Explaining Crisis Onset in India-Pakistan Relations

Sameer Lalwani & Hannah Haegeland

In late November 2008, terrorists besieged two luxury hotels in a posh section of India’s commercial capital, Mumbai, during a three-day assault that resulted in the deaths of more than 170 people. The attack marked India’s 9/11 moment; indeed, the episode was quickly dubbed “26/11.” The fact that a Pakistani-based terrorist organization had conducted the complex assault triggered an immediate interstate crisis between India and Pakistan. Lashkar-e-Taiba (LeT), the terrorist group quickly attributed with responsibility for the attack, had close ties to the Pakistan military and intelligence services, raising suspicions that the attack was state-sponsored. Tensions ran high as the Indian government convened national security meetings and considered military response options, including air strikes on LeT’s headquarters in the middle of the densely populated Punjab province and air strikes on suspected terrorist staging areas in Pakistan-controlled areas of the disputed Kashmir region. Ultimately the Indian government opted not to conduct a military response – a decision attributed to Indian dysfunctional decision-making, weak-willed leadership, limited military capabilities, a robust Pakistani deterrence posture, and U.S. crisis intervention efforts.1See for example Polly Nayak and Michael Krepon, The Unfinished Crisis: US Crisis Management After the 2008 Mumbai Attacks (Washington, D.C.: Stimson Center, 2012); and Vipin Narang, Nuclear Strategy in the Modern Era: Regional Powers and International Conflict (Princeton: Princeton University Press, 2014).

What is striking is that two years earlier, LeT was attributed responsibility for conducting a similar mass casualty attack, but without triggering expectations of a crisis. During evening rush hour on Mumbai’s heavily trafficked commuter rail system, seven pressure cooker bombs exploded in tight coordination over an 11-minute span killing over 200 people and injuring over 700.2“Mumbai Bombers ‘Will Never Win,’” BBC News, July 12, 2006. The attack on July 11, 2006, was dubbed India’s “7/11.” Although blame quickly fell on LeT and Pakistan for the attack, there appeared to have been little discussion of Indian military options against Pakistan, no convening of national security officials to discuss a response to Pakistan, and no serious consideration of a display or use of force. Furthermore, despite higher casualties than 26/11, the Mumbai commuter rail attack is not included in a traditional accounting of major India-Pakistan crises.3Nayak and Krepon, The Unfinished Crisis, 2.; P.R. Chari, Pervaiz Iqbal Cheema, and Stephen P. Cohen, Four Crises and a Peace Process: American Engagement in South Asia (Washington, D.C.: Brookings Institution Press, 2007); and Vipin Narang, “Posturing for Peace? South Asia’s Nuclear Postures and South Asian Stability,” International Security 34, no. 3 (2009/10): 38-78.

This prompts an interesting puzzle – why did one event trigger a crisis while the other did not? The episodes were roughly matched in terms of location, fatalities, the suspected perpetrating group, and type of trigger, yet the 2008 event resulted in a “crisis episode” while the 2006 event was treated merely as a terrorist incident. Eerily similar serial bomb blasts had been set off in Mumbai in 1993 that killed more than 250 and injured over 700. Within days of the bombing, the group responsible was publicly linked to Pakistan, yet no crises ensued.4Chari, Cheema, and Cohen, Four Crises and a Peace Process, 28.

This is not the first time this empirical puzzle has emerged. This pattern of “non-crises” or “almost-crises” has recurred in the Indian-Pakistan relationship. In 2001, a terrorist attack on the Indian Parliament that only resulted in the loss of a few lives triggered the first part of the 2001-2 Twin Peaks crisis. The crisis involved massive military mobilization on both sides and intensified shows of force including artillery exchanges across the Line of Control (LoC) but stopped short of war. Before this event, however, there was a strikingly similar terrorist attack just a year earlier on another iconic symbol – the historic Red Fort – in the heart of New Delhi.5The Red Fort is no mere artifact. Though no longer in use as a government building, it is the iconic site closely associated with the first major Indian rebellion against the British in 1857 and where every year on India’s independence day the prime minister hoists the Indian flag and delivers a speech to the nation. The Parliament attack erupted into a major crisis while the Red Fort attack did not trigger such escalatory pressures and went comparatively unnoticed.

In January 2016, a Pakistan-based terrorist group attacked an Indian airbase in Pathankot, Punjab, killing seven Indian soldiers. Instead of a crisis or India-Pakistan standoff, both countries sought to collaborate on a joint investigation of the attack. By contrast, in September 2016, another cross-border attack on an army installation in Uri, Kashmir, was believed to have violated a red line. This attack triggered a crisis episode for India and Pakistan and eventually resulted in limited cross-border retaliatory strikes by India, which the Indian Army described as “surgical strikes.”6“Surgical Strikes: Full Text of Indian Army DGMO Lt Gen Ranbir Singh’s Press Conference,” The Indian Express, September 29, 2016.

In addition to terrorist attacks, military exercises and cross-border incursions appear to also have varying effects on crisis onset. At times, mass military exercises have triggered crisis episodes, like India’s Brasstacks exercise that trigged the 1987 Brasstacks Crisis or Pakistan’s Zarb-e-Momin that precipitated the 1990 Compound crisis. While there has been much consternation over recent Indian military exercises in 2012 and 2015, they have not resulted in crises.7Manu Pubby, “Pakistan Seeks Information on Indian Army’s Exercise in Rajasthan,” The Economic Times, November 6, 2015. An unexpected second major Indian strike corps exercise in one year (2015) offered an uncharacteristically low level of information besides pre-notification but did not trigger a crisis in the same manner as in decades prior.8Ali Ahmed, “The Strange Silence Surrounding an Indian Military Exercise,” The Diplomat, November 2, 2015.

Pakistan’s military incursion to take over Loonda Post in 2002 bears striking similarity to its incursion into the Kargil sector in 1999.9Praveen Swami, “When Pakistan Took Loonda Post,” Frontline, 2002. The latter event triggered an actual limited war between two nuclear powers for only the second time in history.10Peter R. Lavoy, ed., Asymmetric Warfare in South Asia: The Causes and Consequences of the Kargil Conflict (New York: Cambridge University Press, 2009).
Both incidents began with a quiet Pakistani troop movement to capture Indian-controlled territory in the Kargil heights. In each incursion, India learned of Pakistan’s infiltration only after assaults on Indian forces. Both ultimately resulted in an Indian military response and recapture of territory, although only the Kargil incident escalated to the level of crisis. South Asian history is littered with similarly patterned provocations that were prime candidates for an international crisis, with only a few actualizing into crisis onset.

In summary, during the past few decades several potential crisis events have punctuated the India-Pakistan rivalry, but only some of these have escalated to the level of a crisis. Why then do some events trigger a crisis episode while others do not? This narrow empirical question introduces some broader theoretical questions: What is an international or interstate crisis? What triggers crisis onset? How can scholars, analysts, and policymakers better anticipate these episodes? The essay seeks to address these questions.

India and Pakistan have fought four wars and are the only two states with nuclear weapons that regularly exchange fire. Beyond providing rich data for crisis study, understanding how crises between India and Pakistan are triggered – and which characteristics of a provocation have the most escalatory potential – can have critical “red flag” utility for crisis managers in India and Pakistan and in third-party countries with a high stake in limiting escalation short of the nuclear threshold.11Traditionally, the United States has played this third-party manager role. In particular, crisis management has been a key priority for the United States since the 1998 nuclear tests added a nuclear tinge to all India-Pakistan crises. In the future, however, China may play a larger role in preventing escalation. For more on this, see the essay in this volume by Yun Sun and Hannah Haegeland, “China and Crisis Management in South Asia.” Thus far, the literature has made inferences about the causes of crises by studying a handful of the major ones while neglecting minor crises and altogether ignoring the “dogs that do not bark” – that is, the observations of candidate crises or provocations that did not actually trigger a crisis episode.

This essay proceeds as follows. First, it offers a definition of an interstate crisis and its essential properties. Second, it details how we approach crisis onset and the importance of studying it specifically in the context of India and Pakistan. Third, it unpacks how a state retains agency to “select” into crises (or not) and proposes a simple model for the stages of crisis onset.12James D. Fearon, “Selection Effects and Deterrence,” International Interactions 28 (2002): 5-29. Fourth, it briefly describes our research design including data sources and how we identify crises. Fifth, it draws on the literature on international crises and India-Pakistan rivalry to distill some plausible hypotheses to explain crisis onset, as well as ways to operationalize and measure them. Finally, the essay presents some analysis of our initial findings and concludes with suggested future avenues for additional research.

What is a Crisis?

Every interstate war and conflict starts as a crisis. To better anticipate crises and understand their causes, this essay first seeks clarity on the meaning of crisis and its properties. For this, we turn to the extensive literature on the subject, mostly written during the Cold War, to unpack essential properties and dynamics. Snyder and Diesing describe a crisis as an international conflict episode – a sequence of interactions between two states – lying in an “intermediate zone” between peace and war.13Glenn H. Snyder and Paul Diesing, Conflict Among Nations: Bargaining, Decision Making, and System Structure in International Crises (Princeton: Princeton University Press, 1977), 10. Crisis – from the Greek word krisis, which means a decision point14Chari, Cheema, and Cohen, Four Crises and a Peace Process, 4. – involves a “moment of truth” for decision and action and serves as a critical site of bargaining.15Snyder and Diesing, Conflict Among Nations, 4. International crises are distinct from normal decisions affecting interstate relations because of the following perceptions: threats to core interests, abnormal intensity, higher stakes, uncertainty, and time constraints.

