating the number of years since the last rebel split occurred in a given conflict dyad in order to control for temporal dependence (Beck et al. 1998), because one rebel faction breaking away might lead to more splits in the future. For the same reason, we cluster the robust standard errors on the dyad level.

Findings

Table 2 shows our models that estimate the impact of peace processes on the likelihood of splinter factions breaking away from a rebel group. Since the occurrence of a rebel split is a binary variable, we make use of logit models. Robust standard errors, clustered on the conflict level, are used to account for the clustering of rebel splits within a given conflict. Model 1 estimates the likelihood of a split occurring in an active conflict dyad-year, while Model 2 is based on observations in both active and inactive conflict dyad-years. Both models suggest that peace negotiations have a positive and statistically significant impact on the likelihood of a rebel split. The substantial impact of peace processes is quite strong. All else equal, the probability of a splinter faction breaking away from a parent rebel group is only 0.5% if no peace process takes place in an active conflict dyad-year. This probability increases by 496.3%, to 3.3%, if the conflict parties engage in peace negotiations. The probability of a rebel faction splitting away in active and inactive conflict years is 0.6% without peace negotiations, but increases to 4.0% when a peace process is taking place. This is an increase of 577.6%. This provides support for hypothesis 1 that the occurrence of a peace process in a given conflict dyad-year increases the likelihood of a rebel split.

Table 2.

Logit Estimates on the Likelihood of Rebel Splits in Intrastate Armed Conflict, 1975-2011.

VARIABLES	(1)	(2)	(3)	(4)
VARIABLES	Active conflict	Active and inactive conflict	Active conflict	Active and inactive conflict
Peace process	1.863**	1.970**
	(0.439)	(0.412)
Start of peace process			1.128	1.571**
			(0.592)	(0.473)
Peace process ongoing			2.837**	3.047**
			(0.545)	(0.442)
Peace agreement concluded			1.307*	0.954*
			(0.608)	(0.483)
Implementation of peace agreement			0.987	1.111
			(0.737)	(0.630)
Multiple rebel parties	0.184	0.191	0.108	0.115
	(0.107)	(0.0978)	(0.107)	(0.0971)
Territorial conflict	−0.506	−0.226	−0.737	−0.502
	(0.429)	(0.351)	(0.422)	(0.352)
Conflict intensity	−0.0245	−0.127	−0.116	−0.209
	(0.553)	(0.529)	(0.593)	(0.551)
Conflict duration	−0.0901	−0.0463	−0.0938*	−0.0500
	(0.0504)	(0.0458)	(0.0432)	(0.0394)
Internationalized conflict	−0.407	−0.0752	−0.209	0.0205
	(0.676)	(0.619)	(0.699)	(0.596)
Social network ties	1.033	0.698	1.080	0.775
	(0.554)	(0.446)	(0.570)	(0.460)
Democratizing move	0.700	0.409	0.795	0.496
	(0.487)	(0.479)	(0.509)	(0.493)
Level of democracy	−0.0463	0.284	−0.114	0.114
	(0.451)	(0.428)	(0.466)	(0.405)
Time since last rebel split	0.0148	−0.00656	0.0248	0.00885
	(0.0380)	(0.0412)	(0.0330)	(0.0342)
Constant	−5.683**	−5.631**	−5.415**	−5.377**
	(0.627)	(0.602)	(0.575)	(0.562)
Wald chi²	37.35**	39.01**	64.15**	86.90**
Pseudo R²	0.1248	0.1076	0.1590	0.1609
Observations	1,818	2,103	1,818	2,103

Note: Conflict years without peace processes are the reference category. Robust standard errors, clustered on the conflict level, are in parentheses. *significant at 5%; **significant at 1%.

Model 3 and 4 estimate the impact of the different phases of a peace process on the likelihood of a rebel split, with model 3 only based on data on active conflict dyad-years and model 4 based on data on both active and inactive conflict dyad-years. We find that conflict dyad-years in which a peace process is started are only significantly more likely to experience a split when we look at both active and inactive years. We thus find mixed support for Hypothesis 2. It follows from Model 4 that, all else equal, the probability of a rebel split in active and inactive conflict dyad years increases from 0.8% to 4.0% when a peace process is initiated, which is a 577.6% increase. This finding is in line with the argument that subcommanders wishing to break away from a rebel parent group may see a peace process in which the fighting is paused as a window of opportunity to break away. If fighting stops because of the start of a peace process, breakaway factions are less at risk of attack by either government or rebel forces, increasing their chances of survival.

