The Institutional Origins of Organized Repression: Sub-State Capacity and State Violence during the Punjab Crisis 1

Transcription

1 The Institutional Origins of Organized Repression: Sub-State Capacity and State Violence during the Punjab Crisis 1 Philip Hultquist Ph.D Candidate Department of Political Science University of New Mexico MSC University of New Mexico Albuquerque, NM philhult@unm.edu Abstract. The literature on state repression and insurgency argues that the quality of state repression targeting is key in determining whether repression has a deterrent effect or a backlash effect. Selective targeting is found to be more effective at counterinsurgency than indiscriminate targeting. Why, then, are states so seldom selective in their use of repression? Why do states ever pursue repression that is counterproductive and strengthens the armed opposition? Building on the state capacity and counterinsurgency literatures, I argue that the variation in the capacity of different state agents (i.e., sub-state institutions) helps explain observed state violence and its degree of divergence from preferred selective violence. The Indian response to the Sikh insurgency in Punjab provides a useful illustration due to its experience with different types and levels of repression and its effect on the strength of the Khalistan movement at different stages of the insurgency. 1 This is a draft of a dissertation chapter that was originally prepared for the International Studies Association s Annual Meeting held in Montreal, Quebec, Canada, March 16-19, It remains a work in progress. Please contact the author before citing. There may be a more current version available. Comments are welcome. I would like to acknowledge Chris Butler, Mark Peceny, Bill Stanley, and Michele Leiby for help with theory on this topic. I would also like to acknowledge the Political Science Study Group at the University of New Mexico for their helpful comments. All errors in execution are entirely my own. For funding, I would like to gratefully acknowledge The Dylan Balch-Lindsay Fellowship and the Dean s Dissertation Scholarship.

2 Philip Hultquist 2 The Sikh Insurgency and the Puzzle of State Repression In the late 1970s and early 1980s, the Indian central government and the Punjab state government saw what they deemed a troubling rise of Sikh ethnonationalism. This ethnonationalism manifested itself in a number of Sikh groups, some of which called for an independent Sikh religious state to be named Khalistan. 2 These groups began an escalating number of agitation campaigns of protest and civil disobedience for political concessions. As these agitations threatened the interests of the government, Prime Minister Indira Gandhi s Congress (I)-led central government responded with repression and other anti-agitation measures that strengthened the extremists and separatists within the Sikh political system vis-à-vis the moderates (Chima, 2010). This effectively turned a movement for political and economic concessions into a national struggle for freedom (Pettigrew, 1995, 4). Then, after nearly a decade of high-intensity warfare and a brutally repressive counterinsurgency, the insurgency was defeated and the government declared victory in The state repression during the Sikh insurgency presents a familiar puzzle. Why does state repression sometimes deter the opposition and other times strengthen it? State repression is thought to be both the mechanism through which conflicts escalate (Mason and Krane, 1989; Goodwin, 2001; Kalyvas, 2006) and the means by which opposition movements are defeated, thwarted, or deterred (Lichbach, 1987; Hegre and Sambanis, 2006, 522). Three related literatures what I call the state repression literature, the counterinsurgency literature, and the civil war literature provide a general consensus that the quality of state repression is key in determining whether repression has a deterrent effect or a backlash effect. 3 Specifically, the use of selective 2 Khalistan literally means land of the pure. I use the terms Khalistan movement and Sikh insurgency interchangeably in this paper. 3 The terminology of deterrent effect and backlash effect comes from De Nardo (1985).

3 Philip Hultquist 3 targeting is found to be more effective at counterinsurgency than indiscriminate violence, or what I will call collective violence (e.g., Mason and Krane, 1989; Goodwin, 2001; Kalyvas, 2006). Indeed, the initial repression measures taken by the government in 1984 did not target extremists selectively but rather the Sikh ethnoreligious group, a highly collective level that includes Sikhs from all across the political spectrum. By targeting Sikhs as a collective group, the Indian government played into the rhetoric of the extremists, which built upon the suspicions of average Sikhs fearing they could not keep a separate ethnoreligious identity in a Hindudominated country. Collective repression, then, strengthened the armed opposition both in the numbers and commitment of militants but also increased the militant sympathies of the Sikh public. The corollary also appears to be true in the Sikh case. The use of selective or targeted repression played a major role in defeating the insurgency in the later stages of conflict. Christine Fair (2005) finds: The state police also found ways to improve their relations with the residents, which created means for gathering more effective intelligence leads. This generated a positive cycle: access to better information about the militants and their operations enabled the police to take more targeted action, enhanced their ability to protect citizens, and decreased arbitrary or indiscriminate action (p ). The difference in state repression strategies impacted both the rise of militancy and its ultimate demise. This leads to the second puzzle and the focus of this paper. What accounts for the differences in state repression? Why do states ever pursue repression that is counterproductive and strengthens the armed opposition? This paper weighs in on these questions from the second puzzle with evidence from the Indian response to the Sikh insurgency in Punjab, India. I advance a theoretical account for variation in state violence that builds on the state capacity literature. In short, I argue that the

