Settling with Autonomy after Civil Wars: Lessons from Aceh, Indonesia

Autonomy arrangements short of secession have been popular among international actors as a solution to deadly self-deter- mination con ﬂ icts. However, the number of peace agreements incorporating autonomy, while clearly eliminating the possibility of secession, is limited in practice. According to the literature, this is because: (1) ethnic rebels are too weak militarily to extract substantial concessions from national governments; and (2) because the governmental offer of autonomy is not credi-ble to rebels. Drawing upon the single case study of Aceh, this research examines how warring parties overcome these two obstacles and successfully settle ethnonational con ﬂ icts through autonomy. First, although ethnonational con ﬂ icts do not tend to be costly over a single year, they often last for decades, and the accumulation of this cost over years can be substantial. In light of this long-term cost, national governments have incentives to offer substantial concessions to stop the con ﬂ ict from persisting inde ﬁ nitely. Second, rebels could sign the agreement if they perceive that their autonomy is internationally guaranteed in the long term. It is imperative for international policy makers to consider these points when they try to help resolve ethnonational con ﬂ icts through autonomy.

• Rebels can overcome their fear that the national government would later rescind the autonomy arrangement if they believe it is internationally guaranteed.
• International actors are encouraged not only to help implement the agreement immediately after the conflict, but also to consider becoming a guarantor of the autonomy arrangement in the long term.
Self-determination conflicts are one of the most intractable conflicts in the world, with the potential to last for decades. Many of them take the form of violent ethnonational conflicts, namely conflicts between an ethnic minority rebel and a national government over the territory the former considers as their homeland. To resolve these conflicts, autonomy arrangements short of secession have been advocated among practitioners as a compromise between rebels demanding self-determination and national governments unwilling to relinquish their sovereignty (Caspersen, 2017;Wolff and Weller, 2005).
Yet according to the civil war literature, warring parties are unlikely to reach autonomy arrangements because of two obstacles. First, national governments are unwilling to offer substantial concessions. This is because rebels in ethnonational conflicts tend to be poorly armed and incapable of inflicting heavy casualties on governments (Cunningham et al., 2009). As a result, the latter do not face a hurting stalemate. The issue is compounded because rebels believing the territory as theirs tend to demand significant concessions from the government (Toft, 2006). As a result, no mutually agreeable solution is available. Second, when national governments offer autonomy, rebels have difficulty believing that the government is truly committed to this offer. They fear that the autonomy arrangement might, one day, be unilaterally abrogated by the national government (Fearon, 2004). Hence, rebels find it difficult to accept the offer of autonomy.
However, in contrast to our clear understanding on these two obstacles to an autonomy arrangement, specific ways to overcome these obstacles have not been explicitly explored in the literature. When seeking a way out, how do warring parties overcome these two obstacles and reach an autonomy arrangement? This paper aims at filling this gap in the literature (Caspersen, 2017;da Rocha, 2019;Zartman, 2000).
Exploring this question requires hypothesis generation and examination of the causal mechanisms operating in the conflict resolution process by analysing secondary and primary evidence. For both, case studies have advantages over quantitative research since the former allow us to observe what caused an outcome in a specific case (George and Bennett, 2005;Gerring, 2007). Drawing on the single case study of Aceh, Indonesia, I will theorize ways to overcome the two obstacles identified above. First, while rebels in a typical ethnonational conflict can neither pose a serious military threat nor inflict serious economic and military cost to the national government (Cunningham et al., 2009), the cost for the government would accumulate over years as the conflict tends to continue for decades. In light of this potentially infinite future cost, government leaders with a longtime horizon have reasons to substantially concede to indigenous rebels, even if they are not facing a hurting stalemate at present. This argument extends Zartman's (2000) ripeness theory which posits that warring parties need to face 'pain' from the conflict to pursue conflict resolution through negotiation. This paper shows that a weaker 'chronic ache' from the conflict is equally worth getting rid of as 'pain' from the conflict when one realizes that they have to live with this dull ache, namely small annual costs, for a long time unless the conflict is resolved. Second, ethnonational rebels can overcome their fear that the government would later unilaterally abolish autonomy if they are convinced that international actors which have sufficient leverage over the government will guarantee autonomy through their commitment to the peace agreement.
This paper proceeds as follows. After offering definitions of relevant concepts, I will first show that the number of autonomy arrangements emerging out of negotiations to settle violent ethnonational conflicts is limited in practice despite its popularity among international policy makers. Second, to explain this, I will examine the civil war literature, identifying the two obstacles to reach an autonomy arrangement. Third, I will provide my arguments to overcome these obstacles before exploring them in a case study. The case of Aceh is analysed to identify how actors were able to overcome these two obstacles. The conclusion discusses the academic and policy implications of the arguments.