The first feature of a crisis is a threat or challenge to existing structure or a state’s vital interests, goals, or values.16Richard Ned Lebow, Between Peace and War: The Nature of International Crisis (Baltimore: Johns Hopkins University Press, 1981), 10-11; Michael Brecher and Jonathan Wilkenfeld, International Crisis Behavior Project, 1918-2001 (Ann Arbor: Inter-University Consortium for Political and Social Research, 2003); and Charles F. Herrmann, “Indicators of International Political Crises: Some Initial Steps toward Prediction,” in Theory and Practices of Events Research: Studies in International Actions and Interactions, ed. Edward Azar and D. Ben Dak (New York: Gordon and Breach, 1974), 233-43. Without “skin in the game” a state would not be provoked by what otherwise might be considered reckless or dangerous actions. The second feature is a surprising intensification or seemingly sudden deviation from the mean – “a distortion in the type and an increase in the intensity of disruptive interactions between two or more adversaries.”17Brecher and Wilkernfeld, International Crisis Behavior Project. The concern is that intensification of events and forces “substantially above ‘normal’ (i.e., average) level…increases the likelihood of violence occurring in the system.”18Oran R. Young, The Intermediaries: Third Parties in International Crises (Princeton: Princeton University Press, 1967), 10. Related of course is a third central feature of international crisis interactions, which is “the perception of a dangerously high probability of war.”19Snyder and Diesing, Conflict Among Nations, 6.

A fourth feature is incomplete information and uncertainty. Schelling writes:

The essence of the crisis is its unpredictability. The ‘crisis’ that is confidently believed to involve no danger of things getting out of hand is no crisis; no matter how energetic the activity, as long as things are believed safe there is no crisis. And a ‘crisis’ that is known to entail disaster or large losses or great changes of some sort that are completely foreseeable, is also no crisis; it is over as it begins, there is no suspense. It is the essence of a crisis that the participants are not fully in control of events; they take steps and make decisions that raise or lower the danger, but in a realm of risk and uncertainty.20Thomas C. Schelling, Arms and Influence (New Haven: Yale University Press, 1966), 97.

Finally, a crisis involves a temporal dimension where “policy-makers perceive themselves to be acting under time constraints.”21Lebow, Between Peace and War, 12. Studies employing simulations or experiments in laboratory-like environments to evaluate which components of a crisis have the most discernible effect on decision processes found that the amount of time available to make a decision was the most impactful variable.22Charles F. Herrmann, “Threat Time and Surprise: A Simulation of International Crisis,” in International Crises: Insights from Behavioral Research, ed. Charles F. Herrmann (New York: Free Press, 1972), 187-211. Standard information, coordination, and miscalculation problems inherent to state decision-making can be exacerbated under time pressure, intensifying uncertainty and the risk of war.

Based on this analysis, we define an interstate crisis as a decision point between peace and war in which a state perceives an intensification of a cross-border threat to national interests, heightened uncertainty, and time constraints and at least considers retaliation by force. We use this definition to evaluate whether empirical events between 1998 and 2016 can be categorized as actual “crises” (as perceived by India).

In the context of South Asia, misperceptions about military exercises or mobilization, fears of pre-emption, gray zone incursions, and cross-border attacks have precipitated crises. In recent years, the most commonly feared and analyzed South Asia crisis scenario arises from an attack by a violent nonstate group.23For example, see Toby Dalton and George Perkovich, Not War, Not Peace: Motivating Pakistan to Prevent Cross-Border Terrorism (New Delhi: Oxford University Press, 2016). Though India averages three terrorist incidents per day,24According to the Global Terrorism Database, India had 1,019 terrorist incidents in 2016. the Indian government generally considers itself to be in a crisis when it suffers an abnormally significant terrorist attack, which it often attributes (rightly or wrongly) to Pakistan. Beyond the inherent destruction and loss of life in an attack, major incidents are perceived as breakdowns in immediate and general deterrence that could invite further harm.25Shivshankar Menon, Choices: Inside the Making of India’s Foreign Policy (Washington, D.C.: Brookings Institution Press, 2016), 62, 67-70. These events galvanize Indian decision-makers to consider retaliatory responses but are shadowed by uncertainty and the risk of escalation to war between two nuclear powers. Pressure for making rapid decisions emerges from a desire to satiate the domestic audience, avoid crisis interveners, and to effectively communicate to external audiences its retaliation responses to the initial provocation. This particular type of crisis scenario is the subject of this study, though there are other potential crisis scenarios worth examining.

Why Study Crisis Onset?

A crisis offers a unique window for scholars and practitioners into international relations behavior because it “tends to galvanize, clarify, and concentrate many important elements in international politics, and to reveal the interaction between them more explicitly than in other empirical contexts.”26Snyder and Diesing, Conflict Among Nations, 5. Studying crises also helps practitioners draw lessons on how to manage, intervene, and potentially preempt future crises. If it can be rendered discernible, the threshold of “onset” in a crisis would inherently serve as the very first focal point to reduce risks of escalation.27For more on focal points, see Dan Altman, “Advancing without Attacking: The Strategic Game around the Use of Force,” Security Studies, forthcoming.

Scholarship on international crises disaggregates four distinct phases: onset, escalation, de-escalation, and impact.28Michael Brecher and Patrick James, “Patterns of Crisis Management,” Journal of Conflict Resolution 32, no. 3 (1988): 426-456; and Michael Brecher, “Crisis Escalation: Model and Findings,” International Political Science Review 17, no. 2 (April 1996): 215-230. While much contemporary work has covered the last three, onset studies appear less frequently, potentially because of their difficulty. The question for governments is knowing when they are in a crisis. Sometimes these events present themselves starkly – such as an attack like 9/11 – but sometimes they present themselves only gradually, such as the slow-moving refugee crisis of 2015 spurred by convulsive violence and collapsing regimes in the Middle East.

Crisis onset was thoroughly studied in the Cold War 1980s and 1990s by scholars trying to identify system effects because any international crisis could have become a flashpoint for great power intervention, competition, and conflict. Even if the United States and the Soviet Union were not the primary actors in a crisis, they still risked being drawn into a shooting war and all the risks of escalation to total nuclear war. Crises nested prior to and within the Korean War offer good examples of where secondary parties to the conflict chose to operate in secrecy for escalation control.29Austin Carson, “Facing Off and Saving Face: Covert Intervention and Escalation Management in the Korean War,” International Organization 70, no. 1 (2016): 103-31. The superpowers sought to anticipate and prevent the outbreak of major international crises to stave off a more serious confrontation.

In the post–Cold War era, interest in crisis onset decreased, at least for those entities concerned with major power war. The number of conflicts increased, but the risk of such events drawing in nuclear powers in a unipolar world seemed remote. Comparatively less attention is paid to the causes of crisis onset in international security politics than, for instance, the causes of international financial crises. The interests in systemic analysis of international crisis onset shifted to early warning and predictions of civil war onset,30James D. Fearon and David D. Laitin, “Ethnicity, Insurgency, and Civil War,” American Political Science Review 97, no. 1 (2003): 75-90. potentially for humanitarian concerns such as the U.S.-sponsored Political Instability Taskforce.

Today, the study of interstate crisis explores dynamics like escalation, crisis bargaining, management, coercion outcomes, and consequences but tends to treat the crisis itself as exogenous. Crises seem to emerge from the vicissitudes of international politics, and the work that does seek to explain onset tends to look for immediate triggers specific to individual cases rather than more systematic approaches. Thus, this essay seeks to account for crisis onset by examining the conditions under which seemingly inert military-political events transform into crises.

India-Pakistan Crises

In addition to analytical and conceptual value, there is a more practical application for unpacking crisis onset. The risks of escalation in nuclearized South Asia are severe. The United States has been involved in de-escalating and defusing nearly every major India-Pakistan crisis over the past 30 years.31See Sumit Ganguly and Devin T. Hagerty, Fearful Symmetry: India-Pakistan Crises in the Shadow of Nuclear Weapons (Seattle: University of Washington Press, 2005); Chari, Cheema, and Cohen, Four Crises and a Peace Process; and Nayak and Krepon, The Unfinished Crisis. In fact, some scholars have argued that triggering U.S. involvement in a crisis is inherently part of Pakistan’s strategic deterrent posture.32Narang, Nuclear Strategy in the Modern Era, 57-76.

The region’s history of close calls coupled with the stakes of nuclear-tinged crises provides sound justification to reexamine crisis onset, specifically in the dyad of India and Pakistan. While attempts at crisis prediction may be a fraught exercise, scholars and policymakers would profit from a closer understanding of why crises erupt in South Asia. Now that the prospect for nuclear-tinged crises are (re)emerging in other dyads like the United States and Russia and the United States and China, the findings from South Asia might yield useful mechanisms and process insights applicable to other regions anticipating crises.

The Subjectivity of Crisis: Perceptions and Political Incentives

India’s divergent reactions to the 26/11 and 7/11 attacks described earlier in the essay suggest a degree of subjectivity and agency in identifying a crisis. In fact, perception (and misperception) conditions all these components of a crisis and therefore constitutes an essential, if obvious, intervening variable between a provocation and crisis. Snyder and Diesing term this intervening stage a “precipitant,” which is when a state perceives an “intolerable situation” or an “especially provocative act” brought upon by an adversary.33Snyder and Diesing, Conflict Among Nations, 11.

The sense of tension within a crisis is subjectively felt within states even if it derives from some objective conditions or events.34Ibid., 9. There is an inherent subjectivity in that the precipitant functions as a convenient legitimizer and only a proximate cause. Snyder and Diesing observe:

When McGeorge Bundy was asked about the importance of the Communist raid on the Pleiku base triggering the start of U.S. bombing of north Vietnam, he replied “Pleikus are like streetcars” – one will come along eventually if you wait long enough.35Ibid., 11-12.

Different parties to a crisis can maintain different perceptions of the timing and intensity of it, or whether there was even a crisis to begin with.36Ibid., 17. For instance, some scholars and practitioners cast doubt on whether India and Pakistan were really concerned during the 1990 Compound crisis, while the United States perceived it as very serious and containing nuclear risks.37Chari, Cheema, Cohen, Four Crises and a Peace Process, 80-117 and Karthika Sasikumar, “Crisis and Opportunity: The 1990 Nuclear Crisis in South Asia” Nuclear Proliferation in South Asia: Crisis Behaviour and the Bomb, ed. Sumit Ganguly and S. Paul Kapur (New York: Routledge, 2009), 76-99. In other words, a provocative incident may not be perceived as abnormal or threaten national interests or values equally. Moreover, the sense of urgency for a decision or the beliefs about the degree of unpredictability or risk of war may vary tremendously.

