Self-Defense Militias, Death Squads, and State Outsourcing of Violence in India and Turkey

ABSTRACT What explains the variation in states’ nonstate partners in civil warfare? States often use nonstate actors to do what their regular military forces cannot do well – navigate the local population. Some of their nonstate partners are ordinary civilians, while others are battle-hardened fighters with a rebellious or criminal past. The choice of proxy carries serious implications for the patterns and effects of violence during civil war, human rights, and international security. This article is the first to disaggregate the nonstate counterinsurgents and offer an explanation for why and how states use each type. It brings together the politics of collaboration with the politics of exploitation. The article shows that the state’s use of nonstate proxies is shaped by the supply of willing collaborators, the state’s ability to exercise control over them, and the trade-offs underlying the use of the different types of nonstate actors. The empirical evidence used to support this argument comes from a novel, comparative study of Turkey’s counterinsurgency campaign against Kurdish separatists and India’s counterinsurgency against Kashmiri separatists. The original data were collected through fieldwork in the disputed territories of each country.

comprise inexperienced civilians. This article disaggregates nonstate proxies and offers an explanation of why and how states use each type. It focuses on the supply of collaborators, the state's capacity to impose control over them, and the tradeoffs in working with different types of proxies.
The state's choice of proxy depends in part on the capabilities of prospective recruits and the degree of control the state can reasonably expect to establish over them. Nonstate actors with significant experience wielding extra-institutional violence ("veterans") are more militarily skilled than their counterparts who have little or no previous combat experience ("rookies"). For the same reason, veterans are more dangerous and difficult to rein in. Veterans are, therefore, more likely to be used in areas where the battleground needs call for highly targeted offensive operations and where the state possesses the means to control them. They often constitute "death squads" and work alongside special operations forces of the military or police. 11 The militarily unskilled rookies are more likely to be used in areas requiring static defensei.e. incorporated into "self-defense militias" 12and where the state exercises relatively less control. While the comparative advantages and disadvantages of veteran and rookie proxies shape the "demand" for one over another, the "supply" of collaborators is a necessary condition of outsourcing. Where willing collaborators are scarce, states may resort to coercion for their recruitment and management.
The proposed theory is developed by combining the "most-different systems" research design and within-case analysis of India's (1988India's ( -2003 and Turkey's (1984Turkey's ( -2002 counterinsurgency campaigns. This article is the first work that systematically compares these two important cases. The fine-grained analysis of within-case variation enables systematic process tracing, which boosts the internal validity of the argument. By focusing our attention on the similar features of an otherwise dissimilar pair of cases, the most-different systems design helps to eliminate other possible explanations. 13 It allows us to identify the common logic of violence outsourcing that plays out in countries that have relatively few similarities. The observation of an analogous violenceoutsourcing process across India and Turkey provides a robust confirmation of the argument and suggests its broader applicability for other non-weak states.
Turkey and India have distinct histories of state-building: the former was the heartland of the Ottoman Empire while the latter was a British colony. At the time of the insurgency, Turkey was a semi-democratic state under military tutelage, while India had a functioning civilian democracy. A long-time member of NATO, Turkey has generally maintained a strong alliance with the United States, while India has historically pursued nonalignment and relative distance from Washington. Since the early Kurdish uprisings of the 1920s and 1930s, Turkey had not faced significant rebellion until the rise of Kurdish ethnic separatism in the late 1970s. India experienced rebellion almost continuously: in the northeast, beginning soon after the country gained independence in 1947, in the Punjab in the 1980s and early 1990s, and in Kashmir in the 1990s and 2000s. Despite these differences, Turkey and India responded very similarly to their respective insurgencies. The strikingly similar outsourcing patterns across India and Turkey and the variation within each country across space and time provide strong support for the proposed framework.
The empirical findings are based on fieldwork in Srinagar (in Jammu and Kashmir) and Delhi, in India; Islamabad, in Pakistan; and Diyarbakır, Ankara, and Istanbul, in Turkey. Interviews and archival research were conducted during summer and winter research trips between 2012 and 2016 at these locations, as well as in Washington and London. Interviewees include state and military officials, journalists, local experts, civilian victims and witnesses of the conflict, and former rebels. In addition to providing first-hand accounts, the interviews and archival research helped to identify and discern the credible secondary materials from a plethora of partial, inaccurate, and ideologically biased accounts. The interview and archival sources are supplemented with reports from non-governmental organizations, newspapers, and scholarly works. Due to the highly sensitive nature of the subject, many of the interviewees are kept anonymous.

Explaining Proxy Selection
The terms "pro-government militia," 14 "partisan," 15 "paramilitary," 16 and "civilian defense force" 17 are currently used in the literature to designate state-sponsored armed nonstate agents. There is considerable overlap and ambiguity among the existing terms, as the scholarship has not yet reached a consensus on these concepts. Nevertheless, what all of these groups have in common is that they comprise agents who are employed outside the regular military channels and have no prior professional ties to the state's military or security apparatus. 18 Figure 1 displays the global prevalence and distribution of pro-government militias between the years 1981 and 2007. 19 At least 64 percent of the 332 organizations identified had direct links to a state institution. 20 Considering the challenges of collecting accurate data on violent groups' relationships with state institutions, the number likely underestimates the incidence of state outsourcing of violence.
While the existing literature does not directly tackle the problem of proxy variation, emerging scholarship on rebel cooptation suggests two potentially useful mechanisms: fragmentation and patronage. I outline each and then propose a more comprehensive approach that combines the strategic logic of exploitation with the politics of collaboration.