4 Philip Hultquist 4 variation in the repression-related capacity of different state agents (e.g., army, police) helps explain observed state violence and its degree of divergence from preferred selective violence. State agents with high coercive capacity tend to lack local intelligence capacity, which is necessary to distinguish between actual insurgents and the general population (that is, to target selectively). The lack of accurate intelligence leads to collective targeting and a backlash effect of strengthening the insurgency. When facing clandestine insurgencies, state agents that are able to leverage local intelligence capacity, even at the expense of demonstrating their coercive capacity, are found to be more effective at selective targeting, hence effective counterinsurgency. This finding is conditioned by the need for high organizational capacity of state agents. The Sikh insurgency provides a useful illustration due to its experience with different types and levels of repression and its effect on the strength of the Khalistan movement at different stages of the insurgency. In 1984, after the initial rise in militancy, the central government chose to use the Indian Army as the primary counterinsurgency institution against a fairly limited extremist movement. The Indian Army, despite its high coercive capacity, was unable to exploit the actionable intelligence required to target selectively. Instead, Operation Bluestar and Woodrose (discussed in more detail below) diverged from the state s selective targeting preference to target Sikhs at the collective level, leading to a clear backlash effect. In later years of the insurgency, the Army took a backseat role to the Punjab police, which, after developing much stronger organizational capacities, was able to leverage its local intelligence capacity to selectively target actual insurgents, including most high-ranking militants. The paper is organized as follows. I begin with the theoretical account, building on counterinsurgency, civil war, and state capacity literature. I then provide a brief narrative of the insurgency, over viewing the origins of the conflict. The next sections illustrate how the theory

5 Philip Hultquist 5 performs with evidence from different points in the insurgency. In the process, I discuss how different agents vary in their execution from the preferred selective repression strategies of their respective principals in light of their varied institutional capacities. I end with a discussion of how well the theory performs, how well it is expected to travel, and how I will extend this argument in future research. Literature and Theory The general consensus in the literature argues that states that are interested in defeating an insurgency or an armed challenger should prefer the use of selective repression or state violence aimed at individual insurgents or their collaborators. Selective state violence is found to be more effective at deterring potential rebel recruits or collaborators by affecting their decisionmaking (Mason and Krane, 1989; Kalyvas, 2006). When the state uses selective targeting, potential rebel recruits or collaborators can believe their chances of survival are best by siding with the government or at least not siding with the insurgency (Mason and Krane, 1989). Collective state violence, on the other hand, often has a backlash effect. 4 When states use violence at highly collective levels, such as entire villages or ethnic groups, civilians (i.e. potential rebels or rebel collaborators) find that their chances of survival can actually increase by supporting or joining the rebels (De Nardo, 1985; Mason and Krane, 1989; Goodwin, 2001; Kalyvas, 2006). This, of course, strengthens the insurgency and presents a greater threat to the state. 4 I use the term collective rather than indiscriminate to more accurately convey the level of aggregation at which states target. The term indiscriminate connotes random violence, which is not accurate. Very little, if any, violence in civil war is random or actually indiscriminate. Most violence discriminates at some level, whether it is targeted at a specific village thought to be a stronghold of rebel support or at some ethnic or other identity level (see Hultquist, n.d. for further conceptual issues regarding the nature of repressive targeting).

6 Philip Hultquist 6 If selective violence is more effective at thwarting an insurgency, why are states so seldom selective in their use of violence? What constrains states from using violence selectively? The preliminary answer to these questions is that the use of selective violence is extremely difficult. A selective repression strategy prefers to target insurgent leaders, rank and file insurgents, and civilians that collaborate with the insurgents in that order. More importantly, it must distinguish between the people that belong to these groups from those that either support the government or are neutral. To make any of these distinctions, the state must rely on information or intelligence from the local civilian population (Kalyvas, 2006). Someone must denounce the insurgents and their collaborators. The question is, How does the state get the civilians to denounce their neighbors? Kalyvas (2006) provides a strong account for civilian denunciation. He argues that civilian denunciation and intelligence gathering is best achieved under the condition of sufficient territorial control by the state military. When the military sufficiently controls an area, it can provide much needed protection from insurgents. It can also provide selective benefits for its supporters and informants, and equally important, selective punishments for defectors or insurgent collaborators. For Kalyvas, physical occupation and territorial control are necessary to achieve the adequate intelligence to target insurgents and their collaborators selectively. This paper both builds on and departs from the Kalyvas (2006) account. I follow Kalyvas (2006) in arguing that intelligence gathering through civilian denunciation is key to selective targeting, effective counterinsurgency, and eventual victory. I depart in two ways. First, I consider a broader set of cases. Kalyvas (2006) focuses on cases of high-intensity civil war where insurgents hold at least some territory. The variation he addresses is in who holds territory and to what degree. Much of the evidence comes from not only full-fledged civil wars like the