India and Pakistan routinely engage in a choreography of cross-border fire across the disputed LoC without triggering a crisis. The thousands of incidents of annual cross-border fire between 1998 and 2002 slowed to a trickle after the 2003 ceasefire agreement. That agreement slowly eroded and over the past five years annual incidents have climbed again into the hundreds. Nevertheless, it is unclear if any India-Pakistan interstate crisis has erupted because of this cross-border firing activity, whereas on nearly any other border, this might constitute an act of war.

Some provocations might undeniably thrust a state into crisis, such as a militarized surprise attack like Japan’s on Pearl Harbor or Pakistan’s incursion into the Indian-controlled Kargil region in 1999. Other provocations might have no risk of generating a crisis, such as the routine incursion of animals across the Indian and Pakistani border, which are sometimes suspected by local police of espionage.38Elizabeth Frock, “Pakistan ‘Arrests’ Monkey for Crossing India Border,” Washington Post, December 5, 2011; and “‘Spy Pigeon’ Detained in India After Crossing Border from Pakistan,” Dawn, May 29, 2015. But between these extremes, a wide range of agency exists to select into a crisis, both due to varying perceptions of seriousness as well as varying political incentive structures.

Perception is a critical yet capricious variable that affords agency to one or more of the states in the strategic interaction of a potential international crisis. If a state does not perceive any or all elements central to a crisis, it may not be “alarmed.” This would render the candidate triggering event “inert” and indistinguishable from routine international interactions.

In addition to subjective perceptions, rational political logic may treat events and episodes like a crisis. For instance, if a provocation is public and galvanizes a domestic audience, states may have incentives to signal to their public they are taking the challenge seriously and at least deliberating over a response, if not executing some retaliatory actions.

By contrast, if a state subject to a provocation believes any reaction would generate a high probability of war and certain defeat because the other side possesses escalation dominance, it might sublimate its normative concerns about sovereignty and strategically choose not to acknowledge or select into a crisis so as not to incur the international reputation and domestic audience costs of “backing down.” Weaker actors may do this on a regular basis in response to aggressive moves by more powerful actors. Though the 2007 Israeli airstrike on Syria’s Al Kibar nuclear facility could have sparked a crisis, the Syrian government’s conspicuously muted response might have been based on a calculation that it had nothing to gain and much to lose from an aggressive retaliatory response.39Avner Cohen and Leonard Spector, “Israel’s Airstrike on Syria’s Reactor: Implications for the Nonproliferation Regime,” Arms Control Today, July/August 2008.

Knowing that perception and strategic calculations can mediate the identification or selection into a “crisis,” the question that follows is: When might a candidate crisis become an actual crisis? The puzzle described in the beginning of the essay suggests both South Asian states, but particularly India, have exercised a degree of agency in selecting into crises.

To explain this, a decision tree (Figure 1) attempts to map the entire process of how a crisis might unfold. What is generally treated as an exogenous condition can actually be conceived as a choice that states “enter into” by deeming an event transgressive. Triggers are treated by the literature as self-evident ex ante when in fact episodes like the anecdotes in the introduction suggest this is not the case. States may choose – whether publicly or privately – to identify events or actions as transgressive, potentially based on a strategic or political calculus. It is this first choice or first move that constitutes crisis onset. If and when the other party (or parties) acknowledges this first move, the states enter into a crisis.

Figure 1: Stages of Pre-Crisis, Crisis Onset, and Escalation

By examining a wide range of provocations or candidate trigger events – some of which result in crisis onset and some of which remain dormant – this essay (and larger project) aims to identify what features of a provocation correlate to crisis onset. With this finding we hope to unpack the political or strategic logics at work.

Research Design

The dependent variable this essay seeks to investigate is the well-studied event of an interstate crisis and, specifically, interstate crises between India and Pakistan. Many studies of interstate crises treat the crisis itself as an independent variable to explain other phenomena in international politics like interstate war,40Brecher, Crisis Escalation. as a unit of analysis to explain phenomena like crisis bargaining or compellence,41Todd S. Sechser and Matthew Fuhrmann, Nuclear Weapons and Coercive Diplomacy (Cambridge: Cambridge University Press, 2017). or focus on explaining the dynamics of historical crises.42Lebow, Between Peace and War. This study follows the third strain and specifically seeks to explain the onset of an interstate crisis. While most rich qualitative studies of historic crises generate a number of plausible explanations for crisis onset, dynamics, and termination, they tend to focus on what are widely regarded as crises in retrospect. The purposeful probing of known crises reveals a lot about the dynamics within crises but not much about their origins.43This is commonly referred to in social science as “selection on the dependent variable.” See Gary King, Robert O. Keohane, and Sidney Verba, Designing Social Inquiry: Scientific Inference in Qualitative Research (Princeton: Princeton University Press, 1994), 12937. To study crisis onset, we examined what we identify as a prior stage, the “provocation,” which may or may not become what Snyder and Diesing term a “precipitant” – a country’s active response to a provocation that then generates a crisis. The provocation serves as the unit of analysis for this study. All precipitants involve provocations, but not all provocations precipitate crises. We are interested in understanding what types or under what conditions provocations precipitate crises and when provocations remain dormant.


In an ideal world, this study would examine the correlates of precipitants that yield crises but also evaluate how they stack up against all the “dogs that do not bark” – episodes similar in most ways to crisis-yielding events that do not become crises. However, while such an approach would be methodologically sound, it is empirically very difficult as it would require uncovering a vast range of events like diplomatic spats and routine military exercises that are generally kept secret or remain unobservable for data collection.

Consequently, we bound the observations in our dataset to a set of cross-border provocations, mostly by violent nonstate actors, because of their conspicuousness.We generated a list of 51 observable provocations based on actions perpetrated against Indian targets from 1998-2016 by militant organizations with distinctly cross-border origins and links.44These include the Haqqani Network, Harkat ul-Ansar, Harkat-ul-Jihad Islami (HUJI), Harkut-ul-Mujahideen, Indian Mujahideen (IM), Jaish-e-Mohammed, and Lashkar-e-Taiba (LeT). It is important to clarify why we consider a number of observations where IM was believed to be the perpetrator of the attack as “cross-border.” There is some debate over whether IM is an affiliate of LeT, interlinked, or one and the same. (Praveen Swami, “The Indian Mujahidin and Lashkar-i-Tayyiba’s Transnational Networks,” CTC Sentinel 2, no. 6, June 2009.) At times IM was acting at the behest of LeT, at other times, they acted semi-autonomously. (Stephen Tankel, Storming the World Stage: The Story of Lashkar-e-Taiba (Oxford: Oxford University Press, 2013), 143.) What is clear is that Indian intelligence (and consequently Indian media) either portrays them as intertwined or simply attributes all provocations by IM to LeT, especially in the immediate aftermath of an attack. (C. Christine Fair, “Students Islamic Movement of India and the Indian Mujahideen: An Assessment,” Asia Policy, no. 9 (2010): 101-19.) Even if in retrospect, these groups have been re-identified as distinct, the treatment and perception in the immediate aftermath of a provocation is what is most relevant, because it is that period that can easily trigger crisis onset. For instance, in the immediate aftermath of the 2008 IM attacks in Jaipur, Ahmdebad, and Bangalore, these were publicly considered by Indian Intelligence Bureau officers to be jointly planned by LeT and HUJI and executed by IM. (Vicky Nanjappa, “How the Indian Mujahideen Was Formed,” Rediff News, July 29, 2008.) IM is considered “the Indian arm of the Lashkar-e-Taiba.” (Diana George, ed., “Yasin Bhatkal Allegedly Behind These Deadly Attacks in India,” NDTV, August 30, 2013.) This is because Indian intelligence believed it had uncovered evidence of the “Karachi Project,” in which groups of Indian nationals like IM were conducting attacks on Indian urban centers, essentially as arms of LeT. (Animesh Roul, “After Pune, Details Emerge on the Karachi Project and its Threat to India,” CTC Sentinel 3, no. 4, April 2010.) Even if they considered IM an independent organization, another reason Indian intelligence officials believed these were cross-border attacks was that they suspected the IM leadership had been trained by Pakistan’s Inter-Services Intelligence at LeT camps (“Yasin Bhatkal Trained by Lashkar in Pakistan,” The New Indian Express, September 1, 2013.), and because the perpetrators immediately fled to Pakistan after the attacks. (Nanjappa, “How the Indian Mujahideen Was Formed.”) Even though domestic organizations like the Communist Party of India (Marxist) or United Liberation Front of Assam perpetrate terrorist attacks that certainly serve as provocations to the Indian state, they are not international provocations that generate the risk of an interstate crisis. In the dataset generated, all but one of the observations were attacks perpetrated by nonstate groups believed by India to be operating with material support from the Pakistani state; thus, they all bore the potential for generating an India-Pakistan interstate crisis.45Attribution is a difficult factor to determine with complete confidence and several provocations in our dataset are still disputed or being tried in Indian courts. For the purpose of this study, however, the Indian government’s belief in probable attribution in the immediate aftermath of a crisis is the salient factor. One former U.S. official commented to us that at some point the Indian government stopped worrying about the attribution problem because it believed “as a matter of faith” the attacks stemmed from Pakistan. That makes all these incidents like enough units to be included in a single dataset.

Identifying India-Pakistan “Crises”

While this study treats “provocation” as the unit of analysis, the dependent variable of interest is interstate “crisis.” To identify specific instances of crisis (at least from India’s vantage point) from this broad set of provocations, we looked for event details that met our criteria for a crisis described above – that is, perceptions of a challenge or threat, abnormal intensity, anxiety about the chance of violence or war, and time and information constraints. Crisis onset is meant to be distinct from crisis intensity, which derives in part from the choice for and type of retaliation.46Thus crisis intensity is the subject of Krepon and Dowling’s essay in this volume.