Fragmentation and Patronage
Recent studies of imperial warfare suggest that when local elites are politically fragmented, states enlist collaborators more easily. 21 The finding that fragmentation facilitates the recruitment of collaborators is also supported by the literature on insurgent fratricide. 22 The logic is that deadly internal  -government Militias, 1981-government Militias, -2007  disputes, be they among the local elites or the insurgents themselves, incentivize the weaker factions to seek an alliance with the state. This argument predicts higher rates of collaboration with the state among nonstate actors that need physical protection. A related supply-side argument points to the role of patronage in proxy selection. The relationship between the state and the proxy involves the latter's provision of irregular military service in exchange for privileges supplied by the former on mutually beneficial terms. 23 This argument predicts higher rates of state-proxy collaboration where the state is able to provide selective benefits to potential collaborators. 24 The classic patronage-based relationship is illustrated by the British strategy in the tribal territory of the North-West Frontier of colonial India. The British presence in the Punjab beginning in the mid nineteenth century deprived the North-West Frontier Pathan tribes of "a field for raids and booty." 25 Cognizant of the impending security and economic threat, the British created an irregular armed force comprising local Pathan tribesmen. Their salary and pension were designed to keep the tribes from raiding and looting "the pride and backbone of British India" (i.e. the Punjab). 26 Fragmentation and patronage certainly enable state-proxy collaboration. However, neither of these mechanisms adequately accounts for the type of proxies states will be willing and able to work with in times of war. Existing research is also silent on the role of state coercion in proxy creation and use. Ranging from torture to a promise of a reduced prison sentenceor in combination with positive inducementscoercion often allows the state to recruit the more reluctant adjuncts 27 and to ensure their compliance during the counterinsurgency campaign.

Proposed Framework: Control, Collaboration, and Comparative Advantage
States' selection of nonstate proxies is a function of the former's capacity to impose control over the latter, the supply of willing collaborators, and the comparative advantages and disadvantages of working with veterans and rookies. In confronting an insurgency, the state's ideal partner is a highly skilled fighter who possesses deep knowledge of the insurgent network as 23 Idean Salehyan, "The Delegation of War to Rebel Organizations," Journal of Conflict Resolution 54/3 (2010), 493-515. 24 Lee J.M. Seymour, "Why Factions Switch Sides in Civil Wars," International Security 39/2 (2014), 105. 25 Mason in Charles Chenevix Trench, The Frontier Scouts (London: Jonathan Cape 1985), xiv. 26 Trench,The Frontier Scouts,1. 27 The reluctance to serve the state may be the result of ideological or very practical considerations. The nonstate proxies often become the favored targets of the insurgents, who see them as a serious threat to the insurgent operations and legitimacy yet, at the same time, easier targets than their regular counterparts due to inferior equipment and training.
well as local "cultures, perceptions, values, beliefs, interests and decisionmaking processes." 28 Regular soldiers are highly skilled fighters, but possess limited local and insurgent knowledge. The disconnect between the soldiers and the local population may be traced back to the increased mechanization of militaries since the First World War, 29 as well as modern military training practices that involve detaching the soldier from his or her familiar environment. When a state partners with a group that is not part of its military or police forces, the former confronts the classic principal-agent problem. 30 The state's capacity to manage this problem is significantly shaped by its spatial reachterritorial controlacross the theater of war. Figure 2 illustrates the different configurations of territorial control. In Zone A, the state exercises full control. Zone B represents the insurgent-dominated area, whereas Zone C is the contested region where both sides actively compete for influence. Zone D represents an area that has been largely neglected in the burgeoning civil war literature. It is a zone that is neither fully controlled nor, for the time being, actively contested. It is a so-called "brown" area 31 where neither side exercises a high degree of penetration. Zone D exists because neither the state nor its rivals can, or necessarily desires to, be present everywhere at all times. Among the nonstate alternatives, the ideal partner for the state is a skilled fighter with insider knowledge of the insurgency and its logistics network. The disadvantage of using this type of a proxy is that their fighting skills are precisely what make them dangerous and difficult to control. The cost of losing control of skilled fighters can be unpalatably high. The global prominence of the standing army model owes much to the havoc wreaked by the skilled fighters-turned-bandits of the Thirty Years' War. The "regular" armed forces represent the shift from the delegation to the centralization and professionalization of violence. Unless they are weak or collapsed, modern states are more likely to use skilled fighters when and where they can supervise them.
Nonstate actors that possess the desired local knowledge and have significant experience wielding extra-institutional violence, whether of a criminal or rebel nature, may be harnessed to counter the insurgents' advantages over the state. These veterans' comparative advantage is in performing specialized, offensive, and highly targeted tasks that take advantage of their high mobility and combat skillse.g. as "death squad" agents. Consequently, they are likely to be used where the insurgents are actually located (Zones B and C). We should expect them in Zone C, where the state can exercise control over their activities. Using death squads in Zone B requires the state to create robust control mechanisms, such as embedding these groups firmly within special operations units of regular military or police forces. Where the state exercises full control (Zone A), or where the insurgents are not yet present (Zone D), states do not require highly targeted offensive operations and, consequently, veterans.
