Abstract

Why do rebel groups often fight each other when confronting a common, and typically stronger, enemy—namely, the government? Inter-rebel aggression is a calculated response by rebel groups to opportunities for expansion and to the threats confronting such groups. In particular, inter-rebel wars occur under the following conditions: (1) when a window of opportunity emerges allowing rebel groups to defeat coethnic rivals at low costs and thus achieve rebel hegemony, or (2) when rebel groups are confronted with a window of vulnerability, in which they experience a radical deterioration of power relative to that of coethnic rivals and then attempt to escape in a desperate military gamble. Both strict military considerations and ethnic identities shape the rebel groups' calculus. Coethnicity influences the threat perception of rebel groups and their ability to grow at the expense of rivals, thus providing both defensive and offensive motives for inter-rebel aggression. Because coethnic rebel groups want to mobilize and control the same ethnic communities, they are highly sensitive to immediate, intense conflicts of interest and see other coethnic rebel groups as serious threats. Moreover, coethnic rebel groups can generally expect to absorb the resources of defeated rivals, which in turn may improve their chances in the fight against the government. In-depth case studies of multiparty insurgencies in Ethiopia's Eritrea and Tigray provinces, based on interviews with former rebel leaders, strongly support the window of opportunity and vulnerability theory of inter-rebel war.

Introduction

In early 2014, the media were abuzz with reports of battles between Sunni rebels taking part in Syria's civil war. Pundits were quick to note that the Alawite-dominated regime of Bashar al-Assad would be the ultimate beneficiary, observing that “infighting is a boon to regime forces and clearly detrimental to the overall strength of the Syrian opposition.”1 Subsequent developments confirmed this initial assessment, as the government made major gains in the following months of inter-rebel war.2 Analyses of the causes of the infighting, however, were not as insightful as the predictions of its consequences. The initial narrative of moderate rebels fighting against the ruthless organization known as the Islamic State of Iraq and Syria (ISIS) became untenable, as evidence emerged that the anti-ISIS coalition included other jihadist organizations.3 Why would rebel groups sharing ethnosectarian identities and a religious ideology divert resources from their common struggle against a powerful foe to fight each other?

Contemporary history is replete with instances of inter-rebel fighting. For example, during the Algerian independence struggle against the French, the National Liberation Front crushed its competitor, the Algerian National Movement. More recently, the Tamil Tigers wiped out rivals fighting against the Sri Lankan government, and Kashmiri insurgent groups clashed while battling India's security forces.4 The struggle between al-Qaida in Iraq and other rebel groups in Anbar Province in the mid-2000s is perhaps the episode of inter-rebel violence that has attracted most attention, given the involvement of U.S. forces in the war and the subsequent reduction of insurgent activity in Iraq.5

Inter-rebel war is puzzling from several theoretical perspectives. Strategic logic should dissuade rebels from diverting scarce resources from the struggle against their common enemy. In particular, balance of power thinking should push rebel groups to ally against the government, as it is usually the strongest belligerent.6 These instrumental considerations should be reinforced by the enhanced social cohesion in the rebel camp created by the shared experience of violence at the hands of government forces.7 When civil wars occur along ethnic lines, this effect should be especially strong, as ethnic violence “hardens” identities and deepens inter-ethnic hostility and fears.8

This article presents a theory of the onset of inter-rebel war—purposeful, large-scale combat between distinct rebel organizations—and conducts an initial test with case studies of insurgencies in Ethiopia's Eritrea and Tigray Provinces in 1972–91.9 This conceptualization distinguishes inter-rebel war from low-level clashes/skirmishes between rebel groups, which often occur at the initiative of foot soldiers and junior commanders rather than the leadership, and which are pervasive in civil wars.10

I argue that inter-rebel war is a calculated response by insurgent groups to opportunities for expansion and threats emanating from their environment. In particular, inter-rebel war tends to occur when rebel organizations face “windows of opportunity” or “windows of vulnerability.” Windows of opportunity are situations in which a rebel group is markedly more powerful than its coethnic rival(s) and in which the government (even if much stronger in terms of overall military power) does not pose an immediate and serious threat. Under these circumstances, the stronger rebel organization will be tempted to launch a “hegemonic bid”—that is, the use of overwhelming force to get rid of a coethnic rival (or rivals). Windows of vulnerability occur when a rebel group that is relatively weak and/or confronts a significant government threat faces the prospect of a dramatic decline in power relative to a coethnic rival. If no other option appears viable, the organization will be tempted to resort to force against the rival in a desperate attempt to overcome its predicament.11 I call this typology of inter-rebel war “gambling for resurrection.”12

By contrast, I argue that explanations of inter-rebel war focusing only on the balance of power or ethnic/ideological compatibility are at best incomplete. In deciding when and whom to fight, rebel groups do not follow simple balancing logic or lash out against groups with different worldviews and goals.

I consider rebel groups coethnic if their announced political aspirations directly relate to the same ethnic group's fate and if the overwhelming majority of their rank and file or the leadership belongs to the same ethnic group (i.e., both ethnic goals and membership criteria need to be met).13 Coethnicity is a key element of both types of windows, as it affects rebel groups' threat perception and ability to grow at the expense of rivals. Coethnic rebel organizations have overlapping ambitions with regard to their ethnic group—they want to mobilize and control the same communities. Thus, they are highly aware of their immediate, intense conflict of interest and see each other as a serious threat. Moreover, coethnic rebel groups can generally expect to absorb the resources of defeated rivals, which in turn would improve their chances in the fight against the government.

When sub-ethnic differences are present, I consider rebel groups coethnic if (besides the basic requirement of the overwhelming majority of their rank and file or leadership belonging to the same ethnic group) their political goals relate to the fate of the ethnic group as a whole and they recruit (or are willing to do so) across sub-ethnic cleavages.14

Window theory challenges two influential views on the effects of ethnic identity on civil war dynamics. One view envisions coethnicity as a powerful source of solidarity and cooperation. In particular, the ethnic security dilemma literature expects individuals to flock automatically to their ethnic side under conditions of state collapse or large-scale violence, and treats ethnic groups, rather than armed organizations, as the relevant units of analysis.15 The other view contends that coethnicity does not have an important effect on civil war behavior, but rather tends to be a façade for more mundane, often purely materialistic or individual survival–maximizing considerations.16 My contention that coethnicity affects the risk of inter-rebel violence by increasing the net benefits of infighting challenges the view of ethnicity as largely epiphenomenal in civil war dynamics. At the same time, it qualifies the earlier view emphasizing the causal role of ethnicity by pointing out that, in some circumstances, shared identity may cause intra-ethnic conflict rather than cooperation.

As examining the validity of window theory requires fine-grained measures of the inter-rebel balance of power and government threat, I present comparative, longitudinal case studies of insurgencies in Ethiopia's Eritrea and Tigray Provinces. Marshaling primary and secondary sources, including my interviews with former rebel leaders, I assess whether variation in the presence of windows of opportunity and windows of vulnerability matches my theoretical expectations and trace rebel decisionmaking processes, while controlling for possible confounders.

Unraveling the puzzle of inter-rebel war not only is important for theoretical reasons; it also has crucial policy implications. When considering whether to intervene in a civil war, policymakers would typically want to have a sense of the probability of inter-rebel fighting and to have tools to induce cooperation or conflict between rebels. A key element of President Barack Obama's strategy against ISIS was trying to promote violent conflict between that organization and other armed groups.17 By contrast, forging an alliance between Muslims and Croats was the lynchpin of the U.S. intervention strategy in Bosnia's civil war. Moreover, inter-rebel violence may lead to insurgent side-switching, as the group being defeated turns to the government for help.18 Insurgent defection not only entails a shift of military resources between opposing sides, but also often provides the government with precious intelligence on the groups that continue to oppose it and thus can have a major impact on counterinsurgency outcomes.19

The article proceeds as follows. The next two sections survey existing explanations for inter-rebel war and introduce my argument, detailing how windows of opportunity and vulnerability prompt insurgents to fight each other. The following two sections present the empirical setting of the article (multiparty ethnic insurgencies in Ethiopia), research design, and case study findings. The final section discusses the argument's contributions to the study of civil wars and policy implications flowing from the analysis.

Existing Explanations for Inter-rebel War

The literature on civil war processes has cast light on several aspects of inter-rebel dynamics, including infighting, civil war alliances, and insurgent side-switching.20 Yet, despite their substantial contributions, these studies do not provide entirely satisfactory answers to the puzzle of inter-rebel war on theoretical or empirical grounds. In the only published study specifically on the topic, Hanne Fjelde and Desirée Nilsson find that insurgent groups are more likely to be involved in inter-rebel fighting when they are either strong or weak relative to other groups; they have exclusive control of some territory; they fight in areas with drug cultivation; and government authority is weak.21 Although important, these are, in a sense, correlations in search of a theory, as the statistical findings are not integrated in an overarching explanation of inter-rebel violence. On the empirical front, the study's exclusive reliance on blunt, largely cross-sectional data raises questions about whether its statistical associations actually reflect causal relations.

Fotini Christia has advanced an elegant theoretical explanation for civil war alliances, a phenomenon closely related to inter-rebel war.22 She argues that alliances follow a minimum winning coalition (MWC) logic in which shared identity plays no role: belligerents strive to be part of the smallest possible alliance sufficient to win the war. When the MWC threshold is passed, one or more warring parties will abandon the dominant coalition in search of an optimally sized one. The implication for inter-rebel war is that rebel groups should fight each other only when one (or a subset) of them is stronger than the government. Many well-known instances of groups fighting each other while facing an unambiguously more powerful government, however, cast serious doubts on the ability of MWC to explain inter-rebel war. For example, in Algeria the National Liberation Front wiped out the Algerian National Movement, despite the overwhelming power of the French army.23 Similarly, the National Front for the Liberation of Angola and the Popular Movement for the Liberation of Angola fought each other, rather than cooperating against the vastly superior Portuguese forces.24 This pattern is not limited to anti-colonial wars. For instance, Kashmir's insurgent groups engaged in inter-rebel war notwithstanding the clear military edge of the Indian government.25

Finally, Peter Krause theorizes about how the distribution of power within self-determination movements affects their prospects of success against the government. Krause supports his argument with evidence from the Algerian and Palestinian national movements, showing that both were most effective when they were hegemonic (i.e., dominated by a single group).26 One of the underlying mechanisms in Krause's theory—that non-hegemonic movements are more likely to experience internecine violence—suggests a prediction about inter-rebel fighting consistent with window theory: inter-rebel war is unlikely once a rebel group has firmly established itself as the hegemon. The theory presented in this article can thus be seen as complementing Krause's argument by further specifying conditions likely associated with rebel infighting and endogenizing the emergence of insurgent hegemons through violent inter-rebel competition.

Window Theory of Inter-rebel War

By definition, rebel groups exist in an anarchic environment—there is no overarching authority enforcing agreements between them, as they have challenged the state's monopoly of violence. Much like states in the world depicted by realism, insurgent organizations have to rely on themselves and cooperation with other self-regarding actors to survive and achieve their goals. This characterization is arguably more fitting for civil war settings than for the international system: in civil wars, the balance of power game is played especially hard and participants frequently “fall by the wayside.”27 By contrast, “state death” in the international system is a rare occurrence.28 As noted, balance of power logic would lead one to expect rebel groups to be natural allies, given that they typically face a stronger common enemy. So why does inter-rebel war occur?

My answer stems from two observations. First, ideological differences, disagreements over strategy, different priorities, and overlapping ambitions frequently create conflicts of interest among rebel groups. Even groups with the same strategic goals (e.g., ending foreign occupation or regime change) can have conflicting organizational goals (e.g., increasing group membership and funding).29 While typically having similar strategic goals, coethnic rebel groups tend to have especially immediate and intense conflicts of interest about organizational goals: they aspire to control the same population and territory and know that they could absorb with relative ease the resources of defeated coethnic rivals, which in turn would improve their chances in the fight against the government. Thus, coethnic rivals may pose obstacles to a group's growth and even represent an existential threat. Second, the government's aggregate military superiority does not automatically translate into an ability or a willingness to launch well-timed, decisive offensives against rebels weakened by infighting. Civil wars therefore may be characterized by phases of low government threat in which inter-rebel war may not be prohibitively costly.

