Behind enemy lines: State-insurgent cooperation on rebel governance in Côte d’Ivoire and Sri Lanka

ABSTRACT This article examines the conditions that foster state-insurgent cooperation in rebel governance. State-insurgent cooperation is puzzling because it can alienate hardliners, undermine the parties’ legitimacy, reveal sensitive information, and cause autonomy losses. We propose that conflict parties are more likely to discount these costs when they have overlapping civilian constituencies with high governance provision expectations. Analysing rebel governance in Côte d’Ivoire and Sri Lanka using original data, we find that civilian expectations prompt cooperation even when the parties appeal to separate constituencies. The article nuances existing theories of rebel governance and contributes new knowledge on state-insurgent interactions in civil war.


Introduction
A counter-intuitive facet of civil war is that enemies may at times also cooperate to provide governance. For example, the regional government in Myanmar worked alongside the Karen National Union (KNU) rebels to deliver health care in rebel-controlled areas after the 2015 ceasefire agreement. 1 In the Philippines, the Moro Islamic Liberation Front (MILF) actively cooperated with the government on disaster relief, 2 and in Afghanistan, certain statefunded schools reopened in Taliban-controlled territory after negotiations began between the Ministry of Education and the Taliban. 3 The above are examples of state-insurgent cooperation on rebel governance, that is, coordinated action towards a shared governance objective in rebel-held areas by a state and a rebel group which are simultaneously involved in armed conflict with one another. 4

This raises the question: Under what conditions do states and insurgents cooperate in providing rebel governance?
Rebel governance is 'the set of actions insurgents engage in to regulate the social, political, and economic life of non-combatants during war'. 5 A burgeoning literature unpacks why some rebel groups engage in governance while others do not. 6 These studies make important contributions to knowledge on civil war dynamics. Moreover, several authors acknowledge the existence of state-insurgent cooperation, 7 but there have been fewer efforts to theorise and examine under what conditions state-insurgent cooperation occurs. 8 States and their contenders are often assumed to be 'locked in a straightforward struggle for a monopoly of violence', 9 meaning that we often miss instances of cooperation. This is an omission given that documentary evidence suggests that states and rebels sometimes actively or passively cooperate in providing governance in rebel-held areas. This study addresses this knowledge gap.
We provide theory and evidence to explain under what conditions states and insurgents cooperate in providing rebel governance. State-insurgent cooperation is puzzling because it is costly. Collaboration behind enemy lines may alienate hardliners, undermine efforts of legitimacy-building through governance, reveal sensitive information to the enemy, and cause a loss of autonomy. We therefore direct our theoretical effort towards uncovering the specific factors that tip the strategic scale in favour of cooperation. The purpose of this study is to build -rather than test -theory across two cases that both experienced state-insurgent cooperation. We argue that state-insurgent cooperation on rebel governance is more likely (1) when civilians have high expectations of governance provision, and (2) the conflict parties appeal to overlapping civilian constituencies. We posit that civilian pressure is critical because it increases the costs of neglecting governance, hence incentivising collaboration. We thereby aim to capture the interplay between rebels, civilians, and the state in the provision of governance in rebel-held territory. 10 We explore these propositions in two cases of state-insurgent cooperation: Côte d'Ivoire and Sri Lanka. In Côte d'Ivoire (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011) and Sri Lanka ), a rebel group challenged the state's authority and controlled large swaths of territory for an extended period. Both the Forces Nouvelles (FN) in Côte d'Ivoire and the Liberation Tigers of Tamil Eelam (LTTE) in Sri Lanka developed rebel governance. However, the cases vary on some key factors held to influence rebel governance, including the technology of rebellion, the insurgents' political objectives, UN peacekeeping presence, and conflict intensity. These differences allow us to assess the drivers of state-insurgent cooperation across very different cases. The empirical analysis is based on rich empirical data collected through extensive field research in both countries.
The study makes two contributions to knowledge on civil war. First, we advance new knowledge on the conditions that foster state-insurgent cooperation on rebel governance. Describing and explaining the emergence of collaborative wartime political orders is critical, as such cooperation may influence both civil war dynamics and counterinsurgency efforts. 11 As emphasized in the introduction to this special issue, assessing how states and insurgents co-produce rebel governance can also shed new light on the wider processes of state formation, the legacies of rebel governance, and post-war state-building. 12 Second, we contribute new knowledge on how civilians shape state-insurgent interactions in civil war. Several studies show that civilian mobilisation can influence rebel behaviour. 13 We add to this knowledge by demonstrating that civilians can also pressure the conflict parties to overcome their differences and cooperate on governance. This is an important finding as it suggests that civilian pressure may open conflict resolution possibilities by bringing the belligerents closer on issues short of the main incompatibility.