We also find support for hypothesis 3. Conflict dyad-years in which a peace process is ongoing and in which a peace agreement is concluded are significantly more likely to experience rebel splits. This association is statistically significant at the 1 percent level for an ongoing peace process and at the 5 percent level for dyad years in which a peace agreement is concluded. Based on Model 3, we find that, all else equal, in conflict dyad-years in which a peace process is ongoing, a rebel split becomes 1,472.5% more likely, increasing from 0.7% to 11.6%. This is 296.3% for conflict dyad-years in which negotiations have led to the conclusion of a peace agreement, from 0.7% to 2.9%. Based on the extended dataset that includes observations on both active and inactive conflict dyad-years, Model 4 further suggest that an ongoing peace process in a given conflict dyad-year makes a rebel split 1697.3% more likely, from 0.9% to 15.4%. All else equal, the conclusion of a peace agreement makes a rebel split 168.5% more likely, from 0.9% to 2.3%. Unexpectedly, the likelihood of a rebel split thus increases most while negotiations are ongoing.

We do not find support for Hypothesis 4, either in Model 3 or 4. A rebel split is not significantly more likely during the implementation phase of a peace agreement, though we do find a positive association.

As a first robustness check, we examine whether the non-random assignment of peace processes influences our findings. It is likely that those conflict years in which negotiations take place have very different characteristics than those in which no negotiations take place. To illustrate the importance of considering the non-random assignment of peace processes consider the split within Sudan Liberation Movement/Army in Darfur during the Abuja peace talks. The ongoing peace talks triggered Minni Minnawi to break away from the parent group led by Abdul Wahid (Duursma 2017b). Yet, at the same time, the lower conflict intensity as a result of the peace talks in Abuja is likely to have affected Minnawi’s decision because breaking away at that point in time increased the chances of his faction surviving. This dynamic is not unique to just Darfur, since ongoing peace processes often lead to a decrease in the conflict intensity (Svensson 2007). This is just one example that illustrates that our dataset is probably imbalanced.

A matching design is used to deal with the methodological problem of the non-random assignment of peace processes and reduce the imbalance in our data (see: Rubin 1979). More specifically, we use the Coarsened Exact Matching (CEM) method introduced by Blackwell et al. to coarsen the independent variables and recode them so that similar values are grouped together. The Exact Matching algorithm is then used to identify the matches. Next, the coarsened values are abandoned and the actual values of the independent variables in the matched data can be used to estimate the causal effect of peace processes (Blackwell et al. 2009). The CEM method thus essentially follows the logic of a most similar systems design.

Following the CEM procedure, we use the Global L1 distance to measure the global balance between the two subsamples: conflict dyad-years in which no peace process takes place and conflict dyad-years in which a peace process does take place. The characteristics of the two subsamples would be completely the same if the L1 statistic has a value of 0, whereas an L1 value of 1 would mean that the characteristics of the two subsamples are completely different. For Model 5, the Global L1 distance is 0.59 before matching, which means that only around 31% of the empirical distributions of the two subsamples overlap. The value 0.59 serves as baseline reference for the unmatched data. After matching, the Global L1 distance for Model 5 is 0.36; 1,085 cases could be matched. For Model 6, the Global L1 distance is 0.58 before and 0.36 after matching, with 1,348 matched cases (see Table A-3 in the appendix). While the CEM procedure has reduced the imbalance in the data, it is not completely removed. We therefore employ a logistic regression in Table 3 to estimate the effect of peace processes based on the matched data.

Table 3.

Logit Estimates on the Likelihood of Rebel Splits in Intrastate Armed Conflict, 1975-2011.

VARIABLES	(5)	(6)
VARIABLES	Active conflict	Active and inactive conflict
Peace process	1.183*	1.692**
	(0.488)	(0.467)
Multiple rebel parties	0.233	0.322
	(0.224)	(0.175)
Territorial conflict	−0.615	−0.345
	(0.596)	(0.498)
Conflict intensity	0.602	0.634
	(0.620)	(0.632)
Conflict duration	−0.0191	0.101
	(0.0866)	(0.0534)
Internationalized conflict	−0.334	0.333
	(0.660)	(0.649)
Social network ties	1.028	0.573
	(0.600)	(0.489)
Democratizing move	0.834	0.575
	(0.695)	(0.680)
Level of democracy	−1.043	0.260
	(0.784)	(0.504)
Time since last rebel split	0.0584	−0.0923
	(0.0712)	(0.0541)
Constant	−5.352**	−5.739**
	(0.649)	(0.598)
Wald chi²	33.95	34.88
Pseudo R²	0.0857	0.1016
Observations	1,085	1,348

Note: Conflict years with no peace processes is the reference category. Robust standard errors, clustered on the conflict level, are in parentheses. *significant at 5%; **significant at 1%.