Definition of relevant concepts and existing literature
This paper examines how warring parties in ethnonational conflicts successfully negotiate an autonomy arrangement. Ethnonational conflicts are defined as conflicts between ethnic minority rebels and national governments where the former demands some kind of self-determination such as autonomy or independence over what they consider their homeland (Gurr, 1993). The minority group the rebels claim to represent 1. is territorially concentrated, usually forming the majority of the population in the disputed territory (Toft, 2003). Typically occurring in peripheral regions of the country, many of these conflicts (have) lasted for decades as is exemplified by the Karen in Myanmar, Xinjiang in China, and Aceh and Papua in Indonesia (Buhaug et al., 2009;Fearon, 2004).
Autonomy, a type of territorial self-governance, refers to a special status for a specific territorially delimited entity with a wider range of powers than other territorial entities in the same country (Wolff, 2013). As such, granting the status of (wider) autonomy to a specific region 'need not affect the general institutional organisation of a state' (Wolff and Weller, 2005, p. 13). Even though autonomy is also granted during the interim period prior to self-determination, this paper focuses on autonomy arrangements that eliminate the possibility of secession, namely autonomy short of independence.

Autonomy as a solution to ethnonational conflicts
There are both proponents and opponents to autonomy as a solution for ethnonational conflicts. According to proponents, autonomy helps fulfill the aspirations of indigenous ethnic groups to control their own homeland while national governments do not need to relinquish sovereignty (Caspersen, 2017). Moreover, autonomy could be reassuring to minorities, who would otherwise worry that the national government would jeopardize their security and control of their areas (Rothchild and Hartzell, 1999). However, according to opponents, territorial self-governance reinforces both the group identity and the governance capacity in each subnational entity. As a result, it becomes much easier to mobilize people along ethnic lines, leading to stronger demands for secession (Bunce, 1998;Cornell, 2002). Facing these contrasting views, Cederman et al. (2015) try to understand under what conditions territorial self-governance is likely to prevent self-determination conflicts. They find that postwar territorial self-governance should be combined with meaningful participation in the state's executive to prevent another war.
In practice, autonomy has been a popular option among policy makers to settle ethnonational conflicts. As Nina Caspersen (2017, p. 16) observes, 'the preferred option [for the international community] remains for self-determination to be realized through various forms of autonomy arrangements'. Autonomy, once seen as 'at best a highly unusual tool of state construction, or at worst a highly dangerous one' during the Cold War became more popular after it ended (Wolff and Weller, 2005, p. 1).
Yet there are only a limited number of peace agreements resolving ethnonational wars through autonomy short of independence. This is not because the number of violent ethnonational conflicts is small. According to a recent dataset for the years 1945-2012, there have been 73 self-determination movements which experienced a level of violence high enough to be classified as civil wars (Sambanis et al., 2018).
But from Nina Caspersen's comprehensive study of peace agreements that were signed between 1990 and 2010 to settle self-determination conflicts, only six autonomy arrangements (without a future, implicit or informal option of independence) were found to be used to settle conflicts whose level of violence reached the threshold of civil wars: Chittagong Hill Tracts, Bosnia, Eastern Slavonia, Aceh, Mali, and Mindanao (Caspersen, 2017;Sambanis et al., 2018). Among them, the settlement in Eastern Slavonia was nothing but a virtual capitulation (Caspersen, 2017). Also, examining the older period between 1945 and 1997, Rothchild and Hartzell (1999) found only seven cases where autonomy was used for civil war settlement: Nagorno-Karabakh, Chechnya, South Ossetia, Abkhazia, Transnistria, Nicaragua, and the Philippines. Among them, Nicaragua was not a selfdetermination war. Additionally, four of these cases are usually considered to be unrecognized states, whose status was achieved based on a capitulation. In sum, there are not many cases where substantial peace negotiations lead to autonomy short of independence. Why?

Obstacles to reach an autonomy agreement
The civil war literature has identified two reasons ethnonational wars are difficult to be resolved through autonomy. First, researchers have argued that the level of concession the government is willing to make does not match ethnonational rebels' expectations. On the one hand, national governments are not willing to offer meaningful concessions to peripheral rebels because these rebels are typically very weak militarily (Cunningham et al., 2009), even though they are likely to endure because the government has more difficulty in operating militarily in the periphery (Buhaug et al., 2009). Furthermore, governments might prefer to merely contain these insurgencies rather than seriously engage with them through war or dialogue. This is because the strategy of containment does not require much political and military capital. As rebellion in peripheral areas does not pose threats to the political survival of politicians, they prefer to devote their time and money to other, more pressing, domestic issues (Mukherjee, 2014). This is particularly so in medium capacity states such as Indonesia, which are neither developed nor failed states (Mukherjee, 2014). Generally speaking, ethnonational conflicts are not painful enough for governments to consider that it faces a hurting stalemate (Zartman, 2000). This lack of incentives to offer substantial concessions is aggravated in a multiethnic state due to the reputational concern that conceding to one ethnic group would set a precedent and induce similar demands from other ethnic groups (Toft, 2003;Walter, 2009) even though the validity of this theory has been contested (Fujikawa, 2017;Nilsson, 2010;Sambanis et al., 2018).