Rebel governance and state-insurgent cooperation
Civil war is characterised by more than the use of violence by the state and rebel groups. A burgeoning new literature explores the determinants of rebel governance. 14 Rebel governance involves the creation of institutions and practices for organising civilians, such as police forces, taxation regimes, and service provision. A recent example of rebel governance was the Islamic State's creation of a Caliphate governing millions by raising taxes and imposing harsh rules on civilians. 15 Existing research focuses on identifying conditions that make rebels more inclined to govern, such as a dependence on civilian cooperation 16 and international legitimacy. 17 Moreover, research suggests that civilians can influence the character of rebel governance by mobilising against the rebel proto-state. 18 Taken together, there is an emerging consensus on the contextual and rebel grouplevel factors that encourage rebel governance.
The above-mentioned studies make important contributions to knowledge of civil war. Yet, we know less about the state's role in the provision of rebel governance. Existing theories tend to either exclude the state or conceive of the state as an actor that rebel governance undermines. 19 The state's absence in existing scholarship stands in stark contrast to the documentary evidence that suggests that state-insurgent relations are 'frequently characterized by bargains, deals, and norms' 20 that shape violence and governance dynamics. 21 Likewise, while scholars have documented instances of interaction between state and non-state actors under labels such as 'mediated states' 22 and 'hybrid political orders', 23 the role of the state remains under-theorised in rebel governance scholarship. 24 This study aims to fill this gap.
The outcome of concern in this study is state-insurgent cooperation on rebel governance. Although civil war is often conceptualised as an outright contest between the state and a rebel group, and rebel governance as part of a process of 'competitive state building', 25 state-insurgent cooperation can vary along a spectrum from non-existent through passive to active cooperation. We are primarily interested in active cooperation, that is, clear and coordinated action towards a shared governance objective in rebel-held areas by a state and a rebel group involved in armed conflict with one another. 26 We further recognise that state-insurgent cooperation can also be less active and include 'live-and-let-live bargains', and informal or covert cooperation. However, as active cooperation is the costliest and therefore most puzzling outcome, this outcome best suits our theory-building ambition. Below, we outline the costs associated with state-insurgent cooperation and the conditions under which the state and rebels should be more willing to overlook those costs.

The costs of state-insurgent cooperation on rebel governance
State-insurgent cooperation on rebel governance is puzzling because it is costly for both warring parties. Active cooperation is especially costly, as it entails open and direct engagement with the enemy. Although the costs of cooperation may vary across contexts and differ for the state and the rebels, we believe that there are at least four important costs. 27 First, state-insurgent cooperation can alienate or upset hardliners in both parties. Most armed actors consist of moderate and extremist factions. 28 Hardliners are likely to oppose any type of collaboration with the adversary and attempt to block cooperation on governance. This opposition may sow seeds of division within the conflict parties. 29 Rebel groups, for instance, are more likely to fragment when the rebel leadership initiates peace negotiations that hardliners perceive as a betrayal of the rebellion. 30 In the worst-case scenario, state and rebel leaders that are seen as colluding with the enemy are overthrown through an internal coup and replaced by hardliners. The Abu Sayyaf Group (ASG) in the Philippines, for instance, split from the moderate MILF following the latter's decision to govern the Mindanao autonomous region in collaboration with the government. 31 Second, state-insurgent cooperation is costly because it undermines efforts to build legitimacy through governance. Civil war is often viewed as an exercise in 'competitive state building' 32 in which the state and the insurgents seek to outperform one another in the provision of public goods to win the population's support. 33 The provision of governance helps the rebels strengthen their legitimacy and popular support, 34 and extending state governance to ungoverned territories is a key counterinsurgency doctrine. 35 State-insurgent cooperation should undermine the purpose of such competitive state-building, both because it signals that neither of the conflict parties is capable of governing alone, and because the potential legitimacy-enhancing effects of governance would be divided between the parties. Moreover, the state may shy away from cooperation because it risks inadvertently strengthening the rebels' legitimacy. The Kurdish Democratic Union Party (PYD) and its military wing the People's Defense Units (YPG), for instance, gained considerable public support through its service provisionservices the Syrian government helped provide. 36 Third, state-insurgent cooperation may reveal sensitive private information to the enemy. This cost should be most severe for insurgents, who may need to allow state employees to operate in rebel-held areas. Information is critical in war, especially in irregular civil wars where the insurgents rely on stealth and civilian immersion. 37 Collaboration on rebel governance often means that state bureaucrats enter hostile territory and operate behind enemy lines, which allows them to collect valuable intelligence on the rebels' military positions, strength, and resolve. Moreover, state bureaucrats could operate as fifth columnists that seek to undermine the rebels from within, which imposes monitoring costs on the rebels. Stephan, for example, proposed that government administrative staff in IS-held areas could use goslow tactics and propaganda subversion to undermine the Caliphate's political project. 38 Fourth, state-insurgent cooperation generates interdependence between the belligerents and thus entails a loss of autonomy. The perceived benefits of cooperation mean that both the state and the insurgents are forced to exercise greater restraint in interactions with one another to preserve those benefits. This loss of autonomy may limit the belligerents' military options and impose additional monitoring costs to ensure that mid-level commanders and the rank-and-file do not inadvertently engage in actions that jeopardise cooperation. For instance, rebels that depend upon the state's provision of basic services can be pressured by the state through economic embargoes. Likewise, state dependence on rebel provision of state services makes the state vulnerable to rebel non-cooperation.