An event was coded as a case of “crisis” if it met two objective criteria.47We believe these two criteria to have inter-coder reliability. First, it was either previously coded as a crisis by the International Crisis Behavior dataset (ICB),48Michael Brecher, Jonathan Wilkenfeld, Kyle Beardsley, Patrick James, and David Quinn, International Crisis Behavior Project, version 12 (Durham and Los Angeles: Duke University and University of Southern California, 2016), data-collections. previously coded as a crisis by reputable South Asian security scholarship,49Chari, Cheema, and Cohen, Four Crises and a Peace Process; Nayak and Krepon, The Unfinished Crisis; and Ganguly and Hagerty, Fearful Symmetry. or intuitively met our definition of crisis delineated above. Second, India’s national security principals were convened through an unplanned Cabinet Committee on Security (CCS) meeting within a week of the event. (from which we infer a consideration of retaliation by force). (See Table 1)

Table 1: Observations of India-Pakistan Interstate Crisis, 1998-2016
Crisis Event Meets “Crisis” DEFINITION International Crisis Behavior (ICB) Database Coded “Crisis” In Past Analysis Crisis-Triggered Cabinet Committee On Security (CCS) Meeting Date* Sources For CCS Meeting
1. Nuclear tests Chari, Cheema, and Cohen, 2007, 19. 5/11/98 India Today, May 25, 1998
2. Kargil crisis Chari, Cheema, and Cohen, 2007, 2. 5/25/99 The Times of India, May 27, 1999
3. IC 814 hijacking Chari, Cheema, and Cohen, 2007, 152-53; Dulat, Kashmir, 36-39. 12/29/99 The Times of India, May 27, 1999
4. Red Fort attack Ganguly and Kapur (eds), 2008, 148. 12/23/00 Frontline, January 6, 2010
5. Kashmir Provincial Assembly attack Krepon and Nayak, 2006, 49. 10/1/2001** The Independent, October 2, 2001
6. Indian Parliament attack Krepon and Nayak, 2006, 10; Chari, Cheema, and Cohen, 2007, 149-53, 12/13/01 The Times of India, December 16, 2001
7. Kaluchak massacre Krepon and Nayak, 2006, 54; Chari, Cheema, and Cohen, 2007, 169. 5/18/02 The Times of India, May 19, 2002
8. Jaipur bombings - 5/14/2008** The Times of India, May 20, 2008
9. “26/11” Mumbai attack Krepon and Nayak, 2012, vii. 11/27/08 Hindustan Times, November 28, 2008
10. Pune Bakery bombing - 2/15/10 Rediff, February 15, 2010; Indo-Asian News Service, February 15, 2010; The Times of India, February 16, 2010
11. Varanasi bombing of Sheetla Ghat - 12/10/10 The Indian Express, December 8, 2010
12. Uri Indian Army base attack NA*** Curtis, 2016. 9/21/16 Mint, September 21, 2016; Financial Express, September 30, 2016
*Only includes first CCS that occurred after attack; **Estimated based on media reports; *** Database ends in 2015


For the second criteria, a key indicator that India perceived a crisis seems to be the emergency convening of a CCS meeting. Analysts and former military officials confirm the CCS to be “the apex body responsible for all matters impinging on India’s security…”50Gurmeet Kanwal, “Command and Control of Nuclear Weapons in India,” Strategic Analysis 23, no. 10 (2000): 1,719. Some have likened this to convenings of principals in the White House Situation Room on sensitive crisis decision-making. The CCS is the locus of consideration for retaliation or response, whether through the use of force or through other means. Though imperfect, the CCS may be the best public indicator that crisis decision-making is in the offing. Brigadier Gurmeet Kanwal writes,

During the May-August 1999 Kargil conflict the CCS was reported to have met quite often. It is not known how often the NSC was convened. It could justifiably be deduced that in practice, the CCS is now discharging the functions of political guidance and oversight in the higher direction of war.51Ibid., 1,720.

A CCS meeting was therefore a plausible indicator of the consideration of punitive or coercive action – e.g., the use of force – that could escalate to war, and the public revelation of a CCS meeting suggests India’s intention to signal to international audiences. Certainly, a CCS meeting is not an infallible indicator of India’s perception of a crisis, but for the sake of this study, we make the assumption it was (and remains) a necessary one.

These criteria were used to generate a list of 12 crises from our initial set of 51 provocations. Several of these were high-profile crises but others have heretofore gone unrecognized by previous studies or cross-national datasets. Some instances like the post-1998 nuclear tests neatly fit the definition of a crisis and are recognized by datasets like the ICB, while others like the 1999 India Airlines hijacking or 2000 attack on the Red Fort are recognized by practitioners and scholars but not closely studied for their escalatory potential.52A.S. Dulat, Kashmir: The Vajpayee Years (Noida: Harper Collins India, 2015), 36-39; and Praveen Swami, “A War to End a War: The Causes and Outcomes of the 2001-2 India-Pakistan Crisis,” Nuclear Proliferation in South Asia: Crisis Behaviour and the Bomb, ed. Sumit Ganguly and S. Paul Kapur, (New York: Routledge, 2009) 148.

The final step in our analysis was to code each provocation for a host of properties to identify which, if any, correlated with the provocations that actually became crises. In the following sections, we review a set of hypotheses generated from the literature on South Asian crises. We also describe how we operationalized these hypotheses with proximate measures of each concept. Because the number of observations is too small for regression analysis, we perform a series of cross-tabulations of different potential explanatory variables as well as Pearson chi-squared tests to identify if there appears to be any relationship between the various correlates of provocations (independent variables) and crisis onset (dependent variable).53Pearson chi-squared tests are essentially tests of independence. Their purpose is to evaluate whether the null hypothesis (that the pair of variables is totally independent, with no relationship) can be falsified. For a more detailed explanation, see our Results section.

Hypotheses on Crisis Onset

Analysis of crises in South Asia tends to focus on the anatomy of already full-blown crises and on how Indian and Pakistani leaders, as well as third parties like the United States, subsequently manage them.54See for example: Ganguly and Hagerty, Fearful Symmetry; Ganguly and Kapur, Nuclear Proliferation in South Asia; Michael Krepon and Nathaniel Cohn, Crises in South Asia: Trends and Potential Consequences (Washington, D.C.: Stimson Center, 2011); and Polly Nayak and Michael Krepon, US Crisis Management in South Asia’s Twin Peaks Crisis (Washington, D.C.: Stimson Center, 2006). Qualitative overviews that contextualize crises in their historical and geopolitical climates have provided valuable contributions to understanding conflict dynamics, but they often employ implicit hypotheses of crisis triggers that become conflated.55See, for example, Nayak and Krepon, The Unfinished Crisis; chapters in Zachary S. Davis, ed., The India-Pakistan Military Standoff: Crisis and Escalation in South Asia (New York: Palgrave Macmillan, 2011); Devin T. Hagerty, “Nuclear Deterrence in South Asia: The 1990 Indo-Pakistani Crisis,” International Security 20, no. 3 (1995): 79-114; and V.R. Raghavan, Siachen: Conflict without End. (New Delhi: Viking, 2002), 33-57. To explain crisis onset, we studied international relations and South Asia conflict literature for candidate explanations on how crises begin. Finding these to be limited in scope and underdeveloped, we also sought accounts by journalists, government officials, and retired government practitioners. Writing on South Asian crises generally fails to account for the empirical puzzle that motivated our research. When asked what causes crisis onset, most regional experts believed the question’s answer was straightforward and proceeded to offer a host of different – and sometimes conflicting – explanations.

The diversity of explanations underscored the need for a more systematic approach to understanding onset. Below we try to distill, make explicit, and unpack the underlying logic of 10 of the more prevalent of these formal and informal hypotheses. We evaluate which ones might offer the greatest explanatory power as to when, why, and – recognizing there are a host of contingent historical and contextual features

– how provocations might precipitate crises. From a review of the components of a crisis and the variables at play, we distill three families of hypotheses that explain when a provocation or candidate “trigger” is more likely to escalate into a crisis. Each family of hypotheses explores variables that affect perceptions of one of three particular features of a crisis: its abnormality, threat intensity, or time sensitivity.

Each of the hypotheses starts from the assumption that a provocation involves or threatens kinetic violence. There are certainly a range of potential other nonviolent triggers, but because of the sheer difficulty evaluating nonviolent candidate triggers like economic or diplomatic aggression, we make this simplifying assumption. First, we posit that the perception of abnormality might vary by the provocation’s choice of target, the level of casualties, the complexity of tactics, and/or the duration of the provocation episode. Second, threat perceptions might vary on the geography of the threat, the perceived degree of the initiator state’s intentionality, and/or the initiator state’s regime type. Finally, perceptions of time sensitivity and risk might vary based on the intensity of media coverage, the type of government, and the type of leader. In the subsections that follow, we describe 10 hypotheses that roughly fit into one of these three aspects of crises.56Ultimately, these are tests of correlation, not causation. Moreover, these variables are not unrelated; indeed, they may interact, and their combined impact on decision-makers may vary. In treating media coverage as a variable shaping a state’s perception of time sensitivity and risk, for example, we recognize that media coverage itself is contingent on other variables surrounding the provocation and how “spectacular” it is – duration, target, fatality levels, etc. Any correlation between high media coverage and crisis onset may be more of a red flag indicator of crisis rather than an independent cause. Future multivariate regression analysis can help to disentangle these effects. We thank Arzan Tarapore for this insight.


Because terrorism and insurgency are frequent phenomenon in South Asia’s political landscape, the use or threat of political violence must be clearly abnormal – the word “spectacular” is often applied57Daniel Markey, No Exit from Pakistan: America’s Tortured Relationship with Islamabad (Cambridge: Cambridge University Press, 2013), 46, 101; Steve Coll, “The Stand Off,” The New Yorker, February 16, 2006; Somini Sengupta, “Crisis May Shift India’s Political Landscape,” The New York Times, November 28, 2008; and Michael Krepon, “The Myth of Deterrence Stability Between Nuclear-Armed Rivals,” in Deterrence Instability and Nuclear Weapons in South Asia, ed. Michael Krepon, Joshua T. White, Julia Thompson, Shane Mason (Washington, D.C.: Stimson Center, 2015), 26-27. – to trigger a crisis. We measure the abnormality or deviance of a crisis provocation by the following variables: casualty levels, target type, complexity of tactics, and the duration of the provocation.