The next-to-ideal nonstate partner for a state is the rookie possessing high local knowledge, albeit poor fighting skills. The rookies' comparative advantage is in static area defense aimed at cutting insurgent logistical links or preventing area infiltratione.g. as members of a self-defense militia. 32 Rather than directly targeting insurgents, self-defense militias typically "nibble" at the insurgency and defend specific areas from rebel incursion. 33 Consequently, rookies are used mainly to facilitate the operation of state forces in contested regions (Zone C) or to deny the insurgency access to areas where neither side has effective control (Zone D). In Zone B, rookies 32 Both rookies and veterans are also sometimes used as auxiliary forces in offensive operations and, consequently, integrated into the regular military formations. Prolonged conflict may "professionalize" rookies, thereby transforming them into veterans. In such cases, they may be used by states for not only defensive but also for offensive operations. Moreover, while the rookies' inferior combat skills make them less effective than their veteran counterparts in offensive operations, they may be easier to mold or manipulate to commit the kinds of violence veterans could refuse. John Mueller, "The Banality of 'Ethnic War'," International Security 25/1 (2000), 42-70. 33 Carl von Clausewitz, On War, trans. James John Graham (London: N. Trübner 1873), 480-81. are helpless against the insurgents. Not only can their government-provided weapons easily fall into rebel hands, but also, given the high rates of defection in insurgent-dominated areas, 34 they may become a fertile source of rebel recruits. Remnants of self-defense militias (once created in Zone C or D) may persist in Zone A, but only because demobilization of nonstate actors is usually slow and costly. 35 The categories of "veteran" and "rookie" are ideal types. Most nonstate actors fall close to, but rarely completely on, each end of the martial prowess spectrum. Both veterans and rookies usually receive basic arms and limited training. They are distinct from mere informants in that the state expects them to engage in armed combat if and when necessary. Counterinsurgency operations involving these proxies can be either covert or overt. The former enables the state to carry out violence that may be deemed illicit or illegitimate by domestic or foreign observers. The latter allows it to signal having the support of at least some elements within the civilian population or the rebel ranks. Typifying state-supported nonstate violence with plausible deniability are the Shabiha forces used by Bashar al-Assad's government during the Syrian Civil War. These irregulars have carried out executions, drive-by shootings, and sectarian attacks. On the other hand, some states purposely publicize their partnership with armed groups. The Russian state openly partnered with select Chechen ex-rebels to demonstrate that it enjoyed the support of some Chechen groups. Moreover, nonstate counterinsurgents may be co-ethnics of the insurgents, as in the aforementioned Russian case, or of the ruling elite, as in the Syrian case. There is some evidence to suggest that the former may be more effective than the latter. 36 Counterinsurgency is a function not merely of the state's battlefield needs, but also of the social conditions in which counterinsurgency takes place. An explanation of the recruitment and management of nonstate counterinsurgents is incomplete without the politics of collaboration. The supply of collaborators is a necessary condition of outsourcing. States usually enjoy highest levels of collaboration from the local population in areas where they have significant territorial control (Zone A). 37 Everywhere else, they must actively seek collaborators to make their strategy work.
Rookies and veterans have a variety of interests that influence their willingness to collaborate. While the availability of veterans usually depends on opportunities for prior combat or criminal experience, states can be creative 34 Kalyvas, The Logic of Violence in Civil War, 196-97. 35 The Civil Defense Patrols, which served as civilian adjuncts to the Guatemalan army, continue to operate across the country and commit abuses two decades after they were supposed to have been discontinued under the 1996 Peace Accords that ended the civil war. Regina Anne Bateson, "Order and Violence in Postwar Guatemala," PhD dissertation, Yale University, 2013. 36 Lyall, "Are Coethnics More Effective Counterinsurgents?" 37 Kalyvas, The Logic of Violence in Civil War, 196. In Kalyvas's "selective violence in civil war" model, Zone A is captured by Zones 1-2.
in finding veterans. For example, during the civil unrest in Ukraine in 2014 the government used sportsmen-criminals, the so-called titushki, for "committing violence either alongside or under the watchful gaze of the notorious special police unit." 38 The state's previous ties to local ethnic groups or kinship and village networks could also lead to collaboration opportunities. Where "voluntary collaboration born of genuine attachment" 39 or "alignment of incentives" 40 is lacking, states may attempt to create it through coercive measures. The political arrangements states make with potential proxies affect the latter's willingness and capacity to collaborate with the central government. The next two sections offer case studies that gauge the explanatory power of the proposed framework.

Village Guards and Islamist Contras in Turkey
In the 1970s, Turkey's major cities were on fire. Right-wing and left-wing political organizations violently clashed on the streets of Ankara and Istanbul. The 1970s were also a time when a wave of Kurdish youth from predominantly rural southeastern Turkey arrived in the major cities to study and work. Among them were the founders of a militant organization that combined Marxism-Leninism with Kurdish separatist nationalismthe Kurdistan Workers' Party (PKK). The goal, as articulated by PKK leader Abdullah Öcalan, was to liberate "Kurdistan" from Turkish colonialism. Its attainment required not only attacking the institutions and agents of Turkish repression, along with their foreign supporters, but also the reorganization of the feudal power structure of Kurdish society. 41 The PKK inaugurated its violent campaign on 15 August 1984 with attacks on two gendarmerie (local police) stations in Siirt and Hakkari. The PKK based its logistics network on local production and relied heavily on the local population for recruits, intelligence, and shelter. The region in which the PKK built territorial control was predominantly rural, characterized by low population density. 42 The difficulty of policing dispersed population settlements, combined with rural norms of solidarity and honor and an economy based on subsistence farming, makes rural areas especially conducive to insurgent influence. 43 The rugged terrain provided the PKK with concealment, while proximity to the Syrian and Iraqi borders offered "strategic depth"the space to withdraw, regroup, and respond to state offensives.
The PKK's rapid advances in southeastern Turkey surprised and initially elicited denial in Ankara. Turkish officials were, at the time, fixated on Greece. Prime Minister Turgut Özal referred to the PKK as a "group of marauders," no more worthy of comment than the old bandits of the Anatolian mountains. As it became evident that the Kurdish separatists constituted a well-organized guerrilla force with local support, the army began to increase its presence in the region. On 19 July 1987, the Turkish Parliament granted a state of emergency rule (Olağanüstü Hal [OHAL]) in 11 provinces that faced substantial PKK activities and violence. An OHAL governor was appointed to coordinate the security forces (military, gendarmerie, and police) while exercising certain quasi-martial law powers, such as restricting press and freedom of expression and association, over roughly six million people. 44 As the Turkish military transferred additional troops to the region and reorganized its activities, the initial insurgent strongholds in the rural peripheries of Diyarbakir, Siirt, and Hakkari became vigorously contested (Zone B à Zone C).
The Turkish army was a conscription-based NATO military designed predominantly for inter-state warfare under the conditions of the Cold War. Its weaknesses vis-à-vis the guerrillas soon became apparent. The PKK was far more skilled at navigating the mountainous terrain and the co-ethnic rural populations. The army overshadowed the guerrillas in terms of sheer numbers and firepower, but found it difficult to isolate and engage them to its advantage. The PKK dictated the terms of military engagementit was able to avoid detection, gain materiel and intelligence support from the local population, and attack military targets of its choosing.

