Abstract

Side switching by armed groups is a prominent feature of many civil wars. Shifts in alignment have far-reaching consequences, influencing key outcomes such as civil war duration and termination, military effectiveness, levels of civilian victimization, and state-building prospects. In Sudan's wars, ideological and ethnic cleavages have not influenced factional alignments nearly as much as one might expect given the prominence of clashing political projects and ethnically organized violence in southern Sudan and Darfur. Recent explanations highlighting the role of territorial control, factional infighting, or relative power considerations also have limited value. In many wars fought in weak states characterized by low barriers to side switching, two mechanisms explain patterns of collaboration and defection: first, political rivalries that lead actors to collaborate in exchange for military support in localized struggles; and second, patronage-based incentives that induce collaboration for material gain. A nested analysis drawing on original data from wars in southern Sudan and Darfur supports this argument. The findings have implications for understanding alignments in civil wars, the role of weak states in counterinsurgency, and ethnic politics more generally, as well as policy relevance for factionalized civil wars.

Introduction

Why do actors switch sides in civil wars? And why do many of these conflicts feature systematic realignments among competing actors? Numerous conflicts include instances of defection with far-reaching implications. In Afghanistan, efforts to build a cohesive army have been hampered by instances of Hazara, Tajik, and Uzbek commanders defecting from the Afghan government and NATO coalition forces to join their erstwhile Taliban enemies.1 The crisis unfolding in the Central African Republic, escalated by the coup against François Bozizé in March 2013, was precipitated by the defection of many of the same “liberators” who brought his government to power in 2003.2 Violence in Somalia has been driven by a repetitive cycle of fragmenting alliances among factions. Al-Shabaab's relative military success in Somalia from 2007 to 2010 owed much to its ability to forge a degree of cohesion and prevent defections.3 In the eastern Democratic Republic of Congo, repeated failures to integrate armed groups into national institutions have led to a cycle of mutinies, sparking humanitarian crises.4 In these and other wars, some factions switch sides as if passing through a revolving door. Unstable alignments among fragmenting factions are more than just a characteristic of these wars—in many ways, they are the war, with repeated side switching fueling cycles of protracted violence.

Alignments in civil wars explain important outcomes, as illustrated by Sudan's troubled history. Fragmentation and side switching have prolonged Sudan's civil wars by frustrating peace efforts and forestalling military victories. A cycle of defections in the western region of Darfur has vastly complicated mediation efforts and degraded the capabilities of insurgent and counterinsurgent organizations, forestalling negotiated settlement or military victory.5 Reliance on tribal militias often drives ethnically motivated violence, rights abuses, and war crimes, all of which complicate postwar state building. In post-independence South Sudan, troubled legacies of collaboration weakened institutions, sowed mistrust, and complicated disarmament. The reintegration of factions that collaborated with the government during the war created an army beset by tribal rivalries and parochial loyalties. Tensions erupted in December 2013, when elite infighting caused a civil war that saw approximately 70 percent of South Sudan's army defect to the opposition. A series of retaliatory massacres and protracted fighting have displaced more than a million people.6

Scholars find similar effects in other conflicts. Side switching and defection have been linked to longer civil war duration; higher numbers of people killed; counterinsurgency dynamics and success; and fundamental changes within war, such as the emergence of new organizations, preferences, and identities.7 The wider literature links variation in fragmentation to important dynamics, including the escalation to violence and outbreak of civil wars.8 Several studies link patterns of infighting among internally divided armed groups to violence against civilians and the diffusion of violent tactics across organizations.9 Other work suggests that preventing defection—whether of armed groups or individuals—is central to the overall political and military effectiveness of armed groups and national movements.10 Variation in the cohesion of armed groups also shapes how wars end, including the likelihood of attaining peace settlements, the scope of concessions within them, and the probability of war recurrence.11

This article provides a theoretical explanation for patterns of alignment and side switching in civil wars. Drawing on evidence from Sudan, it focuses on an important category of civil wars fought in weak and collapsing states characterized by fluid alignments among armed actors. Existing accounts are largely deterministic, emphasizing how relatively fixed ethnic identities and ideological commitments constrain defection, how threats to survival compel the alignment choices of actors with limited options, and how the longer-term aim of maximizing postwar power prompts defections to balance against powerful adversaries. I develop an alternative theory of opportunistic alignments only loosely constrained by identity, ideology, and threats to survival, in which short time horizons emphasize immediate payoffs. Specifically, I identify two key mechanisms that shape alignments: first, political rivalries that prompt collaboration with the side offering weapons, ammunition, and support against local competitors; and second, patronage relations that induce collaboration with the side providing material advantages. These mechanisms account for the surprising frequency of side switching in civil wars fought in states characterized by their fragile control over peripheries; shallow ideological cleavages; and complex relations among multiple ethnic groups, particularly in tribal societies divided along segmented lineages. Under such conditions, factional leaders under pressure to secure resources in violent, patrimonial political systems leverage their ability to collaborate opportunistically to secure military backing and material gains.

The article proceeds as follows. First, I assess existing theories of alignment and develop my argument. Second, I test these hypotheses using a nested analysis of the civil wars in southern Sudan and Darfur. I begin with an analysis of an original dataset of factional alignments in these two wars followed by case studies to reveal causal mechanisms and bolster causal identification. To assess the argument's external validity and scope conditions, I briefly discuss other conflicts. I conclude with implications for theory and policy and questions for future research.

Alignment and Defection in Civil Wars

The question of what drives alignment intersects with several debates in the literature on civil wars, including “ethnic defection,”12 cohesion and fragmentation in armed groups and infighting among them,13 and alliance formation in civil wars and coalition building in their aftermath.14 Alignment begins with actors drawn into a conflict through collaboration with actors on a particular side in a civil war, either at the level of individuals, factions, or the wider alliances they form in multiparty civil wars. I focus here on meso-level factional alignment. A faction is an organized group with an established leadership that acts as an autonomous organization respecting no higher command authority, even though it might join forces with other actors in the conflict or self-identify as part of a larger movement. Focusing on factions shifts attention to the organizational level of analysis and the role of elites whose fortunes are bound to the fate of the armed groups they lead. Factional leaders tend to be outright warlords or ambitious “big men” at the apex of informal networks through which they mobilize followers and command their loyalty.15 They generally draw on forms of tribal, religious, or political authority that facilitate recruitment. Although their fighters do not always follow them across enemy lines when they defect, their authority bolsters loyalty and facilitates recruitment, especially when they have access to a pool of restless young men and generous infusions of weapons and money.

The collaboration inherent in defection takes multiple forms. Factions might privately inform on their allies, arrange informal cease-fires or more formalized nonaggression pacts with enemies, cooperate over zones of control and spheres of influence, or engage in forms of trading and exchange.16 A faction can engage in infighting without crossing over into collaboration with its enemies. Coups or splits can occur with or without subsequent shifts in alignment characteristic of side switching.

Notwithstanding the importance of these forms of infighting and collaboration, I focus here on the relatively understudied phenomena of alignment and defection. I conceptualize and measure factional alignment as collaboration with actors on one side in a conflict through participation in violence, and correspondingly measure defection as actors collaborating with forces they previously fought against by targeting factions with which they were previously aligned. To be aligned, in this sense, is to collaborate with the actors whose conflict structures violence, however loosely, along a war's “master cleavage” that divides two or more sides in a conflict.17 Defection, as defined here, traverses this cleavage: a faction leaves one alliance to collaborate with an alliance on an opposing side.18 To switch sides does not necessarily imply that a faction's preferences are aligned with its new allies—indeed, factions often go out of their way to underscore this point when collaborating with former enemies—but merely that the faction acts in ways that further the alliance's collective interest in coercing mutual adversaries. Collaborating by participating in violence includes integration into joint military structures, loosely coordinated offensives, and more informal cooperation against mutual adversaries. These forms of collaboration contrast with nonaligned stances that include abjuring violence, maintaining a neutral self-defense militia, or attacking factions on all sides of a multiparty civil war.

Side switching poses an interesting set of puzzles, particularly when it becomes widespread. Even if actors have some options in deciding whether to participate in violence and with whom to align once mobilized, the dangers of side switching should make it far less common than forms of more subtle collaboration. Fighting alongside former enemies impedes collective action and exposes factions to charges of betraying the cause, particularly when it contradicts the ideological or ethnic appeals that motivated political mobilization in the first place. Defectors must reckon with the prospect of violent retribution, with disloyalty punished pour encourager les autres. Leaders can fall out with their subordinates, with their cadres and fighters refusing to join them in defecting, leaving them marginalized and vulnerable. Switching sides also tends to burn bridges. Defection engenders mistrust among new allies wary of infiltration and suspicious of residual loyalties. If new allies turn on defectors, they can be left perilously isolated. In sum, high barriers to defection tend to make it a risky proposition.

Yet many civil wars are characterized by high rates of side switching in ways suggesting low barriers to defection at minimal risk. To explain why, I draw observable implications from five major explanations in the literature about what drives alignment behavior and develop an alternative opportunistic theory of alignments.

ETHNICITY

Numerous accounts suggest that ethnicity should determine alignments in civil wars, particularly in weak states. Ethnic boundaries mark salient cleavages, and ascriptive identities are frequently the source of collective grievances.19 Violence tends to reinforce ethnic identities in ways that increase group cohesion.20 Ethnic groups also possess mechanisms of in-group policing that deter defection and sustain collective action.21 The state and armed challengers often use ethnic identity as a marker of underlying loyalties in the face of uncertain allegiances, and then use strategies of ethnic preference to bolster group cohesion in ways that further reinforce ethnicity's salience.22 Indeed, ethnicity should matter most as a determinant of alignment and constraint on side switching in weak states, where actors lacking in resources and organization mobilize through identity networks and are less able to induce ethnic defection.23 To the extent ethnicity shapes alignment, ethnic defection—“whereby individuals join organizations explicitly opposed to the national aspirations of the ethnic group with which they identify and end up fighting against their co-ethnics”24—should be the exception to the rule of ethnically based collective action. When one observes shifts in alignment attributable to ethnic identities, these shifts should be caused by longer-term identity shifts whereby individuals acquire new identities or the boundaries between ethnic groups are redefined over the course of protracted violence.

Hypothesis 1: Ethnicity shapes alignment, with factions (re)aligning with ethnically proximate groups.

IDEOLOGY

Another line of explanation emphasizes ideology's influence on alignment behavior. Whereas fighters often have a loose grasp of ideology, factional elites tend to have deeper ideological commitments and should align according to the compatibility of their political projects.25 The incompatibilities driving conflict tend to be relatively stable, meaning that alignments should be as well. Yet ideological constellations do shift. The literature on spoilers, for example, argues that divergent elite interests shaped by ideology, beliefs, and world-views are a key driver of defection around peace settlements. Peace negotiations often turn moderate adversaries into allies defending agreements against radical holdouts.26 Governments can use limited concessions to co-opt ideologically moderate factions through a policy of “divide and concede.”27 Ideological appeals also play a key role in spurring loyalty shifts and defection from within government security forces.28 Ideology-based arguments therefore expect disparate factions to cohere into relatively stable blocs around shared ideological preferences with common aims reinforcing cohesion. When alignments change, however, it should reflect shifts in the ideological commitments of factional leaders or the ideological ends toward which violence is directed.

Hypothesis 2: Ideology shapes alignment, with factions (re)aligning with ideologically proximate groups.

TERRITORIAL CONTROL AND ORGANIZATIONAL DEMAND

A powerful explanatory variable in recent work on civil wars focuses on territorial control.29 By controlling territory, actors can induce collaboration in spite of underlying ethnic ties or ideological commitments.30 The mechanism of “coercion and survival maximization” translates control into collaboration: “the imposition of control allows the effective use of violence, thus deterring defection; opponents are identified and flee, are neutralized, or switch sides.”31 The “control-collaboration” model emphasizes defections from rebels to state incumbents and focuses on individual, rather than factional, behavior. The basic logic, however, predicts factional alignments in either direction under the same conditions: the willingness of actors to organize collaborationist structures in areas they control; the relative strength of these actors vis-à-vis coerced factions; and atrocities that launch processes of revenge and protection seeking. Factions have the option of retreating to avoid falling behind enemy lines. Many local armed groups are reluctant, however, to leave their land and property behind and their home communities defenseless. This reluctance enhances the prospects for cutting deals with new authorities, particularly when stronger outside organizations lack the capacity to impose control on their own and seek out local collaborators. The prediction here is that when powerful armed groups control territory and seek collaborators, weaker factions operating from that territory should align with them. Defections should occur when changes in territorial control generate organizational demands for collaborators.32

Hypothesis 3: Territorial control induces alignment, with factions (re)aligning with organizations controlling the territory from which they operate if those in control seek collaborators.