On the other hand, rebels conceive of themselves as indigenous to the local area and hence being entitled to retain and control their soil. Typically constructing their territory as indivisible, they demand significant concessions, often even independence (Toft, 2003;. Moreover, when ethnic groups are actively excluded and discriminated against by the national government, rebels are not easily satisfied by governmental concessions as ethnic exclusion reinforces their belief in the zero-sum nature of the conflict (Wucherpfennig et al., 2012). In sum, there is no mutually agreeable solution that both governments and rebels prefer over continuing war. In other words, there is no range for bargaining (Fearon, 1995).
Second, even if warring parties find a mutually agreeable arrangement, rebels would then fear that the national government would later renege the arrangement. It is a type of commitment problem where one party 'would have an incentive to renege on the term' (Fearon, 1995, p. 381). Even if governments offer autonomy to rebels when the former is weak, rebels anticipate that governments would renege on the promise once they are in a stronger position again (Fearon, 2004). After all, 'the ultimate authority for determining jurisdictional disputes is typically vested in the national government, inevitably creating a centripetal force in politics' (Lake and Rothchild, 2005, p. 127). As a result, even when the bargaining range exists, this commitment problem, namely the fear that the government is not committed to autonomy in the long term, prevents rebels from signing a peace agreement.
Overcoming these two obstacles of no bargaining range and the commitment problem would be necessary for conflict resolution. In the next section, I will theorize how to overcome them.

The argument: how to overcome the obstacles to settlement
Drawing on the case of Aceh, I argue that it is possible to overcome the two obstacles for settlements discussed in the previous section. First, extending Zartman's ripeness theory, I argue that accumulating long-term conflict cost gives governments incentives to offer substantial concessions under certain conditions.
Zartman's ripeness theory posits, based on a cost-benefit analysis, that warring parties prefer conflict resolution through negotiation when they perceive that the conflict has reached a painful deadlock ('hurting stalemate') and that there is a way out (Zartman, 2000). This argument superficially suggests that governments would not be interested in negotiations in a typical ethnonational conflict that only incurs a small annual cost since they do not face a hurting stalemate.
However, even if governments incur only small annual conflict cost, this cost would accumulate and could become substantial in the long run if they have to pay it for decades. Conflicts incur accumulating economic, political, or military costs including financial cost for counterinsurgency operations, the death of the population, and international criticism for human rights violations. If national governments do not heavily discount the future, they would have incentives to try to end conflicts with substantial concessions now than pay for the accumulating cost indefinitely. In other words, a continuous chronic ache from a conflict is worth getting rid of, even if the conflict is not causing an acute pain at present. Another image in line with this argument is 'the pebble in the shoe', a term used by Ali Alatas, Indonesia's long-serving foreign minister, to describe the East Timor conflict (Alatas, 2006). Unlike a big stone in the shoe which causes instant pain, a small pebble might only cause some ache. However, if they wear the shoe for a long period, they are likely to sit down and take the pebble out. Thus, national governments do not need to face a hurting stalemate and feel the pain of continuing the war to seek negotiations. However, they need to feel the ache from the conflict and consider that it is likely that the conflict, and thus the ache, would continue. While this point about chronic aches is applicable to any conflict, this is particularly relevant to ethnonational conflicts since many of them last for decades.
In line with Zartman's (2000) theory, the subjective perception of the long-term cost is more important than the objective condition. Since policies on ethnonational conflicts are usually made at the top level, the perception of national leaders is most relevant. Moreover, their subjective perception is potentially influenced by the economic, political, or military conditions they face. For example, a government facing multiple rebels might be willing to accommodate weak rebels to focus on other fronts (Nilsson, 2010). Similarly, a government suffering from an economic crisis must be more sensitive to the conflict cost as was the case with the Habibie administration on East Timor.
Under what conditions is this argument applicable? First, instead of heavily discounting the future, national leaders need to have a long time horizon and care about the longterm conflict cost in addition to recognizing that the conflict will likely be prolonged. 2. Indeed, one of the few autonomy agreements reached in the wake of an ethnonational conflict was in Mindanao in 1996 under Philippine president Fidel Ramos, who had a long time horizon. 3. More generally, one study argues that new democratic leaders are more likely to concede because they 'have a sufficient time horizon to see the prudence of concessions' (Beardsley, 2010, p. 397). In contrast, national leaders retiring soon might want to focus on other more pressing issues rather than ethnonational conflicts which are costly only in the long term. Second, the government needs to be incurring non-negligible costs from the conflict. This condition is not met if the government, or a veto player within the government such as the military, benefits from war economically or politically. For example, they might exploit the territory or benefit from a war economy (Ballentine and Sherman, 2003). Politically, an authoritarian regime, the military, or a nationalist leader might believe that continuing the war would serve their image as a 'strong leader' or 'guardian of the nation'. The government might also rest their legitimacy on fighting and excluding these peripheral ethnic groups as was the case with successive Islamic regimes in Sudan vis-a-vis the Southern Sudanese (Wucherpfennig et al., 2012). Finally, political leaders need to be more concerned about the cost directly deriving from the war over the potential reputational cost of triggering similar demands by other ethnic groups through concession to one ethnic group (Fujikawa, 2017).