Rookie Proxies: The Kurdish Village Guards
In April 1985, the Turkish government amended the Village Law of 1924 authorizing the provincial governors to appoint "temporary" (paid) and "voluntary" (unpaid) village guards 45 in the contested provinces (Zone C). Figure 3 displays  The new legislation allowed the Turkish military to train and arm civilians to guard their villages against PKK militants. Some also served alongside the regular soldiers when the military sent troops near their villages. The village guards received two weeks of training 47 at military bases. 48 Their syllabus covered the use of weapons, wireless communication technology, and basic tactics, such as keeping the heels down while exchanging fire with the enemy, as the PKK snipers notoriously targeted the heels.
The village guards used in operations typically walked in front of the regulars while carrying automatic rifles and radios. 49 Most of them did not wear uniforms, though some preferred the uniform because they were otherwise the first ones to be targeted by the PKK. 50 These rookie proxies were tied to the army and gendarmerie (rural military police), 51 not the police. As one Turkish Ministry of Interior official explained: "The civilian authority often did not even know when, where, and how they [the village guards] were used." 52  When the Turkish state first started recruiting village guards in the late 1980s, they were stationed primarily in the contested areas between the state and the PKK (Zone C). As the OHAL region expanded, the PKK strongholds (Zone B) began to shrink. The Turkish state supported the village guards "both financially and logistically" 53 and made no effort to keep the system a secret. 54 The existence of the village guard system, composed of ordinary Kurdish civilians, was meant to signal the Kurdish people's support of the Turkish state. The PKK, threatened by local collaboration, called for the village guards' "mass destruction" in its 1987 "Decree on Village Guards." Turkish soldiers taken prisoner were often exchanged, while captive village guards faced summary execution. 55 The PKK carried out numerous violent raids and attacks specifically in the villages whose members had volunteered to become part of the village guard system. The majority of the civilian casualties between the years 1986 and 1988 were of individuals living in villages whose members participated in the village guard system. 56 To fill the village guard ranks, the Turkish military initially reached out to local clans that were known to be sympathetic to the state. Ankara had been suspicious of Kurdish political mobilization ever since the Sheikh Said uprisings shook the republic soon after its founding. The state's intelligence agencies closely monitored the local power struggles between the Kurdish clans and noted their pro-state and anti-state sentiments. 57 Army conscripts provided a continuous flow of valuable intelligence about clan elite fragmentation in the region. 58 This intelligence enabled the armed forces to identify the clans most likely to support the central government in its struggle against the separatists. In the late 1980s, state officials made informal deals with influential clan leaders (aghas) from the sympathetic clans to guard their own villages against PKK militants.
The first clans to collaborate with the state were those identified with the right and far-right (ultra-nationalist) political parties. Some of them "were already in conflict either with the PKK directly, or with local clans which enjoyed PKK support." 59 By acting as intermediaries, the aghas made it possible for the Turkish military quickly to recruit thousands of village guards in the contested areas of the region. Among the most notorious was the Jirki clan in Hakkari, whose chief was still wanted for the killing of six gendarmes in 1975. The tribal chief struck a deal with state officials and, after a token court appearance, raised a force of Jirkis as village guards around the Beytussebap region. Another Hakkari chief demanded the 53  release of his son from prison before supplying village guards. 60 The aghas usually collected and were responsible for distributing the monthly salaries of their clan's village guards, which reinforced their local influence.
In order to prevent the insurgency from spreading further to the north and northeast of the conflict zone, the Turkish state decided to widen the scope of the village guard system. By the early 1990s, the clan-based approach to recruitment had been depleted, and so the military began directly approaching villagers about joining the self-defense militias. 61 The state applied this strategy both in the contested regions (Zone C) and in the rest of the Kurdish-majority areas in the north and northeast of the conflict zone which had not been so affected by the fighting (Zone D). Figure 4 maps the geographical expansion of the village guard system in the early 1990s from Zone C to Zone D.
By expanding the village guard system to the periphery of the conflict zone, the state sought to prevent the PKK from acquiring new territory. The main function of the rookie proxies in these areas was to guard the territory against insurgent incursion. They were primarily tasked with protecting their villages as well as the roads and mountain paths the PKK could use to traverse the territory. Zone D served as a buffer zone between the insurgent-filled and the insurgent-free areas, and the village guards served as the buffers.
Patronage played only a partial role in the state-proxy relationship in the 1990s. The official village guard salary was significantly higher than the average per capita income in the region, which was plagued by high unemployment and underemployment. The monthly stipend of a village guard in 1992, for example, was approximately US$250 in areas where the annual per capita income was about US$400. 62 However, when the financial incentives did not work, military commanders threatened the civilians with village evacuation 63 or made village guard service "an informal requirement for return" after the evacuation had taken place. 64 There were cases of villagers being tortured for refusing to collaborate with the state. 65 The patronage hypothesis over-predicts collaboration. The Turkish state used selective material inducements successfully during the late 1980s and the 1990s to recruit thousands of village guards. The Kurdish regions of Turkey were among the poorest in the country. The relative wealth of the central state meant that it could afford to purchase cooperation widely among the Kurdish villages. However, patronage was insufficient to meet the state's proxy needs. The PKK made collaboration with the state very costly by brutally targeting village guard members, their families, and even entire villages.
Fragmentation also provides an important but partial account of collaboration. Kurdish clan elites who were threatened by the PKK's anti-feudal ideology collaborated with the state. However, when the agha channel dried, the state did not shy away from applying coercion. Whereas in the late 1980s the Turkish state relied on fragmentation and patronage to recruit and manage the rookie counterinsurgents, in the 1990s the benefits of fragmentation diminished and patronage was supplemented by intimidation.
The village guards earned "a reputation for being the least disciplined of the Government's security forces." 66 Hundreds of them were involved in crimes, including drug trafficking, corruption, theft, rape, and murder of rivals falsely identified as terrorists. Some, willingly or unwillingly, collaborated with the PKK. 67 The state turned a blind eye to many of these offenses so long as the village guards facilitated its primary military objective of wresting territorial control from the PKK. The impunity for criminal offenses served as a reward for loyalty by the state to its proxy and, thus, facilitated the collaboration.