FRATRICIDAL FLIPPING

Another theory of alignment emphasizes lethal competition within rebel movements and threats to survival driven by the hegemonic agendas of dominant insurgent organizations attempting to eliminate competitors. What Paul Staniland terms “fratricidal flipping” provokes defection after intra-insurgent violence pushes the targeted survivors to defect to the state for protection. “Providing intelligence to the state and operating under its protection,” writes Staniland, “is the most effective way to stay alive if a group is losing its battle within the insurgency.”33 By the same logic, fighting between government-aligned paramilitaries, or between government forces and their proxy militias, could lead targeted factions to defect to rebel organizations when they offer protection that increases survival prospects in the wake of costly infighting. According to this hypothesis, alignment is shaped by patterns of violence characterized by struggles for dominance, deadly fratricide, and credible threats that drive protection-seeking behavior. Defection should correspond to the intensity of violent competition within either the rebel or government camps as incomplete fratricide forces factions into defensive realignments motivated by survival considerations.

Hypothesis 4: Patterns of violence shape alignment, with infighting forcing weaker factions to (re)align with more powerful organizations offering protection.

RELATIVE POWER CONSIDERATIONS

Another theory of alignment stresses relative power considerations. Here, alignment is motivated by survival and the desire to maximize postwar political payoffs through forming alliances that represent minimum winning coalitions. “Groups seek alliances that are powerful enough to secure victory,” Fotini Christia argues, “but small enough to avoid having to share payoffs.”34 In choosing with whom to align, groups balance the desire to be on the side of the alliance that wins the war against commitment problems that exacerbate fears of being marginalized by more powerful allies when the war ends. Only when outright military victory by one side appears inevitable will groups bandwagon with the likely victors. In civil wars fought by fragmented actors in weak states, one should observe unstable alignments, as the government lacks a preponderance of power that would compel rebels to cooperate in balancing against it. Wars fought in weak states thus generate cyclical realignments, as power shifts precipitate defections to balance competing alliances.35 Although the theory operates at the level of ethnically organized warring groups, Christia argues that the alignment decisions of local factional commanders tend to be consistent with alliance patterns at the group level.36 Correspondingly, balance of power considerations among competing groups at the regional or national level should influence factional alignments driven by survival concerns and postwar power maximization.37

Hypothesis 5: Relative power considerations shape alignment, with factions (re)aligning according to the intergroup balance of power.

AN OPPORTUNISTIC THEORY OF ALIGNMENTS: RIVALRY AND PATRONAGE

The above explanations locate important dynamics across a range of conflicts. Ethnicity, ideological commitments, territorial control, patterns of violence, and balancing considerations constrain the alignment options of actors in many civil wars. These same factors, however, all suggest scope conditions. What explains alignments in civil wars when identities are rooted in crosscutting, relational forms of belonging, rather than clearly defined ethnic projects? What determines alignments when territorial control is tenuous, or barriers to broadcasting military power limit the ability of factions to carry out hegemonic campaigns to eliminate competitors? Or when local violence is disconnected from high-stakes power struggles in the wider war?

These conditions characterize the weak states that host most civil wars. Such circumstances invite “opportunistic” behavior in the sense that actors “use immediately damaging means to pursue ends that would be unavailable or forbidden to them under other circumstances.”38 This sort of behavior is more likely in low-capacity regimes, whether under weak states or fragile rebel rule, as coordinating mechanisms that structure wartime violence break down and “invite both small-scale racketeering and forms of political struggle that facilitate private vengeance, pleasure seeking and profit taking.”39 When ethnic or ideological divides are shallow, adversaries cannot credibly threaten survival, and balance of power considerations in the wider war are secondary concerns, a range of options opens up in bargaining over the terms of collaboration. Under such conditions, I argue that two related mechanisms are prominent in driving opportunistic alignments.

First, alignment is driven by local political rivalries that powerful outside actors promote and exploit through offers of support against mutual enemies in exchange for collaboration. I conceptualize rivalry in terms of competition over scarce goods involving the significant likelihood of force being used in the dispute. A rivalry requires a clearly identifiable dispute recognized by both factions as entailing a risk of violence, but not necessarily a militarized dispute.40 The sorts of disputes generating rivalries take many forms, including struggles for leadership and promotions in armed groups, control of political offices and institutions, personal and tribal animosities between long-standing enemies, and turf wars for local influence. Although the connection between political rivalries and alignment is easier to observe ex post, rivalries often leave indicators that facilitate ex ante measurement. Credible and persistent rumors of intense competition, public accusations and recrimination, resurrected historical antagonisms, and escalating threats all provide evidence of rivalry, while the support exchanged for collaboration is evident in arms and ammunition transfers, logistical support, military deployments, and patterns of fighting.

The outbreak of violence increases the intensity of local disputes by militarizing the terms of competition, creating a supply of vulnerable factions threatened by competitors in local and regional politics. Tensions should be particularly high in weak states, where authority vacuums invite contestation by ruthless strongmen, and where weaker outside organizations rely on local intermediaries to enforce their rule. Although outside state or rebel organizations may have difficulties exerting control over actors on their peripheries, they can often manipulate tensions between them. Offering military support can shift local power balances advantageously, engendering local collaboration and deflecting opposition efforts. The theory here cannot explain the origins of the political rivalries that trigger realignments. Some are rooted in long-standing feuds that precede the war, whereas others escalate through wartime processes and internal competition within movements. My argument does suggest, however, that insurgent and counterinsurgent strategies rely heavily on exploiting local rivalries to fragment the opposition and extend influence by gaining the collaboration of local actors. In turn, local actors seek advantages against rivals through instrumental alignments with more powerful outsiders in exchange for arms, ammunition, supplies, and military backing against their mutual enemies.

Hypothesis 6a: Political rivalries shape alignment, with factions (re)aligning with more powerful groups providing weapons, ammunition, and military support against local competitors.

Second, I argue that alignment often reflects material incentives to engage in patronage relations. Whereas the political rivalry hypothesis links alignment to localized struggles for political power, the emphasis here is material gain in political economies transformed by violence. Patronage-based incentives shape alignments through networks connecting local clients to wealthier patrons. I track the patronage relationships behind factional alignment through evidence of material incentives to collaborate, including rewards such as cash payments; salaried positions and promotions; tolerance of corruption; land title, pasture, and grazing rights; and control of lucrative assets and resources, including flows of humanitarian aid and development funds.

The emphasis on patronage highlights several transformations in local political economies during civil wars. Violence opens new avenues of economic mobility, creating chances for ambitious politicians, regime notables, and local strongmen to renegotiate their incorporation into patronage networks on more favorable terms.41 Displacement creates groups of idle young men susceptible to military recruitment by political entrepreneurs seeking to profit from disorder. Patrimonially organized armed groups, however, require sustained flows of resources to function. The loyalty that local leaders command is partly contingent on the material rewards they can deliver to their followers. War also disrupts prewar patterns of production and trade, threatening the material welfare of many communities, particularly in subsistence economies. Outsiders exploit this vulnerability, putting pressure on factional leaders attempting to sustain political mobilization or secure the welfare of their supporters. Communal groups with secure land titles see their claims contested by landless outsiders; traders find property confiscated; and control over aid flows is used to threaten enemies with deprivation and starvation. Together, these factors push factional leaders to switch sides when a more profitable bargain can be struck.

Hypothesis 6b: Patronage relations shape alignment, with factions (re)aligning with wealthier groups providing material rewards and support against local economic competitors.

This argument about opportunistic alignments differs from existing explanations in several ways. First, the opportunism underpinning the rivalry and patronage mechanisms differs from the deterministic logic of ethnicity and ideology insofar as it emphasizes the flexibility of identities and commitments, as well as the ability to defect across shallow and ambiguous identity cleavages. Unlike arguments emphasizing balancing between relatively cohesive ethnic groups focused on maximizing postwar power at the national level, the account here pays more attention to local politics and the fragmentation of actors manifesting a lack of “groupness.”42 The intensity of political and economic competition between closely related actors, combined with their social, geographic, and political distance from what is at stake in the overarching conflict, prioritize local, sub-ethnic struggles. Instead of playing a long game to maximize power after protracted wars, many factions use alignments to make short-term gains in wars with no clear endgame.

Second, the opportunistic logic of alignment incorporates the insights of the collaboration-control model. Fragile territorial control simultaneously lowers the risk of defection to far enemies and increases the areas open to contestation between near enemies; for more powerful outsiders, barriers to extending territorial control create incentives to use collaborating factions to exercise indirect rule. Yet my argument departs from the prevailing wisdom that weak states and rebels are less capable of provoking ethnic defection.43 Stathis Kalyvas and Matthew Kocher, for instance, suggest that weak states direct ethnic conflict toward “ethnic wars” by using indiscriminate violence and mobilizing through ethnic networks in ways that erect barriers to ethnic defection; conversely, strong states direct ethnic conflict toward “irregular wars,” employing discriminate violence and promoting collaboration across identity cleavages in ways that promote ethnic defection.44 Sudan's wars, featuring mobilization through identity networks, high levels of ethnic defection, and indiscriminate violence, have been fought as ethnic wars and irregular wars at the same time.

Third, the argument here underscores the importance of state and rebel policy in courting defectors, unlike the fratricidal flipping mechanism, which suggests that “counterinsurgent manipulation, bribery, and peace processes are unlikely to induce significant defection on their own.”45 Rather than passively making themselves vulnerable to potential defectors, both insurgents and counterinsurgents actively flip them. Far from defection being a last-minute option of factions on the brink of elimination, factional leaders keep lines of communication open with their enemies to maintain a market price for their collaboration.

Research Design: A Nested Analysis of Sudan's Wars

To test the hypotheses above, I employ a nested analysis of factional alignments in Sudan's wars.46 In some ways, Sudan is exceptional. Its diversity is remarkable: by one reckoning prior to the independence of South Sudan, anthropologists divided Sudan into 597 ethnic groups speaking 400 languages and dialects.47 It was Africa's largest state before the secession of South Sudan in 2011, complicating the consolidation of authority over rebellious peripheries and the extension of coercive power.48 In other ways, however, Sudan is a typical war-prone state, scoring high on the risk factors robustly associated with civil wars, including weak statehood, political instability, a small military establishment, a large population, inconsistent democratic institutions, an undemocratic and war-prone neighborhood, and gaping inequalities among ethnic groups.49

To manage persistent rebellions from its peripheries, successive Sudanese governments have followed a colonial pattern of administrative tribalism that seeks to co-opt instrumentally loyal groups.50 With an overstretched, ineffective military, the Sudanese government enlisted tribal militias in a strategy that Alex de Waal has labeled “counterinsurgency on the cheap.”51 In southern Sudan and Darfur, both the government and the rebels deliberately undermined the rural subsistence economy to attract humanitarian aid and render local communities dependent on them. Their actions placed many communities on the edge of survival. Combined with the arming of tribal militias, this initiated cycles of raiding and plundering that forced beleaguered locals to seek outside backing from more powerful patrons eager to mobilize them into war efforts.52 In Darfur, for example, “local chiefs and military commanders assessed their options and sold their allegiance to the highest bidder, constantly defending their autonomy by defrauding and double-crossing faraway patrons who knew little of the realities on the ground.”53 In the market for the allegiances of men who command violence, guns and money are the mediums of exchange.

Far from being a Sudanese phenomenon, this form of governance is characteristic of the sorts of “mediated states” that have emerged in the Horn of Africa and the Sahel. In these states, a weak government has little choice but to partner with a diverse range of local intermediaries and rival authorities to control its hinterlands.54 Conflicts throughout this troubled region have persistently drawn in outsiders seeking to manage violence driven by shifting alliances among factions in multiparty civil wars, whether in Somalia, Côte d'Ivoire, Mali, Libya, the Central African Republic, or Sudan and South Sudan.