If state-insurgent cooperation on rebel governance is costly, why do we sometimes see states and insurgents engage in both active and passive cooperation amidst civil war? Below, we outline two factors that are key in tipping the strategic scale in favour of cooperation: high expectations for governance among civilian constituents and overlapping civilian constituencies.

Civilian constituencies and state-insurgent cooperation
A first condition that should foster state-insurgent cooperation on rebel governance is civilian expectations. Civilians are an important source of food, money, recruits, and information in civil war. 39 Guerrilla warfare and counterinsurgency doctrines hold that winning civilian support is critical to successfully fight a civil war. 40 This dependence on civilians bring noncombatants some leverage over how the belligerents govern civilians. Several studies demonstrate that civilians influence patterns of rebel governance by cooperating with, or resisting, the insurgents. 41 We argue that pressure to appease civilian demands can also promote state-insurgent cooperation.
Because civilians are important for the war effort, both the state and insurgents have incentives to seek civilians' political and material support. Such civilian constituencies can be seen as political opportunity structures that both enable and constrain the conflict parties' behaviour. 42 Civilian constituencies do, however, vary in terms of their expectations of governance. Civilian constraints on the ruling authority are the greatest when civilians have high expectations of the provision of a certain form of governance, as such expectations may result in non-cooperation or resistance. 43 While resistance may not always be feasible, evidence suggests civilians have resisted even the most repressive armed groups, including the Islamic State and al-Qaeda. 44 Moreover, high expectations of governance need not manifest as outright resistance, as everyday forms of resistance, as well as credible threats of non-cooperation or resistance, can sometimes be sufficient for influencing armed actors' decision-making. 45 Civilian expectations of governance in rebel-held areas may also be a problem for the state, as these civilian constituencies may more easily shift their allegiance to the rebels if insurgents out-govern the state. In contrast, when civilian expectations of a certain form of governance are low, the likelihood that civilians will demand investments in governance or resist should be lower. In consequence, the state and insurgents should perceive greater incentives to cooperate in providing the coveted governance when civilian expectations are higher rather than lower. Thus, our first proposition is that state-insurgent cooperation is more likely when the conflict parties appeal to a civilian constituency that has high expectations of governance.
Moreover, state and rebel constituencies vary in the degree to which they overlap, that is, the extent to which the state and insurgents compete for the loyalty of the same group of civilians. 46 For example, conflict parties that appeal to the same ethnic or religious group have a high degree of constituency overlap, while conflict parties that mobilise mutually exclusive ethnic identities have a low degree of constituency overlap. Civilian constituency overlap incentivises the conflict parties to exercise restraint in their interactions with civilians and compete for their favour because a loss of support among overlapping constituencies likely implies a boost in support for the enemy. 47 For the state, cooperating in providing rebel governance may be the last available way of reaching constituencies in rebel-held areas. A loss of support among non-overlapping constituencies, in contrast, is unlikely to generate increased support for the enemy, as such constituents have greater barriers for side-switching. That is, much like core voters, mutually exclusive constituents are unlikely to lend support to the enemy, no matter how badly the armed actor performs in terms of providing wartime governance. 48 Constituency overlap should therefore incentivise stateinsurgent cooperation on rebel governance, as cooperation may prevent a net loss of popular support. Thus, our second proposition is that stateinsurgent cooperation is more likely when the conflict parties appeal to overlapping civilian constituencies.

Data and methods
We examine and refine the proposed argument in two case studies of stateinsurgent cooperation on rebel governance in Côte d'Ivoire and Sri Lanka. These cases are interesting to study because active state-insurgent cooperation took place in both civil wars, despite several differences between the cases. Moreover, while we also observe passive state-insurgent cooperation, our primary focus on active cooperation constitutes a hard test for our theory, as active cooperation is costlier than passive cooperation. As Table 1 highlights, state-insurgent cooperation occurred in Côte d'Ivoire and Sri Lanka despite varying technologies of rebellion, insurgent political objectives, UN peacekeeping presence, conflict intensity, and ceasefire arrangements, conditions plausibly related to cooperation. 49 Thus, the comparison builds on a most-different case selection design that leverages the differences between the cases to identify a shared factor that can explain the outcome.
We use the method of structured and focused comparison to guide the analysis, meaning that we pose the same set of analytical questions to both cases. All analytical questions are disclosed in the Appendix. The analysis draws on original data collected through fieldwork in Côte d'Ivoire (eight months) and Sri Lanka (five months). For Côte d'Ivoire, we build on 93 semistructured interviews, 1,500 newspaper articles collected in Ivorian archives, and a unique database of wartime civilian protests. For Sri Lanka, we build on 96 semi-structured interviews with respondents in north-eastern Sri Lanka. Both data collection efforts focused on interviews with key decision-makers, representatives of civil society organisations (CSOs), community leaders, religious leaders, teachers, doctors, ex-combatants, and local government officials, including relevant decision-makers at the national and local level. 51 Moreover, our analysis builds on a review of relevant secondary sources. Each case study proceeds in the following parts: (1) state-insurgent cooperation on rebel governance, (2) the costs of state-insurgent cooperation, (3) civilian expectations, (4) constituency overlap, and (5) limiting the costs of cooperation.