Regarding the second obstacle, two ways are known to solve commitment problems in conflict resolution: powersharing pacts and third-party guarantees (Walter, 2002). The former allows parties to have some political power to defend themselves. However, a simple autonomy deal, a type of powersharing pact, might not be enough for ethnonational rebels as it could be always unilaterally retracted by national governments (Lake and Rothchild, 2005). Thus, autonomy should ideally be combined with meaningful participation in the state's executive in postconflict settings (Cederman et al., 2015). However, since the ethnic groups that rebels represent are usually peripheral, it is highly unlikely that they would be allowed to have substantial power at the national government level. Even though autonomy alone is insufficient to guarantee the agreement in the long term, meaningful participation in the state's executive is unlikely in practice.
The other solution to commitment problems is third-party guarantees. The role of third parties in resolving self-determination conflicts has been studied extensively in the literature (Caspersen, 2017;da Rocha, 2019;Schneckener, 2008), but there has been much less analysis on their role in guaranteeing autonomy or future referendums in the long term. Caspersen briefly discusses various short-term and long-term international guarantees, praising the Belfast Agreement in Northern Ireland, which took the form of an international treaty between the British and Irish governments, as 'a novel way of guaranteeing autonomy' (Caspersen, 2017, p. 32). Wolff refers to potential forms of international guarantees on the autonomy arrangement in Transnistria. The involvement of other states as guarantors, mediators, or observers is one option, but concluding a bilateral or multilateral treaty to guarantee the agreement 'could prove useful and effective in assuring the parties' (Wolff, 2011, p. 870).
There are precedents where autonomy was guaranteed internationally. Most notable is the Aland Islands, whose autonomy has been protected under an agreement between Finland and Sweden. Another case is Hong Kong, whose autonomy was guaranteed under the Sino-British Joint Declaration. 4. I argue that rebels can overcome commitment problems if they are convinced that their autonomy is guaranteed internationally. As a guarantor, third parties are ideally influential and strong governments or international organizations which have both the capacity and will to deter the national government from rescinding their commitments. In addition to their involvement in the immediate implementation phase of the agreement, guarantors can lay down the agreement in a treaty or sign the agreement as guarantors or witnesses to show their long-term commitment to it. While it is ideal if agreements take the form of treaties, the perception and conviction of ethnonational rebels that their autonomy is guaranteed internationally is more important than the exact form of the international guarantee. This is again in line with Zartman (2000), whose framework emphasizes subjective perception.
The case of Aceh highlights these two dynamics at play. First, president Susilo Bambang Yudhoyono had a long time horizon, thinking the war would persist indefinitely with a military approach. Considering the substantial long-term conflict cost, the administration seriously negotiated with the secessionist rebels, GAM (Free Aceh Movement). Second, GAM believed that the EU and the international community would guarantee the autonomy arrangement and democracy in Aceh.
Aceh is chosen for five other reasons. First, since the peace agreement in 2005, its provisions were largely implemented, and Aceh has remained peaceful. Second, Aceh was able to secure a high degree of autonomy through serious negotiations rather than through capitulation on either side (Caspersen, 2017). Third, it is a 'typical case' of ethnonational conflicts: (1) Aceh is located in Indonesia's periphery; (2) the Acehnese form the majority of the population in the province of Aceh; (3) the conflict had lingered for decades; and (4) GAM had no real odds of liberating the whole of Aceh, let alone threatening the capital.
Fourth, commitment problems were acute for the Acehnese since they already had the experience of their autonomy arrangements being dishonored by the national government. Fifth, Indonesia is a multiethnic state where governments are supposed to be unwilling to offer concessions to rebels, worried about setting a precedent for other regions (Toft, 2003;Walter, 2009). From this standpoint, Aceh is a hard case to settle.
Meanwhile, it is important to note an idiosyncratic condition contributing to the conflict resolution in Aceh: the catastrophic tsunami in December 2004. Yet the conflict was resolved not just because of the tsunami. The literature suggests that the tsunami was relevant, but not necessarily decisive in the conflict resolution of Aceh (Aspinall, 2005;Morfit, 2007). Even after the tsunami, support for negotiating with the rebels was far from unanimous within Indonesia, with substantial opposition to the peace talks among the military and parliamentarians (Aspinall, 2005). On GAM's part, throughout the negotiations after the tsunami, its leadership often seriously considered walking away from the negotiations (Kingsbury, 2006). Moreover, the case of Sri Lanka offers evidence that tsunamis do not automatically end ethnonational conflicts (Tunc ßer-Kilavuz, 2019). Overall, the conflict resolution in Aceh cannot be simply traced back to the tsunami.