Quantitative Analysis

To test these hypotheses on the alignment of factions in civil wars, I use an original dataset on civil wars in southern Sudan (1983–2010) and Darfur (2002–12), gathering evidence from numerous secondary sources, including the biographies of factional commanders, reports of the United Nations, Africa Confidential, Human Rights Watch, the International Crisis Group, Sudanese newspapers, and detailed field reports of the Small Arms Survey's Sudan project.

The dataset is organized around “factional (re)alignment,” a dichotomous variable indicating whether a faction defected or not in a given year. I conceptualize and measure alignment as collaboration with one side in the conflict through active participation in violence, and correspondingly measure defection as actors collaborating with forces they previously fought against by targeting formerly allied factions. The dataset comprises 393 faction alignment years (263 from South Sudan and 130 from Darfur). Data collection recorded 132 factions, but data collection was only feasible for 92 of them.55 Three considerations led factions to be excluded. First, I was unable to gather the detailed evidence required for either small factions operating in obscurity or factions that dropped out of the conflict or merged with others shortly after splits. Insofar as my argument emphasizes local factional struggles, rather than ethnically or ideologically organized alignments at the national level, exclusion of the alignments of smaller actors should not bias the findings in favor of my hypotheses. The data paint a picture of the most important actors and alignments in each war. Exclusion of smaller groups, however, does potentially bias the findings against the territorial control and fratricidal flipping hypotheses highlighting the coercion of the weak by the strong. Second, I omitted factions when there was no evidence of command autonomy (i.e., factions were operationally subordinate to a wider organization, in which case they were considered part of the commanding faction), or when there was no clear evidence of collaboration with one side through participation in violence (in which case they were considered nonaligned). Third, I excluded factions involved primarily in foreign wars, such as Chadian factions in Darfur and the Lord's Resistance Army in southern Sudan.

The independent variables are coded to measure the predicted risk of defection in a given alignment. All variables are dichotomous given data limitations that preclude more fine-grained measures. The first two hypotheses suggest that factions align with ethnically (H1) or ideologically (H2) proximate groups, and predict a higher probability of defection when misaligned in these terms. Indicators for “ethnic proximity” include the ethnic composition of factional leaderships and the identities they claim to represent, while manifestos and public statements of a faction's political aims indicate “ideological proximity.” By way of example, consider the coding for Kerubino Kuanyin's faction, whose career saw him move back and forth between rebel and government camps four times. Between 1992 and 1998, he led a militia aligned with the government after defecting from the Sudan People's Liberation Army (SPLA). Because Kuanyin was a Dinka and a committed separatist, both the ethnicity and ideology variables predict a risk of defection back to southern rebels throughout the period during which he sided with the government. He eventually changed sides in January 1998, only to defect once more to a government-backed militia in 1999.

Hypothesis 3 (H3) suggests that patterns of territorial control coincident with organizational demand for collaborators determine alignments. Indicators of “territorial control” and “organizational demand” include whether a faction operates in territories controlled by a more powerful organization that seeks its collaboration.56 In 1987, for example, the Sudan People's Liberation Movement/Army (SPLM/A) captured areas around Pibor, the stronghold of the government-aligned Murle militia of Ismail Konye. At the time, the SPLA had shifted from a policy of retribution against government-backed southern militia toward seeking alliances with them to consolidate their hold over traditionally hostile areas. Thus, the hypothesis predicts the Murle militia's defection after their territory came under the control of the SPLA and the group sought its collaboration, a switch that occurred later in 1987.

Hypothesis 4 (H4) argues that patterns of violence drive alignment, with internecine violence compelling the defection of weaker factions targeted for elimination. I measure “fratricidal flipping” on the basis of reports of infighting between factions on the same side and predict a risk of defection for the weaker faction when protection is available through side switching. In June 2003, for example, two factions of government-aligned militia under the command of Tito Biel and James Leah clashed with the stronger, pro-government faction of Paulino Matip Nhial. With the SPLA courting government-aligned factions at the time, the fratricidal flipping predicts the realignment of the weaker South Sudan Independence Movement factions for protection; both defected to the SPLA later that year.

The fifth hypothesis (H5) concerns “relative power” considerations, as factions seek to maximize their power by forming minimum winning coalitions or balancing against them. There were moments in both wars when factions aligned in ways consistent with this logic. After Sudan's government signed a peace deal with powerful rebels from Darfur in 2006, for example, many Arab militias that had fought with the government felt betrayed. Sidelined in the postwar Darfur that the agreement envisioned, and fearing the government was turning against them by co-opting their local rebel adversaries, several Arab tribal militias defected.57 The inherent difficulties of measuring relative power complicate a formal statistical test of H5.58 Therefore, I compare the descriptive statistics from Sudan with detailed data from Afghanistan and Bosnia that support Christia's theory before evaluating it qualitatively in the case studies.

To test my hypothesis linking alignment to “political rivalries” (H6a), I look for evidence of intense rivalries and whether outsiders exploited these through offers of military support. State and rebel organizations attempt to maintain secrecy, but it is a fair assumption that the visibility of tensions generally corresponds to their intensity, with more serious rivalries spilling into public view.59 Consider the case of George Athor, a former deputy chief of staff in the SPLA, who defected with his followers in April 2010. His defection followed a failed bid for a governorship in the April 2010 elections in the South, in which he stood as an independent candidate against an opponent backed by the ruling party. The public nature of the rift indicated a political rivalry, whereas confirmation of military backing for Athor's faction from the government of Sudan suggests that Khartoum manipulated it to trigger his defection.60

My hypothesis regarding how “patronage incentives” (H6b) induce alignment is coded to reflect collaboration contingent on material rewards, tracing the patronage politics behind alignments as closely as possible. One example is the government-backed Mahariya Arab militia led by Mohamed Hamdan Dogolo “Hemeti.” After he defected in August 2007, seeking alliances with Darfur's rebels, the government made a deal to buy back his loyalty through a lucrative offer, including high-level positions in the army and government for himself, his uncle, and his brother. The deal reportedly involved “a cash payment of SPD [Sudanese pound] 1 billion (USD 444,000) for himself, and half that amount again for his brother, ostensibly to compensate fighters and their families. UN informants said 3,000 of his men were incorporated into the regular army, with army salaries, and 200 to 300 were promised officer training.”61 DESCRIPTIVE STATISTICS Sudan's wars saw defections in 64 faction years, and 329 years with no change in factional alignment, for an overall side switching risk of 16 percent per faction year (see table 1).62 Relative defection risks were remarkably high, at 18 percent in southern Sudan and 12 percent in Darfur. Defection was widespread from both the rebel and government camps in the two conflicts, with factions aligned on either side defecting at the same rate of 16 percent per faction year. Table 1. Factional Alignments in Sudan Government-Aligned Faction YearsRebel-Aligned Faction YearsGovernment to Rebel DefectionsRebel to Government DefectionsDefection Risk/Year Southern Sudan 193 70 29 19 18% Darfur 26 104 10 12% Full Sample 219 174 35 29 16% Government-Aligned Faction YearsRebel-Aligned Faction YearsGovernment to Rebel DefectionsRebel to Government DefectionsDefection Risk/Year Southern Sudan 193 70 29 19 18% Darfur 26 104 10 12% Full Sample 219 174 35 29 16% In terms of the independent variables, the means in table 2 indicate the proportion of cases in which a given variable predicts an increased risk of defection (i.e., is coded as a 1). The high means for both ethnicity (M=0.56) and ideology (M=0.47) indicate that roughly half of factional alignments were across ethnic and ideological macro cleavages. The low mean for the territorial control variable (M=0.05) suggests how infrequently state and rebel forces were able to compel collaboration through territorial control. Internecine violence among allies was relatively widespread, as indicated by the mean for the fratricidal flipping variable (M=0.13), with more than 52 instances of factions targeted by stronger allies. This undercounts the incidence of infighting by half because the data indicate a defection risk only for the weaker faction(s). Political rivalries and material incentives to defect were observed in 10 percent and 16 percent of faction years, respectively, suggesting just how active insurgent and counterinsurgent organizations were in seeking alliances through the exploitation of local rivalries and offers of material rewards. Table 2. Zero-Order Correlations and Descriptive Statistics AlignmentEthnicityIdeologyTerritorial ControlFratricidal FlippingPolitical RivalryPatronage Incentives Ethnic Identity −0.07 Ideology 0.03 0.80** Territorial Control 0.11* 0.05 0.07 Fratricidal Flipping 0.13** 0.07 0.07 −0.01 Political Rivalry 0.47** −0.19** −0.20** 0.07 0.13* Patronage Incentives 0.57** 0.02 0.09 0.05 −0.09 0.38** Mean 0.16 0.56 0.47 0.05 0.13 0.10 0.16 N=393 AlignmentEthnicityIdeologyTerritorial ControlFratricidal FlippingPolitical RivalryPatronage Incentives Ethnic Identity −0.07 Ideology 0.03 0.80** Territorial Control 0.11* 0.05 0.07 Fratricidal Flipping 0.13** 0.07 0.07 −0.01 Political Rivalry 0.47** −0.19** −0.20** 0.07 0.13* Patronage Incentives 0.57** 0.02 0.09 0.05 −0.09 0.38** Mean 0.16 0.56 0.47 0.05 0.13 0.10 0.16 N=393 * Correlation is significant at the 0.05 level (two-tailed). ** correlation is significant at the 0.01 level (two-tailed). Christia's data on alliances in Afghanistan and Bosnia, which provide some of the only data available for comparison, help to put the above statistics into perspective. Examining commanders in three Afghan provinces, Christia finds relatively little shift in alignments among commanders active across different periods in the Afghan-Soviet war (1978–89) and the Afghan civil war (1992–98). In the first war, only 5 of 114 commanders switched sides, for a defection risk of 4 percent per commander.63 In the second war, 17 of the 64 commanders switched sides in bandwagoning with the Taliban, for a defection risk of 27 percent per commander.64 Christia's Afghan data, however, are coded over multiyear periods of three to seven years from provinces expected to see the most intense local rivalries, suggesting a far lower incidence of side switching than the annualized risk of 12 percent and 18 percent among factions in Darfur and southern Sudan, respectively. In the Sudan sample, 41 of 92 factions switched sides at some point (with 23 of them switching two or more times). A key theoretical implication of this analysis concerns the organization of violence at the local and national levels. In both Afghanistan and Bosnia, Christia finds that “actors appear more or less compliant with dynamics taking place over their heads.”65 For instance, she finds that 26 percent of commanders in Afghanistan's jihad (1978–89) and 14 percent in the country's intramujahedin war (1992–98) engaged in violent clashes incompatible with their higher-order alliances.66 Again, these data are recorded across different multiyear periods from provinces expected to experience infighting, implying a lower rate than the annualized 20 percent and 33 percent risk of infighting observed in Darfur and South Sudan, respectively. Similarly, whereas Christia convincingly demonstrates that local and national ethnic elites in Bosnia acted in concert when coordinating violence against their group's common enemies, the misalignment of Sudanese factions in terms of ethnicity and ideology (see table 1) suggests that factional alignments cut across group cleavages at the regional and national levels. Balancing considerations at the group or subgroup level are thus unlikely to be driving faction-level alignment behavior in southern Sudan or Darfur. RESULTS AND INTERPRETATION To test the hypotheses above, I employ a Cox proportional hazards model, an appropriate test given the structure of the data and hypotheses.67 The survival analysis treats each factional alignment as a case and assesses the durability of alignments and risks for failure in terms of defection. The results are presented in table 3 as hazard ratios, with a ratio greater than 1 indicating an increased risk of defection for a particular explanatory variable. Model 1 includes all cases; subsequent models disaggregate the data. Models 2 and 3 focus on southern Sudan and Darfur to assess conflict-specific effects, and models 4 and 5 examine rebel- and government-aligned factions to examine whether different factors shape defection from rebel or government camps.68 Table 3. Cox Proportional Hazard Ratios for Alignments in Sudan's Civil Wars Hazard Ratios Independent VariableModel 1: Southern Sudan and DarfurModel 2: Southern SudanModel 3: DarfurModel 3: Rebel-AlignedModel 4: Government-Aligned Ethnicity 0.64 (0.40) 0.46 (0.63) 1.65 (0.66) 0.67 (0.51) 0.76 (0.78) Ideology 1.26 (0.40) 1.18 (0.67) 1.54 (0.62) 1.23 (0.65) 0.39 (0.78) Territorial Control 1.67 (0.40) 1.38 (0.45) 2.79 (1.58) 2.06 (0.58) 1.01 (0.62) Fratricidal Flipping 1.65* (0.29) 1.63 (0.33) 1.22 (1.07) 1.52 (0.45) 2.06* (0.43) Political Rivalry 3.72*** (0.32) 3.37** (0.38) 3.50* (0.72) 4.10** (0.53) 1.81 (0.42) Patronage Incentives 7.13*** (0.29) 5.41*** (0.32) 23.96*** (0.91) 8.17** (0.59) 006.80*** (0.36) Factional Alignments (subjects) 131 92 39 69 62 Number of Failures (defections) 64 49 16 29 35 Time At Risk (years) 393 263 130 174 219 Hazard Ratios Independent VariableModel 1: Southern Sudan and DarfurModel 2: Southern SudanModel 3: DarfurModel 3: Rebel-AlignedModel 4: Government-Aligned Ethnicity 0.64 (0.40) 0.46 (0.63) 1.65 (0.66) 0.67 (0.51) 0.76 (0.78) Ideology 1.26 (0.40) 1.18 (0.67) 1.54 (0.62) 1.23 (0.65) 0.39 (0.78) Territorial Control 1.67 (0.40) 1.38 (0.45) 2.79 (1.58) 2.06 (0.58) 1.01 (0.62) Fratricidal Flipping 1.65* (0.29) 1.63 (0.33) 1.22 (1.07) 1.52 (0.45) 2.06* (0.43) Political Rivalry 3.72*** (0.32) 3.37** (0.38) 3.50* (0.72) 4.10** (0.53) 1.81 (0.42) Patronage Incentives 7.13*** (0.29) 5.41*** (0.32) 23.96*** (0.91) 8.17** (0.59) 006.80*** (0.36) Factional Alignments (subjects) 131 92 39 69 62 Number of Failures (defections) 64 49 16 29 35 Time At Risk (years) 393 263 130 174 219 NOTE: Table presents hazard ratios (the exponentiated form of the coefficient). *** Significant at the 0.001 level. ** significant at the 0.01 level. * significant at the 0.10 level. The findings suggest that patterns of political rivalry and patronage shape alignments. The risk of defection increases anywhere from three to four times in the presence of an intense and visible political rivalry that outsiders are willing to exploit through offers of military support; incentives to instrumentalize collaboration in pursuit of patronage-based rewards increase the probability of defection anywhere from four to eight times. The results from models 2 and 3 comparing southern Sudan and Darfur suggest that similar dynamics operate in both wars. Interestingly, models 4 and 5 comparing rebel- and government-aligned factions suggest that whereas the patronage mechanism shapes alignments on both sides, the political rivalry variable is only significant for rebel defections to the government. The difference in the ability of rebel and government forces to manipulate rivalries to gain defections likely reflects the greater military resources and reach of state forces. Sudan's military intelligence has proven notoriously adept at stoking rivalries in both wars, and the army uses means unavailable to rebels, such as airdrops to government-aligned factions behind enemy lines. The outsized hazard ratio for the material incentives variable in the Darfur case reflects the small number of cases and the remarkable success the government and rebels have had in buying off factions on the opposing side. One source claims, for instance, that Abdel Wahid's rebel faction in Darfur paid the militia of the government-aligned Nuwaiba tribe more than1.5 million to enlist its support in factional infighting in early 2010.69