7 Philip Hultquist 7 Greek Civil War and the U.S. Civil War, but external occupation and intervention, including German occupation of Europe during World War II and U.S. extended counterinsurgency in Vietnam, Iraq, and Afghanistan (Kalyvas, 2006; Kalyvas and Kocher, 2009). However, this limits the scope of his argument. Most modern insurgencies never grow large enough to hold territory. Rebel organizations may have territories with greater civilian support, but many never achieve liberated zones. When insurgencies are weak, they remain clandestine. They hide among the population, not in control of them. Still, these insurgencies matter; clandestine insurgencies can be quite formidable and still represent a threat to the state. In fact, a common strategy they use to grow is to antagonize the state into using brutal repression that drives more recruits and money into their welcoming arms. More research into the nature of insurgency and counterinsurgency at all levels of intensity is necessary for a fuller understanding of the phenomenon. Moreover, policymakers may be interested in thwarting small insurgencies before they grow. With a broader set of cases in mind, I also depart from the Kalyvas (2006) argument about the necessity of territorial control through military power. While territorial control may be required for large-scale civil wars, most states facing clandestine insurgencies do not have the capacity to carry out the physical military occupation required to achieve selective targeting in this manner. Further, a military occupation of areas that house clandestine insurgencies may be seen as overkill and actually increase popular dissatisfaction with the state. Rather than focus on the military capacity for territorial control, I advance an alternative account of state capacity that breaks open the unitary actor assumption and examines different types of sub-state capacity. The next section briefly discusses the state capacity approach to civil

8 Philip Hultquist 8 war and then provides an overview of three types of sub-state capacity - coercive, intelligence, and organizational - and their effect on counterinsurgency. State Capacity and Counterinsurgency State capacity has been linked to domestic conflict by numerous studies (Gurr, 1986; Tilly, 2003; Fearon and Laitin, 2003; DeRouen and Sobek, 2004; Buhaug, 2006; Hendrix, 2010). Fearon and Laitin (2003) argue that weak states provide an opportunity structure that favors rebellion because potential rebels view the state as an easy prize. Essentially, potential rebels are undeterred by weak states, whom they view as more easily defeated. Potential rebels will get this signal by observing an incompetent, corrupt, and mismanaged bureaucracy (Fearon and Laitin, 2003; DeRouen and Sobek, 2004). Once an insurgency begins, weak states are less capable of using repression as rationally and effectively as they may prefer. That is, they do not have the capacity for selective violence. Without the capacity to gather adequate intelligence, weak states substitute indiscriminate or collective violence for selective, which can strengthen the insurgency. There has been a recent surge in works on state capacity that attempt to disentangle the variety of types of state capacity (Sobek, 2010; Hendrix, 2010; Gleditsch and Ruggeri, 2010). These new studies recognize that states can vary in different types of capacity. States like North Korea may be administratively and institutionally weak but militarily strong. Still, these studies aggregate at the national level and ignore sub-state capacity or the capacity of state institutions to carry out selective repression. I focus here on three types of sub-state capacity: coercive, intelligence, and organizational.

9 Philip Hultquist 9 Coercive capacity refers to the overall fighting capacity of an institution. It is the capacity to use force and exercise military power. States vary in the coercive capacity in general, but institutions within states vary as well. National armies have a much higher coercive capacity than the local police, for instance. National armies have more advanced weaponry, including automatic weapons, tanks, and a variety of explosives, to name a few. They also have soldiers trained in controlling territory, a critical component for obtaining the information required to target insurgents and their collaborators selectively (Kalyvas, 2006). However, they often lack in what I call local intelligence capacity. Intelligence capacity is the ability to gather useful and actionable information. In this regard, information is actionable if it leads to the identification of selective targets. Actionable information comes from local sources. Agents with high local intelligence capacity may be comprised of members largely from the same ethnic group as the locals, or perhaps, speak the local language. Members may be recruited from the local population, who may individually know useful information or know which civilians have such information. Knowing which civilians hold local power, thus whom the state can ally with, may be obvious in some cases but only apparent to locals in others. This presents an important tradeoff when it comes to selecting the appropriate agent for a repression strategy. Where national armies have the ability to control territory, they may lack local intelligence capacity that facilitates identifying selective targets. Local police may have the opposite strengths: much lower coercive capacity but perhaps stronger local intelligence capacity. The selection of the appropriate agent may depend on the nature of the insurgency and the specific situations that states face. A state facing a strong insurgency that controls significant territory may find the coercive capacity of the national army necessary to take back the territory.

10 Philip Hultquist 10 States facing clandestine insurgencies that do not hold territory may find institutions with local intelligence capacity more suitable for achieving the civilian denunciation required for selective targeting. While the first two dimensions of capacity can be thought of, more or less, as something of a tradeoff (although theoretically an increase in one does not necessarily decrease the other), organizational capacity should linearly increase the likelihood that the agent will carry out the preferred type of repression. Organizational capacity can be thought of as tightness of control from principal to agent. Organizational capacity, then, could be thought of similarly to standard principal-agent arguments. Loosely controlled agents have more leeway to exploit their information advantage in the pursuit of their individual motives, which may vary greatly from the principal s goals and motivations (Mitchell, 2004). This argument has been tested crossnationally. The studies find that a higher level of bureaucratic corruption, a measure for organizational control at the state-level, increases sexual violence and human rights violations in general (Butler, Gluch, and Mitchell, 2007; Bohara, Mitchell, Nepal, and Raheem 2008). Organizational control and corruption should vary by sub-state institution as well. In the paper that follows, I apply these concepts to the case of the Sikh insurgency in Punjab, India, to show how variations in the repression-related capacity of state agents affects the nature of the observed repression and its degree of divergence from the preferred selective repression strategy. First, I provide a brief narrative of the origins of Sikh militancy.