Aceh: ending an ethnonational conflict through autonomy
Aceh's first rebellion against the Indonesian government took place between 1953 and 1962. The rebel's leader, Daud Beureueh, who once served as the head of the All-Aceh Association of Ulama, was disappointed that Sukarno, Indonesia's first president upon independence, retracted his promise in 1947 that Islamic law would be implemented in Aceh (Aspinall, 2009). However, this rebellion was not of a separatist nature, aiming to replace Sukarno's secular Indonesia with an Islamic Indonesia together with forces in other regions. Still, special territorial status was granted to Aceh in 1959, which eventually ended the rebellion in 1962. Yet the status was largely unimplemented (Sulaiman, 2006). Sukarno lost power after an attempted coup in 1965, and the authoritarian and centralized New Order regime, led by Suharto, emerged. In the 1970s, while the economic situation in Aceh was comparable to the national average, large reserves of liquefied natural gas (LNG) was found there, and extraction started. But it did not benefit the Acehnese people as much as they desired (Aspinall, 2009).
Disgruntled by the marginalization of Aceh and accusing Jakarta of solely profiting from Aceh's natural resources, a new rebellion began in Aceh in 1976. It was waged by a separatist movement, GAM, whose leader Hasan di Tiro, an Acehnese intelligentsia who participated in the first rebellion, believed that the Acehnese people were entitled to the right to self-determination (Aspinall, 2009). However, the Indonesian military soon crushed this rebellion. By the 1980s, GAM leaders left Aceh for Sweden as refugees.
But, from 1989, the separatist activities by GAM recurred. This time, the Indonesian military cracked down on separatists harshly. This was accompanied by significant human rights violations.
When Suharto's authoritarian regime collapsed in May 1998, vice president Bacharuddin Jusuf Habibie took over the presidency, initiating a democratization process and relaxing censorship (McGibbon, 2004). Exploiting this new political space, urban citizens and university students in Aceh started demanding justice for past human rights violations. The Habibie administration apologized for the abuses but could not deliver justice because of the still important role the military played in national politics at the time. As a result, the demand of the Acehnese escalated into an insistence on an independence referendum (Aspinall, 2009). GAM also increased its activities.
In addition, as part and parcel of the democratization process, the Habibie administration not only implemented decentralization across Indonesia but also granted a special status to Aceh and an option of independence to East Timor (McGibbon, 2004). This policy shift was a significant departure from Suharto, who rejected even special autonomy for East Timor (Alatas, 2006). Only against this backdrop, was it possible for subsequent governments to offer autonomy and self-governance to GAM.
As Abdurrahman Wahid replaced Habibie in October 1999, huge rallies demanding an independence referendum for Aceh took place in various locations in Aceh (Aspinall, 2009;Miller, 2009). Wahid sought accommodation, investigating human rights violations and becoming the first Indonesian president to negotiate with GAM. The latter led to a ceasefire agreement in May 2000, but it was not effective as both the military and GAM violated it. Towards the end of the Wahid era, 'it [was] often estimated that about 80 per cent of Aceh's villages [were] under GAM control or influence' (International Crisis Group, 2001, p. 5).
After Wahid was ousted in July 2001 due to corruption scandals, nationalist Vice-President Megawati Sukarnoputri took over the presidency (Crouch, 2010). Special autonomy laws for Aceh and Papua, discussed since 1999, were finally adopted in August 2001, but the Megawati administration was not willing to fully implement them (Miller, 2009). After the collapse of another ceasefire agreement in December 2002, the Megawati administration ended the negotiation process, imposing martial law in the province in May 2003. The Indonesian military operations cost almost US$230 million during the one-year martial law period up to May 2004 (Sijabat, 2004).

Yudhoyono and Kalla, GAM, and the resolution of the conflict
Yudhoyono did not face a hurting stalemate in Aceh when he was inaugurated as president of Indonesia in October 2004 (Aspinall, 2005). After his predecessor Megawati imposed martial law on Aceh, the military had clearly gained the upper hand vis-a-vis GAM, regaining control of most villages in Aceh (Crouch, 2010). One of the key government negotiators in the peace process, Sofyan Djalil, even stated that 'militarily GAM was defeated' (quoted in Morfit, 2007, p. 119). Nevertheless, the Yudhoyono administration sought negotiation with GAM even before the tsunami.
This move was, indeed, an extension of the new vice president Jusuf Kalla's previous attempts to contact GAM while he was still a minister under the Megawati administration (International Crisis Group, 2005). Under Kalla's mandate, his deputy, Farid Husain, had the opportunity to meet with Martti Ahtisaari, former president of Finland (Merikallio, 2006). After Yudhoyono became president, Ahtisaari, internationally known as a capable mediator, invited the two opposing parties to a meeting in January 2005, which the Indonesian government accepted before the tsunami (Aspinall, 2005;Morfit, 2007). Against this backdrop, the tsunami struck Aceh in December 2004, killing more than one hundred thousand people there.