These findings cast doubt on the relevance of other explanations for side switching in Sudan's civil wars. Alignments incompatible with ethnic (H1) or ideological (H2) cleavages had no higher risk for defection than identity-compatible alignments. To interpret the findings for ethnicity and ideology, recall that the data measure these variables only in terms of broader identities around which each side mobilizes. There is ample evidence that sub-ethnic identity cleavages and narrow ideological disputes within movements matter for alignment dynamics, particularly as they become entangled in the sorts of political and economic struggles that I argue drive side switching. Moreover, factional alignments do not reflect the size of each faction or its support in the wider population. Although half of factional alignments in southern Sudan cut across ethnic or ideological cleavages, far more southerners fought for and supported the most powerful rebel faction, the SPLM/A, than the government or its southern allies.

Territorial control and organizational demand (H3) finds no support as a mechanism triggering factional defection, reflecting the limited abilities of actors on either side to coerce collaboration. Despite high organizational demand for collaborators, Sudan's vastness provides sanctuary to factions unwilling to be coerced into fighting alongside former enemies. In the face of chronic weakness, collaboration may sometimes be obtained most effectively by renouncing the strict imposition of control. For example, after the SPLA's use of coerced recruitment in administering traditionally hostile tribes in Equatoria generated a backlash from local tribal militia, the SPLA relied on a “mobile strategy” of keeping units on the move to avoid imposing on civilians.70 In Sudan's wars, control follows collaboration more often than it precedes it.71

The fratricidal flipping hypothesis (H4) finds some support in the full set of cases and interestingly among government-aligned factions but not rebel-aligned ones. At key points in the war in the South, Sudan's government promoted infighting among its militia allies to prevent them becoming too strong, and a number of them defected to the SPLA. As with the territorial control mechanism, however, Sudan's geography often limits the ability of factions to credibly threaten the survival of weaker ones when they have the option of retreat into sparsely controlled areas.

The results in table 3 are robust to alternative measurements, controls, and specifications. First, as a faction's identity repertoire contains both ethnic and ideological elements, I created composite variables that predict defection if factions are misaligned in either ethnic or ideological terms, and heightened risk of defection when misaligned along both axes of identity simultaneously. The results for the composite identity variables were insignificant across multiple specifications. Second, the findings were robust when controlling for prior defections to test whether serial defectors were driving results. Finally, there are significant interaction effects between the rivalry and patronage variables, consistent with the close association of power and wealth in war economies.72

A number of factors caution against reading too much into the statistical findings in isolation from the case studies discussed below. First, omitted variable bias is a potential problem, though one mitigated by the inclusion of the main arguments from the literature. Exclusion of the relative power hypothesis is unlikely to skew the results; the tight linkages between competing groups and local commanders on which the theory relies are absent in Sudan.73 Ideally, I would also include variables assessing how the regional and international politics of intervening states affect alignments, but collecting such data proved infeasible.74 Second, a degree of measurement error is inevitable in tracking territorial control, infighting, rivalries, and patronage relations in a data-poor environment. Systematic measurement errors—particularly “false negatives” in the form of unreported offers of military backing or material payouts—could affect the substantive results. Finally, though twenty-eight years of war in in the South and eleven in Darfur provide numerous factional alignments, the sample is still relatively small.

Comparative Case Studies

To provide further evidence and bolster causal identification, I turn to case studies of three factional alignments. The cases are purposively sampled for wide variation on the explanatory variables, inclusion of both rebel- and government-aligned factions, factions ranging in size and power, and historical significance. All three cases are “on the line” in the sense of being well predicted by the model above. I first examine the split in the SPLA in 1991 that led to years of devastating fighting and a legacy that laid the foundations for the current civil war in South Sudan. The second case traces the nondefection of Gordon Kong's faction of the South Sudan Defense Force (SSDF) after the 2005 peace agreement initiated an auction for the loyalties of government-aligned militias. The third case examines the pivotal defection of Minni Minawi's faction of the Sudan Liberation Army (SLA-MM) in Darfur during 2006. This realignment altered the course of the war, frustrated peace efforts, sparked chaotic infighting and displacement that disrupted relief efforts, and created a humanitarian emergency.

The evidence comprises an array of secondary sources on Sudanese armed groups. In the case of the defection of SPLA-Nasir in 1991, secondary sources are supplemented by observations from fieldwork over the course of three research visits to southern Sudan from 2002 to 2005 to SPLA-held areas in Upper Nile and Eastern Equatoria. This fieldwork included interviews with humanitarian relief workers, insurgents, commanders, nongovernmental organizations involved in reconciliation efforts between southern factions, and SPLM/A officials.

The case studies offer a number of findings summarized in table 4. The causal evidence suggests that factional rivalry and patronage powerfully shape factional alignments in these cases. There is little support for other explanations. Complex identity politics, tenuous territorial control, barriers to launching decisive fratricidal assaults, and a disconnect between faction-level behavior and group-level power balances enable frequent side switching in Sudan.

Table 4.

Predictions and Outcomes for Case Studies

Do hypotheses predict side switching?
Place, DateFactionsInitial AlignmentEthnicity H1Ideology H2Territorial Control H3Fratricidal Flipping H4Relative Power H5Political Rivalry H6aPatronage Incentives H6bDefect?
Southern Sudan, 1991 SPLA-Nasir Rebel no no no no no yes no yes
Southern Sudan, 2006 SSDF Kong Government yes yes no no yes no no no
Darfur, 2006 SLA-MM Rebel no no no yes yes yes yes yes
Do hypotheses predict side switching?
Place, DateFactionsInitial AlignmentEthnicity H1Ideology H2Territorial Control H3Fratricidal Flipping H4Relative Power H5Political Rivalry H6aPatronage Incentives H6bDefect?
Southern Sudan, 1991 SPLA-Nasir Rebel no no no no no yes no yes
Southern Sudan, 2006 SSDF Kong Government yes yes no no yes no no no
Darfur, 2006 SLA-MM Rebel no no no yes yes yes yes yes

NOTES: Predictions reference whether the faction was expected to switch sides given the initial alignment. Bolded cells correctly predicted. Factions refer to Sudan People's Liberation Army-Nasir (Riek Machar and Lam Akol), South Sudan Defense Forces (Gordon Kong), and Sudan Liberation Army-MM (Minni Minawi).

THE SPLIT: THE DEFECTION OF SPLA-NASIR, 1991

After years of steady battlefield gains and success in co-opting government-aligned factions throughout southern Sudan, a failed coup in August 1991 divided the SPLM/A into two competing factions and reversed its momentum. The split led to a decade of internecine violence between the SPLA and a loose coalition of militias that emerged out of the breakaway SPLA-Nasir faction that aligned with the government in the North.

The defection had its origins in a failed bid for leadership by senior SPLA commanders who received pledges of weapons, ammunition, and military support from a government eager to weaken the SPLA in ways consistent with the political rivalry mechanism (H6a). The architects of the coup, Lam Akol and Riek Machar, two prominent and politically ambitious SPLA commanders, had become marginalized before defecting. Akol, in particular, had fallen out of favor with leader John Garang after an effective demotion in 1990.75 The coup plotters circulated critiques of Garang's autocratic rule among senior commanders, only to see support for a leadership change dissipate after the fall of the regime of Mengistu Haile Mariam pushed the SPLA out of its vital Ethiopian rear bases. Backed into a corner when Garang called an unprecedented meeting of commanders that the coup leaders feared would lead to their arrest, they prematurely proclaimed Garang's overthrow. Before hatching the plot, however, they had obtained the support of the Sudanese government, which was keen to sow divisions within the SPLA.76 When the bulk of SPLA commanders failed to support the coup, the breakaway faction became increasingly reliant on supplies, ammunition, and weapons from the government and the military backing of other government-aligned factions.