As shown in Model 5 and Model 6 Table 3, the analysis based on the matched data does not alter the main findings. The significant and positive correlation between peace processes and the occurrence of rebel splits remains, even when taking into account the non-random assignment of these peace processes. It should be noted that while the imbalance is reduced in both Models, there still is imbalance. In other words, while the matching exercise provides further support for Hypothesis 1, it is important not to overstate the extent to which we can unquestionably confirm Hypothesis 1 based on this matching exercise.

We conduct five additional robustness checks, the results of which can be found in the appendix (Tables A-4 and A-5). First, to investigate whether there is a difference between the effects that bilateral and mediated negotiations have on rebel splits and to make our results more easily comparable to those of Olson Lounsbery and Cook (2011), we replicated the models above using mediation rather than peace processes in general as our main explanatory variable. We find similar results to the above: when considering mediation in general, there seems to be a strong association between mediation and rebel splits, but this effect seems mainly driven by mediation efforts aimed at concluding a peace agreement.

Second, one limitation of using a logit model is that it is impossible to take into account how peace processes influence the risk of a rebel group split beyond the dyad-year of the observation. We therefore use a Cox proportional hazard model that looks at the time until a rebel group split occurs. Within this model, we include a time-varying interaction term: we interact the time since the last peace process started within a given dyad with the peace process variable. This allows us to examine whether the risk of a split increases as peace talks drag on and factions within the rebel group grow increasingly frustrated. This interaction term is not significant, suggesting that risk of a split is not conditional on how long a peace process has been ongoing.

Third, we run a model in which we also control for rebel-inflicted sexual violence. Previous research has found that rebel-inflicted sexual violence increases the likelihood of rebel groups splits (Nagel and Doctor 2020), but also makes the onset of mediation more likely (Nagel 2019). To this purpose, we follow Nagel (2019) and draw on the Sexual Violence in Armed Conflict (SVAC) dataset to construct a dummy variable that is coded as 1 if there was any sexual violence inflicted by rebels reported by the US State Department, Amnesty International, or Human Rights Watch (Cohen and Nordås 2014). Because of data availability, this analysis is conducted without the observations between 1975 and 1988. We also find that rebel-inflicted sexual violence has a positive impact on the likelihood of rebel splits, but controlling for rebel-inflicted sexual violence does not change our finding that peace processes increase the likelihood of a rebel split.

Fourth, another additional possible confounder we control for is the age of the rebel group. A model in which the duration of the conflict is replaced with the age of the rebel group did not alter our main findings.

Fifth, and lastly, we control for a possible curvilinear relationship between time and rebel splits. Newly erupted conflicts might trigger a rapid mediation effort, while at the same time young rebel groups struggle to unite competing factions. Additionally, rebel groups in conflicts that drag on for years may finally accept peace talks because the various factions within the rebel group are reaching the limits of their ability to remain unified. In order to check whether such a “U-shaped” relationship between time and rebel splits confounds our main findings, we ran a model with the addition of a squared version of conflict duration variable and another model in which we added the squared version of the variable that measures the age of the rebel group. This did not change the main finding that peace processes are strongly associated with rebel groups splits.

Conclusion

Rebel organizations are not homogenous actors, but consist of multiple factions united in war by the pursuit of a common goal. When a rebel group engages in a peace process and this common goal seems within reach, rebels’ individual preferences about if, how, and by whom peace should be pursued tend to gain salience. As we have argued in this article, peace processes thus exacerbate pre-existing structural and substantial divisions within rebel organizations, and can therefore cause rebel fragmentation. More specifically, factions within a rebel group may disagree with the onset of negotiations, the conclusion of a peace agreement, or the implementation of an agreement—and thus subsequently break away during the peace process.