State-insurgent cooperation in Côte d'Ivoire (2002-2011)
The Ivorian civil war broke out in 2002, when the FN rebel coalition captured northern Côte d'Ivoire to overthrow President Laurent Gbagbo, whom they accused of discriminating against northern ethnic groups. The civil war reached a stalemate in mid-2003, which led to the establishment of a UNsupervised demilitarised zone that left 60% of the country under rebel control. 52 Following a period of 'no war, no peace', the civil war ended in massive bloodshed when the 2010 presidential election triggered masselectoral violence and the FN captured the economic capital Abidjan. 53 State-insurgent cooperation on rebel governance in northern Côte d'Ivoire started in 2003 and deepened incrementally. By 2007, state-insurgent cooperation encompassed issues like humanitarian assistance, education, taxation, and economic development. 54 The Ouagadougou Agreement signed in 2007 formalised state-insurgent cooperation under a process of re-deploying the state administration to rebel-held areas, which resulted in the 'co-habitation of two antagonistic powers'. 55 Cooperation was particularly prominent in the education sector. Active state-insurgent cooperation on education began in 2003, when the Ministry of Education (MoE) and the FN coordinated to organise national exams in rebel-controlled areas. 56 Cooperation formalised in 2004, when the government set up a minimal administration in rebel-held areas that worked alongside a rebel education committee known as the Comité de Sauvetage de l'École (C2E). 57 Despite some set-backs, the MoE and C2E continued to coordinate national exams, curricula, and financing for northern students throughout the crisis 58 and helped ensure basic schooling for 400,000 school children. 59 The conflict parties incurred several costs for cooperating. First, government and rebel hardliners attacked their leaders for collaborating with the enemy. Government hardliners like Charles Blé Goudé and Simone Gbagbo, for example, challenged the Minister of Education Michel Amani N'Guessan for bolstering the FN's strength and pouring resources into enemy territory. 60 FN commanders levied similar critique against FN secretary general Guillaume Soro, which sowed seeds of division within an already factionalised rebel movement. 61 Second, by financing part of the FN administration, the government contributed to strengthen the rebellion's perceived legitimacy. The MoE's cooperation with the FN, for instance, helped bolster the rebellion's reputation as a service provider. 62 Third, FN commanders worried that allowing government bureaucrats to enter rebel territory risked revealing sensitive information, which generated monitoring costs. Civil servants relocated to the north testified that the FN often viewed them as spies, 63 and rebel commanders insinuated that redeployed civil servants acted as fifth columnists. 64 Given these costs, what explains state-insurgent cooperation on rebel governance in Côte d'Ivoire?
Citizens in rebel-ruled areas had high expectations for governance. Citizens perceived the pre-war Ivorian state as central to societal development and were accustomed to comparatively high levels of welfare provision. 65 Education, in particular, occupies a central role in the imagined Ivorian nation, as it is seen as the foremost channel for upward social mobility and a political career. 66 The central role of education was further enhanced by the Gbagbo government' ambitious education agenda right before the war. 67 Moreover, while there were regional differences, the pre-war state maintained a prominent presence throughout the entire territory. In fact, before the economic downturn in the 1980s, Côte d'Ivoire presided over an impressive government apparatus. 68 There was also overlap between the government and the FN's constituencies. While the Ivorian crisis is often cast as an ethnic war between northern and southern ethnic groups, there were important nuances to wartime civilian loyalties. Ethnic, religious, political, and regional loyalties jointly determined civilian allegiance in Côte d'Ivoire. Multi-ethnic coalitions were sustained by patronage networks rather than ethnicity, which determined civilian loyalties in unpredictable ways. 69 Overlapping identities meant that there 'existed a strong counternarrative to defining the enemy as a categorical identity group', 70 which incentivised the conflict parties to appeal to broad constituencies including citizens of opposing camps. And despite President Gbagbo's ethnonationalist rhetoric, he presided over a multi-ethnic government, and high-ranking military leaders often emphasised that Côte d'Ivoire was a product of 'a fundamentally mixed nation'. 71 This belief stemmed from Côte d'Ivoire's heterogenous social fabric. As one minister noted, the government had to build a broad ethnic coalition, because 'no single ethnic group is large enough to win the presidency'. 72 High civilian expectations for governance manifested in several ways. First, when fighting abated in 2003, civilians organised to fill the governance vacuum left by the retreating state administration. Local civil society leaders, students, and left-behind state bureaucrats helped organise local schools 73 and the registration of birth and death certificates. 74 Traditional leaders helped resolve local disputes. 75 The most famous civilian initiative was École Pour Tous (EPT), a CSO formed in 2003 to salvage the northern schools. 76 These civilian initiatives provided forums in which government and FN representatives could coordinate governance while deflecting audience costs on CSOs or international partners. EPT, for instance, served a key role in pushing the belligerents to coordinate their education initiatives in the North. 