Several other commanders were later convinced to defect from the SPLA and collaborate with the splinter faction through material inducements (H6b), but these do not appear to have motivated the initial defection by the coup leaders.77 For example, William Nyuon Bany, the SPLA chief of staff, was given financial inducements to defect after meetings with the government at peace talks.78 The process of fragmentation unleashed by the split, however, was sustained by tribal rivalries and the predations of local strongmen.79 A series of retributive raids led to widespread violence and a “steady siphoning of civilian cattle into the byres of individual southern military leaders and warlords.”80

Alternative explanations shed little light on the defection of the commanders who went on to form SPLA-Nasir. The split tends to be seen in terms of ethnic animosity between the Nuer and Dinka (H1), with Machar's Nuer fighting Garang's Dinka. Yet this explanation cannot account for the SPLA-Nasir's alignment with an Arab government, which continued to employ indiscriminate violence against both Dinka and Nuer communities. Indeed, most of the subsequent fighting occurred between rival sections of Nuer aligned on either side.81 The initial coup was ideologically motivated (H2), appealing to the calls for an openly separatist platform, resentment at Garang's leadership, and the SPLA's awful human rights record.82 Yet events showed this to be little more than posturing; the breakaway SPLA-Nasir's deepening alliance with the government belied its separatist aims, while its leaders’ behavior mirrored the worst excesses of Garang's dictatorial control, and its troops committed terrible human rights abuses.83

Nor can the defection be explained in terms of territorial control (H3), fratricidal flipping (H4), or relative power considerations (H5). Despite the Sudanese government's support for the coup, it was launched from territory under rebel control—albeit along the front with lines of supply from government forces. Although the fratricidal flipping mechanism (H4) explains SPLA-Nasir's increasing reliance on the government as fighting with the rump SPLA intensified, the plot was hatched only after government promises of support were in place and the coup preceded any actual infighting. Moreover, government forces had little direct capacity to provide protection to SPLA-Nasir units, though they did direct other government-aligned militias in the area to support them. Finally, the defection did not reflect shifts in relative power (H5) among groups fighting within the ethnic alliance the SPLA/M had constructed. The loss of Ethiopian rear bases had critically weakened the SPLA. Dinka commanders, however dominant within the SPLA, were far from making up a viable minimum winning coalition at that stage of the war. The coup “was an opportunistic move, carefully calculated to secure the leadership from Garang at the weakest moment in the SPLA/M's recent history.”84 Although the coup plotters—Machar, a Nuer, and Lam Akol, a Shilluk—were from groups with strained relations with the Dinka core of the SPLA, many prominent Nuer and Shilluk failed to support the coup.

THE NONDEFECTION OF SSDF GORDON KONG, 2006

After more than two decades of war, the 2005 Comprehensive Peace Agreement formally ended the war—but it did not bring the violence to a close. The agreement stipulated the dissolution of all armed groups in southern Sudan other than the SPLA and the Sudan Armed Forces, leading to intense bargaining over the loyalties of government-aligned factions, many of them organized around the South Sudan Defense Forces umbrella group. Many SSDF factions reconciled with the SPLA, signing the “Juba Declaration” in January 2006, which offered amnesty and integration into the army of the newly autonomous Government of Southern Sudan. A number of prominent commanders refused the offer, however, and continued to receive support from the northern government, including Gordon Kong ‘s SSDF faction.85

Material incentives (H6b) did much to shape the alignment calculus of these factions. The SPLA lured many factions by providing various rewards, tolerating rampant corruption by commanders who accepted the amnesty, and offering high-level army positions for mostly illiterate commanders and jobs in the army for their fighters.86 Many saw the nondefection of Kong as a reflection of the SPLA's unwillingness to provide bribes high enough to induce him to switch sides.87 Prospectively, his faction controlled valuable territories containing oil-processing facilities that he stood to lose if he defected.88 Property holdings in Khartoum also helped to keep him loyal.89

The political rivalries (H6a) that Kong accumulated during the war might also have played a role in keeping him loyal, though the evidence here is more speculative. He led a government-aligned militia in the mid-1980s, before reconciling and reintegrating with the SPLA in 1988, only to defect during the split in 1991. He had also clashed with several other commanders who reintegrated with the SPLA when they were allied in the government camp. This checkered past likely diminished his faction's value to the SPLA, decreasing his leverage in bargaining and raising questions about his prospects in an organization dominated by so many of his enemies.

Neither ethnicity (H1) nor ideology (H2), both of which point toward a realignment with the SPLA, shaped Gordon Kong's collaboration with the government. Many government-aligned commanders who stayed out of the SPLA after 2006 were Nuer, like Kong (a Jikany Nuer), but many Nuer commanders were represented in the SPLM/A at high levels—including Riek Machar, who was offered the vice presidency before he was dismissed from government and launched a rebellion in late 2013. Ideologically, Kong was a committed separatist, but he continued to collaborate with Khartoum even after South Sudan became independent in 2011.

Despite territorial control (H3) and fratricidal flipping (H4) both correctly pointing to no change in the faction's alignment, evidence suggests that neither played a causal role. Although Kong's faction was deployed next to a garrison of government troops during 2006, he effectively controlled his own personal fiefdom and was not coerced to continue collaborating. Other SSDF commanders deployed alongside the Sudanese army defected without incident. In terms of fratricidal flipping, the split between SSDF commanders defecting to the SPLA and those remaining loyal to the government of Sudan was mostly peaceful. The SPLA sought to provide security guarantees to defecting commanders to demonstrate the credibility of its commitments to holdouts, mitigating concerns for survival. Finally, although this particular case cannot be explained by relative power considerations (H5), the theory's prediction of bandwagoning once military victory appears imminent does explain the wider pattern of government-aligned militia defecting to the SPLM/A after the 2005 Comprehensive Peace Agreement signaled likely victory.

DIVIDING THE SLA: THE DEFECTION OF SLA-MM, 2006

In May 2006, Minni Minawi, leader of a faction of the SLA-MM, signed the Darfur Peace Agreement with the government of Sudan. Minawi's defection was a critical juncture in the war in Darfur, shifting allegiances on both sides of the war and leading to a serious escalation in fighting.

Minawi's side switching reflected rivalry for dominance within Darfur's rebel movement (H6a), with his slide into the government camp greased by pledges of cash and powerful government positions (H6b). By the time of peace talks in late 2005, Minawi was politically marginalized within the SLA, but militarily powerful. He had appointed members of his own clan to impose his authority on the Zaghawa wing of a divided SLA, generating resentment of his blatant tribalism. Intense clashes with other factions, particularly with the SLA faction of his rival, Abdel Wahid, led to widespread infighting and pushed both leaders to find external support to tip the balance. Before defecting to the government, Minawi had obtained promises of weapons, vehicles, and logistical support from Sudan's military, gambling that government backing would secure a position of dominance for him in Darfur.90