Thus, windows of opportunity to launch a hegemonic bid may open up, allowing an insurgent group whose power has peaked to lock in the benefits of its dominant position by destroying a weaker coethnic rival, without exposing itself to an unacceptably higher risk of government victory.30 In addition, when no other solution appears possible, rebel groups may resort to force to forestall the deterioration of their power relative to a coethnic rival, gambling for resurrection in the face of windows of vulnerability.31

In both cases, when deciding whether to use force, rebel groups weigh the benefits and costs of inter-rebel war.32 In terms of benefits, hegemonic bids may reduce the long-term threat posed by coethnic groups and allow the absorption of resources previously under their control. Gambles for resurrection entail the benefit of potentially reducing the urgent threat posed by rising coethnic rivals. Coethnicity powerfully affects the benefits of inter-rebel war: coethnic rebel groups represent both serious threats and enticing opportunities for expansion for each other, as they aspire to control the same population and territories, and may expect to do so with relative ease once they achieve a hegemonic position in their ethnic community. In terms of costs, the possibility that inter-rebel war might weaken the rebel group(s) looms large. Moreover, in situations in which the government is faring poorly on the battlefield, infighting could entail the major opportunity cost of forgoing rebel victory. The costs of inter-rebel war can be relatively low, however, for a group enjoying marked superiority over its rival(s) and facing a powerful government that is nonetheless unwilling or unable to rapidly intervene to take advantage of infighting.

Windows of opportunity are situations of likely high benefits and low costs of inter-rebel war. By contrast, windows of vulnerability are characterized by both high expected costs and benefits: the prospect, however remote, for a rebel group to forestall its decline relative to a coethnic rival and thus reduce the urgent threat posed by the latter by attacking it may warrant paying the costs and running the risks involved in fighting without a favorable inter-rebel balance of power and/or government threat environment.33 Gambles for resurrection are responses of later resort, as rebel groups typically would go to considerable lengths in search of a less risky way out of their predicament.34

A rebel group would face windows of opportunity and vulnerability at the same time if it expects to fall from its current position of superiority to helplessness vis-à-vis a coethnic rival. A blend of window of opportunity and vulnerability logics would then drive the group toward a “hybrid” inter-rebel attack:35 the group may reasonably hope to achieve a quick and cheap inter-rebel victory, but it would also be spurred into action by the dangers lying ahead and thus accept significantly more risks and costs in its hegemonic bid than would be warranted in the absence of a window of vulnerability.

Below I discuss in more detail the factors shaping the cost-benefit calculus of rebel groups—coethnicity, government threat, and the inter-rebel balance of power.36

COETHNICITY, CUMULATIVITY, AND BENEFITS OF INTER-REBEL WAR

Coethnicity feeds defensive and aggressive motives for inter-rebel war, as it holds the promise of both an improved threat environment and expanded resources.

As a result of overlapping ambitions to control the same ethnic community, coethnic rebel groups tend to see each other as direct competitors, posing obstacles to their goals and even representing existential threats. Coethnic organizations might interfere with their respective mobilization efforts and derail the struggle against the government by spoiling ongoing negotiations, reducing the clarity of demands and the credibility of threats to the incumbent,37 and even switching to the government's side. A rebel group could therefore use force to reduce the threat posed by a coethnic rival. Refraining from doing so at an opportune moment means not only forgoing a chance to address the problems just discussed, but also exposing the group to future risk of attack by the rival or of having to make unpalatable concessions.

The high level of threat perception that characterizes relations between coethnic rebel groups is compounded by the “cumulativity” of their resources (i.e., resources can be extracted at low cost and their possession allows a group to protect itself and expand its power).38 When a rebel organization wipes out a coethnic group, it can often coopt large segments of the latter's membership. More important, the group that wins the inter-rebel fight is likely to be able to recruit and extract resources at a relatively low cost from the coethnic population previously under its rival's control.39 Achieving rebel hegemony through a quick and cheap fight under these circumstances holds the prospect of strengthening the victor for its ongoing fight against the government.

By contrast, a group perceived by the social base of support of a defeated rival as a “foreign occupier” is likely to experience difficulties with recruitment and extraction and will have to devote substantial resources to policing the population, to the point of experiencing a net drain of its resources.40 For this reason, the existence of profound, long-term conflicts of interest between non-coethnic rebels groups (for example, about the distribution of power among a country's ethnic groups) does not tend to prompt inter-rebel war: their resolution is better left to the postwar phase, as infighting would only benefit the government.

The foregoing discussion raises the question: Why are multiple rebel organizations associated with the same ethnic group? This phenomenon is quite common, as rebel groups are typically built on preexisting social networks, such as political parties, religious organizations, kinship ties, and veteran associations, which are smaller than ethnic communities.41 Sometimes rebel groups initially coalesce around influential individuals or along sub-ethnic cleavages; other times, the existence of multiple groups is the result of processes of organizational fragmentation along similar divides.

Individuals' propensity to join and support coethnic rebel organizations can be traced to ethnic parochialism—the well-documented tendency for people to be particularly cooperative with and favor members of their group—and a vast amount of evidence from the developing world, where much political and economic competition occurs along ethnic lines.42 As a result, in ethnically heterogeneous clientelist democracies and electoral autocracies, elections are often ethnic censuses, as ethnicity provides a remarkably precise predictor of voting behavior;43 and patronage is distributed along ethnic lines.44 Moreover, developing countries' rulers often resort to “ethnic stacking” (i.e., filling key government and military positions with coethnics) to ensure loyalty.45

While these empirical patterns are widely acknowledged, there is much debate over their causes. The dominant view in political science holds that individuals cooperate with coethnics because of the prospect of material gain. In particular, agreements between coethnics are more easily enforced given the latter's especially dense social networks facilitating information flows.46 Moreover, given their relative visibility and stickiness, ethnic markers can be used by individuals as quick and crude information shortcuts to gauge the likely behavior of their counterparts; if people come to believe that coethnics tend to cooperate with each other, ethnic favoritism becomes a self-fulfilling prophecy.47 A second view, rooted in social identity and political socialization theories,48 posits that individuals have genuine, emotionally laden attachments to their ethnic group. These attachments prompt them to support organizations representing the group's interests even at high personal costs, in particular during large-scale ethnic violence.49

Window theory is agnostic about the relative importance of the two families of mechanisms described above: both are likely to make individuals more willing to join and cooperate with coethnic rebel organizations, and thus contribute to cause inter-rebel war. After rebel group A's defeat of coethnic rival B, the latter's supporters would tend to be drawn to group A, with the expectation that it would be a bulwark against a government controlled by ethnic “others” threatening the security, well-being, and dignity of their ethnic community. Moreover, rebel group A's members, because of a shared culture with coethnic rebel B, would be able to tap into ethnic networks, making threats of punishment and promises of rewards credible, which in turn would facilitate control of the population previously under rebel group B's sway.50

BALANCE OF POWER, GOVERNMENT THREAT, AND COSTS OF INTER-REBEL WAR

Inter-rebel war also has its costs. Besides the direct expenditure of scarce resources, these consist of increased vulnerability to the incumbent and forgone opportunities to make gains against it. In particular, the government could take advantage of the situation and attack the squabbling rebels. Moreover, inter-rebel fighting could push the group that is losing to turn to the government for help, strengthening the counterinsurgency effort. In addition, inter-rebel war may entail major opportunity costs if it diverts resources from the anti-government struggle when a concerted insurgent effort could bring about important territorial gains or even victory, thus offering a chance for the incumbent to regroup.

The costs of inter-rebel war may be low, however, when three conditions are jointly present: (1) there is an imbalance of power between rebel organizations, (2) the government does not represent a serious and immediate military threat, but (3) it is not so weak as to suggest major opportunity costs. In the face of these low costs, the powerful group may be tempted to wipe out its rival(s) to become the rebel hegemon.51

The government is typically the strongest belligerent, but this superiority does not mean a constant willingness and ability to unrestrainedly bring to bear its military power. Political, military, and logistical constraints may limit government power projection, so that it may not represent a serious and immediate threat to the rebels. For example, for a long time the prospect of high domestic political costs deterred the Indian government from launching large-scale offensives against the United Liberation Front of Asom and the Naxalite insurgents.52 In some instances, the level of government threat faced by an insurgent movement is largely exogenous, as when security forces are busy dealing with a foreign enemy or paralyzed by internecine struggle. In others, the government may consciously adopt a low-threat stance, as putting maximum pressure on the rebels might not be a top priority.53

Rebel groups can gauge whether the government poses a serious and immediate threat based on recent and ongoing battlefield trends and troop deployments, as well as intelligence on its plans. The typical scenario of serious and immediate threat would be when the government is engaged in (or about to launch) a major offensive. Even if government forces are not making major territorial headway, an intense fight in which the insurgents are stretched thin to resist the government's onslaught does not represent a permissive environment for inter-rebel fighting: diverting resources to inter-rebel war (even if short and decisive) may pave the way for the government's battlefield success. By contrast, in a context of sporadic, limited clashes, insurgent leaders may see the diversion of resources to inter-rebel fighting as entailing an acceptable short-term increase of vulnerability.

The typical scenario of high opportunity costs is one in which the rebels are poised to capture important territory from the government or victory appears within reach if they sustain maximum pressure against the government. In these situations, inter-rebel war should be deterred by the prospect of forgoing major military gains.

The inter-rebel balance of power affects the prospective costs of inter-rebel war, too. Attacking weaker rebel groups promises a cheap victory. By contrast, groups of comparable strength are likely to put up a serious fight. The ensuing long war of attrition would likely weaken the rebel camp as a whole, thus creating the conditions for a successful government offensive. Therefore, a favorable inter-rebel balance of power is a requirement for launching a hegemonic bid. By contrast, rebel groups gamble for resurrection when facing an unfavorable distribution of power and/or a serious and imminent government threat (i.e., when inter-rebel war is likely to be very costly).

Testing Window Theory

This article tests window theory and alternative explanations for inter-rebel war with a comparative study of relations between rebel groups in Ethiopia's Eritrea and Tigray Provinces. Existing datasets do not provide the fine-grained and contextual measures of rebel power and government threat needed to assess the argument. As coding this information for a large number of cases is an extremely time-consuming task, the in-depth study of a small number of rebel pairs is an appropriate way to conduct an initial test.54 This approach allows me to explore whether the causal processes suggested by window theory are at play in rebel leaders' decisionmaking and to conduct structured comparisons in which many possible confounding variables are held constant.55

Windows shape rebel groups' behavior indirectly through the perceptions and beliefs of their leaders. Thus, the ideal evidence for testing my argument is information on rebel leaders' calculus. Window theory would be strongly supported by evidence (1) that rebel decisionmakers launched an attack as a result of either their perception of an opportunity to cheaply eliminate weaker coethnic rivals or an urgent need to forestall a decline in relative power despite the potential steep costs, and (2) that they refrained from attacking when comparable opportunities and vulnerabilities were absent. I gathered relevant data by conducting extensive interviews with individuals who had taken part in (or were directly informed about) rebel groups' decisionmaking. I address concerns about subjects' biases and defective memories by triangulating information across interviewees who belonged to different organizations (and thus presumably had divergent incentive structures) and with accounts provided in rebel publications and the secondary literature.

Information from the secondary literature is useful not only for cross-checking purposes, but also for coding the existence of windows when data on rebel leaders' assessments of the inter-rebel balance of power and government threat are not available. In the absence of information on rebel leaders' perceptions of the defining features of windows of opportunity and windows of vulnerability, I adopt the following coding criteria to identify them (independently of whether inter-rebel war occurs). I measure the inter-rebel balance of power along four dimensions: (1) number of fighters in each rebel group, (2) their access to weapons, (3) the organizational cohesion of each rebel group (measured as the relative absence of splits and feuds among each group's leaders and pervasive indiscipline among the rank and file), and (4) tactical-operational skills (proxied by previous battlefield experience and reports of battlefield prowess).56 As extant theory does not provide clear indications about the relative importance of these aspects of rebel power, I consider a rebel group stronger if it starkly outranks rivals in a net number of dimensions of power for which I have information.57 I code a sharp shift in rebel relative power—a key element of windows of vulnerability—if there is evidence of a marked trend in flows of recruits or weapons or an impending international intervention in favor of specific rebel groups.

In cases in which rebellion takes the form of guerrilla warfare, I code a low level of government threat if insurgents typically initiate contacts with security forces (by launching hit-and-run attacks), which implies that the rebels can choose when and where to fight and thus control the pace of their losses. In conventional warfare, a permissive threat environment typically would be characterized by static battle lines, which, based on previous interactions, the insurgents know they can comfortably defend in the absence of a significant escalation of government attacks.

Finally, I code high opportunity costs of infighting if (1) the rebels have just made substantial territorial gains or conquered several towns (suggesting a clear favorable battlefield trend, which the insurgents should be wary to jeopardize), or (2) they are engaged in an offensive to take a major town or an area on an international border (suggesting the prospect of making strategic gains against the incumbent).