In the only quantitative study that directly looks at how peace negotiations influences rebel fragmentation, Olson Lounsbery and Cook (2011) find that mediation in a given conflict makes rebel splintering 99.3% more likely. Looking at peace negotiations in general and the time when a rebel splinter group breaks away rather than when it first engages in armed conflict, we find that a peace process makes a rebel group 240% or 656.5% more likely to splinter, depending, respectively, on whether we restrict the analysis solely to active conflict dyad-years or not. We thus find that studying rebel fragmentation through looking at when rebel splinters first emerge in a conflict dataset has led to an underestimation of the impact of peace processes on rebel splits.

Investigating in which phase(s) of a peace process rebel fragmentation is particularly likely, we find that the likelihood of a rebel split significantly increases in conflict dyad-years that experience the start of a peace process, ongoing peace negotiations, or the signing of a peace agreement. Surprisingly, while we expected the onset and conclusion of a peace process to be particularly critical junctures, the likelihood of a rebel split actually increases most in conflict dyad-years in which negotiations are ongoing. The likelihood of rebel splintering does not significantly increase during the implementation of a peace agreement. These findings are not only important from a theoretical point of view, but also from a practitioner’s perspective: knowing that peace processes increase the likelihood of rebel fragmentation, it is imperative that the actors engaged in peace efforts address the potential grievances and goals of different rebel factions during negotiations in order to prevent rebel splits. At the same time, it is important to recognize the limitations of our findings. While peace negotiations drastically increase the likelihood of rebel fragmentation in a given conflict dyad-year, peace processes probably interact with many other factors that contribute to rebel group fragmentation. Examining these interactions is a promising avenue for future research.

Supplemental Material

Supplemental Material, sj-pdf-2-jcr-10.1177_0022002720958062 - Fueling Factionalism? The Impact of Peace Processes on Rebel Group Fragmentation in Civil Wars

Supplemental Material, sj-pdf-2-jcr-10.1177_0022002720958062 for Fueling Factionalism? The Impact of Peace Processes on Rebel Group Fragmentation in Civil Wars by Allard Duursma and Feike Fliervoet in Journal of Conflict Resolution

Supplemental Material

Supplemental Material, sj-zip-1-jcr-10.1177_0022002720958062 - Fueling Factionalism? The Impact of Peace Processes on Rebel Group Fragmentation in Civil Wars

Supplemental Material, sj-zip-1-jcr-10.1177_0022002720958062 for Fueling Factionalism? The Impact of Peace Processes on Rebel Group Fragmentation in Civil Wars by Allard Duursma and Feike Fliervoet in Journal of Conflict Resolution

Footnotes

Declaration of Conflicting Interests

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

Funding

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

ORCID iD

Allard Duursma

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Abrahms

2008. “What Terrorists Really Want: Terrorist Motives and Counterterrorism Strategy.” International Security 32:78–105.

Achen

C. H.

2005. “Let’s Put Garbage-can Regressions and Garbage-can Probits Where They Belong.” Conflict Management and Peace Science 22:327–39.

Asal

Victor

Phillips

Brian J.

. 2018. “What Explains Ethnic Organizational Violence? Evidence from Eastern Europe and Russia.” Conflict Management and Peace Science 35 (2): 111–31.

Asal

Brown

Dalton

. 2012. “Why Split? Organizational Splits among Ethnopolitical Organizations in the Middle East.” Journal of Conflict Resolution 56:94–117.

Beck

Katz

J. N.

Tucker

. 1998. “Taking Time Seriously: Time-series-cross-section Analysis with a Binary Dependent Variable.” American Journal of Political Science 42:1260–88.

Blackwell

Iacus

S. M.

King

Porro

. 2009. “cem: Coarsened Exact Matching in Stata.” The Stata Journal 9:524–46.

Braithwaite

J. M.

Cunningham

K. G.

. 2020. “When Organizations Rebel: Introducing the Foundations of Rebel Group Emergence Dataset.” International Studies Quarterly 64 (1): 183–93.

Chalk

2013. “The Bangsamoro Islamic Freedom Fighters: The Newest Obstacles to Peace in the Southern Philippines?” CTC Sentinel. Combating Terrorism Center at West Point. https://www.ctc.usma.edu/the-bangsamoro-islamic-freedom-fighters-the-newest-obstacles-to-peace-in-the-southern-philippines/.

Christia

2012. Alliance Formation in Civil Wars. New York: Cambridge University Press.

10.

Cohen

D. K.

Nordås

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

11.

Coppedge

Michael

Gerring

John

Knutsen

Carl Henrik

Lindberg

Staffan I.