Minawi's side switching also reflected Khartoum's strategic use of patronage (H6b). The government's approach at the peace talks was “to compile a file on every single individual in the opposing camp and try to buy them off one by one, with money or positions—or both.”91 By signing the Darfur Peace Agreement, Minawi became the most powerful official in Darfur, at least on paper. His followers also obtained lucrative appointments, and the government reportedly financed his office in Khartoum with funds of approximately 1 million a month.92 To be clear about the causal sequence, these rewards were offered as an incentive prior to the faction's defection. Minawi's defection was a miscalculation, however; rather than consolidating his leadership over the resistance, within months he had lost territory, fighters, and vehicles, as well as political credibility.93 Patterns of infighting might have played a role in the defection, with the heavy fighting between rebel factions preceding Minawi's realignment suggestive of fratricidal flipping (H4). Several factors complicate this reading, however. SLA-MM largely initiated the internecine violence and was arguably militarily dominant at the time of his defection. Moreover, government efforts to divide the rebels at the peace talks were highly conspicuous.94 Minawi's defection did not involve relative power considerations and balancing behavior (H5), whether at the national level or the regional level of competing Fur and Zaghawa. Minawi was very much the junior partner in the coalition he forged with the government through the Darfur Peace Agreement. The likely prospect that the government would freeze Minawi's faction out of a meaningful share of the spoils should have given him pause if relative power balances and commitment problems were determining alliance choices. Indeed, Minawi and his supporters were increasingly marginalized, including from Minawi's own Zaghawa kin, leading SLA-MM back into rebellion in late 2010. The other hypotheses provide incorrect or inadequate explanations for Minawi's defection. There was no clear relationship between broader ethnic identities and alignment around the defection (H1). By co-opting SLA-MM, the government divided Zaghawa loyalties, with Zaghawa commanders and fighters prominent in several factions on the rebel side, including the powerful Justice and Equality Movement. Contrary to interpretations emphasizing ideologically motivated spoiling behavior around a peace agreement (H2), Minawi's negotiating position was actually more hard-line than his rival Abdel Wahid's; the intense rivalry between the two meant that if one signed, the other would not.95 Lastly, the government lacked territorial control over the parts of Darfur where Minawi operated, meaning it was unable to coerce his defection (H3). Instead, his defection put government-aligned forces in the heart of rebel-held areas. Beyond Sudan Other conflicts suggest the broader relevance of the dynamics highlighted in the Sudan case for civil war alignments. The opportunistic factional alignments observed above reflect Sudan's nested and ambiguous identity cleavages, constraints on exercising coercive power over such a vast area, and the sharp disjuncture between violence at the local and national levels. These conditions are common in weak states, particularly in Africa, where strategies of governance and counterinsurgency rely on alliances forged through mutual rivalry and patronage. In Chad, for instance, fighters and commanders follow a “trajectory of ‘rebellion-reintegration-defection,'” as they switch sides repeatedly. “Political and military actors are continuously reconsidering their tactical loyalty to the regime,” writes Marielle Debos. “Their decision to join a rebellion or not depends on political calculations based on their perception of the situation and on their own ambition to achieve a better position within the security forces.”96 The rebellion in Mali in 2012 saw the defection of mostly Tuareg soldiers previously integrated into the national army. Among them were commanders from elite U.S.-trained units who switched allegiances to fight alongside Ansar-Dine, the Movement for Unity and Jihad in West Africa, and al-Qaida in the Islamic Maghreb.97 Rather than ideology or Islam being a factor, recruitment into these groups tended to occur along the lines of clanship and integration into regional trafficking networks.98 The restoration of order in Mali has seen the government in Bamako resurrect the same clientelist system that has fueled a cycle of revolts, using bribes to underwrite alliances with select Arab and Tuareg groups in the north while exploiting divisions within them by selectively backing local notables.99 The alignments of armed clan militias in Somalia constantly change in response to local rivalries or outright bribes in ways that perpetuate violence as factions chase “crisis rents.”100 Similarly, in Congo the fragmentation of militias with loose allegiances has been driven by local conflicts over land and control of natural resources, taxes, and social status in ways evocative of the opportunistic political economy discussed above.101 In other weak states, one also sees alignment following the contours of rivalry and patronage. Akbar Ahmed's work on tribal Islam notes countless examples of this pattern. Rulers unable to control fractious societies seek the collaboration of peripheral tribal elites through elaborate patronage relations, using offers of support to allow weaker actors to check stronger ones.102 Variations on this theme can be seen in contemporary Afghanistan. Both the Taliban and the Afghan government have proven able to induce commanders from the opposing side to defect through offers of cash and weapons.103 Similarly, the “Anbar Awakening” in Iraq saw Sunni tribes realign to fight alongside their former enemies as U.S. and coalition forces employed them to hunt down al-Qaida in Iraq. The realignment of Sunni tribes helped to stem the insurgency through transfers of cash and weapons to former insurgents before the U.S. withdrawal, with many militias subsequently defecting back to the insurgency.104 The opportunistic theory of alignments stresses the conditions that provide actors wider latitude to instrumentally switch allegiances. By implication, a different pattern of alignments should prevail with more salient ethnic and ideological cleavages, tighter territorial control, higher capacities to credibly threaten factional rivals, and a closer coupling of local factions with national alliances in ways that promote balancing behavior. Nonetheless, one occasionally sees opportunistic alignments under different conditions than those emphasized here. Economic incentives were at the root of the split that launched the intra-Muslim civil war in Bosnia, where violence forged deep ethnic cleavages.105 In Afghanistan, where local commanders tended to align according to group demands, Christia notes that those alignments deviating from this pattern tended to involve local scrambles for power and monetary payments to tribal militias.106 In Chechnya, the transformation of the Kadyrovtsy militia into government-backed counterinsurgents reflected competition among rival Chechen factions vying for Russian support in localized power contests. Moscow used generous infusions of cash and weapons to rent the loyalties of local strongmen as it reasserted control over Chechnya.107 Conclusion After a meeting with a powerful Arab militia leader in Darfur, one American diplomat noted the “the shallow, mercurial nature of Khartoum's relationship with their most trusted allies in the field.”108 This observation, I argue, reflects the centrality of guns and money in sustaining opportunistic collaboration. Rather than discounting existing explanations, I argue that their scope conditions limit their applicability to an important category of civil wars fought in weak states characterized by low barriers to defection that generate highly fluid alignments. The findings advance existing research in several ways. Whereas existing arguments highlight immediate threats to survival and concerns with postwar power maximization, I find lower threats to survival and alignments motivated by short-term, instrumental calculations. Recent work suggests a limited role for states generally, and weak states in particular, in manipulating the internal politics of armed groups to trigger defections.109 Yet in Sudan, a disorganized, resource-poor “weak state” and the fragmented rebel movements challenging it have proven able to repeatedly gain the instrumental collaboration of factions on the opposing side. Contrary to work on alliances at the group level and subgroup level, this study suggests that important dynamics play out at lower levels as ethnic groups and tribes are split into warring camps aligned on either side. The findings contribute to important theoretical and policy debates and raise questions for future research. One set of theoretical implications from the findings above concerns ethnic politics. Prominent accounts of Sudan's wars emphasize identity politics. “The state of affairs now prevailing in Sudan,” argues Francis Deng, “is the culmination of a long historical process in which northerners and southerners were the principal antagonists in the war of racial, cultural, and religious identities.”110 There is substantial truth in this view, which mirrors influential scholarship on the role of ethnic exclusion in causing civil wars and increasing ethnic cohesion.111 Yet the identities structuring Sudan's wartime cleavages provide little guidance in mapping loyalties or patterns of violence on the ground. Rather than relying on ethnic exclusion, Sudan's government has simultaneously included and excluded different subgroups.112 Indeed, African rulers have long frustrated ideologically minded reform rebels or pan-ethnic opposition through selective access to patronage opportunities and encouraging local conflicts to divide opponents.113 This implies, first, that identities arguably matter most for alignments when they are channeled through the subgroup rivalries at the core of the opportunistic theory of alignments outlined above. Second, it suggests that the ethnic polarization is more a consequence than a cause of divergent alignments in civil wars. Third, it calls for shifting the focus to tribal and kinship relations, rather than ethnicity per se. Unlike categorical ethnic identities that consider groups as neatly bounded sets of individuals competing for hierarchical rank and control of the state, tribal identities are relational and situate shifting groups within social orders defined by the sorts of conflicts emphasized here.114 A second set of implications concerns counterinsurgency and state building. Counterinsurgents frequently exploit rivalries and use patronage to assert a modicum of control, however tenuous, over rebellious peripheries.115 Triggering defections poses considerable risks, however. Without a capable army to consolidate state authority in areas under the control of militia temporarily brought over to its side, Khartoum's reliance on them spiraled out of control. Regional elites alternated between loyalty and rebellion to drive up the price of their incorporation into the state's privilege structure, rendering large parts of the country ungovernable. The government ultimately lost South Sudan and has little to show for more than a decade of fighting in Darfur. Putting former adversaries on their payrolls and fueling local rivalries might serve the short-term purposes of counterinsurgency campaigns or military interventions. From a state-building perspective, however, such practices exact a high price, fragmenting organizations, depleting social capital, and weakening state institutions. In terms of policy, the account here suggests a number of implications, particularly given the rise in multiparty civil wars fought by fragmented armed groups. The analysis underscores the need for conflict resolution to mend rifts within local communities, particularly those where groups were divided into factions fighting on opposite sides. International mediation efforts in Sudan, as elsewhere, have neglected opaque peripheral conflicts to concentrate on those at the center, turning peace negotiations into an extension of elite bargaining over the distribution of lucrative sinecures.116 This study also suggests the importance of interventions targeting local wartime processes that militarize political competition and devastate the economies of the communities from which armed groups mobilize.117 The finding that cash substitutes for conviction in many wars is not surprising given the economic damage wrought by protracted violence. One study found that almost half of the households in one community in North Darfur were dependent on militia salaries, for example.118 Finally, this study yields three topics for future research. First, it suggests the need for greater attention to the phenomenon of nonalignment, particularly in multiparty civil wars. The leaders of several tribes in Darfur, for instance, managed to resist government pressure to fight as paramilitaries, even as their neighbors were drawn into the war. The theory of opportunistic alignments advanced above suggests that remaining neutral should correlate with the ability of a faction to manage rivalries with neighbors and a degree of economic security. In Darfur, for example, many of the groups that managed to avoid being pulled into the war possessed secure land claims, good relations with neighboring tribes, and the political power to defend their autonomy against state pressure; conversely, paramilitaries were recruited from marginalized and impoverished groups, particularly the nomadic Arab tribes lacking land title, landless migrants from Chad, and those targeted by rebel raiding.119 Second, the inner social workings of the networks through which factions mobilize also require further research.120 The analysis above largely takes factions as a given, but they have diverse origins that represent the confluence of multiple interests and agendas. Factional leaders often occupy positions of traditional authority with attendant obligations to their followers. Side switching is a means to power and wealth for these men, but it can also be a potential means of upward mobility for the communities backing them. At other times, commanders’ social ties can act as a break on opportunism. The social embeddedness of factional leaders and outsiders’ reliance on them as intermediaries in forms of indirect rule also help to explain the puzzling leniency with which they are treated after double-crossing former allies. For instance, “Hemeti,” the aforementioned militia commander who led a costly mutiny against Sudan's government in 2007–08, has since been promoted to the rank of brigadier general in Sudan's army, partly owing to his ability to command the allegiances of the Mahariya section of the Northern Rizeigat, one of the largest Arab tribes in South Darfur. Lastly, future research should seek to synthesize research on mobilization and participation in war, fragmentation and cohesion in armed groups, infighting among them, and their alignment decisions to understand how these related dynamics fit together. This study suggests that splits in armed groups and infighting lead to side switching when the risks are relatively low and organizations on the opposing side are willing to transfer resources to underwrite mutually advantageous collaboration. The next step is to devise a coherent theoretical account of behavior across these processes as actors seek power, profit, and survival amid the violence around them. Appendix 1 Hazard Ratios Independent VariableModel 5: Full SetModel 6: Full Set Political Rivalry 10.1* (0.27) Patronage Incentives 10.7* (0.26) Factional alignments (subjects) 134 134 Number of defections (failures) 64 64 Time at risk (years) 393 393 Hazard Ratios Independent VariableModel 5: Full SetModel 6: Full Set Political Rivalry 10.1* (0.27) Patronage Incentives 10.7* (0.26) Factional alignments (subjects) 134 134 Number of defections (failures) 64 64 Time at risk (years) 393 393 NOTE: Table presents hazard ratios (the exponentiated form of the coefficient). * Significant at the 0.001 level. Acknowledgments The research for this article was funded by a Veni research grant from the Dutch Scientific Organization. The author is grateful for research assistance from Anna Morath and comments from Ariel Ahram, Kristin Bakke, Kathleen Gallagher Cunningham, Ursula Daxecker, Lotje de Vries, Maria Koinova, Jorg Kustermans, Bethany Lacina, Romain Malejacq, Theodore McLauchlin, Sabine Otto, Costantino Pischedda, Andrea Ruggeri, Marieke Schomerus, Henning Tamm, Rebecca Tromble, and the anonymous reviewers. Notes 1. Antonio Giustozzi, “The Taliban beyond the Pashtuns,” Afghanistan Papers No. 5 (Waterloo, Canada: Centre for International Governance Innovation, July 2010). 2. International Crisis Group, “Central African Republic: Priorities of the Transition,” Africa Report No. 203 (Brussels: International Crisis Group, June 11, 2013). 3. Stig Jarle Hansen, Al Shabaab: The History and Ideology of a Militant Islamist Group 2005–2012 (Oxford: Oxford University Press, 2013). 4. International Crisis Group, “Eastern Congo: Why Stabilization Failed,” Africa Briefing No. 91 (Brussels: International Crisis Group, October 4, 2012); and Timothy Raeymaekers, “Post-War Conflict and the Market for Protection: The Challenges to Congo's Hybrid Peace,” International Peacekeeping, Vol. 20, No. 5 (November 2013), pp. 600–617. 5. Victor Tanner and Jérôme Tubiana, “Divided They Fall: The Fragmentation of Darfur's Rebel Groups,” Small Arms Survey Working Paper No. 6 (Geneva: Small Arms Survey, July 2007). 6. International Crisis Group, “South Sudan: A Civil War by Any Other Name,” Africa Report No. 217 (Brussels: International Crisis Group, April 10, 2014); and Alex de Waal, “When Kleptocracy Becomes Insolvent: Brute Causes of the Civil War in South Sudan,” African Affairs, July 2014, pp. 347–369. 7. Stathis N. Kalyvas, “Ethnic Defection in Civil War,” Comparative Political Studies, Vol. 41, No. 8 (May 2008) pp. 1043–1068; Paul Staniland, “Between a Rock and a Hard Place: Insurgent Fratricide, Ethnic Defection, and the Rise of Pro-State Paramilitaries,” Journal of Conflict Resolution, Vol. 56, No. 1 (February 2012), pp. 16–40; Stephen Biddle, Jeffrey A. Friedman, and Jacob N. Shapiro, “Testing the Surge: Why Did Violence Decline in Iraq in 2007?” International Security, Vol. 37, No. 1 (Summer 2012), pp. 7–40; and Fotini Christia, Alliance Formation in Civil Wars (Cambridge: Cambridge University Press, 2012). 8. Adria Lawrence, “Triggering Nationalist Violence: Competition and Conflict in Uprisings against Colonial Rule,” International Security, Vol. 35, No. 2 (Fall 2010), pp. 88–122; Kathleen Gallagher Cunningham, “Understanding Strategic Choice: The Determinants of Civil War and Nonviolent Campaign in Self-Determination Disputes,” Journal of Peace Research, Vol. 50, No. 3 (May 2013), pp. 291–304; and Michael Woldemariam, “Battlefield Outcomes and Rebel Cohesion: Lessons from the Eritrean Independence War,” Terrorism and Political Violence, online edition, July 15, 2014. 9. Hanne Fjelde and Desirée Nilsson, “Rebels against Rebels: Explaining Violence between Rebel Groups,” Journal of Conflict Resolution, Vol. 56, No. 4 (August 2012), pp. 604–628; and Kathleen Gallagher Cunningham, Kristin M. Bakke, and Lee J.M. Seymour, “Shirts Today, Skins Tomorrow: Dual Contests and the Effects of Fragmentation in Self-Determination Disputes,” Journal of Conflict Resolution, Vol. 56, No. 1 (February 2012), pp. 67–93. 