Insurgencies in Eritrea and Tigray

Northern Ethiopia is an ideal setting for a preliminary test of window theory, as multiparty ethnic rebellions occurred at the same time in two adjacent provinces. That Eritrean and Tigrayan rebels had contacts—and thus opportunities for violent conflict—within and across ethnic lines allows me to assess whether interactions between coethnic rebel groups differ from those between non-coethnics. Moreover, the insurgencies in Eritrea and Tigray display substantial variation on the dependent variable (inter-rebel war occurred in some instances but not in others, and rebel pairs differed in their propensity for infighting), thus offering multiple opportunities to observe the theory's causal mechanisms.

Eritrean and Tigrayan ethnic identities emerged from complex historical processes, involving Italy's colonization of Eritrea, large-scale intrastate violence, and ethnonational mythmaking.58 Both identities coexist with other nested and overlapping identities. The Eritrean identity, in particular, encompasses sub-ethnic religious (Muslims vs. Christians), linguistic (Tigrigna speakers vs. speakers of other languages), and regional (highlanders vs. lowlanders) cleavages, while many of Eritrea's and Tigray's inhabitants share a Tigrigna-speaking identity. As is common in civil wars, Ethiopia's rebel groups formed around sub-ethnic social networks (along largely overlapping religious, linguistic, and regional lines in Eritrea, and along a primarily regional divide in Tigray). As noted, I consider rebel groups with distinct sub-ethnic bases coethnic if their political goals relate to the fate of the ethnic group as a whole and if they recruit (or are willing to do so) across sub-ethnic cleavages.

Table 1 reports all dyadic interactions (i.e., instances of inter-rebel war or peaceful coexistence) between rebel groups that operated in adjacent areas in Eritrea and Tigray.59 For reasons of space, the following analysis focuses on dyadic relations in 1972–91 involving the two organizations that eventually defeated the Ethiopian government in 1991—the Eritrean People's Liberation Front (EPLF) and the Tigray People's Liberation Front (TPLF)—and other Eritrean and Tigrayan groups active in the same period (i.e., the EPLF, the TPLF, or both are members of all the pairs examined).60 Focusing on these two groups is especially useful, as each at some point cooperated with and later fought against the same coethnic rival; this approach allows me to control for dyad-specific features, thus substantially reducing the number of confounders.

Table 1.

Inter-rebel Relations in Eritrea and Tigray

Observed Outcome (war/coexistence)Type of WindowCorrect Prediction?
War ELF-ELM, 1965 window of opportunity (coethnic, imbalance of power, low government threat, low opportunity cost) ✓ 
War EPLF-ELF, 1972–74 window of opportunity (coethnic imbalance of power, low government threat, low opportunity cost) ✓ 
Coexistence EPLF-ELF, 1975–79 (coethnic but high opportunity cost, 1975–77; high government threat, 1978–79) no window ✓ 
War TPLF-TLF, 1975 (coethnic, imbalance of power, low government threat, low opportunity cost) window of opportunity ✓ 
War TPLF-Teranafit, 1976 (coethnic, eroding imbalance of power, low government threat, low opportunity cost) window of opportunity & vulnerability ✓ 
Coexistence TPLF-ELF, 1975–78 no window (non-coethnic) ✓ 
Coexistence TPLF-EPRP, 1975–77 no window (coethnic but no imbalance of power) ✓ 
Coexistence ELF-Teranafit, 1976 no window (non-coethnic) ✓ 
War TPLF-EDU, 1976–78 no window(non-coethnic) 
Coexistence ELF-EDU, 1976–78 no window (non-coethnic) 
War ELF-ELF-PLF, 1978 window of opportunity (coethnic, imbalance of power, low government threat, low opportunity cost) ✓ 
War TPLF-EPRP, 1978–79 window of opportunity (coethnic, imbalance of power, low government threat, low opportunity cost) ✓ 
War TPLF-ELF, 1979–81 no window (non-coethnic) 
War EPLF-ELF, 1980–81 window of opportunity (coethnic, imbalance of power, low goernment threat, low opportunity cost) ✓ 
Coexistence TPLF-EPLF, 1981–91 no window (non-coethnic) ✓ 
Observed Outcome (war/coexistence)Type of WindowCorrect Prediction?
War ELF-ELM, 1965 window of opportunity (coethnic, imbalance of power, low government threat, low opportunity cost) ✓ 
War EPLF-ELF, 1972–74 window of opportunity (coethnic imbalance of power, low government threat, low opportunity cost) ✓ 
Coexistence EPLF-ELF, 1975–79 (coethnic but high opportunity cost, 1975–77; high government threat, 1978–79) no window ✓ 
War TPLF-TLF, 1975 (coethnic, imbalance of power, low government threat, low opportunity cost) window of opportunity ✓ 
War TPLF-Teranafit, 1976 (coethnic, eroding imbalance of power, low government threat, low opportunity cost) window of opportunity & vulnerability ✓ 
Coexistence TPLF-ELF, 1975–78 no window (non-coethnic) ✓ 
Coexistence TPLF-EPRP, 1975–77 no window (coethnic but no imbalance of power) ✓ 
Coexistence ELF-Teranafit, 1976 no window (non-coethnic) ✓ 
War TPLF-EDU, 1976–78 no window(non-coethnic) 
Coexistence ELF-EDU, 1976–78 no window (non-coethnic) 
War ELF-ELF-PLF, 1978 window of opportunity (coethnic, imbalance of power, low government threat, low opportunity cost) ✓ 
War TPLF-EPRP, 1978–79 window of opportunity (coethnic, imbalance of power, low government threat, low opportunity cost) ✓ 
War TPLF-ELF, 1979–81 no window (non-coethnic) 
War EPLF-ELF, 1980–81 window of opportunity (coethnic, imbalance of power, low goernment threat, low opportunity cost) ✓ 
Coexistence TPLF-EPLF, 1981–91 no window (non-coethnic) ✓ 

NOTE: Each row indicates an episode of inter-rebel war (or its absence) involving rebel groups operating in adjacent areas (dyads whose members could not have fought each other, as they did not operate in adjacent areas or were not active at the same time are not reported). Shaded rows indicate interactions analyzed in the article.

ELF = Eritrean Liberation Front; ELM = Eritrean Liberation Movement; EPLF = Eritrean People's Liberation Front; TPFL = Tigray People's Liberation Front; TLF = Tigray Liberation Front; EPRP = Ethiopian People's Revolutionary Party; EDU = Ethiopian Democratic Union; ELF-PLF = Eritrean Liberation Front-People's Liberation Front

Tracing Trajectories of Inter-rebel War in Ethiopia

In 1961, the Eritrean Liberation Front (ELF) launched its rebellion against the government. The ELF dominated the Eritrean movement until 1971, when three splinter groups emerged from a drawn-out process of organizational implosion. In early 1972, shortly after the splinter groups had announced their decision to merge into the EPLF, the ELF attacked them. The fight continued until 1974, when the fall of Ethiopia's emperor radically altered the political-military landscape. Amid strikes, student protests, and army mutinies, a group of left-leaning officials known as the “Derg” took control of the government. In the following three years, the ELF and the EPLF cooperated in wresting most of Eritrea's territory from government forces, weakened as they were by turmoil in the capital, anti-government rebellions throughout the country, and Somalia's attack in the southeast. ELF-EPLF cooperation continued in the face of a radical reversal of military fortunes in 1978. Having consolidated its hold on power in the capital and benefiting from massive Soviet military support, the Derg repelled the Somali invasion and then launched a major offensive in Eritrea. The ELF and the EPLF lost virtually all of their territory, but managed to jointly defend their positions in the Sahel. After the last unsuccessful government offensive in 1980, the EPLF, in cooperation with the TPLF, attacked and defeated the ELF, thus establishing itself as the rebel hegemon in the Eritrean camp.

In neighboring Tigray, several insurgent groups started operating in 1975–76. The TPLF easily crushed its coethnic rivals, the Tigray Liberation Front (TLF) and the Tigray People's Liberation Movement Coordinating Committee—better known as “Teranafit” (Tigrigna for “coordinating”)—in 1975–76. The subsequent fight against the Ethiopian Democratic Union (EDU)—a pan-Ethiopian, rather than Tigrayan, organization—proved more difficult, but the TPLF prevailed in 1978. The group was now in a position to deal with its last remaining coethnic competitor, the Ethiopian People's Revolutionary Party (EPRP), which it quickly eliminated.

Once the EPLF and the TPLF were firmly in control of their respective ethnic communities, they argued bitterly over the conduct of the struggle against the regime, but refrained from fighting each other. Following a renewal of their alliance, the EPLF and the TPLF inflicted a decisive blow on the government in 1991. The TPLF took control of the central government while the EPLF achieved its goal of Eritrean independence.

Below I show that window theory sheds much light on this complex pattern of inter-rebel relations, whereas alternative explanations fall short. The EPLF and TPLF emerged as hegemons in their respective ethnonational camps from a process of violent selection, through which they wiped out weaker coethnic rivals when the government posed a limited threat. Consistent with my argument, the bulk of the inter-rebel fighting took place within, rather than across, ethnic lines; crucially, despite serious disagreements, the EPLF and the TPLF did not fight each other.

INTRA-ERITREAN RELATIONS: FROM WAR TO COOPERATION AND BACK

Window theory explains the alternation between armed clashes and peaceful coexistence that characterized the relationship between Eritrea's main rebel groups. The ELF and the EPLF were coethnic; both struggled for Eritrean independence, and both had exclusively Eritrean membership.61 In 1972, the ELF launched a hegemonic bid in a moment of relative strength and low government threat. The two, by contrast, cooperated in 1974–79, when the costs of infighting were prohibitively high. Eventually, the EPLF took advantage of a fleeting moment of limited government threat to crush its weaker rival in 1980–81.

the first eplf-elf war. Window of opportunity logic fits the ELF's 1972 attack on the EPLF. The ELF acted in a moment of marked military superiority and limited government threat, with the objective of reasserting its hegemony in the Eritrean liberation struggle.

The stark imbalance of power between the ELF and the three splinter groups that would merge into the EPLF is beyond doubt. As EPLF senior commander Mesfin Hagos noted, “It was a one-sided war at the start. The EPLF was at an infantile stage as an organization … The balance of power was five to one.”62 The secondary literature paints a consistent picture: the fledgling organizations operated in remote areas, had limited access to external supplies, and together could marshal only 500 fighters, compared to 2,000 under the ELF's control.63

It is also clear that the Ethiopian government did not pose a serious and immediate threat to the ELF when the group started the inter-rebel war. In a 2013 interview with the author, Gime Ahmed (then ELF counterintelligence officer) noted that intercepted communications at the time revealed that the government had no intention of escalating its limited military efforts to take advantage of the infighting.64 In fact, during the 1972–74 inter-rebel war, government forces maintained a low profile, as they had done in 1971.65 The last major government offensive had occurred early that year; and in the period June 1971–June 1972, the insurgents initiated the vast majority of clashes.66

Decisionmaking evidence confirms that the ELF tried to eliminate the splinter groups' challenge to its hegemonic position. In late 1971, the ELF passed a resolution authorizing the use of force to “preserve the unity of the revolution” in case the groups refused to come back to the fold, and then tried to enforce it.67

eplf-elf cooperation. Despite its military superiority, the ELF fell short of crushing its competitor, probably because of tactical blunders and the interference of the Sudanese army. By late 1974, a combination of pressure from the Eritrean population to stop the fratricide, the emergence of opportunities for major battlefield gains against the government, and the substantial strengthening of the EPLF had brought the inter-rebel war to an end.68 Window theory correctly predicts the absence of a new burst of inter-rebel fighting in the following five years, given its prohibitively high costs—the opportunity cost of forgoing victory at first, and then the risk of outright defeat.

In 1974, while the central government was in disarray, the leaderships of the EPLF and the EFL saw a unique opportunity to act to achieve independence and understood that infighting could spoil it. ELF leader Ahmed Nasser noted: “We concluded that fighting [with the EPLF] must stop to exploit the new developments; the regime was weak, this was a moment of opportunity to achieve independence.”69 EPLF leader Osman Saleh Sabbe revealed his conviction that an opportunity for the rebel movement had emerged, declaring that the separatist groups, now stronger than ever, would switch from hit-and-run attacks in the countryside to open offensives against army camps and in urban centers.70 These expectations proved largely correct: by 1977, the two groups controlled 90 percent of Eritrea and all but three main urban centers.71

Window theory correctly predicts that rebel groups would not fight each other under these circumstances lest they undermine their rapid military progress. Moreover, the power advantage of the ELF that had prompted its attack in 1972 had all but vanished by late 1974: the EPLF had substantially grown in strength, and ELF leaders had concluded that its rival could not be easily defeated, making inter-rebel aggression too costly.72

The year 1978 saw a drastic reversal of battlefield trends. With massive Soviet aid, the Derg launched a major offensive in Eritrea. To survive, the EPLF and ELF had to withdraw to remote areas in northwestern and northern Eritrea. The insurgents then jointly resisted a rapid succession of four additional large-scale offensives against their main remaining strongholds in the Sahel.73 As window theory would predict, the rebel groups did not fight each other in 1978–79 when dealing with a serious and immediate government threat: inter-rebel war would have been suicidal in the face of “one of the biggest setbacks Eritrean independence fighters endured.”74

the second eplf-elf war. Window theory fits the EPLF's 1980 attack on the ELF, too. The EPLF exploited a fleeting window of opportunity to crush the ELF and become the Eritrean hegemon.