11 Philip Hultquist 11 Brief narrative on the origins of Sikh militancy 5 National unity has been an obvious and difficult problem in India both before and after its independence from the United Kingdom and remains a difficult task to this day. India is home to a great number of ethnic, linguistic, and religious minorities, each always fearful of losing their identity in a constitutionally secular but Hindu-dominated state. Due, in part, to this long-term and underlying fear (as well as a great number of other factors), India has experienced several insurgencies aimed at greater autonomy or secession. The Sikh and Kashmiri insurgency stand out as the most violent. In each case, both the state and the opposition contributed heavily to the violence. This section briefly overviews the Sikh insurgency and serves as a point of reference for the following evidence section. Although I am not specifically concerned with causes of the insurgency in this paper, I spend some time on them to show that the rise of militancy was not inevitable. Joyce Pettigrew (1995) argues that the Sikh insurgency should be understood through four contexts. The first two, agrarian policies and military recruitment policies, represent "predisposing circumstances leading to political violence," while the second two, the state terror of 1984 and the police activities in the villages, represent "direct, immediate causes that turned a farmers' movement, initially concerned with specific socio-economic issues, into a national struggle for freedom" (Pettigrew, 1995, 4). I deal with the second two, the impact of state repression policies, in the following sections. As Pettigrew (1995) notes, the origins of the conflict lie in agrarian policies of the Indian central government toward the riparian Punjab's water rights. 6 Punjab is a largely agricultural 5 I will not cover the origins of the Sikh religion, the evolution of Sikh consciousness, or other Sikh history in any detail. There are, however, many works on these subjects, including, for example, Singh, Mansukhani, and Mann (1992); Kapur (1986); and Singh (2008).

12 Philip Hultquist 12 state where Sikh farmers produce mostly wheat on small, owner-occupied farms (ibid, 5). The Indian central government (or Union government) mandated that 75% of Punjab water be allocated to non-riparian states (ibid, 5). Complementary to the issue of water rights, farmers in Punjab experienced a significant rise in the cost of production (due to diesel and power shortages) accompanied by a low price for wheat. 7 Despite impressive gains in agricultural productivity during the 1960s, the Green Revolution brought rising indebtedness and increasing economic grievances based on the fact that Punjabi farmers could not control their own destiny, which looked quite promising from the 1960s through the mid-1970s. The indebtedness might not have been a huge problem if the benefits of the Green Revolution continued. However, the output during the Green Revolution could not keep up with rising expectations, which follow a classic J-curve theory of grievance leading to violence (Singh, 1984; Davies, 1962). These economic grievances were augmented by Union policies on Army recruitment (Pettigrew, 1995). Sikhs make up approximately 2% of the total Indian population but have, since British times, been recruited into the military and government service at much higher proportions (Cohen, 1988). Access to government jobs, and in particular Army jobs, was a significant source of income for Punjabi households (Bhalla and Chadra, 1982, 874 cited in Pettigrew, 1995, 7). The discontent came from a growing perception that these job opportunities were drying up as the Union government sought to recruit for the Army proportionately by region (Cohen, 1988). This would effectively end the Punjabis comparative advantage in obtaining Army jobs, which it felt were distributed justly based on the Punjab having endured the brunt of Pakistani land and air attacks during the Indo-Pakistan wars of 1965 and See also Singh, Daljeet (1992) for a discussion of the Punjab river-waters dispute. 7 Wheat production costs rose 84% between 1971 and 1978 (Kahlon and Kurian, 1981, cited in Pettigrew, 1995, 5).

13 Philip Hultquist 13 (Pettigrew, 1995, 6). 8 Coupling the new recruitment policies with Green Revolution agrarian grievances added to the growing discontent with the Union. The result was a feeling of deliberate isolation and discrimination towards the Sikh community. 9 These policy grievances hardly seem irreconcilable, and Pettigrew (1995) notes Sikh leaders originally sought redress through democratic means. The first and most noted example of this is the passing of the Anandpur Sahib resolution in 1973 by a working committee of the Akali Dal, the primary political party claiming to represent Sikhs and maintain Sikh culture in Punjab. The Anandpur Sahib resolution demanded more devolution of power to the Punjab state, which, after the 1966 Punjab State Reorganization Bill 10 was enacted, represented a Sikh majority state (approx Sikh to Hindu) (Chima, 2010, 32). 11 It was not until 1978, however, that the Anandpur Sahib Resolution became a symbol for national sovereignty. Between 1978 and 1984, there was a recurring pattern of 1) Center Sikh (led by the Akali Dal political party) negotiations to implement the Anandpur Sahib Resolution, 2) an agreement to implement most of it, 3) reneging by the Center, and 4) renewed Sikh agitation protests and strikes, leading to new negotiations and the pattern starting over again. In addition to large-scale agitations, the years saw increased violent interactions between Sikh extremists and Punjab police as well as communal violence between Sikhs and 8 The Punjab region was split during partition with the capital Lahore going to Pakistan and, hence, serves as an important border area for both Pakistan and India. The "P" in Pakistan stands for its Punjab province. 9 See also Singh, Kharak (1992) on Sikh grievances and demands. 10 The Reorganization Bill was a success of the Akali Dal s long-term strategy of a truncated Punjab where Sikh could represent a majority, thus enabling the Sikh party to exert more influence in Sikh populated areas (Chima, 2010, 29-34). The truncated Punjab did not include all parts that were demanded and left many unhappy with the solution. Notable grievances were the status of Chandigarh as Union Territory rather than sole capital of Punjab and several adjoining Punjabi-populated territories that were not included in the new Punjab. 11 Kapur (1986) cites the figure at approximately Sikh to Hindu.