Teorell

Jan

Altman

David

Bernhard

Michael

, et al. 2020. “V-Dem Codebook v10”. Varieties of Democracy (V-Dem) Project. https://www.v-dem.net/en/data/reference-materials-v10/.

12.

Cronin

A. K.

2006. “How al-Qaida Ends: The Decline and Demise of Terrorist Groups.” International Security 31:7–48.

13.

Cunningham

D. E.

2006. “Veto Players and Civil War Duration.” American Journal of Political Science 50:875–92.

14.

Cunningham

K. G.

2011. “Divide and Conquer or Divide and Concede: How Do States Respond to Internally Divided Separatists?” American Political Science Review 105:275.

15.

Cunningham

K. G.

2014. Inside the Politics of Self-determination. New York: Oxford University Press.

16.

Cunningham

K. G.

Bakke

K. M.

Seymour

L. J. M.

. 2012. “Shirts Today, Skins Tomorrow.” Journal of Conflict Resolution 56:67–93.

17.

De Waal

2015. The Real Politics of the Horn of Africa: Money, War and the Business of Power. Cambridge, UK: Polity.

18.

DeRouen

Bercovitch

Pospieszna

. 2011. “Introducing the Civil Wars Mediation (CWM) Dataset.” Journal of Peace Research 48:663–72.

19.

Doctor

A. C.

2020. “A Motion of No Confidence: Leadership and Rebel Fragmentation.” Journal of Global Security Studies 5 (4): 598-616. https://doi.org/10.1093/jogss/ogz060.

20.

Duursma

2014. “A Current Literature Review of International Mediation.” International Journal of Conflict Management 25:81–98.

21.

Duursma

2017a. “Partnering to Make Peace: The Effectiveness of Joint African and non-African Mediation Efforts.” International Peacekeeping 24:590–615.

22.

Duursma

2017b. “When to Get Out of the Trench: Using Smart Pressure to Resolve Civil Wars.” Civil Wars 17:43–61.

23.

Duursma

Svensson

. 2019. “Introducing an Agenda-based Measurement of Mediation Success: The Divergent Effects of the Manipulation Strategy in African Civil Wars.” International Negotiation 24 (2): 296–323.

24.

Fjelde

Nilsson

. 2018. “The Rise of Rebel Contenders: Barriers to Entry and Fragmentation in Civil Wars.” Journal of Peace Research 55:551–65.

25.

Hacene-Djaballah

1985. “Conflict in Western Sahara: A Study of Polisario as an Insurgency Movement.” Wahington, DC: Faculty of the School of Arts and Sciences, The Catholic University of America.

26.

Johnston

2008. “The Geography of Insurgent Organization and its Consequences for Civil Wars: Evidence from Liberia and Sierra Leone.” Security Studies 17:107–37.

27.

Khadiagala

G. M.

2007. Meddlers or Mediators? African Interveners in Civil Conflicts in Eastern Africa. Leiden: Martinus Nijhoff.

28.

Kydd

A. H.

Walter

B. F.

. 2002. “Sabotaging the Peace: The Politics of Extremist Violence.” International Organization 56:263–96.

29.

Lidow

N. H.

2016. Violent Order: Understanding Rebel Governance through Liberia’s Civil War. New York: Cambridge University Press.

30.

Mahoney

C. W.

2017. “Splinters and Schisms: Rebel Group Fragmentation and the Durability of Insurgencies.” Terrorism and Political Violence 32:345–64.

31.

Marshall

M. G.

Gurr

T. R.

Jaggers

. 2016. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2015. Dataset Users’ Manual. Vienna, VA: Center for Systemic Peace.

32.

McLauchlin

Pearlman

. 2012. “Out-group Conflict, In-group Unity? Exploring the Effect of Repression on Intramovement Cooperation.” Journal of Conflict Resolution 56:41–66.

33.

Melander

Pettersson

Themnér

. 2016. “Organized Violence, 1989-2015.” Journal of Peace Research 53 (5): 727–42.

34.

Mosinger

E. S.

2019. “Balance of Loyalties: Explaining Rebel Factional Struggles in the Nicaraguan Revolution.” Security Studies 28:935–75.

35.

Nagel

R. U.

2019. “Talking to the Shameless?: Sexual Violence and Mediation in Intrastate Conflicts.” Journal of Conflict Resolution 63:1832–59.

36.