Fatalities. A first potential contributor to a provocation’s abnormality is the number of fatalities in a single incident. This last aspect is important for unpacking why concentrated loss of life can have a greater effect. After all, car accidents dispersed over time and area kill over a million people annually, but a concentrated event like a natural disaster that kills a few hundred people tends to elicit a crisis atmosphere and strong reactions. It is intuitive to then expect that a high level of fatalities in a provocation would constitute an aberration or deviation from the norm. It is important, however, to understand that terrorist incidents are a frequent occurrence in India. From 2011-15, the Global Terrorism Database estimates India annually averaged 738 terror incidents and 370 nonmilitant fatalities.58Most scholars of South Asian political violence believe this systematically undercounts the level of nonstate actor violence, and other datasets like the South Asia Terrorism Portal estimate higher annual averages. Thus, deviation from the mean would require a departure from this baseline of two terror incidents and one terror fatality daily.

Countless accounts by journalists, former government officials, and analysts use deviance in casualty levels to explain past India-Pakistan crises and implicitly suggest that in future provocations, a high death toll will automatically trigger a crisis.59See for example, Toby Dalton and George Perkovich, “Is a Pakistan-India War Just One Terrorist Attack Away?” Herald, January 23, 2017. For example, one scholar contends the 2001 Parliament attack was “a lot less provocative than Mumbai” because “that attack failed” and “about five people died.”60S. Paul Kapur quoted in Adam Gorlick, “Q&A: South Asia Security Expert Discusses Terrorist Attacks in Mumbai,” Stanford Report, December 4, 2008, By contrast, the scholar contended there was “a lot of pressure domestically for the government to act in a forceful way” after Mumbai because “this attack killed almost 200 people, wounded hundreds more, lasted almost three days and targeted the financial hub of India.” Because of the attack’s lethality, he argued, “[t]here’s going to be a lot of pressure domestically for the government to act in a forceful way.”61Gorlick, “Q&A: South Asia Security Expert Discusses Terrorist Attacks in Mumbai”

The provocativeness of high fatality levels seems intuitive yet merits closer examination given anecdotal inconsistencies. A number of widely reported, high-fatality events in India occurred in the early 2000s without triggering a crisis. Among these “almost crisis events” are the above-cited example of the 2006 Mumbai train bombings, which resulted in more civilian casualties than the infamous 2008 Mumbai attack on November 26 that sparked a crisis.

Hypothesis 1: A provocation resulting in high fatalities should be associated with a higher risk of crisis than a provocation resulting in low fatalities.

All the variables for this study were treated in dichotomous terms. Lethality was measured to be “high” with 10 or more nonmilitant fatalities and “low” with under 10 nonmilitant fatalities.

Target Type. While some targets of provocations are considered “fair play,” such as the choreographed exchange of fire along the LoC, other targets might be perceived as exacting an unacceptable cost and result in escalation toward a crisis. One would anticipate provocations like militant attacks that target civilians or soft targets to be more deviant compared to violence against hardened or military targets.62See for example, Dalton and Perkovich, “Is a Pakistan-India War Just One Terrorist Attack Away?” Many analysts have argued that the reason provocations in 2001, 2002, and 2008 escalated into the 2001-2 Twin Peaks and 2008 Mumbai crises is because targeting a civilian government installation, military families, and tourists in a luxury hotel were beyond the pale. One veteran reporter explains:

Indian investigators had traced the Pakistani hand in both the March 1993 Mumbai serial blasts and the July 2006 suburban train bombings. These attacks had killed more people –257 and 187 respectively. But 26/11 was different. It was the first attack carried out by Pakistani nationals who hit civilian targets and foreign nationals with calculated brutality.63Sandeep Unnithan, “Why India Didn’t Strike Pakistan after 26/11,” India Today, October 14, 2015.

We apply a dichotomous measure of whether a provocation’s target was a civilian or a military/security force target to assess whether this factor is associated with escalation to crisis.

Hypothesis 2a: A provocation that targets civilians should be associated with a higher risk of crisis than a provocation that targets military or security forces.

Another variant of targeting creating a sense of abnormality is if an iconic feature of the country is targeted. In such a case, even an attack lacking in other potentially escalatory attributes might prove immensely provocative. Attacks on iconic national targets can challenge the very bulwark of the country and the state. The 2001 attack on the Indian Parliament is said to have been “the most audacious and most alarming act of terrorism” on India in over two decades64Nayak and Krepon, US Crisis Management in South Asia’s Twin Peaks Crisis, 16. because it constituted “an assault upon the very foundations of Indian democracy.”65Chari, Cheema, and Cohen, Four Crises and a Peace Process, 151. Regarding the similar symbolic importance of the 2008 Mumbai attack, Riedel argues, “Mumbai is the symbol of India’s economic and cultural life. Mumbai is as valuable a target as Washington, London or Berlin and has been a target on many occasions. Foreigners, diplomats and financial entrepreneurs were the targets.”66“Terror Expert and Obama Advisor Bruce Riedel: ‘A Nightmare We Cannot Afford in the 21st Century,’” Spiegel, December 8, 2008. Some analysts have even argued that provocations that threaten or destroy core national military assets (like combat aircraft) constitute much greater deviance from the mean.67Shashank Joshi and Praveen Swami independently suggest that Pathankot was not a crisis like Kulchak because the militants didn’t hit major assets (like combat aircraft) or kill civilians like family of military personnel. See Shashank Joshi, “Pathankot Attack: India-Pakistan Peace Talks Derailed?” BBC News, January 7, 2016; and Deeptiman Tiwary, Sagnik Chowdhury, Pranav Kulkarni, and Praveen Swami, “Probing Pathankot Terrorist Attack: How Wires Got Crossed in Delhi,” The Indian Express , January 9, 2016.

Hypothesis 2b: A provocation that targets iconic sites or symbols of the nation should be associated with a higher risk of crisis than a provocation targeting other sites.

Target types were measured in two ways – attacks that hit soft civilian rather than security force targets (military and police) and those that targeted iconic rather than non-iconic targets in India.

Geography. The location of a provocation – whether it strikes peripheral or core territory – can impact the probability of a crisis. International relations and civil war scholarship finds geography to be a core determinant of threat perception.68Stephen M. Walt, The Origins of Alliance (Ithaca: Cornell University Press, 1987); Monica Duffy Toft, The Geography of Ethnic Violence: Identity, Interests, and the Indivisibility of Territory (Princeton: Princeton University Press, 2003); and Sameer Lalwani, “Selective Leviathans: Explaining State Strategies of Counterinsurgency and Consolidation” (Ph.D. Dissertation, Massachusetts Institute of Technology, 2014). Some interstate crisis scholarship has found geographic distance, measured as “close to home” versus “more distant,” strongly influenced states’ crisis proneness.69Michael Brecher, “Crisis Escalation: Model and Findings,” International Political Science Review 17, no. 2 (1996): 215-30.

The 2001 and 2008 attacks were particularly impactful because they attacked India’s core rather than its periphery. Krepon explains:

The Twin Peaks and 2008 crises differed from the 1990 and Kargil crises in that the triggering events occurred far from the Indian state of Jammu and Kashmir. Instead, the flashpoints for these two crises were mass-casualty assaults directed against high-profile targets in New Delhi and Mumbai that represented India’s political integration, economic advancement and connectivity to the globe.70Krepon and Cohn, Crises in South Asia, 5.

Part of the reason attacks on core geography are so incendiary is the number of people they can affect and the erosion of trust in the state. One scholar writes:

The loud message was that a small group of individuals could turn a megalopolis of almost 15 million inhabitants into a battlefield for at least a day. All the time, the intention is to terrorise the largest number of people, eroding the ordinary man’s confidence in the ability of the authorities to protect him, and, in the long run, persuading a majority of the people, who just want to live their lives, to trade their freedom for the security that the terrorist promises in his utopia.71Amir Taheri, “Mumbai Attacks: The Terrorists’ Tactics,” The Telegraph, November 27, 2008.

Some Indian observers have given serious thought to how the geographic location of a provocation might influence the perception of crisis and dictate India’s response.72Ranjan Roy, “Where’s the Battle Cry?: Terrifying Pacifism; Does our government suffer from passive transient reactivism?” Times of India, November 6, 2005, 12. Some research suggests that Indian territory can be divided into spaces of order and disorder: core areas where violence is more costly and scary, and peripheral areas where it is expected and discounted.73Paul Staniland, “States, Insurgents, and Wartime Political Orders,” Perspectives on Politics 10, no. 2 (2012): 243–64; and Sanjib Baruah, Durable Disorder: Understanding the Politics of Northeast India (New Delhi: Oxford University Press, 2005). Contested regions of India like the Kashmir Valley or volatile Northeast India regularly exhibit disorder and violent contestation, while areas in the mainland of India or the metropolitan zones might be considered safe spaces. These perceptions might give rise to a belief that attacks on ordered spaces will be particularly provocative while those in disordered spaces will not.74Between 2012 and 2015, 12.4 percent (378 of 3,047) of the terror attacks in India recorded by the Global Terrorism Database took place in Jammu and Kashmir. Other concentrations were in the northeast, Bihar, Jharkhand, West Bengal, Chhattisgarh, and Odisha. In 2015, only 14 attacks took place in Punjab. Gurdaspur and Pathankot, the locations for two of the most recent cross-border attacks in 2015 and 2016, are in Punjab but right on the border of Jammu and Kashmir. See “Global Terrorism Database,” National Consortium for the Study of Terrorism and Responses to Terrorism, 2016.

Hypothesis 3: A provocation on core geography should be associated with a higher risk of crisis than a provocation on peripheral geography.

Provocations in peripheral, contested geography were measured as attacks in Jammu & Kashmir or India’s Northeast75There were no observations from the Northeast, but this is identified for the sake of clarity. while provocations in core geography included major urban centers and much of the Indian heartland and Hindi belt.

Provocation Type.

Aside from location, targeting, and lethality, provocations may precipitate crises due to their style or type. Specifically, complex, coordinated militant assaults require significant communication, training, planning, organization, and resources. These attacks are much harder to carry out than bombings and therefore much more abnormal and threatening than a single shooter or a car bombing.76See for example comparative analysis of attack types and tactics, Brian J. Phillips, “This Is Why the Paris Attacks Have Gotten More Coverage Than Other Attacks,” The Washington Post, November 16, 2015. Similarly, a complex gray zone infiltration will be much more provocative than a single mortar round fired across a border because it signals a distinct capability.