The EPLF enjoyed a clear edge over the ELF. As Awet Weldemichael put it when referring to the first government offensive in 1978 and the EPLF's 1980 attack, “The ELF was an already collapsing edifice, waiting for an Ethiopian onslaught and a push from its domestic rival.”75 The imbalance of power between the two groups was primarily a function of the ELF's lack of internal cohesion, as several ELF interviewees stressed.76 The EPLF leadership was fully aware of this situation. In the words of commander Mesfin Hagos: “The ELF organizationally was very weak … We knew that very well … The ELF did not have a comparable [to the EPLF] organization.”77

Although a marked imbalance of power between the EPLF and the ELF had existed for some time, the EPLF did not face a permissive threat environment until 1980. In late 1979, an EPLF counterattack at the end of the fifth government offensive disrupted preparations for a subsequent operation. The government continued to actively plan for a massive attack in the Sahel, but would not be able to launch operations of comparable intensity in Eritrea until 1982.78 EPLF sources suggest that the leadership understood the constraints preventing the government from immediately launching a new all-out offensive and decided to exploit them to eliminate the ELF.79 In interviews with the author, Mesfin Hagos noted that the attack on the ELF was timed so that inter-rebel fighting would be over by the time the government threat intensified once again: “We needed to kick the ELF out before the new offensive. We were following the propaganda, the military mobilization, troop movements. We knew we would not be able to sustain a war on two fronts. … The Ethiopians were saying: ‘The next offensive will be decisive.’ So, they were trying to organize a massive offensive, but their capacity had been weakened in the fifth offensive in July 1979. So, we knew we had a long time to solve our issues with the ELF.”80

Finally, as window theory would lead one to expect, the EPLF did not experience any serious trouble in operating in areas previously under its coethnic counterpart's sway, and many ELF fighters joined it after the ELF's defeat. As Adhanom Gebremariam recalls, “The Eritrean people were willing to support any organization that would have the upper hand. They thought that the government was the primary enemy and the EPLF was a formidable force, which could defeat the enemy.”81

INTRA-TIGRAYAN RELATIONS: THE TPLF'S MARCH TO HEGEMONY

Window theory provides a key to understanding rebel relations in Tigray, too. As its leader, Aregawi Berhe, notes: “[The] TPLF saw it as imperative to claim and realize a power monopoly in Tigrai [sic],” which it achieved by sequentially crushing the weaker TLF, Teranafit, and EPRP in a period of limited government threat.82

tplf-tlf war. Although I lack incontrovertible decisionmaking evidence, the broad outlines of the 1975 attack by the TPLF against the TLF are consistent with window-of-opportunity logic: the TPLF was stronger than its coethnic rival (both organizations had Tigrayan membership and ethnonational agendas), and the government posed a limited threat, thus promising the TPLF a cheap and quick victory.83

The TPLF launched a surprise attack on the TLF when the two were camped together for talks aimed at forming a united Tigrayan front. Many TLF members joined the TPLF's ranks after the dissolution of their group.84 TPLF interviewees unanimously point to the numerical inferiority of the TLF, which had also been experiencing internal problems and a membership hemorrhage.85 The government did not pose a serious and immediate threat, because, as Berhe notes in describing the Derg's activities in the province at that time, “in the vast rural areas of Tigrai, there was no meaningful government hold or influence to deter the mobility of the TPLF.”86

tplf-teranafit war. In 1976, the TPLF faced both a short-term window of opportunity and a long-term window of vulnerability: it was stronger than its coethnic rival—Teranafit—and the government still posed a limited threat, but the EDU—allied with Teranafit—was organizing a formidable force across the border in Sudan in preparation for an offensive into Tigray. Consistent with my argument, the TPLF wiped out its weaker coethnic rival before EDU reinforcements could tilt the balance.

Teranafit was a loose coalition of Tigray's landlords and bandits at the head of a peasant army that took up arms against the government in 1976. It did not have an articulated ideological program, besides wanting to restore traditional Tigrayan authority in the wake of the Derg revolution. In particular, Teranafit professed allegiance to Ras Mengesha, a former provincial governor and symbol of Tigrayan nationalism (he was the heir in Tigray's royal line), who had fled to Sudan amid a roundup by the revolutionary regime of high-ranking officials and royal family members. As John Young notes, “The Derg's dismissal of Tigrayan governor Ras Mengesha appeared to herald an era of even more harsh Amhara rule. In such a climate, appeals to Tigrayan national sentiments were essential for any political group wanting peasant support.”87 Teranafit and TPLF should thus be considered coethnic, as both had Tigrayan membership and leadership and professed some form of Tigrayan ethno-nationalism (in the case of Teranafit, related to the restoration of Mengesha). Local peasants saw both TPLF and Teranafit as “sons of Tigray” and urged them to cooperate against the Amhara-dominated Derg.88

In Sudan, Mengesha joined other members of the old regime and set up the EDU, with the plan of launching a sweeping offensive across the border into Tigray and then overthrowing the Derg. Supporters of Mengesha in Tigray started coalescing into Teranafit even before the announcement of the creation of the EDU in early 1976.89 As TPLF founding member Ghidey Zeratsion pointed out, the TPLF's leadership “knew that Teranafit would grow stronger as it established relations with the EDU. We knew that the EDU was being organized in Sudan and was getting weapons. Its launch base would be Tigray.”90 While the threat gathered across the border, the TPLF enjoyed a margin of superiority over Teranafit. According to TPLF sources, the two organizations had a comparable number of fighters, but the TPLF possessed superior cohesion and discipline.91 As a result, the TPLF expected to crush its rival when it eventually attacked Teranafit.92

The TPLF also believed that for some time it could grow at Teranafit's expense by attracting its peasant members with better mobilizing techniques and a political program resonating with their class and national aspirations. As Zeratsion put it, “Teranafit did not have a good political orientation; they were mostly made up of peasants and feudal lords. So, we thought we could attract their rank and file over time.”93 Faced with these two competing dynamics (on the one hand, the creation of the EDU, which would eventually strengthen Teranafit and directly intervene in Tigray; on the other, the prospect of outsmarting Teranafit in mobilization), the TPLF leadership initially opted for tactical cooperation, thus postponing the inevitable showdown. Negotiations led to the signing of a cooperative arrangement in June 1976.94

Shortly afterward, however, the killing of “Sihul”—a TPLF leader—during one of the occasional skirmishes between TPLF and Teranafit fighters overturned the TPLF's calculus. As Sihul had served as Tigray's representative in the Parliament and enjoyed a reputation as the province's staunch defender, the TPLF had depended on him to operate among the conservative peasantry. As Berhe observes, “[Tigrayans'] compliance [with TPLF's mobilization efforts] was granted not because they understood the objectives of the emerging front or because of the young revolutionary students, but simply because [of] Sihul … without him the unknown TPLF would have found it difficult to survive and expand.”95 Sihul's death undermined the TPLF's short-term strategy of growing larger than its rival while avoiding open confrontation. In an interview with the author, Berhe suggested as much: “For us Sihul was key, because he represented a link between the younger and older generations as well as between the rural and urban environments.”96 Not long after, the TPLF attacked and quickly defeated Teranafit, which allowed the TPLF to tap into its rival's pool of resources. A large number of Teranafit rank and file joined the TPLF, which did not experience any organized resistance in areas where Teranafit previously held sway.97

The limited threat posed by the government at the time comes across from interviews with TPLF leaders. As Zeratsion recalled, “We knew that … the government was very weak at that time, so it would not launch a major campaign.”98

tplf-eprp: from coexistence to war. Window theory explains the alternation of peaceful coexistence and inter-rebel war characterizing the TPLF-EPRP relationship. The two groups should be considered coethnic, as the EPRP's program emphasized the need to end the oppression of ethnic groups—including Tigrayans'—by the Amhara-dominated central government,99 and the bulk of its leadership was from Tigray.100 From the moment the TPLF and EPRP established a military presence in eastern Tigray in late 1975 until their armed conflict in 1978, ideological disagreements and competition for the support of Tigray's peasants created tensions between them.101 Until the spring of 1978, no window emerged; and consistent with my argument, no inter-rebel war occurred. Throughout this period, the government had very limited power projection in rural Tigray, but the TPLF did not enjoy a marked military edge, as it was engaged in fights against Teranafit and the stronger EDU in its base area in the west of the province.102

The situation changed in the spring of 1978, however. By then, while EPRP forces in eastern Tigray had languished amid internal turmoil and lack of significant military engagement with the Derg, the TPLF had defeated its rivals in the west, acquiring in the process valuable battlefield experience and weapons. At that point, the TPLF took advantage of an opportunity to establish a hegemonic position in the province by attacking its weaker coethnic rival's base area in mid-March 1978. By 1979, the EPRP was relegated to an inconsequential role in the armed struggle and could only operate far from Tigray.103 Consistent with window theory, the TPLF was able to expand to areas previously under the EPRP's influence and intensify its mobilization efforts.104

TPLF and EPRP sources agree that the TPLF was militarily stronger than the EPRP when war broke out. The two organizations had roughly comparable numbers of fighters and levels of armaments, but the TPLF possessed superior internal cohesion, discipline, and fighting skills.105 Berhe summarized well the general perception of the EPRP among TPLF leaders: “We considered them militarily ineffective. There was a lot of rhetoric on their part, but just that. They had no military experience, unlike us. We had fought against Teranafit and EDU for a long time. We had also fought the Derg, even if not in major battles … We understood we were in a better position than the EPRP if war broke out.”106 In addition to lacking significant battlefield experience, the TPLF leaders knew that the EPRP was riven with factionalism.107

As with the other episodes of inter-rebel war in Tigray, the government posed a limited threat. Once again Berhe's view is illustrative: “Our assessment was that the government was weak in Tigray. They could not penetrate rural areas; they had limited forces; they would not dare going to the countryside … They had most of their forces there [Eritrea], and in Tigray they were mostly watching the fight between TPLF and EPRP, the way they had done with Teranafit.”108

REBEL INTERACTIONS ACROSS THE ERITREA-TIGRAY DIVIDE

While both the Eritrean and Tigrayan insurgent movements experienced multiple episodes of infighting, there was remarkably little inter-rebel violence between Eritrean and Tigrayan rebel groups. In fact, the Eritrean fronts provided vital help to fledgling Tigrayan groups in 1975–76. The EPLF trained and armed the TPLF and the EPRP, while the ELF provided similar support to the TLF and refrained from fighting Teranafit.109 Relations between Eritrean and Tigrayan groups were not always harmonious: mutual suspicion, fear of exploitation and abandonment, and tough bargaining were pervasive, as the groups' interests were far from perfectly aligned.110 Yet disputes did not tend to escalate to all-out fights, as the TPLF-EPLF relationship illustrates.

eplf-tplf peaceful coexistence. After the TPLF and EPLF consolidated their hegemonic positions in the respective movements, relations between the two allies soured. The main issue concerned Ethiopia's ethnonational question. The TPLF maintained an ambiguous position on whether its war aims were limited to the liberation of Tigray or extended to overthrowing the government. By contrast, the EPLF opposed Tigray's independence, as it wanted a friendly TPLF-led central government to legitimize Eritrean independence after battlefield victory. Thus, the EPLF insisted that Tigrayan insurgents should seek to rule a multiethnic Ethiopia (minus Eritrea), rather than a breakaway Tigrayan state.111

In 1985, tensions boiled over, with the two groups engaging in public recriminations through publications and radio broadcasts. After the TPLF labeled the EPLF an enemy, the latter severed all contacts and even denied its Tigrayan counterpart access to Sudan through Eritrean territory to deliver relief to victims of the famine ravaging Tigray. Relations between the groups remained strained for three years, until an opportunity emerged in the spring of 1988 to inflict a decisive defeat on the Derg with a coordinated TPLF-EPLF effort. Then, in a nod to the EPLF's position, the TPLF publicly stated that regime change was a prerequisite for peace and for any act of national self-determination. The declaration led to the resumption of joint operations and ultimately rebel victory.112

Notwithstanding the acrimony between the TPLF and the EPLF in 1985–88 and the opportunity for military confrontation provided by the physical proximity of their base areas, no large-scale fight ensued. Consistent with window theory, there is no indication that the EPLF leadership ever thought it could defeat the TPLF and then mobilize Tigray's population the way it had done with the ELF and the population under the latter's control. On the contrary, the EPLF continued to believe that the defeat of the Derg could be achieved only with a multiethnic rebel alliance. As EPLF commander Mesfin Hagos noted, “We were aware of our interdependence with the TPLF. We knew that the enemy wanted to destroy one [rebel group] at the time, and we would be the next target if it defeated the TPLF.”113 The TPLF followed a similar multiethnic approach to move beyond Tigrayan territory and take power in Addis Ababa: it groomed the Ethiopian People's Democratic Movement and the Oromo People's Democratic Organization, with the objective of mobilizing the Amhara and Oromo populations, and then in 1989 brought them into an umbrella organization under its control.114

tplf-elf war. By contrast, interactions between the TPLF and the ELF do not fit window theory. Besides occasional skirmishes related to competing claims over the Eritrea-Tigray border, major clashes between the two occurred in 1979–81. In the fall of 1979 and in the spring of 1980, the TPLF fought ELF contingents trying to escort EPRP elements to neighboring Gondar Province through TPLF-held territory; and in 1980–81, the TPLF took part (on the EPLF's side) in the intra-Eritrean war. A closer look at these episodes, however, reveals mechanisms that, albeit not part of window theory, are not at odds with it. The TPLF-ELF fight thus represents a less damaging falsification blow than if dynamics similar to those that characterize cases of intra-ethnic war, which are explicitly ruled out by window theory, were in evidence.