10. Scott Gates, “Recruitment and Allegiance: The Microfoundations of Rebellion,” Journal of Conflict Resolution, Vol. 46, No. 1 (February 2002), pp. 111–130; Paul D. Kenny, “Structural Integrity and Cohesion in Armed Organizations: Evidence from Protracted Conflicts in Ireland and Burma,” International Studies Review, Vol. 12, No. 4 (December 2010), pp. 533–555; Peter Krause, “The Structure of Success: How the Internal Distribution of Power Drives Armed Group Behavior and National Movement Effectiveness,” International Security, Vol. 38, No. 3 (Winter 2013/14), pp. 72–116; Jeffrey A. Friedman, “Group Structure and Political Violence: Theory and Evidence from the American Indian Wars,” Harvard University, 2014; and Theodore McLauchlin, “Desertion, Terrain, and Control of the Home Front in Civil Wars,” Journal of Conflict Resolution, forthcoming. 11. David E. Cunningham, “Veto Players and Civil War Duration,” American Journal of Political Science, Vol. 50, No. 4 (October 2006), pp. 875–892; Wendy Pearlman, “Spoiling Inside and Out: Internal Political Contestation and the Middle East Peace Process,” International Security, Vol. 33, No. 3 (Winter 2008/09), pp. 79–109; and Kathleen Gallagher Cunningham, Inside the Politics of Self-Determination (Oxford: Oxford University Press, 2014). 12. Kalyvas, “Ethnic Defection in Civil War”; and Staniland, “Between a Rock and a Hard Place.” 13. See, for example, Abdulkader H. Sinno, Organizations at War in Afghanistan and Beyond (Ithaca, N.Y.: Cornell University Press, 2008); Paul Staniland, “Organizing Insurgency: Networks, Resources, and Rebellion in South Asia,” International Security, Vol. 37, No. 1 (Summer 2012), pp. 142–177; Kristin M. Bakke, Kathleen Gallagher Cunningham, and Lee J.M. Seymour, “A Plague of Initials: Fragmentation, Cohesion, and Infighting in Civil Wars,” Perspectives on Politics, Vol. 10, No. 2 (June 2012), pp. 265–283; and Michael Findley and Peter J. Rudloff, “Combatant Fragmentation and the Dynamics of Civil War,” British Journal of Political Science, Vol. 42, No. 4 (October 2012), pp. 879–901. 14. Christia, Alliance Formation in Civil Wars; Jesse Driscoll, “Commitment Problems or Bidding Wars? Rebel Fragmentation as Peace Building,” Journal of Conflict Resolution, Vol. 56, No. 1 (February 2012), pp. 118–149; Seden Akcinaroglu, “Rebel Interdependencies and Civil War Outcomes,” Journal of Conflict Resolution, Vol. 56, No. 5 (October 2012), pp. 879–905; and Håvard Mokleiv Nygård and Michael Weintraub, “Bargaining between Rebel Groups and the Outside Option of Violence,” Terrorism and Political Violence, online edition, April 25, 2014, http://www.tandfonline.com/doi/abs/10.1080/09546553.2013.829459?tab=permissions#tabModule. 15. William Reno, Warlord Politics in African States (Boulder, Colo.: Lynne Rienner, 1998); Kimberly Marten, Warlords: Strong-Arm Brokers in Weak States (Ithaca, N.Y.: Cornell University Press, 2012); and Mats Utas, ed., African Conflicts and Informal Power: Big Men and Networks (London: Zed, 2012). 16. Paul Staniland, “States, Insurgents, and Wartime Orders,” Perspective on Politics, Vol. 10, No. 2 (June 2012), pp. 243–264. 17. Stathis N. Kalyvas, “Ontology of Political Violence: Action and Identity in Civil Wars,” Perspectives on Politics, Vol. 1, No. 3 (September 2003), pp. 475–494. 18. This definition excludes factions that switch between organizations and alliances that fight one another in addition to the government. 19. Donald L. Horowitz, Ethnic Groups in Conflict (Berkeley: University of California Press, 1985); Walker Connor, Ethnonationalism: The Quest for Understanding (Princeton, N.J.: Princeton University Press, 1993); Roger D. Petersen, Understanding Ethnic Violence: Fear, Hatred, and Resentment in Twentieth-Century Eastern Europe (Cambridge: Cambridge University Press, 2002); Stuart J. Kaufman, Modern Hatreds: The Symbolic Politics of Ethnic War (Ithaca, N.Y.: Cornell Univeristy Press, 2001); and Lars-Erik Cederman, Kristian Skrede Gleditsch, and Halvard Buhaug, Inequality, Grievances, and Civil War (Cambridge: Cambridge University Press, 2012). 20. Barry R. Posen, “The Security Dilemma and Ethnic Conflict,” Survival, Vol. 35, No. 1 (Spring 1993), pp. 127–147; and Chaim D. Kaufmann, “When All Else Fails: Ethnic Population Transfers and Partitions in the Twentieth Century,” International Security, Vol. 23, No. 2 (Fall 1998), pp. 120–156. 21. James D. Fearon and David D. Laitin, “Explaining Interethnic Cooperation,” American Political Science Review, Vol. 90, No. 4 (December 1996), pp. 715–735. 22. Theodore McLauchlin, “Loyalty Strategies and Military Defection in Rebellion,” Comparative Politics, Vol. 42, No. 3 (April 2010), pp. 333–350; and Philip Roessler, “The Enemy Within: Personal Rule, Coups, and Civil War in Africa,” World Politics, Vol. 63, No. 2 (April 2011), pp. 300–346. 23. Stathis N. Kalvyas and Matthew Adam Kocher, “Ethnic Cleavages and Irregular War: Iraq and Vietnam,” Politics and Society, Vol. 35, No. 2 (June 2007), pp. 183–223, at p. 186. 24. Kalyvas, “Ethnic Defection in Civil War,” p. 1045. 25. James C. Scott, “Revolution in the Revolution: Peasants and Commissars,” Theory and Society, Vol. 7, Nos. 1–2 (January/March 1979), pp. 97–134; and Kai M. Thaler, “Ideology and Violence in Civil Wars: Theory and Evidence from Mozambique and Angola,” Civil Wars, Vol. 14, No. 4 (December 2012), pp. 546–567. 26. Stephen John Stedman, “Spoiler Problems in Peace Processes,” International Security, Vol. 22, No. 2 (Fall 1997), pp. 5–53; Andrew Kydd and Barbara F. Walter, “Sabotaging the Peace: The Politics of Extremist Violence,” International Organization, Vol. 56, No. 2 (Spring 2002), pp. 263–296; and Desirée Nilsson and Mimmi Söderberg Kovacs, “Revisiting an Elusive Concept: A Review of the Debate on Spoilers in Peace Processes,” International Studies Review, Vol. 13, No. 4 (December 2011), pp. 606–626. 27. Kathleen Gallagher Cunningham, “Divide and Conquer or Divide and Concede: How Do States Respond to Internally Divided Separatists?” American Political Science Review, Vol. 105, No. 2 (May 2011), pp. 275–297. 28. Maria J. Stephan and Erica Chenoweth, “Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict,” International Security, Vol. 33, No. 1 (Summer 2008), pp. 7–44. 29. Stathis N. Kalyvas, The Logic of Violence in Civil War (Cambridge: Cambridge University Press, 2006); and Ana M. Arjona and Stathis N. Kalyvas, “Recruitment into Armed Groups in Colombia: A Survey of Demobilized Fighters,” in Yvan Guichaoua, ed., Understanding Collective Political Violence: Conflict, Inequality, and Ethnicity (New York: Palgrave Macmillan, 2011), pp. 143–171. 30. See Jason Lyall, “Are Coethnics More Effective Counterinsurgents? Evidence from the Second Chechen War,” American Political Science Review, Vol. 104, No. 1 (February 2010), pp. 1–20. 31. Kalyvas, The Logic of Violence in Civil War, p. 124. 32. Although Kalyvas's theory of ethnic defection focuses on individual recruitment into organizations, my hypothesis is consistent with the literature's wider focus on the role of alliances between the central state and local elites. 33. Staniland, “Between a Rock and a Hard Place,” p. 24. 34. Christia, Alliance Formation in Civil Wars, p. 32. 35. Ibid., pp. 35, 49. 36. Ibid., pp. 127, 138–140. 37. Another implication of relative power theory is that in the wake of fractionalization, a “splinter faction may join up with an opposing group or it may strike out on its own.” See ibid., p. 53. The prediction here is indeterminate concerning the likelihood of side switching in the wake of a split, however. 38. Charles Tilly, The Politics of Collective Violence (New York: Cambridge University Press, 2003), p. 17. 39. Ibid., p. 134. 40. Relying on violence to indicate rivalry ignores deterrence dynamics and truncates the data from below by missing the emergence of rivalry prior to its manifestation in violence. See Gary Goertz and Paul F. Diehl, “Enduring Rivalries: Theoretical Constructs and Empirical Patterns,” International Studies Quarterly, Vol. 37, No. 2 (June 1993), pp. 147–171; and Karl R. Derouen Jr. and Jacob Bercovitch, “Enduring Internal Rivalries: A New Framework for the Study of Civil War,” Journal of Peace Research, Vol. 45, No. 1 (January 2008), pp. 55–74. 41. William Reno, “Patronage Politics and the Behavior of Armed Groups,” Civil Wars, Vol. 9, No. 4 (December 2007), pp. 324–342; and Christopher R. Day and William S. Reno, “In Harm's Way: African Counter-Insurgency and Patronage Politics,” Civil Wars, Vol. 16, No. 2 (June 2014), pp. 105–126. 42. Rogers Brubaker, Ethnicity without Groups (Cambridge, Mass.: Harvard University Press, 2006). 43. Kalyvas, “Ethnic Defection in Civil War,” p. 1051. 44. Kalyvas and Kocher, “Ethnic Cleavages and Irregular War.” 45. Staniland, “Between a Rock and a Hard Place,” p. 37. 46. Evan S. Lieberman, “Nested Analysis as a Mixed-Method Strategy for Comparative Analysis,” American Political Science Review, Vol. 99, No. 3 (August 2005), pp. 435–452. 47. Matthew LeRiche and Matthew Arnold, South Sudan: From Revolution to Independence (Oxford: Oxford University Press, 2012), p. 4. 48. Jeffrey Herbst, States and Power in Africa: Comparative Lessons in Authority and Control (Ithaca, N.Y.: Cornell University Press, 2000). Following Herbst's argument about the relationship between road building and the consolidation of authority, South Sudan had just 110 kilometers of paved road at independence in 2011. 49. Håvard Hegre and Nicolas Sambanis, “Sensitivity Analysis of Empirical Results on Civil War Onset,” Journal of Conflict Resolution, Vol. 50, No. 4 (August 2006), pp. 508–535; and Lars-Erik Cederman, Nils B. Weidmann, and Kristian Skrede Gledistch, “Horizontal Inequalities and Ethnonationalist Civil War: A Global Comparison,” American Political Science Review, Vol. 105, No. 3 (August 2011), pp. 478–495. 50. Sharif Harir and Mohammed Salih, “Tribal Militias: The Genesis of National Disintegration,” in Terje Tvedt and Sharif Harir, eds., Shortcut to Decay: The Case of Sudan (Uppsala: Nordiska Afrikainstitutet, 1994), pp. 186–203. See also Mahmood Mamdani, Citizen and Subject: Contemporary Africa and the Legacy of Late Colonialism (Princeton, N.J.: Princeton University Press, 1996); and Karen Barkey, Bandits and Bureaucrats: The Ottoman Route to State Centralization (Ithaca, N.Y.: Cornell University Press, 1996). 51. Alex de Waal, “Counterinsurgency on the Cheap,” London Review of Books, August 5, 2004, pp. 25–27. 52. LeRiche and Arnold, South Sudan, p. 67. 53. Julie Flint and Alex de Waal, Darfur: A New History of a Long War (London: Zed, 2008), pp. 151–152. 54. Ken Menkhaus, “Governance without Government in Somalia: Spoilers, State Building, and the Politics of Coping,” International Security, Vol. 31, No. 3 (Winter 2006/07), pp. 74–106. 55. The data exclude the institutionalized branches of the Sudanese Armed Forces, police forces, and the Sudanese People's Liberation Army after its transformation into the army of the autonomous Government of South Sudan after 2005, but they do include the splinter factions emerging from them. 56. Kalyvas uses a five-point scale to measure territorial control. I collapse this into a dichotomous measure of whether a given faction operates in an area where a stronger actor could compel collaboration (i.e., Kalyvas's zones 4 and 5, secure but incomplete or full control of the other side, respectively, coded as 1) from areas where they cannot (i.e., Kalyvas's zones 1 to 3). 57. Julie Flint, “Beyond ‘Janjaweed': Understanding the Militias of Darfur,” Small Arms Survey Working Paper No. 17 (Geneva: Small Arms Survey, June 2009). 58. Territorial control, Christia's proxy for relative power, performs poorly when applied to southern Sudan and Darfur. Compared with the conflicts in Afghanistan and Bosnia, Sudan's wars involve weaker territorial control, more fluid front lines, and in Darfur, more mobile combatants. Moreover, both wars were largely regional struggles, meaning that Sudan's government could abjure control over most of both regions yet still remain relatively more powerful than the rebels. 59. These rivalries may involve violence, as per the fratricidal flipping hypothesis, but the two variables seldom covary The correlation between the fratricidal flipping and political rivalries variables is weak but significant (r=0.13, p < 0.01; see table 2). 60. Small Arms Survey, “Reaching for the Gun: Arms Flows and Holdings in South Sudan,” Human Security Baseline Assessment No. 19 (Geneva: Small Arms Survey, April 2012). 61. Flint, “Beyond ‘Janjaweed,'” p. 38. 62. In the four instances in which factions switched sides twice within one year, there are two observations of factional alignments for that year. 63. The Afghan data focus on commanders, whereas the Sudan data in this article focus on factions. Although factions of different size in Sudan operate at the commander, subgroup, and group levels, Christia notes the close overlap of commander alignments with those at the subgroup and group levels, suggesting a basis for comparison. See Christia, Alliance Formation in Civil Wars, pp. 127, 138. 64. Ibid., pp. 137–139. 65. Ibid., p. 140. 66. Ibid., p. 136. 67. Given the absence of prior assumptions about survival times (i.e., the baseline hazard of side switching), the Cox model with nonparametric duration makes more sense than parametric alternatives. I tested the Cox model's proportional hazards assumption graphically, plotting Schoenfeld residuals against time. The residuals exhibit an unsystematic pattern at each failure time, indicating that the covariate effect does not change with time and the proportional hazards assumption fits. See Janet M. Box-Steffensmeier and Bradford S. Jones, Event History Modeling: A Guide for Social Scientists (Cambridge: Cambridge University Press, 2004). 68. Models 6 and 7 (in appendix 1) give hazard ratios for the two robustly significant variables, political rivalry and patronage incentives. 69. Small Arms Survey, “Sudan Liberation Army Abdul Wahid (SLA-AW),” Human Security Baseline Assessment for Sudan and South Sudan (Geneva: Small Arms Survey, September 6, 2011), www.smallarmssurveysudan.org/fileadmin/docs/facts-figures/sudan/darfur/armed-groups/opposition/HSBA-Armed-Groups-SLA-AW.pdf. 70. Douglas H. Johnson, The Root Causes of Sudan's Civil Wars (Bloomington: Indiana University Press, 2003), p. 86; and Mareike Schomerus, “Violent Legacies: Insecurity in Sudan's Central and Eastern Equatoria,” Small Arms Survey Working Paper No. 13 (Geneva: Small Arms Survey, June 2008). 71. On authority in South Sudan, see Øystein H. Rolandsen, Guerrilla Government: Political Changes in Southern Sudan during the 1990s (Uppsala: Nordiska Afrikainstitutet, 2005); Adam Branch and Zachariah Cherian Mampilly, “Winning the War, but Losing the Peace? The Dilemma of SPLM/A Civil Administration,” Journal of Modern African Studies, Vol. 43, No. 1 (March 2005), pp. 1–20; and Christopher Vaughan, Mareike Schomerus, and Lotje de Vries, eds., The Borderlands of South Sudan: Authority and Identity in Contemporary and Historical Perspectives (London: Palgrave, 2013). 72. There were no interaction effects between other theoretically plausible combinations of variables, however. 73. Christia, Alliance Formation in Civil Wars, pp. 125–144. 74. See, for example, Gérard Prunier, “Rebel Movements and Proxy Warfare: Uganda, Sudan, and the Congo (1986–99),” African Affairs, July 2004, pp. 359–383; and Lee J.M. Seymour, “The Regional Politics of the Darfur Crisis,” in David R. Black and Paul D. Williams, eds., The International Politics of Mass Atrocities: The Case of Darfur (New York: Routledge, 2010), pp. 49–67. 75. Robert O. Collins, A History of Modern Sudan (Cambridge: Cambridge University Press, 2008), p. 204. 76. Johnson, The Root Causes of Sudan's Civil Wars, p. 99; and Peter Nyaba, Politics of Liberation in South Sudan (Kampala: Fountain, 1997). 77. Johnson, The Root Causes of Sudan's Civil Wars, p. 80. A participant in the events cited the “sheer opportunism” of those who supported the coup because they had not been promoted. See Nyaba, Politics of Liberation, p. 90. 78. Johnson, The Root Causes of Sudan's Civil Wars, p. 113. 79. Interviews with humanitarian agencies and SPLM/A officials, Nairobi and Lokichokkio, August–September, 2002; and interviews with commanders and administrators in Western Upper Nile, South Sudan, August–September, 2002. 80. Sharon E. Hutchinson, “A Curse from God? Religious and Political Dimensions of the Rise of Ethnic Violence in South Sudan,” Journal of Modern African Studies, Vol. 39, No. 2 (June 2001), p. 319. 81. Interviews, Western Upper Nile, South Sudan, August 2002. 82. See “Why John Garang Must Go Now” and the “Nasir Declaration” (announcing the coup) in Lam Akol, SPLM/SPLA: The Nasir Declaration (Lincoln: iUniverse, 2003), pp. 306–313. 83. Human Rights Watch Africa, Civilian Devastation: Abuses by All Parties in the War in Southern Sudan (New York: Human Rights Watch, 1994). 84. LeRiche and Arnold, South Sudan, p. 79. 85. Several factions were split by disputes over realignment. Gordon Kong, for example, was left with fewer than 200 fighters after refusing to reintegrate, but had the ability to call on hundreds of demobilized civilians with residual loyalties. See John Young, The Fate of Sudan: The Origins and Consequences of a Flawed Peace Process (London: Zed, 2012), pp. 305–325. 86. LeRiche and Arnold, South Sudan, p. 161. Many collaborating commanders were rewarded with promotions unjustified by their credentials to senior ranks of the Sudanese Armed Forces. 87. Pecuniary interests and career prospects were central to bargaining during negotiations leading to the reintegration of many commanders, including Riek Machar, and the peace agreement. Interviews, Nairobi, February 2004. 88. Matthew B. Arnold, “The South Sudan Defence Force: Patriots, Collaborators, or Spoilers?” Journal of Modern African Studies, Vol. 45, No. 4 (December 2007), pp. 501–507. 89. John Young, “The South Sudan Defense Forces in the Wake of the Juba Declaration,” Small Arms Survey Working Paper No. 1 (Geneva: Small Arms Survey, November 2006). When Gordon belatedly signed a peace deal with the government of South Sudan in April 2013 as part of a coalition of Khartoum-backed holdout factions, he and other leaders had their property and cars confiscated by Sudan's government. See “Ex-South Sudan Rebels Claim Khartoum Arrested Its Members, Confiscated Properties,” Sudan Tribune, May 20, 2013. 90. International Crisis Group, “Darfur: Revitalizing the Peace Process,” Africa Report No. 125 (Brussels: International Crisis Group, April 30, 2007), p. 9. 91. Flint and de Waal, Darfur, p. 201. 92. Small Arms Survey, “Sudan Liberation Army–Minni Minawi” (Geneva: Small Arms Survey, September 6, 2011). 93. Abdul-Jabbar Fadul and Victor Tanner, “Darfur after Abuja: AView from the Ground,” in Alex de Waal, ed., War in Darfur and the Search for Peace (Cambridge, Mass.: Harvard University Press, 2007), p. 289. 94. See Patrick B. Johnston, “Negotiated Settlements and Government Strategy in Civil Wars: Evidence from Darfur,” Civil Wars, Vol. 9, No. 4 (December 2007), pp. 359–377. 95. Alex de Waal, a participant in the peace mediation process, notes that Abdel Wahid demanded a personal payoff of as much as5 million as a “signing fee.” See de Waal, “Dollarised,” London Review of Books, June 24, 2010, pp. 40–41.