77 Second, civilians voiced their demands on governance by engaging in nonviolent actions like protest marches, strikes, and sit-ins. Figure 1 shows the number of wartime governance protests in Côte d'Ivoire per year. 78 These data confirm that wartime governance protests were common, and peaked in years when the conflict parties deepened their cooperation (in 2003 and 2007-2008). Protest actions reflected Ivorians' high expectations of governance, despite the war. Moreover, protest leaders often called on the conflict parties to work together to ensure basic service provision. For example, in 2004, 2,000 school children in Korhogo protested to pressure the government and the FN to ensure basic schooling. 79 Civilian protests promoted stateinsurgent cooperation because they increased the conflict parties' costs of continued governance failure. Interviewees explained that such costs mattered because both parties depended on popular support; the FN to remain a viable movement and avoid civilian revolts 80 ; the government to prevent the FN from capitalising on civilian dissent to further push northern secession. 81 Both parties took measures to limit the negative fall-out of working with the enemy. A common government strategy was to denounce the FN's creation of a parallel administration, while simultaneously instructing lowerranking bureaucrats to coordinate local governance. State-insurgent cooperation on education constitutes an illustrative example. The Minister of Education Michel Amani N'Guessan frequently appeared in the media to slant the FN's obstruction of the peace process, and made a number of decisions to halt further reintegration of the state and rebel administrations. 82 Nevertheless, N'Guessan continued to approve the MoE's efforts to coordinate school exams with the FN and to redeploy formal teachers to rebel-run schools in the north. Both parties also engaged in camouflaging their involvement, the government by allowing the FN to take credit for what were state-funded rebel schools, and the FN by portraying state school administrators as rebel-affiliated volunteers. 83

State-insurgent cooperation in Sri Lanka (1990-2009)
Vellupilai Prabhakaran established the LTTE as a separate organisation in 1976. The LTTE coerced other armed groups in north-eastern Sri Lanka into subjugation and claimed to be 'the sole representative of the Tamil-speaking population'. 84 The conflict between the government of Sri Lanka and Tamil opposition groups surfaced internationally as an armed conflict in July 1983, and the LTTE started to govern in northern Sri Lanka in the early 1990s. Their objective was to break away from the Sri Lankan state and create an independent state in north-eastern Sri Lanka, known as Tamil Eelam. The LTTE transformed from a small guerrilla organisation in its inception phase to a conventional fighting force throughout the 1990s and 2000s. 85 The LTTE governed north-eastern Sri Lanka until the civil war's end in 2009, but variations existed over time, across locations, and across governance sectors. 86 The LTTE monopolised governance practices in the justice and security sectors. In other sectors, the LTTE allowed other actors -both state and non-state -to provide basic services. Active state-insurgent cooperation on LTTE governance occurred in both the health care and education sectors. 87 The Sri Lankan Ministry of Health continued to facilitate health care services through government hospitals, even though the LTTE made the final decisions in terms of policy implementation. 88 School principals and teachers had to follow the LTTE's instructions but continued to work for the Sri Lankan government. 89 The Sri Lankan government suffered costs for its cooperation with the LTTE. For instance, President Chandrika Bandaranaike Kumaratunga (1994-2001) alienated some hardliners within her government on the issue of service provision in the northeast, which left the government divided. 90 Some government factions pushed for a financial cut-off on service provision in LTTE territories, but these factions did not possess the political clout to force her hand. 91 Nevertheless, they managed to undermine Kumaratunga's efforts, and service provision in the northeast was less substantial than in the other provinces. 92 State-insurgent cooperation also imposed costs on the LTTE. As the rebels became increasingly dependent upon government services, they became vulnerable to political pressure through service embargoes. 93 For instance, in the aftermath of the withdrawal of Indian peacekeepers in 1990, the government imposed an economic embargo on Sri Lanka's northeast encompassing over 60 items of basic consumer, medicinal, and other daily goods. 94 Sri Lankan citizens had high expectations of governance. Pre-war Sri Lanka in several ways resembled a developed welfare state and enjoyed relatively high levels of socio-economic development. 95 Expectations in terms of education and health care prior to the civil war were substantial, including in Sri Lanka's northeast. 96 Sri Lankan citizens had grown accustomed to the relatively stable provision of public services since independence. 97 Citizens of the northern city of Jaffna, for instance, held good education in high esteem. As Bavinck notes about LTTE-controlled Jaffna in 1994: 'Education has always been the main industry in Jaffna but because of the ethnic conflict it seems that this thirst for education and the determination to succeed at exams has intensified'. 98 Little civilian constituency overlap existed in Sri Lanka. During the decades of civil war, ethnicity and nationalism constituted salient aspects of the discourses and identities in Sri Lanka. The different imagined communities were deeply ingrained into the Sri Lankan population's conscience. Therefore, it was very rare for an ethnic Tamil to support the Sri Lankan government, or vice versa. Yet, the LTTE's nationalistic master narrative resonated more among the ethnic Tamil population (who are mostly Hindu or Christian), and far less among the Tamil-speaking Muslims. 99 The Muslim community in some ways resembled an overlapping constituency, but was forcefully expelled from the main LTTE territories in 1990.