14 Philip Hultquist 14 Hindus. On the side of Sikh extremist violence, the most notable player in the early stage was an extremist preacher, Sant Jarnail Singh Bhindranwale. Bhindranwale and his followers were responsible for several assassinations, including police and dominant local Hindu leaders. His interactions with police got off to a poor start in Following the deaths of the Akali workers by Haryana police, Sikh extremists assassinated a prominent Hindu leader. Bhindranwale was to be arrested for the crime. Bhrindranwale escaped and the Punjab police burned his organization s busses and holy texts, and physically assaulted followers (Chima, 2010, 63). Bhindrawale later courted arrest, and at the scene, there was a clash between his supporters and police and paramilitary forces resulting in the death of 18 protesters (ibid, 64). The Punjab police began to extra-judicially execute Sikhs in custody and claim the deaths were the result of armed fighting. These are referred to as faked encounters (ibid, 69). This unleashed a violent cycle of killing and counter-killing as Bhindranwale s supporters assassinated police officers suspected of killing or torturing Sikhs. By 1984, Bhindranwale had fortified the Golden Temple Complex and the Akal Takht the holiest Sikh shrine, where he was able to carry out extremist violence with relative impunity. Both the central and Punjab governments were reluctant to pursue Bhindranwale or his men in the Golden Temple for fear of it being viewed as an attack on Sikhs at the collective level. The Army, Coercive Capacity, and Collective Targeting This paper began with the premise that collective targeting in state repressive policies at the early stages of conflict had a backlash effect, which strengthened the militants in numbers and support. This section examines the role of the state agents who carried out the two most prominent collective targeting events, Operation Bluestar and Operation Woodrose.

15 Philip Hultquist 15 Operation Bluestar The Army and the Golden Temple When Mrs. Gandhi was told that Operation Blue Star had started, she must have wondered whether it would provide the decisive inspiration for the Sikh independence movement, a movement which at that time had very little support outside of Bhindranwale s entourage and small groups of Sikhs living in Britain, Canada, and the United States - Tully and Jacob (1985, 155) The infamous Operation Bluestar, where the Indian Army raided the Golden Temple Complex and severely damaged the Akal Takht, had the most obvious and long-lasting backlash effect on the trajectory of the conflict. It not only had a clear local effect that strengthened the militancy of the movement, it caused the mutiny and desertion to militancy of many Sikhs in the Army (Tully and Jacob, 1985, 194), helped create and reinforce a dedicated Sikh diaspora, and ultimately resulted in the assassination of Prime Minister Gandhi herself. Based on his seemingly invulnerable position, Bhindranwale ratcheted up the action in late 1982 through early Bhindranwale s men murdered a senior police officer outside the Golden Temple (in addition to many non-senior officers throughout the previous years) and appeared to be responsible for the shooting of seven Hindus (six fatally), who were separated from Sikhs on a bus before being shot. These incidents led to imposition of President s Rule, or rule by the central government, which required the dismissal of the Congress-led government in Punjab and gave the Punjab police unquestioned authority to shoot who they wanted when they wanted and search where they wanted (Tully and Jacob, 1985, 107). 13 Bhindranwale was 12 Bhindranwale was invincible earlier in part because he was used by Congress to drive a wedge in the Akali Dal. Later, as he became popular in his own right, he appeared invincible in spite of Congress. For more information see especially, Brass (1988) and Tully and Jacob (1985). 13 These preceding events and the actual Operation Bluestar are documented with much greater detail in other works, including, but not limited to: (Tully and Jacob, 1985; Chima, 2010; Brass, 1988; Gill, 2001).