The first TPLF-ELF clashes occurred in the context of the intra-Tigrayan TPLF-EPRP fight. Bereft of allies in Tigray (as the TPLF had wiped out its local partner—the TLF—and had gotten closer to the EPLF), the ELF tried to prop up the EPRP after its first defeat at the hands of the TPLF. Evidence on the ELF's decisionmaking is limited, but it seems that the ELF intended to escort EPRP survivors to Gondar, where they would reunite with other EPRP units.115 The TPLF was alarmed by the possibility of its rival's return to Tigray, and thus a battle ensued when its forces encountered the ELF-EPRP contingent in the fall of 1979. Similarly, in the spring of 1980 the TPLF clashed with an ELF contingent escorting EPRP elements expelled by the TPLF from Gondar.

After these battles, in early 1981 the TPLF eagerly accepted the EPLF's proposal to join it in its ongoing all-out offensive against the ELF.116 As Gebru Asrat noted, there was “a transmission of the intra-Eritrean conflict in Ethiopia through this system of alliances”—a mirror image of the ELF's earlier entanglement in the TPLF-EPRP feud.117

The TPLF wanted to eliminate an organization that had threatened its interests in Tigray—the ELF—but there is no indication that it expected to take over the resources previously under the latter's control, as in instances of intra-ethnic war. In fact, the concern driving the TPLF to help crush the ELF was the survival of its ally, the EPLF, rather than absorption of the ELF's resources. Having noted that the ELF could hinder the EPLF's defensive efforts against the massive Derg offensive then under preparation and thus bring about government victory, Tedros Hagos summed up the TPLF's rationale for helping the EPLF against the ELF: “The collapse of the Eritrean revolution was not going to have a pleasant effect on our self-interests.”118

The implied counterfactual here is that the TPLF likely would not have launched a large-scale attack to expel the ELF from Eritrea had there not been another Eritrean organization ready to fill the vacuum. Although providing supporting process evidence is exceedingly hard, given that rebel decisionmakers tend not to indulge in counterfactual analysis, EPLF-TPLF relations after the ELF's defeat represent a helpful comparison: notwithstanding the tensions between the EPLF and the TPLF discussed above, the two groups refrained from fighting each other. In fact, the TPLF continued to be deeply interested in the survival of its Eritrean counterpart, as the alternative would not have been the takeover by the TPLF of the insurgency in Eritrea, but rather an opportunity for the Derg to focus its full resources on the Tigrayan insurgents.119

ALTERNATIVE EXPLANATIONS AND ENDOGENEITY CONCERNS

Below I show that alternative explanations of inter-rebel war perform poorly, focusing in particular on Christia's minimum winning coalition theory, group and dyad characteristics, as well as third-party states' and incumbent's actions. In doing so, I also address endogeneity concerns.

Christia's MWC theory suggests that infighting should occur when one rebel group (or coalition) is sufficiently strong to take on both the government and other rebels. Her theory, however, predicts inter-rebel cooperation in all instances in which inter-rebel war occurred in Eritrea and Tigray: the rebels were much weaker than the Ethiopian government in terms of troop numbers, armaments, and territorial control, so they should have refrained from infighting.

In Eritrea, when the ELF attacked its rebel splinter groups in 1972, the government-rebel balance of power was skewed in the government's favor. As Woldemariam notes when describing military conditions in Eritrea in 1972, “The position of the Ethiopian military in the province was simply too strong to be challenged … The chance that the Ethiopian military would incur significant losses … was remote.”120 The government still controlled the bulk of Eritrean territory, and its forces outnumbered and outgunned the rebels by even wider margins in 1980–81, when the EPLF attacked the ELF. Like window theory, Christia's MWC theory correctly predicts the absence of inter-rebel war in Eritrea in 1978–79, when a reinvigorated army launched a series of major offensives. Moreover, her theory would correctly lead one to expect the EPLF and the TPLF not to fight after establishing their hegemonic positions in the respective rebel movements, given that both groups were still dwarfed by the government's forces for a long time. MWC theory cannot explain, however, why the rebels continued to cooperate past the point when a government defeat appeared inevitable.

MWC theory's explanatory power is comparably limited in Tigray. When the TPLF wiped out the TLF, the two combined had 200 fighters and controlled no territory. Similarly, at the time of their fight, Teranafit and the TPLF were lightly armed, controlled no territory, and had roughly 1,000 fighters each, compared to 45,000 government troops. An even more skewed balance of power prevailed in 1978, when the EPRP and the TPLF fought each other: each group had 1,000 fighters, controlled little territory, and mostly executed hit-and-run attacks, while government forces had significantly expanded in number and firepower with Soviet support.121 In these three cases of infighting, MWC theory would predict inter-rebel cooperation.

Another set of alternative explanations focuses on rebel groups' characteristics. Rebel groups with a certain ideological leaning (e.g., communist or jihadist) could be more likely to fight against organizations with different worldviews or be generally more prone to inter-rebel fighting. A variant of the ideological argument focuses on the distinction between hard-liners and moderates with regard to nationalist goals (i.e., autonomy vs. independence) and willingness to compromise with the government. The empirical evidence does not support these ideological arguments. In Eritrea, the EPLF had a more straightforward Marxist-Leninist orientation than the ELF, where a communist core competed with conservative elements. These marginal ideological differences, however, cannot explain the pattern of alternating cooperation and violent conflict, given that their ideologies were relatively stable features of the organizations. Moreover, despite these differences, both groups resorted to force only when facing a window of opportunity. Also, both organizations advocated Eritrean independence, and there is no indication of either one being more willing to compromise with the government. In Tigray, the TPLF fought the conservative Teranafit as well as groups with which it shared a Marxist-Leninist outlook (EPRP) and a desire for self-determination (TLF). The groups' different agendas for the province did not result in any of them trying to reach a settlement with the government.

Organizational structure/cohesion could affect the risk of infighting, too. Decentralized or undisciplined groups may be especially prone to inter-rebel war, as skirmishes initiated by foot soldiers or low-level commanders could draw in the entire organization through a process of inadvertent escalation. As with ideology, however, organizational structure cannot explain variation over time, as the phases of inter-rebel fighting were not characterized by especially high levels of decentralization or indiscipline. Moreover, both relatively cohesive and disciplined organizations (EPLF and TPLF) and low-cohesion and undisciplined ones (ELF) launched attacks against their rivals. In addition, the evidence on rebel decisionmaking presented above shows that inter-rebel war typically resulted from explicit decisions to use force, even if skirmishes may have influenced those decisions (by intensifying threat perception) or operated as immediate triggers for war.

Rebel leaders' characteristics (e.g., their personality, beliefs, and even mental health) could also influence the risk of inter-rebel war. Accounts of the Eritrean independence struggle are often dominated by the figure of EPLF leader Isaias Afewerki, variously described as power-hungry, ruthless against opponents, and a farsighted military strategist.122 One could therefore hypothesize that his ambition and aggressiveness set the EPLF on a collision course with its Eritrean rival or that inter-rebel war resulted from his peculiar ability to grasp the strategic logic envisioned by window theory. Afewerki's characteristics, however, are an unlikely explanation for the overall pattern of inter-rebel fighting in Eritrea, as the EPLF initiated only one of the episodes of intra-Eritrean rebel war. The ELF, whose leadership changed over time and was less tightly dominated by a single individual, initiated multiple attacks. Moreover, EPLF interviewees suggest that the 1980 decision to eliminate the ELF once and for all was driven by widely understood situational factors (i.e., the favorable balance of power and the limited government threat) rather than Afewerki's individual initiatives. In addition, the same ELF leaders who vowed to destroy the splinter groups in 1971 opted for peaceful coexistence from late 1974; similarly, the EPLF leaders who decided to wipe out the ELF in 1980 refrained from this course of action in previous years and avoided military confrontation with the TPLF. As some accounts of the TPLF's military struggle stress, the ambition and cunning of Meles Zenawi—the late TPLF leader and subsequently Ethiopian prime minister—one could conceive of similar leadership-level explanations of the TPLF's behavior toward other rebel groups.123 Zenawi's characteristics, however, cannot explain the episodes of inter-rebel war in which the TPLF was involved, as they occurred when the organization had a collective leadership (Zenawi consolidated his hold on power in the mid-1980s).

A thoughtful reader might wonder whether the observed association between coethnicity and inter-rebel war is a reflection of preexisting tensions correlated with, but causally unrelated to, ethnicity. In particular, antagonism between their respective leaderships could cause coethnic rebel groups to fight each other. The available evidence suggests this is unlikely to be the case. Whether a rebel group emerged as a splinter from another can be used as a proxy for preexisting tension between organizations' leaders; but in Ethiopia, both groups with a common “mother” organization (ELF and EPLF) and others with separate origins (TPLF, TLF, Teranafit, and EPPR) fought each other. Conversely, as noted, there were tensions between the EPLF's and TPLF's leaders, but these did not erupt in violence (at least not until the Ethiopia-Eritrea interstate war, years after rebel victory).

A final set of alternative explanations emphasizes the role of foreign governments and the incumbent in inciting or restraining inter-rebel war (beyond their impact on the balance of power and the threat environment, as envisioned by window theory). Third parties could induce infighting or cooperation with positive and negative incentives (e.g., offers of aid and threats of abandonment). The incumbent could pit groups against each other by stirring up fears that a rival was cooperating with the government or by cajoling one organization to attack another.

The available evidence in the Eritrea case does not indicate any attempt on the part of insurgents' foreign patrons to ignite inter-rebel violence. To the contrary, they repeatedly tried to promote cooperation against the Ethiopian government. For example, EPLF leader Saleh Sabbe reported mediation attempts by Sudan in 1975–77 to create a joint Eritrean rebel front.124 A skeptic could argue that Eritrea is a peculiar case of exceptionally limited foreign involvement and leverage, in which the insurgents could follow window logic given the absence of external constraints to their behavior. Indeed, some observers of the Eritrean insurgency depict a herculean, autarkic effort that brought about independence despite the world's neglect or even hostility.125 But although Eritrea represents an exception to the Cold War pattern of superpower confrontation by proxy (the Soviet Union provided military aid to the Derg, but the United States did not support the insurgents), this interpretation is overstated. Both the ELF and the EPLF received vital support from Arab countries, Somalia, and, in particular, Sudan.126 As Weldemichael notes, “Sudan provided what any rebellion needs for survival and success: cross-border sanctuaries, secure and reliable supply routes beyond Ethiopian reach, and shelter to waves of Eritrean refugees who, among other things, replenished the guerrillas' ranks.”127 The TPLF enjoyed similar support from Sudan, and there is no indication that, in Tigray, the government in Khartoum departed from the approach it followed in Eritrea of promoting cooperation among opponents of the Ethiopian government.