From the 1990s onwards, the LTTE worked with the pre-existing institutions of the Sri Lankan state. The Sri Lankan government continued to pay civil servant salaries in LTTE-controlled territory, including for doctors and teachers. 100 Given the LTTE's limited financial resources, the rebels would have struggled to finance such services themselves. 101 Reportedly, insurgent leaders approached government counterparts after the withdrawal of the Indian peacekeepers to ask the government to continue its service provision in the northeast. 102 Hence, it was common for government officials in the northeast to work for two superiors: the formal government and the LTTE. 103 Government hospitals and schools continued to function under LTTE rule. As a result, both the government and the LTTE influenced these governance sectors. For instance, school principals were appointed by the government but controlled by the LTTE. 104 A respondent in northern Sri Lanka explained: '[. . .] my wife, she was working as a teacher, so a government job, so she got paid by the government, but we were living under the instructions from the LTTE. So that was the special situation'. 105 The LTTE overcame the costs of cooperation in several ways. First, the LTTE controlled and claimed ownership over public service provision even though it was funded by government resources. As one respondent put it: 'The bureaucratic and organisational system was (. . .) paid by the government, but they for example were not allowed to put up a Sri Lankan flag. The LTTE only allowed them to put up the LTTE flag'. 106 Thus, the LTTE used symbolism and performances to camouflage its dependence on services paid for by the government. Second, the LTTE leadership's reasoned that they had to accept certain services because of the scarcity of essentially everything at several times during the civil war, particularly medical supplies. 107 In this sense, the LTTE was able to portray the Sri Lankan state both as a provider of basic goods and as the enemy. In sum, the LTTE's quasi-state 'integrally depended on the Sri Lankan state's presence as a ghoulish other -manifested in Jaffna Tamil life as a bomber, soldier, and banner of basic goods'. 108

Comparative analysis
Comparing Côte d'Ivoire and Sri Lanka provides several interesting conclusions about how civilians shape state-insurgent cooperation on rebel governance. Table 2 summarises the findings from the within-case analyses and several alternative explanations.
We find support for the first proposition that high civilian expectations for governance provision are associated with state-insurgent cooperation. Civilian constituencies in Côte d'Ivoire and Sri Lanka generally held relatively high expectations of governance. Both countries experienced a period of considerable state investment in services in the decades before their civil wars that raised civilian expectations, and those expectations persisted during the war. Expectations for education provision (the governance sector with most active cooperation) were particularly high, both because the pre-war Ivorian and Sri Lankan states were quite adept at providing education, and because civilians viewed education as key to upward social mobility. Thus, despite many differences between the cases, high civilian expectations for governance provision constituted a shared characteristic. Furthermore, the within-case analyses demonstrate that civilian initiatives pushed key decision-makers to cooperate with the enemy through two main pathways.
First, high civilian expectations triggered civilian action to demand that the conflict parties cooperate to provide governance. Civilian action served as an accountability mechanism that raised the costs of governance failure and thereby made cooperation behind enemy lines seem like a necessity. This mechanism operated most clearly in Côte d'Ivoire, where civilians in rebelheld areas petitioned the conflict parties to encourage cooperation on service provision. In addition, civilian stakeholders often resorted to nonviolent action in the streets to make their voices heard. In Sri Lanka, high civilian expectations manifested in more subtle ways. Civilians in LTTE-held areas regarded the possibility of sending children to school highly, particularly as a means of social mobility and some future prospects despite the ongoing war. 109 A proactive community of educators and parents continued to bring the importance of education to the LTTE's attention. 110 That is, while Sri Lankans relied primarily on less confrontational forms of voice, such as dialogue and advocacy, Ivorians expressed their expectations of governance in more confrontational ways, including through protest action. A plausible reason for this difference is that the FN were less repressive towards civilians than the LTTE, leaving more space for civilian voice. Nevertheless, we find that high civilian expectations of governance can make inaction more costly and thereby promote state-insurgent cooperation.