16 Philip Hultquist 16 undeterred by the imposition of President s Rule and continued his militancy, the assassinations, and most importantly his fortification of the Golden Temple Complex and the Akal Takht in particular the holiest of Sikh shrines. These actions were tantamount to taunting the Prime Minister to actually act beyond imposing President s Rule, which was a clear failure. After escalating violence and Bhindranwale s taunting, Mrs. Gandhi finally gave the order to invade the Golden Temple, which occurred June 5, She was fully aware of the risk she was taking. The Center and the Punjab police were reluctant to enter the Guru Nanak Niwas a rest house in the hostel part of the larger complex but not part of a holy shrine to arrest Bhindranwale in fear of a backlash effect (Tully and Jacob, 1985, 98-99). As Bhindranwale became less secure in the rest house, he sought and achieved permission from the High Priests to live in and fortify the Akal Takht the holiest of Sikh shrines within the larger Complex. This made entering the Complex much more difficult politically since most Sikhs would have understood raiding the rest house but the Akal Takht was strictly sacrosanct. Operation Bluestar was an Army operation that raided the Golden Temple Complex to flush out Bhindranwale and his followers. The goal of flushing out quickly evolved into a lethal mission. The complex was well fortified through the help of former Army Major-General Shahbeg Singh and well defended. The operation took three days and nights and cost substantial casualties but ultimately was successful in defeating Bhindranwale and those of his followers who stayed to fight. The casualty numbers are still disputed, but, at minimum, the operation was responsible for 83 military lives and 493 militant or civilian lives. Tully and Jacob (1985, ) note that these official numbers, published by the central government s White Paper, leave at least 1,600 people unaccounted for. Pettigrew (1995, 8) cites, from human rights groups, figures as high as 4,000 killed. The Operation also left the sacred Akal Takht in rubble.

17 Philip Hultquist 17 Mrs. Gandhi s counterinsurgency strategy, in this case, did not specifically target the Sikhs collectively. While it resulted in high civilian (and especially high Sikh) casualties, both intentional and unintentional, it seems quite clear that the objective was Bhindranwale and his followers, a perfectly selective target. However, there were three reasons that the Operation appeared to be targeting the collective Sikh community. First, the day of the assault was the celebration of Guru Arjun s martyrdom, so it had religious significance. Second, there were a fmuch higher number of Sikh civilians (i.e., noncombatants) in the Complex. Approximately 950 pilgrims were present for the Guru Alrjun celebration (Tully and Jacob, 1985, 185; Chima, 2010, 93). Approximately 1,700 Akali Dal supporters were present to begin the new phase of the agitation (Tully and Jacob, 1985, 185). Third, the Army destroyed the Akal Takht with heavy artillery. The last of these was one part due to an intelligence failure and the other due to a failure of the government to win the battle of public information. It is clear that the government truly did not want to damage the Akal Takht. One of the main objectives was to get Bhindranwale and his armed followers without damaging the Akal Takht or the actual Golden Temple. The mission was given the go-ahead based on the estimation that the Army could achieve their objective without damaging these shrines. Soldiers were ordered not to damage each shrine, and during the first stages of the Operation, they refrained from doing so. After repeated attempts to storm the Akal Takht to reach Bhindranwale s position, the Army asked for permission to use tanks and heavy artillery to achieve their goal. It took two hours for the central government to approve it, reflecting their reluctance to damage the Akal Takht (Tully and Jacob, 1985, 165). This was an intelligence failure because there were plenty of opportunities for the Army to find out how fortified Bhindranwale s position was in the Akal Takht. The Complex was not closed and many people, including those in the Army, were

18 Philip Hultquist 18 able to go in and out of the Complex before the Operation. They should have known the damage that was necessary to complete the mission. 14 The second failure of the government was over information. Many Sikhs did not believe that Bhindranwale had fortified the Temple, especially the Akal Takht. They did not believe he operated a grenade factory in the Temple and was giving military training to young Sikhs. The foreign press was not allowed. Numerous times, before and after the event, Mrs. Gandhi framed her justification for the Operation against Sikhs - the collective - and not against Bhindranwale and his militants (Pettigrew, 1995; Tully and Jacob, 1985; Chima, 2010). Instead of convincing Sikhs that the assault on the Golden Temple was a necessary cost to remove a militant threat that was abusing the Sikh religious symbols, many believed that Bhindranwale was now a martyr who died defending the religion s holiest site (Chima, 2010, 95). In terms of the theoretical propositions above, how did Operation Bluestar happen? The main principal was Prime Minister Indira Gandhi, whose institutional agent was the Indian Army. The decision to enter the Golden Temple Complex was certainly a difficult one for Mrs. Gandhi. In the face of rising violence in Punjab, she was under increasing pressure from the opposition, the media, and the rest of India to act. 15 While she was selectively targeting an armed extremist, she was fully aware that it might come across as collectively targeting Sikhs. So, the preferred repression, in this case, was selective but had the backlash effect as if it were collective, due largely to execution. The Army has high coercive capacity, but in the case of India, acts much like an external intervener, a group that typically lacks local intelligence 14 See Tully and Jacob (1985, ) for more detail on intelligence and operational problems. 15 Her political predicament is explained further in Tully and Jacob (1985, ch , especially 143).