Veteran Proxies: The Kurdish Hizbullah and Ex-rebels in Gendarmerie Intelligence and Counter-Terrorism (JITEM)
On 16 December 1990, the OHAL governor was granted the authority to evacuate residential areas and transfer the local population. An official from the Diyarbakır branch of the Human Rights Association estimates that roughly 4000 villages were forcefully evacuated in the 1990s. 68 OHAL officials report 905 villages and 2523 hamlets (a total of 378,335 forced migrants) evacuated in and around the provinces under emergency rule by 1997, while some nongovernmental organizations estimate the number of internally displaced persons in the region to be between one and four million in 2005. 69 Population resettlement is a classic tool of counterinsurgency. The logic is that, "since the government does not have the capability of dispersing and projecting its military strength to cover these important remote settlements . . . it must bring the inhabitants of these villages to areas which it can directly control." 70 The cities and large towns of southeastern Turkey were the strongholds of the Turkish government (Zone A), and consequently the Kurdish villagers were moved into the urban areas where they could be more easily monitored. 71 The village evacuation policy accelerated the urbanization of the Kurdistan region. 72 Along with the heavy inflow of Kurdish villagers into the cities of southeastern Turkey was a corresponding inflow of PKK followers and sympathizers. Beginning in 1990, the PKK began contesting the state's authority in what was formerly a state stronghold (Zone A à Zone C). 73 A former subprovincial governor and official with the Turkish Ministry of Interior describes: "As a result of the village evacuations, the PKK's military power in the rural areas was broken. But, it resulted in the urbanization of the PKK. The migrants from the villages formed the support bases of the PKK in the cities. They took over areas where the government previously had full control." 74 The towns of Cizre and Bismil (Diyarbakır Province) had several PKKoccupied "liberated zones." 75 In Cizre, the Nur and Cudi neighborhoods were under PKK control. 76 Multiple armed confrontations involving the PKK took place in the Bağlar and Suriçi districts of the city of Diyarbakır. 77 The Petrol district of the city of Batman had a strong PKK presence. 78 The PKK was highly active in Hakkari Province, especially in the towns of Yüksekova and Şırnak. 79 The latter was transformed into a city as the army built up its capacity there against the PKK with a large military installation. About 950 houses in Şırnak became unusable after the military assaulted them while targeting PKK supporters. 80 As the insurgency spread to the cities, the Turkish military began collaborating with individuals who were highly skilled at wielding illicit violencemembers of the militant Islamist organization called Hizbullah. 81 The Kurdish Hizbullah in Turkey was a Sunni Islamist group that traced its origins to the Iranian Revolution. 82 Some youths in Diyarbakır and the surrounding provinces had traveled to the Iranian city of Qom for short-term religious training. The pro-Iranian Kurdish youths began to organize in Diyarbakır at the Menzil bookstore. However, after the 1980 Turkish military coup, the organization splintered. The moderate faction pursued political and social activities, while its radical counterpart, now organizing at the Ilim bookstore, argued that armed warfare was the only way to bring about an Islamic revolution in Turkey. What also separated the radicals from the moderates was that the former deemed religion, not ethnic identity, to be the main source of unity. Incidentally, the Ilim faction leader, Huseyin Velioglu, is known to have studied at the Ankara Political Science Faulty at the same time as Öcalan. 83 The Ilim faction of the Hizbullah (henceforth referred to as "the Hizbullah") was "a mainly urban phenomenon" 84 operating in cities such as Nusaybin, Batman, Diyarbakır, and Van. 85 It began its violent urban campaign against PKK members and sympathizers in October 1991. By 1992, it had become "the second most violent and ruthless organizationafter the outlawed PKK." 86 The locals began to refer to the organization as "Hizbul-contra" due to its suspected connection to the state, as evidenced by the high degree of immunity it enjoyed from the secular Turkish government and military. 87 78 An official from the Turkish Ministry of the Interior. 79  The state used the Hizbullah for death squad activity, pitting "the Kurdish Islamists against the atheist PKK." 88 The secular Turkish military and the Islamist organization made scandalous partners. Consequently, the Turkish military pursued covert collaboration with select Hizbullah leaders who agreed to work with the state on the common goal of defeating the PKK. State agents also infiltrated the organization. The rank-and-file members are unlikely to have known of their leaders' collaboration with the state. 89 Much of the Hizbullah violence was skillfully coordinated. Each operation was carried out by two to four militants, some of whom were under the age of 18. 90 Their targets were PKK insurgents and suspected PKK sympathizers. 91 The militants "operated in broad daylight in the mainly Kurdish cities of southeastern Turkey. People who opposed the state's policy were being killed at the rate of two a day; in all, more than a thousand people were killed in street shootings from 1992 to 1995." 92 The state allowed the death squads to operate with impunity in public areas such as busy streets and coffee houses. Even when the assassins were apprehended by the public and taken to police custody, "they were let go with bogus claims, like they ran away from the second floor of the police station. The people who reported assassins to the police were often then assassinated by the Hizbullah." 93 The year 1993 witnessed the largest number of incidents caused by Hizbullah militants. In the town of Silvan, about 300 people were assassinated that year. In the city of Diyarbakır, about 800 individuals and, in the city of Batman, approximately 300 people were killed. 94 As a former parliamentarian who headed the investigation of the state's relationship with the Hizbullah put it, "This is at a time when the area was so securitized that if a kid threw a stone at somebody, he would be immediately surrounded by plainclothes police. But these assassins could walk in broad daylight, shoot someone in the head, and walk away." 95 The Hizbullah militants were notorious for their execution and torture methods. Their signature execution style was shouting the Tekbir before shooting their victims with one bullet in the head. Their weapon of choice was usually a Makarov pistol. 96 The Hizbullah's signature method of torture was the domuz bağı (hogtie). Corpses were frequently found in refrigerators. 97  usually took place in the morning or evening, when the targets were traveling to or from work. 98 Hizbullah "contras" operated in the urban areas comprising Zone C. However, where the PKK had particularly strong control (Zone B)specific neighborhoods, villages, and in the mountainsthe Turkish military employed ex-PKK fighters. The Turkish state never officially admitted the existence of Jandarma İstihbarat ve Terörle Mücadele (JITEM), but the organization is widely viewed in the country as an important element in the counterinsurgency campaign. Soner Yalçın, a controversial journalist who wrote an early and influential account of JITEM, argues that the organization was initially founded as part of the Özel Harp Dairesi (Unconventional War Directorate) in the General Staff of the Turkish Armed Forces. 