While nothing substantial was agreed in the first round of the mediation process in late January, the second round in late February became a milestone in ending the conflict because GAM made it clear that they were willing to consider a solution within the framework of Indonesia. Until then, GAM had always aimed at independence (Aspinall, 2005).
When the second round began on 21 February, Malik Mahmud, GAM's 'prime minister', insisted that the term 'special autonomy' was not acceptable to GAM. In the eyes of Ahtisaari, the dialogue seemed to be stalled (Merikallio, 2006). Yet for GAM, the term 'special autonomy' had the connotation of broken promises in the past where various forms of autonomy had been granted to Aceh without substance (Djuli and Rahman, 2008;Kingsbury, 2006;Merikallio, 2006). As an alternative, within the GAM leadership, the concept of 'self-government', akin to the Aland Islands and Hong Kong, as more genuine autonomy without such a connotation had been discussed for some time (Crouch, 2010;Kingsbury, 2006;Merikallio, 2006).
Facing a deadlock in the negotiations, GAM decided during this second round to adopt 'self-government' within Indonesia rather than outright independence as their goal, at least within the framework of the present talks (Aspinall, 2005;Djuli and Rahman, 2008). Ahtisaari agreed to this new term, and the biggest impasse was resolved (Kingsbury, 2006;Merikallio, 2006).
What motivated this concession? Battlefield losses on the ground and the effect of the tsunami were certainly important (Aspinall, 2005;Merikallio, 2006), but these factors had existed at the beginning of the negotiations. Another crucial factor might have been a meeting on 19 February between the GAM leaders and ambassadors from several countries including Australia, Malaysia, and the United States. The ambassadors pressed them to accept the territorial integrity of Indonesia (Kingsbury, 2006). Since independence is meaningful only when it is recognized by other states, this event dashed any hope GAM had for independence (Aspinall, 2009). Moreover, through the negotiations over the past few years and international assistance after the tsunami, GAM was successful in internationalizing the conflict and was now unwilling to give up the fruits of their labour by discontinuing the talk (Aspinall, 2009). The importance of having the agreement guaranteed internationally can be seen as an extension of this reasoning.
Most contentious between the third and fifth rounds of negotiations was the issue of local political parties to enable GAM to transform into a political party. Political parties in Indonesia are required by law to be nationwide precisely in order to deter separatists such as GAM to gain any political hold (Aspinall, 2005). The government initially claimed that local parties were not possible, partially because the executives could not change the law unilaterally (Aspinall, 2005;Kingsbury, 2006;Merikallio, 2006). However, GAM was not willing to settle with anything less than the right for the Acehnese to establish local political parties. GAM seriously considered ending the negotiations without a settlement, but an agreement was ultimately reached (Djuli and Rahman, 2008;Kingsbury, 2006;Merikallio, 2006). The comprehensive peace agreement, the Memorandum of Understanding (MoU), stipulates that the 'GoI [Government of Indonesia] will create . . . the political and legal conditions for the establishment of local political parties in Aceh in consultation with Parliament' (MoU: 1.2.1). 5. The MoU was signed on 15 August 2005. Under the MoU, the government of Aceh would have extensive authority 'except in the fields of foreign affairs, external defence, national security, monetary and fiscal matters, justice and freedom of religion' (MoU: 1.1.2.a). Furthermore, the implementation of this agreement would be monitored by an Aceh Monitoring Mission (AMM) composed of the European Union and some ASEAN member states. While the agreement did not take the form of a treaty, Ahtisaari signed the agreement as a witness, possessing authority to make a binding decision on disputes arising from implementation (MoU: 6.1.c).
Unfortunately, the Law on the Governance of Aceh, which was to incorporate the MoU provisions into Indonesian laws, did not fully reflect the peace agreement since its draft was weakened by bureaucrats and politicians in Jakarta (Crouch, 2010;International Crisis Group, 2006). Yet the most impor- Settling with Autonomy elections. The MoU conclusively settled the 30-year conflict which had an overall estimated death toll of 12,000 to 20,000 (Aspinall, 2009).

The Indonesian government: why offer significant autonomy?
Contrary to the theoretical arguments that governments are unwilling to offer any substantial concessions to peripheral ethnic groups, the Indonesian government conceded significantly to GAM by allowing the formation of local political parties. This was 'an offer no other Indonesian government had ever made to any other dissenting group in post-independence history' (Mietzner, 2012, p. 99). The literature on the Aceh conflict suggests that the success of the peace agreement was the result of the leadership and commitment to peace by Yudhoyono and Kalla (Fujikawa, 2017;Miller, 2009;Morfit, 2007). Yudhoyono and Kalla, unlike the previous nationalist administration of Megawati, were mainly concerned about the cost of the Aceh conflict rather than the reputational costs of concession (Fujikawa, 2017). Both of them were known to prefer nonmilitary approaches to solving conflicts (Aspinall, 2005;Mietzner, 2012). After the tsunami, they were further motivated to seek peace not only personally but also due to international pressure to settle the war (Mietzner, 2012;Miller, 2009). For Yudhoyono, whether the government approached the war in Aceh through dialogue or through military measures, it was merely a means to an end to maintain the territorial integrity of Indonesia (Yudhoyono, 2014). In his memoir, he defended himself twice vis-a-vis those opposed to peaceful settlements in Aceh. In the first instance, to those who insisted that GAM should have been crushed, Yudhoyono points out that 'they forget that military operation had already gone on for more than 30 years and the problem of Aceh did not end. Victims fell down for both parties' (Yudhoyono, 2014, p. 191 (Yudhoyono, 2014, p. 359). Elsewhere, Yudhoyono argued that the 'conflict had gone on too long; there were too many victims on both sides. And it was expensive, costing us about $130 million per year in security operations' (quoted in Morfit, 2007, p. 125).