What about the role of the incumbent? Fears that one group might reach a deal with the government at the other's expense were pervasive in Eritrea. The Derg probably played up those fears to drive a wedge between its opponents. For example, in 1976 it did not include the EPLF on its list of public enemies, and it held meetings with representatives of the organization. Similarly, in subsequent years both the ELF and the EPLF had separate contacts with the government.128 As noted, the fact that fear of defection to the government may affect rebel threat perception and thus contribute to motivate inter-rebel war is fully consistent with window theory. The argument would be falsified only if fear of defection were sufficient to cause infighting, regardless of the presence of windows of opportunity and vulnerability. The evidence presented above suggests that this is not the case: inter-rebel war did not occur without windows of opportunity; the ELF and the EPLF may have been alarmed by news of contacts between their rival and the Derg, but infighting occurred only when an opportunity for the EPLF to launch a hegemonic bid materialized in 1980. Analogously, in the mid-1980s the TPLF feared that the EPLF might reach a separate agreement with the government, but this did not lead to inter-rebel war in the absence of a window of opportunity or vulnerability. On the other hand, TPLF sources do not reveal specific fears of defection driving the group's actions toward other organizations operating in Tigray. In fact, TPLF leaders thought that the Derg saw the group as less threatening than others and thus would not provide support to its rivals in Tigray.129

This observation points to the potential endogeneity of government threat. A full-fledged theory of the determinants of government threat is beyond the scope of this article, but it is plausible that, in some cases, the incumbent would strategically manipulate the rebels' threat environment to induce infighting. This possibility is not problematic for window theory as long as changes in rebel group behavior are responses to changes in threat levels. By contrast, the association between threat environment and infighting would be spurious if some factor that the government, but not the researcher, can observe caused both. For example, the government could have intelligence on two rebel groups indicating that infighting is imminent and thus decide to withdraw forces from the area to redeploy them in locations where insurgents are more actively engaged in anti-government operations; in this scenario, the ensuing infighting would be unrelated to the change in government threat. In practice, however, this is not a problem for the Ethiopian cases, as the changes in government threat over time were not the result of the incumbent's anticipation of infighting. In particular, the collapse of central authority in 1974–77, the series of offensives in 1978–79, and the subsequent lull in 1980–81 reflected the tightening and untightening of resource constraints faced by the incumbent for counterinsurgency operations: when the government had fewer resources at its disposal, it adopted a defensive stance, when more weapons and disciplined soldiers were available, it went on the offensive.

Conclusion

This article provides strong preliminary support for window theory. The vast majority of instances of inter-rebel war in Ethiopia followed window of opportunity logic or a combination of window of opportunity and vulnerability logics. The argument also generally correctly predicts the absence of inter-rebel war when costs would be prohibitively high or benefits too low (as when rebel groups did not share an ethnic base).

The article makes four main contributions to the study of civil wars. First, it shows that both strictly military considerations and social-ideational factors matter: a full understanding of civil war dynamics requires integrating them into rich theoretical frameworks.

Second, my argument and findings on coethnicity are consistent with the emphasis of a recent wave of civil war studies on the powerful effects of ethno-nationalism.130 Window theory reminds us, however, that politicized ethnic identity does not necessarily produce intra-ethnic cooperation; under some circumstances, inter-rebel violence may ensue. Initially limiting the scope of the argument to coethnicity (rather than extending it to the broader concept of co-constituency) is appropriate given that the empirical finding of in-group bias inspiring window theory relates to ethnic groups. Future research will need to assess empirically whether rebel groups with a common social base other than ethnicity (i.e., ideology) display similar patterns of interaction.

Third, window theory conceptualizes power in civil war as both contextual and dynamic. Insurgents start operating from a position of weakness, and strive to grow by conducting political and military activities. Inter-rebel war is one of the paths to growth (through accumulation of resources controlled by defeated coethnic rivals). On their part, even mighty governments may be unable in the short run to intensify military operations to take advantage of insurgent infighting. Moreover, the argument suggests that future studies of civil war processes may benefit from incorporating insights from the literature on power dynamics and preventive war in international relations.

Fourth, in addressing the origins of the structure of insurgent movements, the article contributes to an important body of research showing that the structure of these movements (i.e., whether they are dominated by a hegemon) affects their prospect of success.131

This study has also several policy implications. In principle, counterinsurgents could affect the risk of inter-rebel war by manipulating windows. They could create windows of opportunity by providing weapons or money to some groups and reducing military pressure on them, thus generating permissive circumstances for them to attack other groups. Rebel groups may reject this aid, however, as they risk being branded as collaborators. Moreover, whether the beneficiaries of government support would take the bait and exploit the incumbent-engineered window of opportunity remains an open question, as they may fear an opportunistic government attack when the insurgent movement is in the throes of infighting. The limited government threat associated with hegemonic bids in Ethiopia tended to be a function of hard-to-fake political or military constraints to power projection, rather than clever maneuvers. Creating windows of vulnerability for some rebel groups by providing support to a subset of other groups risks running into similar problems of limited rebel willingness to be seen as colluding with the enemy.

While the foregoing discussion suggests limited leeway for governments to induce infighting, window theory also implies substantial constraints to external interveners' ability to foster the growth of favorite rebel groups and weaken unpalatable ones. Besides operational difficulties involved in preventing spillover of aid to other organizations, selectively providing support to a subset of rebel groups (i.e., “picking winners”) risks creating windows of vulnerability for other groups, prompting them to attack the beneficiaries. Similarly, the prospect of direct external military intervention (whether with airpower or ground forces) in support of a rebel group may be perceived by other groups as a window of vulnerability, triggering an attack to weaken the local allies of the intervener. This dynamic was likely at play in the jihadist al-Nusra Front's attacks on the main recipients of U.S. military aid among Syrian insurgents—the Hazzm Movement and the Syrian Revolutionaries Front—as well as on Division 30, the ill-fated rebel unit created by the Pentagon's train-and-equip program.132 Picking winners is not doomed to failure, but would-be interveners should incorporate in their plans measures to counter the unintended effects suggested by window theory, such as avoiding slow and gradual interventions in support of local allies (thus reducing the risk that local opponents would have enough time to crush them) and providing reliable protection for the fledgling armed organizations.

Acknowledgments

The author thanks the anonymous reviewers, Richard Betts, Page Fortna, Robert Jervis, Morgan Kaplan, Peter Krause, Austin Long, Sara Moller, Massimo Morelli, Joseph Parent, Günter Schröder, Lee Seymour, Jack Snyder, Paul Staniland, and Michael Woldemariam for helpful comments. He acknowledges financial support from the Smith Richardson Foundation and Princeton University's Niehaus Center.

Notes

1. 

Bill Roggio and Lisa Lundquist, “Analysis: Shifting Dynamics of Rebel Infighting in Syria,” Long War Journal, January 17, 2014, http://www.longwarjournal.org/archives/2014/01/analysis_shifting_dy.php.

2. 

Liz Sly, “On Third Anniversary of Syrian Rebellion, Assad Is Steadily Winning the War,” Washington Post, March 14, 2014.

3. 

Sam Dagher and Maria Abi-Habib, “Fighting among Rebels Boosts Syrian Regime,” Wall Street Journal, January 13, 2014.

4. 

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, doi:10.1177/0022002711429681.

5. 

Stephen A. Biddle, Jeffrey 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, doi:10.1162/ISEC_a_00087.

6. 

Kenneth N. Waltz, Theory of International Politics (Reading, Mass.: Addison-Wesley, 1979).

7. 

Arthur A. Stein, “Conflict and Cohesion: A Review of the Literature,” Journal of Conflict Resolution, Vol. 20, No. 1 (March 1976), pp. 143–172, doi:10.1177/002200277602000106.

8. 

Chaim D. Kaufmann, “Possible and Impossible Solutions to Ethnic Civil Wars,” International Security, Vol. 20, No. 4 (Spring 1996), pp. 136–175, doi:10.1162/isec.20.4.136; and Omar Shahabudin McDoom, “The Psychology of Threat in Intergroup Conflict: Emotions, Rationality, and Opportunity in the Rwandan Genocide,” International Security, Vol. 37, No. 2 (Fall 2012), pp. 119–155, doi:10.1162/ISEC_a_00100.

9. 

My dependent variable is dichotomous—inter-rebel war/peaceful coexistence. Future studies may focus on more granular variation along the spectrum of inter-rebel relations, including both passive cooperation (mutual avoidance) and active cooperation (alliances).

10. 

Inter-rebel clashes qualify as large-scale combat if repeated battles between units of different organizations occur, a major battle with hundreds of fighters takes place, or the main headquarters of one of the rebel groups is the target of a major attack.

11. 

For an analogous conceptualization of “windows” in interstate relations, see Stephen Van Evera, Causes of War: Power and the Roots of Conflict (Ithaca, N.Y.: Cornell University Press, 1999).

12. 

The term has been popularized in political science by George W. Downs and David M. Rocke, Optimal Imperfection? Domestic Uncertainty and Institutions in International Relations (Princeton, N.J.: Princeton University Press, 1995).

13. 

For a similar approach to the association between rebel organizations and ethnic groups, see Julian Wucherpfennig et al., “Ethnicity, the State, and the Duration of Civil War,” World Politics, Vol. 64, No. 1 (January 2012), pp. 79–115, doi:10.1017/S004388711100030X. Following Donald L. Horowitz, I define ethnicity broadly, as embracing “differences identified by color, language, religion, or some other attribute of common origin” (real or imagined). See Horowitz, Ethnic Groups in Conflict (Berkeley: University of California Press, 1985), pp. 41, 52. This conceptualization is compatible with two pillars of the constructivist research agenda on ethnic politics—constrained malleability and the multidimensional nature of identities. Individuals' identity attributes and the relative salience of various identity dimensions can change, but these processes are rarely instantaneous or costless.

14. 

For example, in the 1980s Iraq's Kurdistan Democratic Party and Patriotic Union of Kurdistan were coethnic even if the former had its stronghold in Iraq's Kurmanji-speaking territories, whereas the latter recruited mostly from Sorani-speaking areas.

15. 

Kaufmann, “Possible and Impossible Solutions to Ethnic Civil Wars.”

16. 

Stathis N. Kalyvas influentially argued that ideologies and ethnic identities are much less important predictors of civil war behavior than coercion and local-level relationships and disputes. See Kalyvas, The Logic of Violence in Civil War (New York: Cambridge University Press, 2006).

17. 

Ben Hubbard, “U.S. Goal Is to Make Syrian Rebels Viable,” New York Times, September 18, 2014.

18. 

Staniland, “Between a Rock and a Hard Place.”

19. 

On the cases of Sri Lanka and Iraq, see Ahmed S. Ashim, When Counterinsurgency Wins: Sri Lanka's Defeat of the Tamil Tigers (Philadelphia: University of Pennsylvania Press, 2013); and Biddle, Friedman, and Shapiro, “Testing the Surge.”

20. 

Navin A. Bapat and Kanisha D. Bond, “Alliances between Militant Groups,” British Journal of Political Science, Vol. 42, No. 4 (October 2012), pp. 793–824, doi:10.1017/S0007123412000075; Kristin M. Bakke, Kathleen G. 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, doi:10.1017/S1537592712000667; Lee J.M. Seymour, “Why Factions Switch Sides in Civil Wars: Rivalry, Patronage, and Realignment in Sudan,” International Security, Vol. 39, No. 2 (Fall 2014), pp. 92–131, doi:10.1162/ISEC_a_00179; and Håvard Mokleiv Nygård and Michael Weintraub, “Bargaining between Rebel Groups and the Outside Option of Violence,” Terrorism and Political Violence, Vol. 27, No. 3 (2015), pp. 557–580, doi:10.1080/09546553.2013.829459.

21. 

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, doi:10.1177/0022002712439496.

22. 

Fotini Christia, Alliance Formation in Civil Wars (New York: Cambridge University Press, 2012). Allied rebel groups, by definition, do not fight each other, but organizations that are not in an alliance may eschew infighting. Thus, arguments about alliances imply predictions about inter-rebel fighting, too.

23. 

Alistair Horne, A Savage War of Peace: Algeria 1954–1962 (London: Macmillan, 1977), pp. 136, 222.

24. 

John A. Marcum, The Angolan Revolution, Vol. 2: Exile Politics and Guerrilla Warfare (1962–1976) (Cambridge, Mass.: MIT Press, 1978), pp. 9–61.

25. 

Staniland, “Between a Rock and a Hard Place.”

26. 

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, doi:10.1162/ISEC_a_00148.

27. 

Waltz, Theory of International Politics, p. 118.

28. 