Second, we also find that high civilian expectations of governance prompted state-insurgent cooperation through a less direct pathway. High expectations inspired early civilian initiatives of self-governance that helped demonstrate the potential gains of cooperation while providing the conflict parties with a way to absorb audience costs. These civilian initiatives were often driven by CSOs and allowed the conflict parties to claim that they supported non-governmental initiatives when they were in fact coordinating with the enemy. In Côte d'Ivoire, professional organisations like cooperatives, labour unions, and student unions with a long history of mediating governance on the ground led these initiatives. In Sri Lanka, the LTTE established the Tamil Eelam Education Council (TEEC) to coordinate the provision of education with provincial representatives. 111 The purpose was to connect with civil society-based advisory committees composed of parents and educators at the district level and to regulate and supplement the provision of education. 112 Thus, high civilian expectations can help produce stateinsurgent cooperation on governance by lowering the associated costs. We do not find support for the proposition that overlapping civilian constituencies help explain state-insurgent cooperation. While the parties in Côte d'Ivoire sought to appeal to partly overlapping constituencies, this was not the case in Sri Lanka, where the government and the LTTE appealed to separate constituencies based on clear-cut ethnic identities. This suggests that overlapping civilian constituencies may be a sufficient, but not necessary, condition for state-insurgent cooperation on rebel governance. In addition, our analysis suggests an important nuance to this point. Although the Sri Lankan government and the LTTE relied on ethnonationalist group identities, the government still attempted to retain Tamils as part of a united Sri Lanka. In fact, the government perceived state-insurgent cooperation as a key way to remain in touch with Tamils in rebel-occupied areas. 113 Thus, although further research is needed to determine if overlapping civilian constituencies drive state-insurgent cooperation, our analysis suggests that the conflict parties' belief that keeping non-constituencies on board is important may push some conflict parties to cooperate in civil war.
The comparative analysis reveals several additional findings that shed light on the scope conditions of our argument regarding state-insurgent cooperation on rebel governance. First, we find that the costliness of cooperation was higher for the government than the rebel side. In both Côte d'Ivoire and Sri Lanka, state involvement in rebel governance alienated hardliners and helped bolster the rebels' legitimacy. In particular, the state's financing of FN and LTTE-provided services allowed the respective rebel movements to strengthen their parallel governance structures and build a reputation as capable service providers. The costs of cooperation were less manifest on the rebel side. While both the FN and the LTTE cast the government as the enemy, the benefits of technical and financial governance support as a means of procuring legitimacy far outweighed the costs of cooperating with the state. In fact, the LTTE did not necessarily try to hide state-insurgent cooperation. Rather, the arrangements on welfare delivery were framed as a matter of a technocratic development administration. 114 As such, it was presented as delinked from the civil war, which likely helped ease any perceived costs. 115 The same held true in Côte d'Ivoire, where the FN won public recognition for their governance accomplishments but incurred few real costs for co-opting the state's efforts to re-establish itself in rebel-held areas. 116 Thus, civilian expectations of governance provision should be more likely to sway the government than the rebels towards cooperation.
Second, there are reasons to believe that the likelihood of state-insurgent cooperation varies depending on the degree to which the conflict parties hold incompatible positions on the purpose and content of a particular governance activity. Active state-insurgent cooperation in Côte d'Ivoire and Sri Lanka primarily concerned education and health care provision, two governance sectors where the conflict parties did not have competing aims. This allowed the belligerents to frame cooperation as a bureaucratic necessity rather than a concession. Collaborating to provide these services would in all likelihood have been more difficult if the rebels had sought to challenge the government by conforming school curricula to sharia law (as did the Taliban 117 ) or undermine a particular discourse of nationhood (as did ethnic armed organisations in Myanmar 118 ). Collaboration on governance issues that threatened the conflict parties' claims to sovereignty, such as security and justice, or for which the conflict parties held incompatible positions, like citizenship issues, was uncommon in both Côte d'Ivoire and Sri Lanka. Thus, civilian expectations of governance provision should be more likely to elicit state-insurgent cooperation on governance issues on which the conflict parties do not hold incompatible positions.
Third, the FN and LTTE's full territorial control over large areas contributed to push the state and rebels towards cooperation. An important addition to our argument is therefore that the belligerents believe that they cannot provide what citizens desire on their own and consequently prefer cooperation over competition. Rebel territorial control, which made it nearimpossible for the state to provide services in rebel-held areas without insurgent acquiesce, provided such an incentive. Hence, civil wars fought along clear frontlines may provide greater opportunities for state-insurgent cooperation than wars fought along fragmented guerrilla fronts.
This study's purpose is to build rather than test theory across two cases that both experienced state-insurgent cooperation. Several potential alternative explanations, such as insurgent political objectives 119 and conflict intensity 120 vary across the cases and hence cannot constitute necessary conditions for state-insurgent cooperation. However, there are two similarities across the cases that may provide an alternative explanation of the outcome.
A first alternative explanation is that an active civil society spurred stateinsurgent cooperation. CSOs played a critical role in rebel governance in both Côte d'Ivoire and Sri Lanka. In Côte d'Ivoire, CSOs like student and teacher unions were the first to fill the governance vacuum. These initiatives provided a stepping-stone for later state-insurgent cooperation. In addition, there was a large influx of international humanitarian organisations that provided the conflict parties with both incentives for cooperation and a smokescreen behind which to coordinate action on governance. Similar dynamics took place in Sri Lanka, where CSOs took up education and health care governance. Particularly, in response to the 2004 tsunami, more humanitarian organisations entered Sri Lanka to provide essential services in terms of disaster relief. 121 While the comparison thus suggests that a vibrant civil society may help facilitate state-insurgent cooperation, we also find evidence of a feedback effect, whereby cooperation enabled civil society action. In both countries, the government often viewed CSOs operating in rebel-controlled territories with suspicion and accused them of acting as rebel-affiliates. EPTsupported schools in Côte d'Ivoire, for instance, were pejoratively termed 'Malian schools' by the press to highlight their lower quality and alleged propaganda role. 122 Likewise, in Sri Lanka, successive governments were suspicious of CSOs that operated in rebel-held territory. The Tamil Rehabilitation Organisation (TRO), for instance, was first applauded by President Chandrika Kumaratunga Bandaranaike for its relief work after the 2004 tsunami disaster, 123 yet was also scrutinised as a suspected front organisation for the LTTE's political and military wings. 124 Incipient stateinsurgent cooperation helped ease such suspicions and gave both domestic and international CSOs greater operational freedom to work in rebelheld areas. Thus, while we do not fully discount the role of civil society in shaping state-insurgent cooperation, we conclude that civil society reinforced a process towards cooperation driven in large part by grassroots expectations and demands.