Here, we can see that Yudhoyono emphasizes two points: the impossibility of eliminating GAM, and the cost of the conflict. However, each alone seems to be an insufficient reason for extensive concessions. First, as Mukherjee (2014) argues, the national government could simply opt for a policy of containment if a military stalemate was the sole problem.
Second, objectively speaking, the yearly cost alone does not seem to be sufficient to justify the extensive offer of having local political parties in Aceh. The annual cost of $130 million is not negligible, but it is not significant for a country whose GDP in 2005 was more than $285 billion, 13% ($37 billion) of which was equivalent to the national government expenditure. 6. Indeed, Yudhoyono and Kalla acknowledged that the conflict in Aceh did not receive attention during the presidential election in 2004 (Morfit, 2007). After all, there were many other pressing issues in Indonesia (Mietzner, 2012;Morfit, 2007). Furthermore, public opinion in Indonesia was in favour of Megawati's military approach (Crouch, 2010;Mietzner, 2012). As was mentioned already, even after the tsunami, there was staunch criticism of the peace talks among parliamentarians from various parties and the military, and the latter did not stop its military operations in Aceh after the tsunami. The situation seems to have been exactly what Mukherjee describes for medium capacity states: too many other important issues to care about small internal wars (Mukherjee, 2014). The government was not facing a hurting stalemate (Aspinall, 2005).
Why was the government serious in resolving the conflict despite the lack of a hurting stalemate? The conflict, even though not hurting, incurred non-negligible cost every year. Considering it had continued for almost three decades and was likely to continue for the foreseeable future, the accumulating cost would be a significant burden for the state in the long run. Those who discount the future heavily might not care about such a long-term cost, but Yudhoyono, who had a long time horizon, did. It is telling that he was citing both the length and the cost of the conflict together when he defended his decision to settle the war peacefully. It was this accumulation of yearly costs that matches the big concessions offered by the government. Even though Yudhoyono seemed to have a long time horizon even before he came to the office, his hope to govern the country for the next ten years might have been another reason to have a long time horizon (Beardsley, 2010). Overall, the combination of a military stalemate and non-negligible yearly cost made the Yudhoyono administration seek conflict resolution. Even without 'pain' from the conflict, the government was willing to concede substantially to the rebel to get rid of the continuous 'chronic ache'.

GAM: how to overcome commitment problems
GAM had always clung to independence as their aim. Yet they finally decided in February 2005 that they might be willing to settle for something less than independence. It was a key decision that led to the conflict resolution in Aceh, but GAM immediately faced commitment problems.
In the third round of the peace talks, they expressed concerns that there was no guarantee that future Indonesian parliaments or dictators would not renege on the agreement (Merikallio, 2006). After the agreement was converted to domestic law, Malik Mahmud was quoted as saying that 'they [the next national government] can begin to demand new laws and prevent us from implementing our own wishes regarding natural reserves' (quoted in Merikallio, 2006, p. 243).
GAM leaders overcame this commitment problem by believing that the peace agreement was guaranteed internationally. Most notably, the model of 'self-government' for GAM was the wide-ranging and internationally-guaranteed autonomy arrangements in the Aland Islands and Hong Kong (Crouch, 2010;Merikallio, 2006). As we have seen, GAM conceptualized 'self-government' as qualitatively different from the past autonomy arrangements. Correspondingly, two local GAM leaders in Aceh, in an interview with Edward Aspinall, a prominent political scientist specializing in Indonesia, told him how they were persuaded to obey the decisions by the exiled leadership in Sweden. The GAM leadership in Sweden mentioned the effect of the tsunami and the international community's expectation for conflict resolution. Additionally, according to the summary of interview notes by Aspinall, the local leaders explained: Moreover, we heard from Sweden 7. that the EU had guaranteed that Aceh would have democracy, although it would remain in the framework of Indonesia. If it really was going to be a democracy, then we would accept it (quoted in Aspinall 2009, p. 235).