Tanisha M. Fazal, “State Death in the International System,” International Organization, Vol. 58, No. 2 (April 2004), pp. 311–344, doi:10.1017/S0020818304582048. By contrast, rebel groups' demise is frequent. See Daniel Byman, Understanding Proto-Insurgencies (Santa Monica, Calif.: RAND Corporation, 2007), p. 1.

29. 

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, doi:10.1177/0022002711429697; and Peter Krause, “The Political Effectiveness of Non-State Violence: A Two-Level Framework to Transform a Deceptive Debate,” Security Studies, Vol. 22, No. 2 (2013), pp. 259–294, doi:10.1080/09636412.2013.786914.

30. 

The commitment problem under anarchy explains why a war-avoiding agreement is not reached. Robert Jervis, “Cooperation under the Security Dilemma,” World Politics, Vol. 30, No. 2 (January 1978), pp. 167–214; and James D. Fearon, “Rationalist Explanations for War,” International Organization, Vol. 49, No. 3 (Summer 1995), pp. 379–414.

31. 

Hegemonic bids and gambles for resurrection follow a better-now-than-later logic. See Jack S. Levy, “Declining Power and the Preventive Motivation for War,” World Politics, Vol. 40, No. 1 (October 1987), pp. 82–107, doi:10.2307/2010195. Whereas gambles for resurrection represent responses to declining power, hegemonic bids attempt to consolidate a group's dominant position before a window of opportunity closes, even if there is no indication of imminent deterioration of the balance of power or intensification of government threat.

32. 

The argument should be especially relevant to relatively cohesive and disciplined rebel organizations, as significant leadership divisions or indiscipline among the rank and file may hinder a response to the imperatives of window theory.

33. 

Windows of opportunity for one group do not necessarily correspond to windows of vulnerability for others. Windows of vulnerability are not merely situations in which a weak group faces a serious existential threat; they entail a clear prospect of a drastic deterioration of the rebel balance of power. Under these circumstances (and in the absence of a major first-strike advantage), the group whose power is growing does not face a window of opportunity, as it has an incentive to wait to use force until its power has peaked.

34. 

Insurgent organizations would engage in diplomatic initiatives to break hostile encirclements and gain new allies, experiment with different mobilization strategies in the face of dwindling popular support, or strive to improve their cohesion and battlefield effectiveness.

35. 

I define this type of inter-rebel aggression in terms of its causes, rather than any specific form of violence.

36. 

I present an extended version of window theory in Costantino Pischedda, “Wars within Wars: Why Rebel Groups Fight Each Other,” University of Miami, 2018.

37. 

Krause, “The Structure of Success.”

38. 

On cumulativity and interstate war, see Van Evera, Causes of War.

39. 

Undisputed access to a rebel group's ethnic support base is likely more important than the absorption of rivals' fighters, as waging guerrilla warfare typically does not require large armies. A pliable support base can provide intelligence, safe havens, money, and logistical support, in addition to recruits. Moreover, some elements of defeated groups may be considered irreconcilable, making less experienced but more reliable new recruits preferable.

40. 

Peter Liberman, Does Conquest Pay? The Exploitation of Occupied Industrial Societies (Princeton, N.J.: Princeton University Press, 1996). The point is not that foreign occupiers are typically unable to extract resources from occupied territories (Stathis N. Kalyvas, for example, reports instances of pro-government militias recruited from rebellious ethnic groups), but rather that extraction costs are likely to be relatively high. See Kalyvas, “Ethnic Defection in Civil Wars,” Comparative Political Studies, Vol. 41, No. 3 (August 2008), pp. 1043–1068, doi:10.1177/0010414008317949.

41. 

Paul Staniland, “Organizing Insurgency: Networks, Resources, and Rebellion in South Asia,” International Security, Vol. 37, No. 1 (Summer 2012), pp. 142–177, doi:10.1162/ISEC_a_00091; and Lee J.M. Seymour, Kristin M. Bakke, and Kathleen Gallagher Cunningham, “E Pluribus Unum, Ex Uno Plures: Competition, Violence, and Fragmentation in Ethnopolitical Movements,” Journal of Conflict Resolution, Vol. 53, No. 1 (January 2016), pp. 3–18, doi:10.1177/0022343315605571.

42. 

Nicholas Sambanis, Jonah Schulhofer-Wohl, and Moses Shayo, “Parochialism as a Central Challenge in Counterinsurgency,” Science, May 18, 2012, pp. 805–808, doi:10.1126/science.1222304.

43. 

Robert H. Bates, “Ethnic Competition and Modernization in Contemporary Africa,” Comparative Political Studies, Vol. 6, No. 4 (January 1974), pp. 457–484; Nicolas van de Walle, “Presidentialism and Clientelism in Africa's Emerging Party Systems,” Journal of Modern African Studies, Vol. 41, No. 2 (June 2003), pp. 297–321, doi:10.1017/S0022278X03004269; Kanchan Chandra, Why Ethnic Parties Succeed: Patronage and Ethnic Head Counts in India (New York: Cambridge University Press, 2004); and Karen E. Ferree, “Explaining South Africa's Racial Census,” Journal of Politics, Vol. 68, No. 4 (November 2006), pp. 803–815, doi:10.1111/j.1468-2508.2006.00471.x.

44. 

Daniel Corstange, The Price of a Vote in the Middle East: Clientelism and Communal Politics in Lebanon and Yemen (New York: Cambridge University Press, 2016); and Raphaël Franck and Ilia Rainer, “Does the Leader's Ethnicity Matter? Ethnic Favoritism, Education, and Health in Sub-Saharan Africa,” American Political Science Review, Vol. 106, No. 2 (May 2012), pp. 294–325, doi:10.1017/S0003055412000172.

45. 

Philip Roessler, “The Enemy Within: Personal Rule, Coups, and Civil War in Africa,” World Politics, Vol. 63, No. 2 (April 2011), pp. 300–346, doi:10.1017/S0043887111000049.

46. 

James D. Fearon and David D. Laitin, “Explaining Interethnic Cooperation,” American Political Science Review, Vol. 90, No. 4 (December 1996), pp. 715–735, doi:10.2307/2945838; and James Habyariamana et al., “Why Does Ethnic Diversity Undermine Public Goods Provision?” American Political Science Review, Vol. 101, No. 4 (November 2007), pp. 709–725, doi:10.1017/S0003055407070499.

47. 

Daniel N. Posner, Institutions and Ethnic Politics in Africa (New York: Cambridge University Press, 2005); Chandra, Why Ethnic Parties Succeed; Henry E. Hale, “Explaining Ethnicity,” Comparative Political Studies, Vol. 37, No. 4 (2004), pp. 458–485, doi:10.1177/0010414003262906; and Corstange, The Price of a Vote in the Middle East.

48. 

On social identity theory, see Henri Tajfel, Human Groups and Social Categories: Studies in Social Psychology (New York: Cambridge University Press, 1981). On political socialization theory, see Keith Darden, Resisting Occupation: Mass Schooling and the Creation of Durable National Loyalties (Cambridge: Cambridge University Press, forthcoming).

49. 

Roger D. Petersen, Understanding Ethnic Violence: Fear, Hatred, and Resentment in Twentieth-Century Eastern Europe (New York: Cambridge University Press, 2002).

50. 

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, doi:10.1017/S0003055409990323.

51. 

Under these circumstances, the image costs of aggression in the eyes of the relevant ethnic community are likely also low: the decisive fight will soon be in the past, and the hegemon will be in a favorable position to emphasize the importance of unity against the government.

52. 

Paul Staniland, “States, Insurgents, and Wartime Political Orders,” Perspectives on Politics, Vol. 10, No. 2 (June 2012), pp. 243–264, doi:10.1017/S1537592712000655.

53. 

As Paul Staniland writes, “States are often content not to devote their full resources to internal war. Deployed state capacity is endogenous to political relationships.” Ibid., p. 253.

54. 

I used Iraq's Kurdish insurgencies as a theory development case study.

55. 

The limitations of this qualitative case study approach are well known: it cannot precisely quantify the impact of independent variables, identify confidence intervals, or assess generalizability beyond the cases examined. Future studies will need to further probe the explanatory power of window theory and assess its broader validity.

56. 

On organizational cohesion as a cause of rebel military effectiveness, see Staniland, “Organizing Insurgency.” On tactical-operational skills as drivers of battlefield outcomes (in both interstate wars and conflicts involving nonstate actors), see Stephen Biddle, Military Power: Explaining Victory and Defeat in Modern Battle (Princeton, N.J.: Princeton University Press, 2004).

57. 

For example, if information is available on two dimensions of relative power, I code a group as stronger if it is starkly superior in at least one dimension and at least roughly equal in the other dimension.

58. 

Alemseged Abbay, Identity Jilted or Re-imagining Identity? The Divergent Paths of the Eritrean and Tigrayan Nationalist Struggles (Lawrenceville, N.J.: Red Sea, 1998).

59. 

See the map depicting the various rebel groups' initial base areas in the online appendix at doi:10.7910/DVN/FIKFCQ/.

60. 

The insurgency in Eritrea started in 1961, but the EPLF made its appearance only in 1972—hence, the 1972–91 period of analysis.

61. 

Initially, the ELF drew most of its members from the province's Muslim-dominated lowlands. The organization always had a substantial Christian component, however; and during the 1972–74 ELF-EPLF armed conflict, Eritrean civilians consistently appealed to both sides to set aside their differences and fight together for the nationalist cause. Moreover, by 1975 the bulk of the rank and file of both groups were Christian highlanders.

62. 

Author interviews with Mesfin Hagos, July and August 2013, Frankfurt, Germany. This imbalance of power clearly emerges from several other EPLF and ELF subjects' accounts.

63. 

John Markakis, National and Class Conflict in the Horn of Africa (New York: Cambridge University Press, 1987), pp. 133–134; and David Pool, From Guerrillas to Government: The Eritrean People's Liberation Front (Oxford: James Currey, 2001), pp. 64–68.

64. 

Author interview with Gime Ahmed, July 2013, Addis Ababa.

65. 

Michael Woldemariam, “Battlefield Outcomes and Rebel Cohesion: Lessons from the Eritrean Independence War,” Terrorism and Political Violence, Vol. 28, No. 1 (2015), pp. 135–156, doi:10.1080/09546553.2014.886575.

66. 

Richard Sherman, Eritrea: The Unfinished Revolution (New York: Praeger, 1980), pp. 80–82; Roy Pateman, Eritrea: Even the Stones Are Burning (Lawrenceville, N.J.: Red Sea, 1998), pp. 132–133; and Bowyer Bell, “Endemic Insurgency and International Order: The Eritrean Experience,” Orbis, Vol. 18, No. 2 (Summer 1974), pp. 427–450.

67. 

Author interview with Ahmed Nasser (an ELF leader at the time of the events), August 2013, Stockholm. The text of the resolution is reported in Gaim Kibreab, Critical Reflections on the Eritrean War of Independence: Social Capital, Associational Life, Religion, Ethnicity, and Sowing Seeds of Dictatorship (Trenton, N.J.: Red Sea, 2008), pp. 173–174. The secondary literature offers a consistent account. See, for example, Pool, From Guerrillas to Government, p. 71.

68. 

Markakis, National and Class Conflict in the Horn of Africa, pp. 134–135; Pool, From Guerrillas to Government, pp. 133–136; and Bereket Habte Selassie, The Crown and the Pen: The Memoirs of a Lawyer Turned Rebel (Trenton, N.J.: Red Sea, 2007), pp. 306–309.

69. 

Author interview with Nasser. Tesfay “Degiga” (ELF member from 1973) made a similar observation in an interview with the author, July 2013, Frankfurt, Germany. See also the ELF's Eritrean Newsletter, September 1, 1981, pp. 11–13.

70. 

Interview with Ma'ariv, October 21, 1974, reported in Haggai Elrich, The Struggle over Eritrea, 1962–1978: War and Revolution in the Horn of Africa (Stanford, Calif.: Hoover Institution Press, 1983), p. 52. Sabbe was an ELF founding member and influential EPLF figure until 1976. Haile Menkerios (EPLF member from 1973) also pointed to the widespread perception that government weakness opened the possibility of achieving independence. Author interview with Menkerios, July 2013, Addis Ababa.

71. 

Elrich, The Struggle over Eritrea, pp. 71–78; Markakis, National and Class Conflict in the Horn of Africa, pp. 136–144; and Woldemariam, “Battlefield Outcomes and Rebel Cohesion,” p. 149.

72. 

Markakis, National and Class Conflict in the Horn of Africa, p. 134. Wolde-Yesus Ammar (ELF member from 1965 and subsequently head of the group's Foreign Office) noted that the realization that defeating the EPLF would be very difficult convinced the ELF to stop the infighting. Author interview with Ammar, August 2013, Frankfurt, Germany.