There is a growing body of literature supporting the idea that diverse and complex tactics are more effective and perceived by a state as more challenging and threatening.77Michael C. Horowitz, Evan Perkoski, Philip B. K. Potter, “Tactical Diversity in Militant Violence,” International Organization 72, no. 1 (2018). Thus, the 2008 Mumbai attack (26/11) involving a complex assault with five sets of two-man teams coordinated by an operations cell is roundly perceived as much more abnormal than the June 2006 Mumbai train bombings (7/11) that killed nearly as many people. Veteran intelligence analyst Bruce Riedel explained the reason the 2008 Mumbai attack was so provocative was because of its complexity. “This kind of attack does not appear to be the work of amateurs,” he said in an interview soon after the attack. “The sophistication of the use of multiple teams of very well-trained killers, as well as the choice of targets – Americans, Brits, Israelis – has all the hallmarks of the global jihad...”78“Terror Expert and Obama Advisor Bruce Riedel.”

In addition to signaling sophisticated planning and organization, the complexity of an attack can augment fear. A review of terrorist attacks in India between 1998 and 2004 found that, on average, armed assaults resulted in higher fatalities than other tactics, including bombings and imposed higher social loss.79Vani K. Borooah, “Terrorist Incidents in India, 1998-2004: A Quantitative Analysis of Fatality Rates,” Terrorism & Political Violence 21, no. 3 (2009): 476-98. Because of the level of fear and potential social impacts a complex attack can generate (such as depressing social interactions and public gatherings, which can negatively impact economic activity), a state might expect that complex attacks are qualitatively different and more provocative regardless of the actual body count. Another analyst explained that Mumbai differed from bombings or suicide attacks because, “[t]his time…the approach was ‘symphonic,’ in the sense that it involved different types of operations blended together. Involved in the operations were men who had placed explosives at selected points. But there were also gunmen operating in classic military style by seizing control of territory at symbolically significant locations along with hostages.”80Taheri, “Mumbai Attacks.”

Hypothesis 4: A complex provocation should be associated with a higher risk of crisis than a simple provocation.

Complex provocations involved multiple assailants engaged in coordinated, small-unit assault tactics, while simple provocations included remote bombings, single shooters, and suicide bombers.

Duration. Closely related to the complexity hypothesis is one related to the duration of a provocation. Scholars of interstate crises argue that the longer the duration of a crisis the more likely it is to get out of control.81Snyder and Diesing, Conflict Among Nations, 492. Following from this, one can also expect that the longer a provocation the higher the prospect that it transforms into a crisis. The longer a provocation, the greater the conspicuousness of the event and the more likely a wider domestic audience will observe it and contribute to escalatory pressures from fear, outrage, or embarrassment.

One of the most salient and distinct features of the Mumbai attack was its length. Noting the difference in duration between the 2006 and 2008 attacks, former Indian National Security Advisor Shivshankar Menon noted, “[o]ver three days [LeT] killed 166 people and wounded at least 308” whereas “seven bomb that exploded in eleven minutes on Mumbai suburban trains…had killed 209 and injured more than 650 people.”82Menon, Choices, 60-61. One Washington D.C.-based South Asia analyst observed that “it wasn’t a surprise that there was an attack…but the length and extent – [that] was all surprising.”83Nayak and Krepon, The Unfinished Crisis, 7.

26/11 unfolded on live television over 60 hours of what one journalist described as a “slow-motion nightmare of methodical and remorseless violence.”84David Randall, “Mumbai: Sixty Hours of Terror and Chaos,” The Independent, November 30, 2008. The crisis’ length raised public and international attention, frustration, anger, and even pressure. It also served as a “major embarrassment for India’s security establishment” and the Indian state because its ham-handed and dysfunctional response were televised to both the general public and, worse, to Indian adversaries.85Erika Kinetz, “India Executes Surviving Gunman of 2008 Terrorist Attacks,”>, November 21, 2012. While all terrorist attacks expose state vulnerabilities and intelligence breakdowns, a longer provocation heightens general public alarm and generates both fear and outrage at the inadequacy of the government response.

Duration might be closely correlated with complexity, but it need not always be. Complexity captures something about the provocation alone, while duration is a dyadic product that results not only from the instigator but also the target’s ability to respond. Quick, effective responses to complex provocations can reduce duration. In a way, duration can also correlate with the level of embarrassment of the host government, adding further fuel to the potential for a crisis.

Hypothesis 5: A provocation of long duration should be associated with a higher risk of crisis than a provocation of short duration.

Duration was measured as “long” if the provocation lasted longer than an hour and “short” if less than an hour. To be clear, duration was only capturing the length of the provocation itself, distinct from the potential crisis that may have followed.

Threat to National Values

A second feature of a crisis is whether or not it constitutes a threat to national values. The key variables that can shape threat perceptions are assessments of the provocateur’s intentions based on their level of control or regime type, as well as the relative hawkishness of the Indian government and its leadership.

Intentionality via Control. A key factor in a potential crisis’ threat to national values is whether the provocation was perceived as deliberate and intentional. States are often aware of one of the most dangerous risks of misperception where an action is seen as “more centralized, disciplined, and coordinated than it is”86Robert Jervis, “Hypotheses on Misperception,” World Politics 20, no. 3 (1968): 454-79. and may discount provocations if they believe them to be inadvertent or stemming from a potential principal-agent problem.

The distinction may be between actions and entities that can be directly attributed to the state and those that are plausibly deniable. For instance, a deliberate probe of a border by regular military units might create pressures for a state to confront and retaliate against the incursion, but an irregular unit that patrols aggressively and happens to cross a border by accident or because of a hotheaded platoon commander might precipitate a stern rebuke but not a full crisis.

There is always a level of uncertainty in the attribution of intent or principal control, but states often rely on certain cues from their intelligence to assess the deliberateness of a provocation. States must maintain a tricky balance of maintaining awareness of salami slicing tactics by an adversary that are intentionally designed to look too small, accidental, or benign to warrant action; however, they also cannot overreact to every incident. Overreactions are costly and can lead to a “boy who cried wolf” syndrome that desensitizes domestic and international audiences.

While there has always remained a question about the degree of influence Pakistan wields with respect to various militant groups that operate from its soil, many analysts tend to believe the Pakistan strategic establishment maintains “reasonably tight control” over LeT.87C. Christine Fair, Fighting to the End: The Pakistan Army’s Way of War (Oxford: Oxford University Press, 2014), 251. Even former members acknowledge that LeT is “tamed by the ISI [Pakistan’s Inter-Services Intelligence]” and thus functions as the closest thing to a direct instrument of policy because it was built and trained by the Pakistan military.88Tankel, Storming the World Stage, 2013, 55-61, 265. (quotation from 265.) Other groups, while patronized by Pakistan, operate with greater distance and autonomy (like the arguably local Hizbul Mujahideen) or have been known to go rogue and even turn on the state– like Jaish-e-Mohammad (JeM). For instance, the displeasure signaled by the ISI after JeM’s 2001 Parliament attack (inferred from the high degree of signals traffic) as well as assassination attempts on Gen. Pervez Musharraf, then the Pakistan Army chief and president, suggested that the Pakistani state wielded less control over those groups than expected.89Fair, Fighting to the End, 252.

Indian strategists in particular tend to consider the Pakistan state responsible for actions perpetrated by LeT. While U.S. analysts and intelligence officials cautioned that LeT involvement and Pakistan’s direct responsibility were distinct, India perceived direct control by the Pakistan Army. During the Mumbai attack, Indian officials suggested the “close relationship between the Lashkar and the ISI” implied “clear and incontrovertible proof” that the provocation was planned and directed by the Pakistan government.90Emily Wax and Rama Lakshmi, “Indian Official Says Pakistan’s ISI Trained, Supported Mumbai Attackers,” The Washington Post, December 6, 2008; and Adam Gorlick, “Q&A.” After the 2008 LeT attack, Home Secretary

G.K. Pillai alleged that Pakistan directly controlled LeT’s actions: “It was not just a peripheral role…They [the ISI] were literally controlling and coordinating it from the beginning till the end.”91“Pakistan ISI Behind Mumbai Attacks: Indian Official,” Reuters, July 14, 2010. Consequently, National Security Advisor Menon called for “immediate visible retaliation of some sort, either against the LeT in Muridke, in Pakistan’s Punjab province, or their camps in Pakistan-occupied Kashmir, or against the ISI,” implying the LeT and ISI were equivalent.92Menon, Choices, 61

Hypothesis 6a: A provocation by a group perceived to be directly controlledby the Pakistani state should be associated with a higher risk of crisis thana group that is not perceived to be directly controlled by the Pakistani state.

Control was measured as high if the perception of LeT’s involvement in an attack was observed.

Intentionality via Regime Type. A second way that threat to national interest might be measured and intentionality inferred is whether the perceived initiator of the provocation is a democracy or autocracy. Hewitt finds that like with the study of militarized interstate disputes, the nature of the two regimes has a statistically significant effect on the onset of an international crisis. Two democratic regimes are less likely to enter into a crisis with each other.93J. Joseph Hewitt, “Dyadic Processes and International Crises,” Journal of Conflict Resolution 47, no. 5 (October 2003): 669-92. States might have more faith in the intentions of democracies and be more inclined to treat provocations as arising from a lack of control or accident.

In the case of South Asia, Indian democracy has been a constant, but scholars have long described the vacillation between Pakistan’s explicitly and overtly military authoritarian regimes and periods when the Pakistani state is divided between nominally civilian governments with the military operating as the deep state. In the latter periods, civilians still exercise some power but without full control.

A prevailing theory is that authoritarian regimes – especially personalist regimes – are more prone to conflict initiation, because they face fewer constraints than democratic regimes with large coalitions.94Mark Peceny and Christopher K. Butler, “The Conflict Behavior of Authoritarian Regimes.” International Politics 41, no. 4 (2004): 565-81; Jessica Weeks, “Strongmen and Straw Men: Authoritarian Regimes and the Initiation of International Conflict,” American Political Science Review 106, no. 2 (May 2012): 326-47. Additionally, military-led governments may also lean toward conflict initiation due to the beliefs, incentives, and routines that make militaries inherently prefer offensive strategies.95Barry Posen, The Sources of Military Doctrine: France, Britain, and Germany between the World Wars (Ithaca: Cornell University Press, 1984); and Brian Lai and Dan Slater, “Institutions of the Offensive: Domestic Sources of Dispute Initiation in Authoritarian Regimes, 1950-1992,” American Journal of Political Science 50, no. 1 (2006): 113-26.