96.

Marielle Debos, “Fluid Loyalties in a Regional Crisis: Chadian ‘Ex-Liberators’ in the Central African Republic,” African Affairs, April 2008, p. 236. See also Marielle Debos, Le métier des arms au Tchad: Le gouvernement de l'entre-guerres [Soldiering in Chad: The interwar government] (Paris: Karthala, 2012).

97.

International Crisis Group, “Mali: Security, Dialogue, and Meaningful Reform,” Africa Report No. 201 (Brussels: International Crisis Group, April 11, 2013).

98.

Roland Marchal, “Briefing: Military (Mis)adventures in Mali,” African Affairs, Vol. 112, No. 448 (July 2013), pp. 486–497.

99.

International Crisis Group, “Mali: Reform or Relapse,” Africa Report No. 210 (Brussels: International Crisis Group, January 2014).

100.

Matt Bryden, Somalia Redux? Assessing the New Somalia Federal Government (Washington, D.C.: Center for Strategic and International Studies, 2013).

101.

Séverine Autesserre, “Hobbes and the Congo: Frames, Local Violence, and International Intervention,” International Organization, Vol. 63, No. 2 (April 2009), p. 259; and Henning Tamm, “UPC in Ituri: The External Militarization of Local Politics in North-Eastern Congo” (London: Rift Valley Institute, 2013).

102.

Akbar Ahmed, The Thistle and the Drone: How America's War on Terror Became a Global War on Tribal Islam (Washington. D.C.: Brookings Institution Press, 2013).

103.

Giustozzi, “The Taliban beyond the Pashtuns,” p. 8.

104.

Biddle, Friedman, and Shapiro, “Testing the Surge”; John A. McCary, “The Anbar Awakening: An Alliance of Incentives,” Washington Quarterly, Vol. 32, No. 1 (Winter 2009), pp. 43–59; and Timothy Williams and Duraid Adnan, “Sunnis in Iraq Allied with U.S. Rejoin Rebels,” New York Times, October 16, 2010.

105.

Fotini Christia, “Following the Money: Muslim versus Muslim in Bosnia's Civil War,” Comparative Politics, Vol. 40, No. 4 (July 2008), pp. 461–480.

106.

Christia, Alliance Formation in Civil Wars, pp. 140–145.

107.

John Russell, “Kadyrov's Chechnya: Template, Test, or Trouble for Russia's Regional Policy?” Europe-Asia Studies, Vol. 63, No. 3 (May 2011), pp. 509–528.

108.

U.S. State Department, “Iftar with the Janjaweed” (Washington, D.C.: U.S. State Department, September 25, 2008), http://www.cablegatesearch.net/cable.php?id=08KHARTOUM1450.

109.

Staniland, “Between a Rock and a Hard Place,” p. 37; Kalyvas, “Ethnic Defection in Civil War”; and Kalyvas and Kocher, “Ethnic Cleavages and Irregular War.”

110.

Francis M. Deng, War of Visions: Conflict of Identities in Sudan (Washington, D.C: Brookings Institution Press, 1995), p. 4.

111.

Cederman, Gleditsch, and Buhaug, Inequality, Grievances, and Civil War; and Roessler, “The Enemy Within.”

112.

Luka Biong Deng, “Social Capital and Civil War: The Dinka Communities in Sudan's Civil War,” African Affairs, April 2010, pp. 231–250.

113.

William Reno, Warfare in Independent Africa (Cambridge: Cambridge University Press, 2011), p. 246.

114.

Rogers Brubaker and Frederick Cooper, “Beyond ‘Identity,'” Theory and Society, Vol. 29 (2000), pp. 1–47; and Sharon E. Hutchinson, Nuer Dilemmas: Coping with Money, War, and the State (Berkeley: University of California Press, 1995).

115.

Fotini Christia and Michael Semple, “Flipping the Taliban: How to Win in Afghanistan,” Foreign Affairs, Vol. 88, No. 24 (July/August 2009), pp. 34–45.

116.

Pierre Englebert and Denis M. Tull, “Postconflict Reconstruction in Africa: Flawed Ideas about Failed States,” International Security, Vol. 32, No. 4 (Spring 2008), pp. 106–139; Alex de Waal, “Mission without End? Peacekeeping in the African Political Marketplace,” International Affairs, Vol. 85, No. 1 (January 2009), pp. 99–113; and Séverine Autesserre, The Trouble with the Congo: Local Violence and the Failure of International Peacebuilding (Cambridge: Cambridge University Press, 2010).

117.

Elisabeth Jean Wood, “The Social Processes of Civil War: The Wartime Transformation of Social Networks,” Annual Review of Political Science, Vol. 11 (2008), pp. 539–561.

118.

Helen Young et al., “Livelihoods, Power, and Choice: The Vulnerability of the Northern Rizaygat,” Tufts University, 2008.

119.

Jérôme Tubiana, “Darfur: A Conflict for Land?” in de Waal, War in Darfur and the Search for Peace, pp. 68–91.

120.

Sarah Elizabeth Parkinson, “Organizing Rebellion: Rethinking High-Risk Mobilization and Social Networks in War,” American Political Science Review, Vol. 107, No. 3 (August 2013), pp. 418–432; and Paul Staniland, Networks of Rebellion: Explaining Insurgent Cohesion and Collapse (Ithaca, N.Y.: Cornell University Press, 2014).