73. 

By 1979, the Ethiopian army had grown to include more than 200,000 soldiers, compared to fewer than 50,000 when the Derg took power; 120,000 heavily armed troops participated in the first offensive. Dan Connell, Against All Odds: A Chronicle of the Eritrean Revolution (Trenton, N.J.: Red Sea, 1993), pp. 157–194; and Awet T. Weldemichael, “The Eritrean Long March: The Strategic Withdrawal of the Eritrean People's Liberation Front (EPLF), 1978–1979,” Journal of Military History, Vol. 73, No. 4 (October 2009), pp. 1231–1271, doi:10.1353/jmh.0.0423.

74. 

Weldemichael, “The Eritrean Long March,” p. 1232.

75. 

Awet Tewelde Weldemichael, Third World Colonialism and Strategies of Liberation: Eritrea and East Timor Compared (New York: Cambridge University Press, 2013), p. 154.

76. 

For example, Ahmed described ELF's “anarchic divisions.” Author interview with Ahmed. This is a recurring theme in the secondary literature, too. See Ruth Iyob, The Eritrean Struggle for Independence: Domination, Resistance, Nationalism, 1941–1993 (New York: Cambridge University Press, 1995), pp. 120–121; and Pool, From Guerrillas to Government, p. 147.

77. 

Author interviews with Mesfin Hagos. Another military commander, Adhanom Gebremariam, provided a similar assessment. Author interview with Gebremariam, April 2014, New York City, New York.

78. 

The much touted sixth offensive would be launched in February 1982, after two years of preparations. See Gebru Tareke, The Ethiopian Revolution: War in the Horn of Africa (New Haven, Conn.: Yale University Press, 2009), pp. 225–246.

79. 

The EPLF-ELF war started the same day as the Ethiopian army launched a large-scale offensive in the southeast against Somali forces that had been under well-publicized preparation for months.

80. 

See also the interview with Isaias Afewerki (president of Eritrea and top EPLF leader at that time) in Adulis, Vol. 1, Nos. 4–5 (1984), pp. 9–15; and EPLF, “Ethiopia's Sixth Offensive and Developments in the Eritrean Struggle,” August 25, 1982.

81. 

Author interview with Gebremariam. Mesfin Hagos made a similar statement. Author interview with Hagos.

82. 

Aregawi Berhe, A Political History of the Tigray People's Liberation Front (1975–1991) (Los Angeles, Calif.: Tsehai, 2009), p. 278 (emphasis in the original).

83. 

Some TPLF sources posit the existence of a plan to eliminate the TLF. See Kahsay Berhe, Ethiopia: Democratization and Unity (Münster, Germany: Monsenstein und Vannerdat, 2005), p. 62; and author interview with Gebru Asrat (TPLF member from 1975 and subsequently in leadership positions), July 2013, Addis Ababa. Others argue that the TPLF leadership was genuinely interested in merging with the TLF, but that the TPLF rank and file's opposition to the merger prevailed. Author interview with Aregawi Berhe (chairman of the TPLF at the time), August 2013, The Hague; and Ghidey Zeratsion (TPLF founding member), August 2013, Oslo.

84. 

Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 81–84.

85. 

Author interview with Berhe; author interview with Asrat; author interview with Tedros Hagos (TPLF member since 1976), July 2013, Mekele, Ethiopia; and author interview with Mokonnen Mokonnen (TPLF member from 1975), September 2013, Silver Spring, Maryland.

86. 

Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), p. 93. The negligible Derg presence in Tigray until 1978 is reported in the secondary literature, too. See John Young, Peasant Revolution in Ethiopia: The Tigray People's Liberation Front, 1975–1991 (New York: Cambridge University Press, 1997), pp. 93, 97.

87. 

Young, Peasant Revolution in Ethiopia, p. 100. The Amhara were the politically dominant group both before and after the Derg took control of the governments.

88. 

Ibid., pp. 102–103; Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), p. 105; author interview with Asrat; and author interview with Mokonnen.

89. 

Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 103–104; author interview with Zeratsion; and Young, Peasant Revolution in Ethiopia, pp. 100–102.

90. 

Author interview with Zeratsion.

91. 

Ibid.; author interview with Berhe; author interview with Tesfay Atsbeha (TPLF senior military figure), August 2013, Cologne, Germany; and author interview with Mulugeta Gebrehiwot (TPLF foot soldier at the time), July 2013, Addis Ababa. Berhe reports that the size of the TPLF in July 1976 was about 1,000 fighters. See Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), p. 170. Historical accounts are replete with references to Teranafit's lack of a coherent organization and the prevalence among its rank and file of criminals who opportunistically joined for looting and peasants who were lured with false economic promises and not deeply committed to the group.

92. 

Author interview with Zeratsion; and author interview with Berhe.

93. 

Author interview with Zeratsion. Other interviewees, in particular Berhe and Gebrehiwot, made similar observations. See author interview with Berhe; and author interview with Gebrehiwot. See also Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), p. 105.

94. 

Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 127–128.

95. 

Ibid., pp. 55–56.

96. 

Author interview with Berhe. Moreover, the TPLF's predicament was exacerbated by news of an EDU contingent inside Ethiopia after Sihul's death. See Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 106–107. Zeratsion also reports that the expectation of the EDU's imminent arrival undermined the TPLF's initial plan. See author interview with Zeratsion.

97. 

Author interview with Tedros Hagos; and Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 107–108.

98. 

Author interview Zeratsion. Aregawi Berhe made a similar observation. Author interview with Berhe.

99. 

Kiflu Tadesse, The Generation, Part 1: The History of the Ethiopian People's Revolutionary Party (Lanham, Md.: University Press of America, 1998), pp. 390–391; and Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 142–145. As EPRP leader Tadesse notes, “Resolving the national question had remained one of the major issues for the organization.” See ibid., p. 229. The TPLF-EPRP ideological debate was about whether class or national/ethnic issues should be prioritized.

100. 

No data on the ethnic background of EPRP fighters are available. EPRP and TPLF interview subjects as well as the secondary literature report that the EPRP leadership was dominated by Tigrayans. See Sarah Vaughan, “Ethnicity and Power in Ethiopia,” Ph.D. dissertation, University of Edinburgh, 2003 p. 166; author interviews with Begasho Ashenafi (EPRP foot soldier), August 2013, Frankfurt, Germany; and author interview with Tekleweini Assefa (early member of TPLF), July 2013, Addis Ababa. That the two groups had overlapping bases of support is clearly illustrated by their competition for that support, reported by TPLF and EPRP sources: both were able to recruit in eastern Tigray, and both tried to convince supporters of their rivals to switch sides. See Tadesse, The Generation, Part 1, p. 394; and Berhe, Ethiopia, p. 75.

101. 

The groups were unsuccessful in negotiating an agreement to cooperate against the government and prevent the escalation of skirmishes involving their fighters and supporters. See Tadesse, The Generation, Part 1, pp. 88, 390–399; Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 143–147; and author interview with Zeratsion, who was the TPLF representative at the negotiations.

102. 

I do not discuss the EDU's attack on the TPLF for reasons of space. As the EDU was not a Tigrayan group—it had a pan-Ethiopian agenda and ethnically mixed composition—the TPLF-EDU fight represents a failed prediction for window theory (see table 1). The TPLF eventually prevailed after the EDU was significantly weakened by a government offensive.

103. 

Tadesse, The Generation, Part 1, pp. 434–437, 471; and Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 149–151.

104. 

Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), p. 281. See also Tadesse, The Generation, Part 1, p. 407; and Berhe, Ethiopia, pp. 75–76.

105. 

Tadesse, The Generation, Part 1, p. 404, reports that in early 1978 the EPRP had about 1,000 fighters in eastern Tigray. Ashenafi provides the same figure. See author interview with Ashenafi. Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), p. 178, notes that in late spring 1978, the TPLF could marshal about 1,000 soldiers. According to Atsbeha (TPLF commander at that time), the two groups were equal in terms of number of soldiers and weapons, but the TPLF was stronger because its fighters were significantly more experienced. Author interview with Atsbeha.

106. 

Author interview with Berhe. Consistently, Kahsay Berhe, Zeratsion, and Mokonnen all stressed that the TPLF emerged battle-hardened from its fight against the EDU, and Asrat reported that the TPLF leadership believed that the group could easily defeat the EPRP by the time the fight against the EDU was ending. See Berhe, Ethiopia, p. 58; author interview with Zeratsion; author interview with Mokonnen; and author interview with Asrat. The EPRP's problems of combat readiness and its limited military experience (often a source of discontent among the organization's rank and file) are a leitmotiv in Tadesse, The Generation, Part 1, pp. 188, 370, 374, 381. Ashenafi and an anonymous EPRP foot soldier also stressed the group's almost complete lack of battlefield experience at the beginning of the inter-rebel war. Author interview with Ashenafi; and author interview with EPRP foot soldier, August 2013, Addis Ababa.

107. 

Author interview with Berhe; and Tadesse, The Generation, Part 1, pp. 374–383. In 1977, the government eradicated the EPRP's urban infrastructure; hundreds of militants took refuge in the group's base area in Tigray, causing internal unrest. See ibid., pp. 269–301, 308–309.

108. 

Asrat and Assefa report similar assessments. Author interview with Asrat; and author interview with Assefa.

109. 

Berhe, Ethiopia, pp. 49–52; Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 251–255; and Awet T. Weldemichael, “Formative Alliances of Northeast African Insurgents: Eritrean Liberation Strategy and Ethiopian Armed Opposition,” Northeast African Studies, Vol. 14, No. 1 (March 2014), pp. 83–122.

110. 

Weldemichael, “Formative Alliances of Northeast African Insurgents.”

111. 

Ibid.; author interview with Mesfin Hagos; author interview with Zeratsion; and author interview with Berhe.

112. 

Weldemichael “Formative Alliances of Northeast African Insurgents,” pp. 103–108; Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 272–273, 315; author interview with Berhe; author interview with Mesfin Hagos; and author interview with Menkerios.

113. 

Author interview with Mesfin Hagos.

114. 

Tareke, The Ethiopian Revolution, p. 109.

115. 

Nasser claims that the battle was unplanned, which appears plausible given that only a relatively small ELF contingent was involved. Author interview with Nasser. See also Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), p. 257.

116. 

Tadesse, The Generation, Part 1, pp. 436–437; and Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 257–259.

117. 

Author interview with Asrat.

118. 

Quoted in Abbay, Identity Jilted or Re-imagining Identity? p. 114.

119. 

As Mokonnen noted, “[The TPLF] wanted the EPLF to stay in power; otherwise, the Derg would focus all its forces on us.” See author interview with Mokonnen.

120. 

Woldemariam, “Battlefield Outcomes and Rebel Cohesion,” p. 147. For details on the Derg's stark advantages in troop numbers, weaponry, and territorial control, see ibid., pp. 140–148; Weldemichael, Third World Colonialism and Strategies of Liberation, pp. 119–125; and The Military Balance, 1971–72 (London: International Institute for Strategic Studies, 1971).

121. 

Berhe, A Political History of the Tigray People's Liberation Front (1975–1991), pp. 100, 170; and Young, Peasant Revolution in Ethiopia, p. 98.

122. 

Connell, Against All Odds, pp. 180–181; and Pool, From Guerrillas to Government, pp. 76–87.

123. 

See, for example, Tareke, The Ethiopian Revolution, pp. 101–102.

124. 

Interview with Sabbe, conducted by Günter Schröder, December 1980, Khartoum. Similar initiatives by Sudan, Somalia, and various Arab countries are also reported by Ahmed Karar, Sudan's deputy chief of security at the time. Karar, interviewed by Günter Schröder, January 1981, Khartoum; author interview with Menkerios; author interview with “Degiga”; and author interview with Nasser.

125. 

Connell, Against All Odds.

126. 

Awet T. Weldemichael, “African Diplomacy of Liberation: The Case of Eritrea's Search for an ‘African India,'” Cahiers d'Études Africaines, Vol. 212, No. 4 (2013), pp. 867–894.

127. 

Ibid., p. 879.

128. 

Tadesse, The Generation, Part 1, p. 130; author interview with “Degiga”; and author interview with Nasser.

129. 

Author interviews with Berhe; author interview with Zeratsion; and author interview with Asrat.

130. 

Lars-Erik Cederman, Kristian Skrede Gleditsch, and Halvard Buhaug, Inequality, Grievances, and Civil War (New York: Cambridge University Press, 2013).

131. 

Krause, “The Structure of Success.”

132. 

Eric Schmitt and Ben Hubbard, “U.S. Revamping Rebel Force Fighting ISIS in Syria,” New York Times, September 6, 2015.