99 With the rise of the PKK's guerrilla activities, JITEM began recruiting captured and surrendered insurgents into special death squads. The mechanisms by which the state used the ex-insurgents show that maintaining control of the veterans in insurgent-controlled areas was a top priority. Those with knowledge of the PKK's hideouts, contacts in the cities and rural areas, and combat experience were considered valuable assets. Ex-insurgent prisoners were used in assassinations of Kurdish rebels and PKK sympathizers. 100 JITEM officers took the ex-insurgents to sting operations, during which time they were closely monitored and supervised. At the end of the operation, the death squad members were returned to their prison cells. Keeping the nonstate counterinsurgents in prison when they were not in use ensured the state's control of their activities. 101 The veterans who worked for JITEM were often promised reduced prison sentences. 102 Neither patronage nor fragmentation fully account for the relationship between the Turkish state and its veteran proxies. To classify prison sentence reduction as patronage would involve significant conceptual stretching. Even then, it does not account for Turkey's relationship with the Hizbullah. The Turkish state found collaborators amongst radical Islamists and captured ex-PKK insurgents. By the late 1990s, it gained the upper ground against the PKK. The Turkish security forces captured Öcalan in 1999. This marked the beginning of a new PKK strategy that emphasized political participation (i.e. working with the legal Kurdish parties) over violence. The goal of establishing an independent Kurdish state was largely abandoned in favor of a negotiated settlement based on Kurdish rights and 98 Lawyer in Diyarbakır. 99 Soner Yalçın, Binbaşı Ersever'in İtirafları (Istanbul: Doğan Kitap 1994). 100 Kurdish expert, interviewed by author, Ankara, August 2013. 101 Human rights activist in Diyarbakir, interviewed by author, August 2013. The mechanism was not, however, abuse-proof. Yalçın claims that an influential officer in JITEM, Major Cem Ersever, formed a team of over forty ex-rebels and village guards that he used to assassinate his rivals as well as those of the state. Yalçın, Binbaşı Ersever'in İtirafları. 102 Kurdish expert. autonomy within Turkey's existing borders. The Turkish government lifted the state of emergency law in 2002. The Islamist Justice and Development Party (JDP) that came to power that year sought to gain support from Turkey's Kurds by expanding social and economic opportunities as well as recruiting local clan elites into the party. 103

Lessons from Turkey
Fragmentation and patronage offer only a partial account of collaboration. The proposed framework, which brings together the politics of collaboration with the strategic logic of exploitation, offers a fuller understanding of Turkey's proxy choices across space and time.
The Turkish military had found its match with the rise of the PKK. The state turned to local collaborators to counter the insurgents' guerrilla tactics. It employed village guards comprising militarily inexperienced rookies in the contested rural areas and, later, in the periphery of the conflict zone to prevent the insurgency from spreading. While some village guards served in offensive operations, most formed essentially static units engaging the PKK if and when it traversed their territory.
The state also recruited veteran proxies from the local populationmilitarily skilled individuals who could mount highly targeted offensive operations against the insurgents. The Hizbullah was mobilized in the contested urban areas to intimidate and assassinate PKK members and sympathizers with the support of state functionaries. JITEM was manned by ex-PKK insurgents, and its members were integrated into special operations forces to launch surgical strikes against the insurgency. While the state relied on a variety of recruitment and control mechanisms, the veterans were used only in the areas where the state forces could monitor, guide, and, if necessary, terminate their activities.

Renegades and Village Defense Committees in India
How far does the proposed framework travel beyond Turkey? This section traces the logic of violence outsourcing during an insurgency in Kashmir, India. The analysis supplements the evidence provided in the case study of Turkey with the intention of teasing out the similarities of proxy incorporation across different contexts.
In the late 1980s, an insurgency broke out in the Indian state of Jammu and Kashmir (henceforth referred to as "Kashmir"). Decades of frustration over the former princely state's unresolved status and the Indian government's manipulation of Kashmiri politics fueled strikes, public protests, and occasional violence. The opposition joined forces under the Muslim United Front (MUF) to run in the 1987 state elections. The rigged outcome and subsequent crackdown on the opposition party 104 roused the onset of an uprising. Hundreds of Kashmiris crossed the Line of Control into neighboring Pakistan to receive arms and training. 105 A militant nationalist organization, the Jammu and Kashmir Liberation Front (JKLF), emerged as the vanguard of the insurgency.
The insurgency was marked by a March 1988 bomb blast at the telegraph office on the main street of the summer capital, Srinagar. The JKLF claimed responsibility for the blast, and the many that followed. "The long, hot summer of 1989 saw languid Srinagar descend into violence," described an observer. 106 In December 1989, the JKLF kidnapped the daughter of India's newly appointed home minister, Mufti Mohammed Sayeed, and then freed her when the government released five detained militants.
Unlike its Kurdish counterpart, the Kashmiri insurgency started as "a largely urban movement." 107 It then spread across the entire Kashmir Valley. The bulk of rebel violence (93 percent) and the vast majority of deaths (95 percent) were reported there. 108 Figure 5 displays the distribution of rebel violence across the entire state of Jammu and Kashmir. 109 Between 1991 and 1992, old-city areas in Srinagar, Baramulla, and Sopore "had in effect become no-go zones for Indian forces." 110 Srinagar was the hotbed of rebel activity. Over half (58 percent) of all Kashmir Valley violent incidents and over one-third (35 percent) of all Kashmir Valley deaths from rebel violence were reported there. 111 However, while most violence was reported in the urban districts, Kashmiri interviewees cited a high prevalence of violence in the rural parts of the Valley as well. "They [the rebels] were everywhere in the Valley, in many villages," recalled one local. 112 Rural violence is typically under-counted due to the "urban bias" in civil war reporting. 113 The insurgents rallied the local population across the Valley, much of which was sympathetic to their cause. The area became an insurgent stronghold (Zone B).