Nagel

R. U.

Doctor

A. C.

. 2020. “Conflict-related Sexual Violence and Rebel Group Fragmentation.” Journal of Conflict Resolution 56 (1): 94–117.

37.

Olson Lounsbery

2016. “Foreign Military Intervention, Power Dynamics, and Rebel Group Cohesion.” Journal of Global Security Studies 1:127–41.

38.

Olson Lounsbery

Cook

A. H.

. 2011. “Rebellion, Mediation, and Group Change: An Empirical Investigation of Competing Hypotheses.” Journal of Peace Research 48:73–84.

39.

Pearlman

2009. “Spoiling Inside and Out: Internal Political Contestation and the Middle East Peace Process.” International Security 33:79–109.

40.

Perkoski

2019. “Internal Politics and the Fragmentation of Armed Groups.” International Studies Quarterly 63:876–89.

41.

Plank

2015. “Not Enough Pieces of the Cake? The Moro National Liberation Front (MNLF) in the Mindanao Final Agreement.” Asian Security 11:154–77.

42.

Plank

2017. “When Peace Leads to Divorce: The Splintering of Rebel Groups in Powersharing Agreements.” Civil Wars 19:176–97.

43.

Regan

P. M.

2002. “Third-party Interventions and the Duration of Intrastate Conflicts.” The Journal of Conflict Resolution 46:55–73.

44.

Reiter

A. G.

2016. Fighting Over Peace: Spoilers, Peace Agreements, and the Strategic Use of Violence. London, UK: Palgrave Macmillan.

45.

Rubin

D. B.

1979. “Using Multivariate Matched Sampling and Regression Adjustment to Control Bias in Observational Studies.” Journal of the American Statistical Association 74:318–28.

46.

Rudloff

Findley

M. G.

. 2016. “The Downstream Effects of Combatant Fragmentation on Civil War Recurrence.” Journal of Peace Research 53:19–32.

47.

Schlichte

2009. In the Shadow of Violence: The Politics of Armed Groups. New York: Campus Verlag.

48.

Schrodt

P. A.

2013. “Seven Deadly Sins of Contemporary Quantitative Political Analysis.” Journal of Peace Research 51:287–300.

49.

Seymour

L. J. M.

Bakke

K. M.

Cunningham

K. G.

. 2016. “E Pluribus Unum, Ex Uno Plures: Competition, Violence, and Fragmentation in Ethnopolitical Movements.” Journal of Peace Research 53:3–18.

50.

Shapiro

J. N.

2013. The Terrorist’s Dilemma: Managing Violent Covert Organizations. Princeton, NJ: Princeton University Press.

51.

Sisk

T. D.

2009. International Mediation in Civil Wars: Bargaining with Bullets. London, UK: Routledge.

52.

Staniland

2012. “Organizing Insurgency: Networks, Resources, and Rebellion in South Asia.” International Security 37:142–77.

53.

Staniland

2014. Networks of Rebellion: Explaining Insurgent Cohesion and Collapse. Ithaca, NY: Cornell University Press.

54.

Stedman

S. J.

1997. “Spoiler Problems in Peace Processes.” International Security 22:5–53.

55.

Svensson

2007. “Bargaining, Bias and Peace Brokers: How Rebels Commit to Peace.” Journal of Peace Research 44:177–94.

56.

Svensson

2014. International Mediation Bias and Peacemaking: Taking Sides in Civil Wars. New York: Routledge.

57.

Tamm

2016. “Rebel Leaders, Internal Rivals, and External Resources: How State Sponsors Affect Insurgent Cohesion.” International Studies Quarterly 60:599–610.

58.

Tamm

2019. “In the Balance: External Troop Support and Rebel Fragmentation in the Second Congo War.” Journal of Strategic Studies. https://doi.org/10.1080/01402390.2019.1701442.

59.

Weinstein

J. M.

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

60.

Woldemariam

2016. “Battlefield Outcomes and Rebel Cohesion: Lessons from the Eritrean Independence War.” Terrorism and Political Violence 28:135–56.

61.

Woldemariam

2018. Insurgent Fragmentation in the Horn of Africa: Rebellion and Its Discontents. New York: Cambridge University Press.

62.

Zartman

I. W.

1995. “Dynamics and Constraints in Negotiations in Internal Conflicts.” In Elusive Peace: Negotiating an End to Civil Wars, edited by Zartman

I. W.

, 3–24. Washington, DC: Brookings Institution.

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.

0.10 MB

0.00 MB

0.20 MB