19 Philip Hultquist 19 capacity. K.P.S. Gill (2001), who later directed the Punjab police during the counterinsurgency, describes the use of the Army in Operation Bluestar and the subsequent Operation Woodrose. The damage Bluestar did was incalculable. This was compounded by Operation Woodrose, the Army s mopping up exercise all over Punjab that sought to capture Bhindranwale s surviving associates and to clear all Gurudwaras in the state of extremist elements. Woodrose suffered from all the classical defects of army intervention in civil strife an extraneous and heavily armed force suddenly transported into unfamiliar territory; mistrustful (in this case, exceptionally so) of the local police and intelligence, but with no independent sources of information; dealing with a population, large segments of which had become hostile; and operating under a political fiat that not only condoned, but emphasised the use of punitive force. Operating blindly, the army arrested large numbers of people, many innocent, others perhaps sympathetic to the militant cause, but by no means associated with any terrorist or criminal activity. Lacking in adequate information to distinguish effectively at the local level, the indiscriminate sweep of Woodrose pushed many a young man across the border into the arms of welcoming Pakistani handlers (Gill 2001, 30). The lack of intelligence capacity during the Army s Operation Woodrose mop up led to the official targeting of Amritdhari (or baptized) Sikhs, a policy published in the July issue of an official Army magazine (Tully and Jacob, 1985, ; Pettigrew, 1995, 10-11, see also endnote 16). This meant they began to target at the collective level in the absence of the capacity to target selectively. Sikhs were baptized as a show of religiosity and commitment to Sikh identity, not militancy or separatism (Pettigrew, 1995, 10-11, 25). The problem of using a coercive instrument like the Army is compounded by perceptions. The Army, being a fairly blunt instrument, perceived Sikhs the collective - as the other. They, therefore, did not trust local police or local intelligence because local sources were Sikh sources. This perception biases against any and all intelligence that could actually distinguish between insurgents and civilians. Without such intelligence, selective targeting is impossible and collective targeting may appear to be the only option.

20 Philip Hultquist 20 Mrs. Gandhi s use of the Army may be understood in light of the dismal status of local institutions at early stages of the militancy. The Punjab police, especially in the early 1980s, had very low organizational capacity. That is to say they were highly ineffective, often corrupt, had low morale, and an even lower will to act against Bhindranwale who was assassinating with impunity police who targeted his followers. To reflect their lack of organizational capacity, I turn to Tully and Jacob (1985, ) who interviewed a senior official after they learned that the police had been arresting petty criminals and passing them off as extremists. He [a senior official in the Punjab government] said, We must appear to be doing something. But the trouble is that the Punjab police are demoralised. So many of their colleagues have been killed by Bhidranwale s men that they have no faith in our ability to protect them. When I asked about the paramilitary police [the Central Reserve Police Force, CRPF] who had been moved into the state to stiffen the backbone of the Punjab police, he replied, What can the paramilitary police do on their own? They do not know the state and have to rely on the local knowledge of the Punjab police. The paramilitary police never believed that the Punjab police were serious about stamping out terrorism. In fact many paramilitary police officers believed that the Punjab police were on Bhindranwale s side. The case of India during the early stages of the Sikh insurgency highlights not only the difficult tradeoff between coercive capacity and local intelligence capacity of sub-state agents but also the importance of organizational capacity. The paramilitaries and the Army were brought in only once the Punjab police displayed its ineffectiveness. Intelligence, Organization, and the Evolution of Selective Repression Operation Bluestar and the other events of 1984 i.e. Operation Woodrose, Mrs. Gandhi s assassination, and the following Congress-led anti-sikh riots set the stage for a brutal insurgency that lasted through While collective and brutal state violence occurred for quite some time (see Pettigrew, 1995, 8-14; Gossman, 1994), eventually the police were able to

21 Philip Hultquist 21 selectively target militants, especially top-ranking ones, which helped bring an end to the bloody conflict. The Punjab police were able to sufficiently suppress the Khalistan movement by the mid- 1990s. Although alternative explanations have been proposed (see Singh, 1996; 2000 for an overview), advancements in the Punjab police s organizational capacity allowed them to leverage their intelligence capacity, which led to the increased use of selective targeting and eventual suppression of the Sikh militancy. Toward the end of the 1980s, the government pursued a much stronger counterinsurgency approach and a signaled clear move away from a negotiated settlement approach. 16 In 1987, the central government imposed President s Rule again and instituted a violent bullet for bullet approach to counterinsurgency. The policy, which continued to evolve through the end of the conflict, sought to seek out and kill highranking insurgents. It used plain-clothes police vigilantes to carryout many of the violent operations (Gossman, 2000) and infiltrate militant operations. The use of petty criminals as informers and infiltrators was also used. Despite the human rights abuses that occurred through this violent policy, by the middle of 1992, many of the top-ranking militants were killed, and by February 1993, most of their organizations had disintegrated (Singh, 1996, 414; Fair, 2005; Gill, 2001). Despite the counterinsurgency success, Singh (1996; 2000) notes that this violent policy was not as perfect as it may appear. [T]he ratio of hardcore terrorists to non-hardcore 16 It is worth noting that it was Indian central government s reneging on the negotiated settlement (the Rajiv Gandhi-Longowal Accord) that helped sustain the conflict, not the negotiated settlement approach per se. Implementing the Accord could have strengthened the moderates within the Sikh political system and further isolated the militants.