Consequently, India may have treated provocations more seriously when Pakistan was under military rule because it may have anticipated perverse incentives in its adversary to intensify and escalate a conflict. This may have accounted for the escalation in 2001 where the Indian government felt the need to show the military-led Pakistan government it would not be bullied.

The Indian government might discount culpability for a provocation that occurs during civilian rule in Pakistan, judging that the aggression was unauthorized by the civilian government and perpetrated by the military to undermine civilian leadership. Under such circumstances, the Indian state might calculate that a crisis and a potential military engagement would further harm the civil-military imbalance in Pakistan in favor of the military, potentially increasing risks to Indian interests. By the 2008 Mumbai attack, the civilian government in Pakistan was so new that one of Prime Minister Manmohan Singh’s advisors suggested that India had a vested interest in not weakening an already weak civilian government in Pakistan that had just emerged from a significant struggle with the military two months before the attack.96Taking punitive military action would only have served to weaken the new civilian government in Pakistan that Singh had worked so hard with to develop trust and cooperation. See Menon, Choices, 63. Thus, the Indian state might be less prone to react to potential provocations when facing a civilian government, but more inclined to escalate a provocation into a crisis when confronting an overtly military government.

Hypothesis 6b: A provocation during a Pakistani military regime should be associated with a higher risk of crisis than a provocation during a Pakistani civilian regime.

To measure intentionality via regime type, we use a dichotomous variable of whether the military regime or an elected civilian government was de jure in charge of the Pakistani government at the time of a provocation.

Domestic Politics. There are two features of domestic politics that may have a direct effect on a state’s perception of the threat level posed by a provocation: the political leaning of the government in power (right or left) and the leader (in India, the prime minister).

For the purpose of our coding, we associate right-leaning governments with more hawkish behavior and left-leaning governments as having more dovish tendencies. Though historically, the Congress party has led during major Indian military offensives – including invasions or annexations of Hyderabad (1948), Goa (1961), and East Pakistan (1971), as well as India’s “peacekeeping” deployment to Sri Lanka (1987-89) – the Bharatiya Janata Party (BJP) is generally recognized to have an “oppositional nationalist” ideology leaning toward a more aggressive and hawkish national security platform.97Vipin Narang, “Pride and Prejudice and Prithvis: Strategic Weapons Behavior in South Asia,” Inside Nuclear South Asia, ed. Scott D. Sagan, (Stanford: Stanford University Press, 2009), 138; and P.R. Chari, Pervaiz Iqbal Cheema, and Stephen P. Cohen, Perception, Politics and Security in South Asia: The Compound Crisis of 1990 (New York: Taylor & Francis, 2003), 75. We hypothesize that provocations arising during a BJP regime may be more likely to precipitate a crisis due to the way in which they interpret such actions through a nationalist lens or pressures from their domestic audiences.

Hypothesis 7a: A provocation during a hawkish government should be associated with a higher risk of crisis than a provocation during a dovish government.

The hawkishness of the Indian government was proxied by whether the BJP was at the reins of India’s coalition government at the time of the provocation. Another level of variation may exist with the party leaders that rise to helm the government as prime minister and their varying worldviews, risk-thresholds, and aggressive temperaments. Menon writes, “[p]ersonalities matter. With a different mix of people at the helm, it is quite possible that India would have chosen differently. In fact, if India is forced to make a similar choice in the future, I am sure it will respond differently.”98Menon, Choices, 81. Over the past two decades, India has had three heads of government: Atal Bihari Vajpayee (BJP) from 1998-2003, who was seen as a pragmatic realist; Manmohan Singh (Congress) from 2004-13, who was seen as a slightly dovish idealist; and Narendra Modi (BJP) from 2014-present, a Hindu nationalist who has cultivated an image of toughness.99Manoj Joshi, “Tough India More Likely under NaMo, Says Manoj Joshi,” India Today, April 2, 2014. Modi in particular has been identified as a “risk taker” possessing “strong nationalist credentials” and feeling the need to “[act] tough.”100Nitin A. Gokhale, Securing India the Modi Way: Pathankot, Surgical Strikes and More (New Delhi: Bloomsbury Publishing, 2017), 4, 32, 135.

Hypothesis 7b: A provocation during a Modi-led government should be associated with a higher risk of crisis than a provocation during a Vajpayee-led government or a Singh-led government.

Measurement for head of government was straightforward with dichotomous variables assigned to the periods when the Indian government was headed by Atal Bihari Vajpayee (1998-2003), Manmohan Singh (2004-13), and Narendra Modi (2014-16).

Time Sensitivity of the Provocation Context

The temporal pressure a government feels contributes to a sense of being within a crisis. This time sensitivity can be shaped by contextual factors of a provocation like pressure from media coverage, pressure from accumulating provocations, and/or countervailing pressure and the potential loss (opportunity costs) of structured, ongoing negotiations or dialogue if the response is delayed.

Media Coverage. One measurement of the psychological impact of a provocation is media attention, which has particular saliency in a large democracy like India. Media coverage can shape the perceived severity of the provocation and the probability of precipitating a crisis. Thus, even before crisis onset, media coverage can hype an event and possibly apply temporal domestic political pressure for a decisive reaction. The core assumption underlying considerations of media coverage as a factor in crisis onset is whether and how media impacts decision-makers.101For a detailed analysis of this question, see the essay by Ruhee Neog in this volume, “Self-Referencing the News: Media, Policymaking, and Public Opinion in India-Pakistan Crises.”

Some analysts speculate that pressure may have mounted after the 2008 Mumbai attack because it was perhaps “the most well-documented terror attack anywhere”102Vikas Bajaj and Lydia Polgreen, “Suspect Stirs Mumbai Court by Confessing,” The New York Times, July 20, 2009. and turned into a “television spectacle.”103Menon, Choices, 61. Riedel explains, “[t]hese attacks dominated global news for 72 hours non-stop. Achieving that amount of media coverage is exactly what the terrorists wanted. With the exception of Sept. 11, we have never really seen such global coverage.”104“Terror Expert and Obama Advisor Bruce Riedel.” Both the 1999 Kargil crisis and the 2001-2 Twin Peaks crisis also elicited “extensive television coverage,” which was believed to have “fueled the public’s anger over the attack” and intensified pressure on the government to contemplate the use of force.105Chari, Cheema, and Cohen, Four Crises and a Peace Process, 154. Recent analysis suggests the Modi government felt pressure from television as well as social media to treat the 2016 Uri attack as a crisis.106Authors’ off-the-record interviews with analysts and practitioners in New Delhi, India, December 2016; Lisa Curtis, “After Attack on Indian Army, U.S. Response to Crisis Must Focus on Pakistani Support of Terror,” Heritage Foundation, Issue Brief, no. 4612, September 21, 2016; and Paul Staniland, “India and Pakistan Clashed Again in Kashmir,” The Washington Post, September 30, 2016.

Past anecdotal analysis ascribes the media a definitive role in the shaping of public opinion.107See for example, Saeed Shafqat, “The Kargil Conflict’s Impact on Pakistani Politics and Society,” in Asymmetric Warfare in South Asia, 280-307. Shafqat asserts that by the 1999 Kargil crisis, “the print media had become an important barometer of public opinion and also an effective instrument used to influence public opinion on specific issues,” 284. One study utilizes limited quantitative data to support this claim. It concludes that assertive Indian press coverage of the crisis in 2001-2 may have “strengthened India’s policy of compellence, as the putative threat to initiate a war with Pakistan was made more credible by intense and widespread press coverage.”108Chari, Cheema, and Cohen, Four Crises and a Peace Process, 187. Even in Pakistan, public opinion “as reflected in media discourse” is a significant factor in foreign policymaking.109Huma Yusuf, “Conspiracy Fever: The US, Pakistan and Its Media,” Survival 4, no. 53 (2011): 95-118, 96.

Hypothesis 8: A provocation that receives high media coverage should be associated with a higher risk of crisis than a provocation that receives low media coverage.

To assess the impact of media coverage as a factor in converting a trigger into a crisis, we sought to measure whether the first week of coverage of the event in India’s most well-known English daily, The Times of India, totaled more than 100 articles and was greater than 5 percent of total news articles. Those events that met both criteria were coded as high media coverage while those that did not were coded as low media coverage.110Coverage of each provocation was measured by the approximate number of articles discussing it in the first week after a provocation (event day plus seven days). We drew upon The Times of India collections of two databases, Proquest Historical Newspapers (for events that occurred prior to and during 2008), and Proquest Global Newsstream (for events that occurred after 2008). Types of articles counted included Proquest’s news databank categories of editorials, letters to the editor, military and war news, front-page news stories, and news articles.

Cumulative Effect. Crises can be the product of a cumulative effect of rising tensions. The Middle East refugee or migrant crisis,111“Migrant Crisis: Migration to Europe Explained in Seven Chairs,” BBC News, March 4, 2016. for example, did not stem from any abrupt incident but rather the sudden realization of a gradual accumulation of pressure that spilled over into a full-blown crisis. In the case of South Asia, one theory is that repeated attacks will generate growing public demand, and/or pressure within the government, for a strong punitive response as a pressure release valve. This belief underpins various explanations of crises in South Asia.112In a recent example, Paul Staniland credited India’s “surgical strike” response to this cumulative factor of escalation: “Modi’s government attempted an opening to Pakistan’s civilian leaders, but the combination of Pathankot and Uri has led Indian policymakers to abandon a policy of conciliation toward Pakistan and instead to combine simultaneous military, economic, and diplomatic coercion.” See Staniland, “India and Pakistan Clashed Again in Kashmir.” Referencing data from the South Asia Terrorism Portal, Praveen Swami’s analysis of the 2001-2 crisis highlights that “both the scale and the frequency of fidayeen targets rose steadily in the build-up to the near-war.” Swami, “A War to End a War,” 148.