The initial response of the Indian government to insurgent violence was to treat it "as a routine law and order problem to be handled through conventional law enforcement techniques by local police forces." 114 However, the 34,000 local policemen proved ineffective, and New Delhi began to grow suspicious of their loyalty. Many of them had social and family ties to the militants. 115 The Indian government responded by bringing in the army and federal security forces, the Central Reserve Police Force, the Border Security Force, and the Indo-Tibetan Border Police. The number of Central Reserve Police Force and the Border Security Force personnel grew roughly tenfold between the years 1989 and 1993. 116 The Indian army's role expanded in 1993 with the introduction of the Rashtriya Rifles, an elite army unit created specifically for counterinsurgency  116 Wirsing,India,145. operations, especially in the rural areas of the Kashmir Valley. In mid-1993, the total strength of regular Indian army troops in the Jammu and Kashmir state was between 300,000 and 400,000. 117 "Indian troops had been ill-prepared for dealing with the forces arrayed against them," explained Indian journalist Praveen Swami. "By 1992, however, the rudiments of a counter-terrorist grid were in place." 118 The Valley transformed from an insurgent stronghold (Zone B) to a contested territory (Zone C).

Veteran Proxies: The Renegades
India's use of death squads in Kashmir may be traced to the December 1992 murder of human rights activist H.N. Wanchoo. It was allegedly ordered by a Border Security Force officer and carried out by former insurgents who were compensated with their release from prison. Other assassinations in 1993 and 1994 also appeared to have been the work of hired gunmen working for the security forces. 119 In 1994, with the transformation of the Valley from an insurgent stronghold to a contested zone, the Indian armed forces began using death squads on a routine basis.
The death squad recruits were a mix of surrendered and captured insurgents. Several factors inspired their collaboration. Insurgent fratricide played an important role in making a large pool of former rebels available to the state. By 1993, the Indian military had gained territorial control over the Valley, and the JKLF lost the leading role in the insurgency to its rival group, the Hizb-ul Mujahideen (Hizb). The JKLF's initial patron, the Pakistani military's Inter-Services Intelligence (ISI), contributed to the JKLF's demise. The Kashmiri nationalist organization's main goal was Kashmir's independence, rather than accession to Pakistan, which was the ISI's objective. The JKLF's rallying cry was "Kashmir banega khudmukhtar" (Kashmir will be sovereign), while the slogan of the Hizb-ul Mujahideen was "Kashmir banega Pakistan" (Kashmir will be one with Pakistan). In 1991, Pakistan began to shift its support from the JKLF to the Hizb and orchestrate splits and defections in the former. In 1994, the debilitated JKLF surrendered and its leader called for peaceful struggle. As the Hizb emerged as the dominant insurgent organization in the Kashmir Valley, it sought to consolidate its power. Much of the insurgent defection occurred among members of the pro-Pakistan armed groups threatened by the Hizb's "bid for dominance." 120 117 Praveen Swami, India, Pakistan and the Secret Jihad: The Covert War in Kashmir, 1947(London: Routledge, 2007 The fragmentation mechanism accounts for much, but not all, of the collaboration. While many of the militants surrendered voluntarily while fleeing the Hizb, some were forcibly recruited. The Indian security forces were known for detaining and torturing many suspected insurgents into collaboration. 121 Equally importantly, the state began to employ the veteran militants after it had gained the ability to control them. It designed two mechanisms for recruiting, monitoring, and managing the death squad members. The first was the special counterinsurgency division of the Jammu and Kashmir police, known as the Special Operations Group (SOG). The SOG was formed on 2 June 1994. It started its operations with just ten personnel who were a mix of surrendered insurgents and Kashmiri police officers. The number of SOG recruits eventually grew to about 4000. Each district had a senior superintendent and a superintendent of police who was in charge of the SOG. "[The] SOG does not exist on papers, yet they are all over the Valley . . . people recount ruthless tales of extortion, rape, molestation, robbery, kidnapping and killing," described a Kashmiri human rights organization, which also estimated that the SOG was involved in about 80 percent of counterinsurgency operations in the Kashmir Valley. 122 The SOG was notorious for "corruption and brutality." 123 It tallied the highest number of custodial killings, after the Border Security Force. 124 The "special police officer" (SPO) system was the second channel through which the Indian state controlled the veteran proxies. Hundreds of exinsurgents were given the status of SPO and attached to paramilitary and army units operating in their localities. They were armed with lathis or canes, but not required to wear regular police uniforms. 125 On 12 March 2012, the Jammu and Kashmir Legislative Assembly reported that the Indian state recruited 23,000 SPOs, 126 though the local estimates range from 26,000 127 to 30,000. 128 The Kashmiri ex-insurgents, whom the Indian state called "renegades," were popularly known as Ikhwanis, a reference to the most notorious statesponsored organization of former insurgents, Ikhwan-ul Muslimoon (Muslim Brotherhood). The Indian state used the renegades to assassinate rebels and their sympathizers. 129 Their main targets were powerful insurgent organizations such as the Hizb, the Harkat-ul-Ansar, and the Lashkar-e-Taiba, as well as members of the banned pro-Pakistan party Jamaat-e Islami. The renegades were also used in joint patrols and as provocateurs and informants. 130 The Indian military forces closely monitored the activities of the renegades. Some were even housed in military compounds. 131 "Many militia members wear civilian clothes but live in military camps, are fully armed and harass people in full view of the personnel," described the International Crisis Group. "There are any number of reports of abusive militia members taking refuge on military bases after fleeing angry locals. If arrested, most militia members are released through the intervention of military personnel." 132 About five renegade outfits operated in the Kashmir Valley. The most notorious of them, the Ikhwan-ul Muslimoon, was headed by the folk singer and former JKLF rebel Kuka Parray. The organization was closely linked to and monitored by the Rashtriya Rifles. It was also involved in illegal timber trade and drug-running, to which the Indian state turned a blind eye. 133 Another prominent former insurgent outfit was called "the Taliban." The name was deliberately chosen to create confusion with the militant Islamic Afghan group of the same name. The Indian Taliban's principal patron was a powerful Congress leader in Kashmir. The organization worked with the Indian army and the SOG. 134