A second alternative explanation is that state-insurgent cooperation occurs when rebels and states need to gain or maintain domestic and international legitimacy. 125 This argument bears much resemblance with our first proposition, but places greater emphasis on the role of external audiences. The role of international legitimacy was visible in both Côte d'Ivoire and Sri Lanka, as both rebel groups were hard-pressed to cultivate a positive international image; the FN to improve a reputation tarnished by mercenary violence against civilians and to prevent further French intervention, 126 and the LTTE to build support for an independent Tamil state. 127 Furthermore, we corroborate existing research on domestic legitimacy as a determinant of state-insurgent cooperation, 128 but add to this conclusion by arguing that the need for domestic legitimacy may vary according to civilian expectations. State-insurgent cooperation in Côte d'Ivoire and Sri Lanka was the result of pre-war expectations of governance that heightened the costs of failing to cooperate, costs that would not have been as pronounced had civilians not had high expectations. So, while we find that legitimacy mattered, it mattered precisely because civilians made their voices heard and demanded the provision of services in the midst of civil war.

Conclusion
This article explores how civilian expectations shaped state-insurgent cooperation on rebel governance in Côte d'Ivoire and Sri Lanka. Our analysis demonstrates that high civilian expectations for governance provision can push the conflict parties (and especially the state) to disregard cooperation costs and collaborate in providing governance behind enemy lines. In highlighting how Ivorian and Sri Lankan civilians organised to demand service provision amidst civil war, this article sheds new light on how civilians shape state-insurgent interactions in civil war.
Our purpose is explorative, so future research should test our main argument's validity across a broader range of civil wars. An important task for future research will be to identify the conditions that push the actors towards cooperation rather than competition to appease civilian expectations of governance. Qualitative approaches can aid in examining the mechanisms whereby civilian expectations shape the conflict parties' perceived costs of cooperating behind enemy lines. Quantitative approaches can examine the cross-national determinants of state-insurgent cooperation on rebel governance. As there are no available datasets on state-insurgent cooperation on rebel governance, an important first step would be to collect new crossnational or subnational data.
The study also has implications for scholarship on civil war more generally. Existing studies tend to conceptualise rebel governance as competitive statebuilding or service provision. 129 This understanding of rebel governance needs to be fine-tuned to account for different forms and degrees of stateinsurgent cooperation. As noted by Schievels and Colley, '[t]hat rebels and state must navigate a complex political landscape with many constraints is too often neglected in explaining rebel and state behaviour'. 130 Bringing civilians into the calculus is an important first step to correct such simplifications.
Our findings stress the importance of research on state-insurgent cooperation in civil war. Getting the conflict parties to put aside their differences to organise school exams, negotiate garbage collection, and coordinate health care provision may provide an avenue towards peace negotiations. It may be easier for adversaries to sit down to discuss opening schools than to find a negotiated solution to the war. In addition, state-insurgent cooperation may increase interdependence between the actors, generate positive experiences, reveal shared priorities, and build negotiation procedures, all of which may help ease later and more difficult peace negotiations. State-insurgent cooperation on national exams in Côte d'Ivoire, for instance, helped signal that the government did not want to exclude the north 'more than necessary'. 131 Further research on when and how state-insurgent cooperation comes about may therefore provide key insights for conflict resolution.
Promoting state-insurgent cooperation on rebel governance is also important, as it may ensure basic service provision to thousands, if not millions, of civilians caught amid civil war. Hence, promoting civilian welfare in civil war demands more knowledge of how the state and insurgents can be involved in service provision. This also warrants a word of caution. Recent research shows that rebel governance can undermine post-war trust in state institutions. 132 Thus, while state-insurgent cooperation can help ensure basic services in civil war, it may also contribute to weaken the state in the long-term. Notes 25. Staniland, "States, Insurgents, and Wartime Political Orders," 243. 26. Ibid., 248. 27. The costs of cooperation may differ for the state and rebels. However, it is outside this explorative article's scope to theorise this in more detail. We thus assume that the actors suffer at least some of these different costs and that these costs discourage cooperation. 28. Stedman, "Spoiler Problems in Peace Processes," 11. 29. Pearlman, "Spoiling Inside and Out," 79. 30. Duursma