Likewise, Irwandi Yusuf, who served as GAM representative in AMM and later won the Aceh gubernatorial election in 2006, insists that 'the peace in Aceh was not generated only by the Indonesian authorities and GAM; it was also generated by the international community, who have [sic] obligations to watch and to warn the parties when they see symptoms of pathologies in the peace process' (Yusuf, 2008, p. 81). Finally, recalling the time when GAM leaders were worried about how to bind future Indonesian governments to this arrangement, Ahtisaari says, 'I tried to explain that no such guarantees could be given. That is how democracy works. But I also said that the agreement would now be a national obligation and that if anyone began trying to alter it, I too would have to intervene' (quoted in Merikallio, 2006, p. 140). The international guarantee was matched by deeds. Peter Feith, a Dutch diplomat serving as the head of the AMM, pressured the government to adopt the Law on the Governance of Aceh reflecting the MoU (Mietzner, 2012). Moreover, Feith and Ahtisaari regularly showed their presence in Aceh over the next few years (Mietzner, 2012). Overall, the GAM leaders, who were unwilling to settle for anything less than self-government, clearly viewed it as essential that their deal was internationally guaranteed.

Conclusions
Through the single case study of Aceh, this paper has investigated specific mechanisms warring parties in ethnonational conflicts could use to overcome the two obstacles for conflict resolution identified in the literature: no bargaining range, and commitment problems. For the first obstacle, I have shown that even if they are not facing a hurting stalemate, national governments with a long time horizon have incentives to seek a way out from ethnonational conflicts and to offer substantial concessions to ethnonational rebels. While Zartman's concept of a 'hurting stalemate' emphasizes pain, I have argued that a weaker chronic ache, even if not painful in the short term, is equally worth getting rid of when one considers its long-term impact on the body politic. This is an important revision and extension of Zartman's theory. Second, commitment problems can be overcome by international guarantees of the agreement. GAM, clearly facing commitment problems, managed to agree to the settlement because they considered that the autonomy agreement was guaranteed internationally.
However, these findings do not imply that every ethnonational conflict can be resolved easily. The ethnonational conflict in Papua, Indonesia, is a case in point. Yudhoyono did not offer substantial concessions to Papuans despite its decades-long conflict. The cost of the Papuan conflict waged by very weak rebels seems too low to be considered costly even in the long run (Fujikawa, 2017).
But why do political leaders have a long time horizon? One simple answer might be that they are transformational leaders, caring about principles rather than just their immediate self-interests (Bass, 1999). Yet one can also speculate that Yudhoyono sought the solution relatively early on in his term because he hoped to govern the country for the next ten years. In other words, if the conflict had not been terminated, it might have continued to affect his own administration (Beardsley, 2010).
Also, is democracy a prerequisite for autonomy arrangements? Peace negotiations in Aceh began against the backdrop of Indonesia's democratization. Similarly, democratization in the Philippines since 1986 was 'an important turning point' for Mindanao's peace process (Talib, 2019, p. 80). In general, one can make a good case that democratic regimes are more likely to accommodate opponents (Walter, 2009). But one recent quantitative study does not find a strong support for this argument (Sambanis et al., 2018). The regime type, however, probably has an important explanatory power in the success and failure of the survival of the autonomy agreement. Southern Sudan's 1972 autonomy arranged under an authoritarian regime in Sudan and Eritrea's 1952 autonomy imposed on the Ethiopian imperial regime by the United Nations were both encroached and subsequently rescinded in 1983 and 1962 respectively by the same authoritarian governments which embraced the agreements (Iyob, 1995;Johnson, 2016). The recent events in Hong Kong cast further doubts to the prospect of autonomy arrangement in authoritarian states.
In light of this paper's findings, what can international policy makers do to help resolve deadly ethnonational conflicts? First, by highlighting the long-term costs, they can encourage governments to consider settling them immediately rather than fighting indefinitely. Second, they should take commitment problems seriously and discuss with rebels what could be done to alleviate their fears. The findings of this paper should not be seen as if rebels were Global Policy: Next Generation (2021)  Settling with Autonomy automatically assured of their autonomy just because peace agreements are transformed into treaties or signed by mediators. These actions signal the commitment of international actors to autonomy only when they have shown their determination to help resolve the conflict through sincere engagement with the parties involved and investment of their time and resources in conflict resolution. This suggests that international policy makers should not only assist in the short-term implementation of peace agreements (Walter, 2002) but also continue to be involved in the long-term sustenance of them.

Data Availability Statement
Data sharing is not applicable to this article as no new data were created or analysed in this study.

Notes
I would like to thank Kazuo Ohgushi, Kiichi Fujiwara, Mark Hoffman, and participants in a seminar at the University of Tokyo and in a Junior Scholar Symposium at ISA 2019 for their helpful comments. I would also like to thank the editors and anonymous reviewers for their detailed and helpful feedback. An earlier version of this paper appeared in Japanese in 2016 in the Journal of the Association of Political and Social Sciences (Kokka-Gakkai-Zasshi), volume 129 (5-6), pp. 526-601 as '"Dochaku-no-tami-gata" Funso to Sono Shuketsu: Aceh Funso no Case Study [The Sons-of-the-Soil Conflicts: Conflict Resolution in Aceh]'.