Rookie Proxies: The Village Defense Committees
The Indian military and police used veterans in Zone C and, with the Rashtriya Rifles, in Zone B. In Zone Dthe mountainous areas of the Jammu regionthe state instituted the rookie proxies: the Village Defense Committee (VDC) system. The VDCs were "self-defense" organizations set up "in remote and inaccessible mountainous parts of the state." 135 Men and women were given arms and trained to fight the insurgents. According to an official statement released by the Jammu and Kashmir Legislative Assembly in 2012, the number of VDC volunteers was roughly 6000. 136 The figure is likely to be a significant underestimate. A reputable Kashmiri NGO puts the number in the Jammu province at over 15,000, 137 and a prominent Kashmiri journalist estimates the number to be as high as 23,000. 138 Most of the recruited villagers were Hindus, while some were Sikhs. The Indian military and security forces had largely focused their efforts on the Valley, where the predominantly Muslim locals were highly sympathetic to the insurgents: "Valley Kashmiris took pride in the militants' exploits because the 'boys' had somehow restored Kashmir's pride." 139 Supply-side politics played an important role in India's choice of rookie proxy. In Jammu, rather than relying on coercion, as did their Turkish counterparts in the north and northeast of Kurdistan, the Indian military worked with ethnic groups that were deemed more likely to collaborate against the Muslim Kashmiri rebels. Rural Jammu had a significant Hindu and Sikh contingent, which was politically motivated to accept arms from the state. In 1996, members of the Hindu nationalist party, the Bharatiya Janata Party (BJP), and the rightwing Hindu nationalist militant organization, the Rashtriya Swayamsevak Sangh (RSS), volunteered for the VDC system, and successfully used it to score "political mileage" for the BJP in the 1996 general election. 140 While the VDC members were "valorized in the Indian media as the vanguard of patriotic resistance to Pakistan and its agents," many of them were "poor, simple villagers, dressed in soiled, tattered clothes and scuffed shoes, clutching antiquated .303 rifles." 141 There were reports of widespread misuse of weapons for settling personal scores. The arming and training of selected ethnic groups also aggravated the existing communal tensions at the local level between the Hindus, Sikhs, and Muslims. The administration of the Doda community, where the measure was first introduced, initially opposed the VDC system for fear of communal violence. "Hindus and Muslims have been living amicably in the areas but with the introduction of the VDCs, a change in the attitude of the people is perceptible," reported a Kashmiri NGO. 142 The Kashmiri insurgency had waned by 2003. Although insurgent violence did not cease completely, it was no longer "the primary means by which Kashmiri political aspirations for political change are manifested." 143

Lessons from India
The Kashmiri insurgents exploited the advantages of guerrilla warfare as they battled the Indian forces. The military and security forces expanded their presence in the Kashmir Valley once state officials recognized the extent of the rebels' influence there. The state began employing veteran proxies, the so-called "renegades," after the military and police forces had built up the mechanisms necessary to recruit and manage them. Similarly to Turkey's use of the Hizbullah in the contested areas (Zone C), India used the renegades under the supervision of its army and police forces across the Kashmir Valley.
The logic of the 1990s village guard system in Turkey and VDCs in India was also similar: the civilian defense units were formed to keep the insurgency from establishing a foothold in the areas where the state was largely absent (Zone D). While the fratricidal flipping mechanism explains the collaboration of the ex-insurgents in the Kashmir Valley, it does not explain the formation of the VDCs in Jammu. Neither does patronage, though the BJP and RSS may have exploited the predominantly Hindu-manned VDCs for propaganda purposes. The state used ethnic divisions in Jammu to recruit reliable proxies, and, in doing so, further aggravated the communal tensions. The supply of collaborators was a necessary condition for forming the death squads and self-defense militias, and, like Turkey, India used all means it could to generate it.

Conclusion
States outsource violence in civil war to counter the advantages insurgents gain through guerrilla warfare. Some of the counter-guerrillas states employ are ordinary civilians; others are experienced in wielding illicit violence. While the former's abilities are often limited to static, defensive functions, the latter may be used for mobile, offensive operations. Ordinary civilians (rookies) are less militarily skilled than their veteran counterparts, but are also less likely to inflict serious damage beyond their immediate field of operation. The comparative advantages and disadvantages of using rookie-manned self-defense militias and veteranmanned death squads shape the state's preferences of when and where to use them. In order for states to be able to recruit proxies, nonstate counterinsurgents must be available. Mechanisms such as fragmentation and patronage play an important role in increasing the supply of collaborators. However, where a supply of willing collaborators is absent, the state may use coercion to generate it.
State-sponsored death squads and self-defense militias are a product of calculated decisions states and collaborators make. Attempts to limit them must, consequently, take these calculations into account. Given the challenges of conducting counter-guerrilla operations, states are unlikely easily to forgo the benefits nonstate actors offer. Considering the willingness of many states to resort to coercion to recruit and manage collaborators, neither rookies nor veterans can easily reject the state's invitation to fight on its behalf.
Appreciating the logic of state outsourcing of violence deepens our understanding of what it means to be a modern state and what constitutes security. Violent nonstate actors do not automatically signal state weakness or collapse. They can, in effect, indicate the presence of a robust coercive apparatus. The Weberian conception of the state is not threatened by the phenomenon of states outsourcing violence to nonstate actors. Just because the modern state has a monopoly on all legitimate violence does not mean that nonstate actors have a monopoly on all illegitimate violence. The state is just as capable of illicit behavior as its nonstate counterparts.