22 Philip Hultquist 22 terrorists killed varied from 1:9 to 1:18, respectively (Singh, 1996, 415). 17 If the second half of these ratios were, in fact, non-hardcore terrorists, the policy would still be considered selective targeting. It is difficult to evaluate how selective the entire police campaign was toward the end of the insurgency. From the beginning, the Punjab police was known to lie about the militant credentials of whom they arrested. They also lied about the nature of how they were killed. Faked encounters where a suspect was killed in detention but officially reported to be killed in an armed encounter with the police were common from the beginning. The police lied about these events but did not appear to make great efforts to conceal the truth, evidenced by the clear signs of torture on the bodies of the dead (Gossman, 1994; 2000). Further, we know that the police were not perfectly selective, which would imply they targeted only known insurgents. They targeted the family member of suspected insurgents as well as informants (Gossman, 1994). However, both of these categories, despite breaching international law, fit with the logic of selective targeting stated above and in the literature (see, for example, Kalyvas, 2006). What we do not know is how many innocent people (i.e., noncombatants, non-rebel collaborators) were targeted and killed during the police campaign. This is crucial theoretically. We know they were effective at killing many top-ranking and not-so-top-ranking militants. However, if, in the process, they killed just as many civilians or noncombatants, we could not call this policy selective and, thus, it would not fit the theoretical predictions from the literature. Nonetheless, this case helps draw the conclusion that local police may be effective counterinsurgency agents, due largely to a higher local intelligence capacity. 18 They speak the 17 This casualty figure covers the range of the first 6 months on 1992 which is within the overall range when counterinsurgency targeting became more selective 18 The Army and paramilitaries - the Central Reserve Police Force (CRPF) and Border Security Force (BSF) played a supporting role to the police toward the end of the conflict (e.g., sealing the Indo-Pakistani border).

23 Philip Hultquist 23 same local language, Punjabi. The police tend to be predominantly Sikh themselves. They are recruited from the local population so they may come to the job with some useful intelligence. Local intelligence capacity is compounded by the public perception of the police compared to central government forces. Fair (2005, 90) notes from her interview with the Director-General of the Punjab police, K.P.S. Gill: [T]he local community saw the Punjab police as a Sikh force largely because police draw from a local, predominantly Sikh, labor force. When the federal entity, the Central Reserve Police Force, was deployed to the Amritsar area, the officers were seen as being there to protect the Hindus. That the forces did not operate together fostered the public perception that they had particular constituencies. The conclusion that police are effective counterinsurgency agents requires some further qualification. While they have some advantages, such as high local intelligence capacity, they have some drawbacks. The Punjab police began the conflict in the early 1980s with very low organizational capacity, with high levels of corruption, and low levels of coercive capacity. This led to a highly ineffective police force, which itself causes a vicious cycle of civilian insecurity, loss of confidence, and loss of intelligence capacity (Fair, 2005, 91). When the government fully committed to the counterinsurgency approach, they reorganized and strengthened the Punjab police. The size of the force was increased from around 35,000 to around 60,000 (Fair, 2005; see also Singh, 1996). According to Fair (2005) and Gill (2001), the police also introduced a number of new strategies to facilitate selective targeting. The police also conducted combing operations in daylight, making use of specific intelligence rather than the general search procedures that typified the previous manner of policing. These operations would typically be done on a Sunday, when all residents were likely to be at home, and were understood to be both preventive as well as detective in nature. This tactic was preventive in the sense that it would discourage militants from taking refuge in particular homes and would discourage families from willingly providing

24 Philip Hultquist 24 such assistance. It was detective in that it afforded the police the opportunity to detain for questioning militants who were identified in this process (Fair, 2005, 96). Over time, the Punjab police increased its coercive capacity and organizational capacity, which allowed it to increase and better utilize its local intelligence capacity to selectively target militants. 19 Rather than overstate my case, I should note that even with the clear increase in organizational capacity over the course of the Sikh insurgency, the Punjab police, like other police in India, still suffers from relatively low organizational capacity and is still perceived to be corrupt (Fair, 2005, 71). Conclusion Discussion and Future Research This paper sought to use the varied Indian responses to the Sikh insurgency to advance a theoretical account for how the repression-related capacity of sub-state agents affects the execution of state repression strategies. The evidence provided above generally supports the theoretical proposition. Local police, with adequate organizational capacity, can utilize their inherent local intelligence advantage to carry out more selective repression strategies. National armies bring higher levels of coercive capacity but often at the expense of local intelligence capacity. The strength of national armies in counterinsurgency is in holding and controlling territory, which itself can facilitate intelligence gathering. However, many, if not most, contemporary conflicts do not reach the stage of full-blown civil war, where each side occupies large territories and then faces off in conventional warfare. Rather, many counterinsurgencies are fought against much weaker opponents whose strategy is to remain clandestine and use intermittent terror tactics. By default, territory remains in government hands. The use of local 19 See Gill (2001) and Fair (2005) for a detailed explanation of how they increased their coercive capacity with updated technology, weaponry, force size, and force multipliers.