The October 2001 attack in Srinagar on the Kashmir Parliament building is believed to have precipitated a small crisis because of the cumulative pressure of multiple prior attacks. After the attack, Prime Minister Vajpayee sent a letter to U.S. President George W. Bush in which he identified the mounting and cumulative frustration from such high-profile attacks and “noted that Pakistan must understand that there was a limit to the patience of the people of India.”113Chidanand Rajghatta, “Pak Must Back Off, PM Warns in Letter to Bush,” The Times of India, October 3, 2001. Krepon and Nayak note that this ramp up of accumulating provocations helped build momentum for the major crisis that evolved in the aftermath of the December 2001 Parliament attack.114Nayak and Krepon, US Crisis Management in South Asia’s Twin Peaks Crisis, 19-23.

At work here is a psychological sense of acceleration. Few studies, however, evaluate public responses to series of similar terrorist attacks with more than anecdotal evidence. A 2016 psychological study in the United States found that trajectory (increase or decrease in frequency) is a more salient factor than frequency (in this study, the number of attacks per week).115Jinshu Cui, Heather Rosoff, and Richard S. John, “Cumulative Response to Sequences of Terror Attacks Varying in Frequency and Trajectory,” Risk Analysis 36, no. 12 (2016): 2,272-84, 2,281. The authors recognize that this study does not account for variation in the rate of change in risk over time, the degree of severity of attacks, or the level of coordination between attacks (2,283). Evolving risk environments have particular salience in South Asia where both India and Pakistan continue to produce fissile material and nuclear-capable delivery systems. In varying the trajectory of successive terror attacks, “respondents experience more negative affect, greater risk perception, and are more likely to engage in avoidance behavior if the number of similar attacks increases over time than if the number of attacks decreases over time.”116Cui, Rosoff, and John, “Cumulative Response,” 2,283. In the context of South Asia, analysts have pointed out that the 2016 Uri attack felt like a last straw after pressure accumulated from a series of attacks in Gurdaspur, Pathankot, and Pampore, so this may explain why Uri resulted in crisis onset.117“Uri, Pathankot, Pampore: 35 Soldiers Killed, over 1,740 Injured in Attacks on Indian Army in 2016,” Firstpost, September 19, 2016; and “Uri Attack: The Last Straw That Broke the Camel’s Back,” The New Delhi Times, October 7, 2016.

Hypothesis 9: A provocation that follows soon after a previous provocation(s) should be associated with a higher risk of crisis than a provocation that is not preceded by prior incident(s).

To measure cumulative effects on a candidate crisis, we identified whether a provocation occurred within 30 days of a prior provocation.

Shadow of the Future. A final hypothesis maintains that in the event of a provocation, a background condition of ongoing diplomatic engagement increases the opportunity costs of a crisis and therefore reduces the probability of one arising. A provocation that could easily turn into a crisis might be purposefully downplayed by one or both parties so as not to derail a larger diplomatic agenda with higher stakes. The ongoing dialogue effectively extends the “shadow of the future” such that both sides care more about expected payoffs from continuing diplomatic engagement than the immediate gains or losses presented by a provocation, reducing the pressure to treat it as a crisis.118We draw on the international institutionalists’ logic as summarized by James D. Fearon, “Bargaining, Enforcement, and International Cooperation,” International Organization 52, no. 2 (1998): 270.

Some researchers allege that this dynamic was at work in January 2016 when the seizure of a U.S. naval vessel by Iran’s Islamic Revolutionary Guard Corps resulted in a curiously muted U.S. response, likely because of a desire to avoid derailing the Joint Comprehensive Plan of Action nuclear deal with Iran that was set to go into effect that week.119Bozorgmehr Sharafedin and Phil Stewart, “Iran Frees U.S. Sailors with Start of Nuclear Deal in Sight,” Reuters, January 13, 2016; and Carlos Munoz, “Navy Commander Surrendered to Iran to Protect Obama’s Nuclear Deal,” The Washington Times, June 30, 2016. Secretary of State John Kerry noted the unique diplomatic context within which this provocation arose and was defused, stating, “I think we can all imagine how a similar situation might have played out three or four years ago, and the fact that today this kind of issue can be resolved peacefully and efficiently is a testament to the critical role diplomacy plays.”120Sharafedin and Stewart, “Iran Frees U.S. Sailors.”

A cursory look at our dataset suggests India suffered multiple major attacks between 2003-7, and yet none of these escalated into crises, potentially because of the ongoing Composite Dialogue. This dampening effect may have also played out in India on a number of occasions. The LeT attack on India’s Red Fort in December 2000, a full year before the Parliament attack, gets considerably less attention despite having been a complex attack on an iconic target. However, some analysts note that the attack occurred just a month after the Vajpayee government ceasefire to facilitate a Kashmir peace process, and overreacting to the event would have been costly both politically and to the policy agenda.121Swami, “A War to End a War,” 148.

Hypothesis 10: A provocation that occurs during a period of significant bilateral dialogue should be associated with a lower risk of crisis than a provocation that occurs without such a background.

To measure whether India might have felt a reprieve from ordinary time sensitivity after a provocation, we identified periods of significant ongoing dialogue, notably the Composite Dialogue from October 2003 to August 2008 (when Musharraf resigned) and the Comprehensive Dialogue between December 2015 and April 2016. 122Ankit Panda, “Back to Square One: Pakistan Calls Off Peace Talks with India,” The Diplomat, April 9, 2016.

There are certainly other variables we could consider in a future study, but they have been bracketed for the sake of manageability.123Additional variables we hope to examine in a future study include: the domestic political environment (e.g., upcoming elections); India’s response toolkit (e.g., state of military readiness or doctrine); Pakistan’s nuclear deterrent and doctrine; perceptions of a spoiler attempt on planned bilateral engagements; foreign judgments of the provocation; and epochal shifts (post-9/11 and/or post 26/11).

Preliminary Quantitative Findings

This essay offers a new dataset on candidate crisis provocations between India and Pakistan as perceived by India to probe the various hypotheses outlined above. Below we review the data and detail preliminary findings.124It might be possible to conduct a similar analysis on provocation and crisis onset as perceived by Pakistan, but this adds an additional layer of complexity, particularly when trying to collect observations of inert provocations (i.e. dogs that don’t bark). Furthermore, since official statements reveal the United States tends to tilt towards Indian perceptions of South Asian crises (rightly or wrongly), we have started with India’s vantage point.

After compiling the dataset of 51 provocations – 12 of which resulted in crisis onset – from 1998 through 2016 (nearly all of which are attacks by violent nonstate actors on Indian targets with perceived support from Pakistan), we coded the variables based on the measurements described above, tabulated the results, and then analyzed the predictions of our 10 hypotheses. Since the dataset as currently constituted was not large enough for multivariate regression analysis, in our first cut at statistical analysis of the data, we cross-tabulate each hypothesized independent variable (e.g., high vs. low lethality) with the dependent variable (crisis onset or not) and then conduct a Pearson chi-squared test.

A Pearson chi-squared test is useful for measuring association between categorical data; it evaluates a null hypothesis that the two variables are independent. If the probability of independence is below a designated threshold (say 5 percent or 10 percent), then the variable is identified as statistically significant and the null hypothesis can be considered incorrect. In other words, the probability of independence is low and the likelihood of association between the two variables is high. The test cannot specify the nature of association – neither degree nor causal direction – but these “sniff tests” can be useful in developing a theory of crisis onset. In the case of India, we found a number of traditionally cited variables did not even pass this first sniff test. The results are presented as cross-tabulations in Table 2.

Table 2: Results of Cross-Tabs and Pearson Chi-Squared Tests



We find that many of the hypotheses and variables posited in fact appeared to have no association when evaluated against the dogs that did not bark – that is, the provocations that did not escalate into a crisis. Based on this data, lethality, target type, geography, adversary control, adversary regime type, and government/party type do not appear to have a correlation with India-Pakistan crises. This is surprising and noteworthy considering these have been some of the most common explanations as to why crises begin and escalate.

Employing a conventional p-value cutoff of .05 (p<.05), the only variables that seem to pass the test of independence are duration, domestic political leader, media coverage, and the shadow of the future shaped by the presence/absence of ongoing dialogue. The small number of observations in this test might warrant a more relaxed p-value threshold of p<.10, in which case the other variables that survive the chi-squared test of independence are complexity and cumulative effects.

These findings suggest that certain properties of the provocation (namely duration and complexity) as well as perceptions filtered by the media, decision-making by leadership, and political context shaped by prior incidents or opportunity costs of dialogue shape a state’s political choice to opt into a crisis after a provocation occurs. That said, the findings for leadership, cumulative effects, and media are still thin.

It is intuitive that leaders and their dispositions play a central role in the treatment or labeling of an event as a crisis, and on the surface, this seems validated by the chi-squared tests. However, the findings of the leader variable do not quite conform to predictions. Instead of Modi’s leadership resulting in a higher risk of crisis, it is in fact the realist pragmatist Vajpayee who was associated with a much higher rate of crises. This may have something to do with the fact that most of these crises are clustered around the time of India and Pakistan moving from a recessed to an overt nuclear deterrent and the adjustments of “nuclear learning” to the risks and uncertainty of this environment.125Feroz Khan, Ryan Jacobs, and Emily Burke, eds., Nuclear Learning in South Asia: The Next Decade (Monterey: Naval Postgraduate School, 2014). Excluding the Vajpayee years from the analysis makes the result go away. When just looking at the Modi and Singh years, the chi-squared value is statistically insignificant and the null hypothesis of independence between leadership and crisis onset cannot be rejected.

Cumulative effects did not pass the chi-squared tests when operationalized at a 60-day threshold and while they did at a 30-day threshold, this result seems inappropriate because there were no provocations within 30 days of a previous provocation that resulted in a crisis.

High levels of media attention on an incident might press a government into treating the provocation as a crisis.126The correlation between media coverage and crisis onset remains statistically significant at the .05 level regardless of how it is operationalized -- as a 100 article threshold (.045), as 5 percent of coverage (.026), as either (.036), and as both (.025). However, it is equally plausible the reverse occurs where the onset of a crisis generates intensive media coverage and thus the coverage is an indicator more than a driver of crisis onset. (It is worth noting that in previous studies, media coverage of major India-Pakistan crisis peaked not in the first week but instead later in the crisis, sometimes several weeks into it.127See graphs in Chari, Cheema, and Cohen, Four Crises and a Peace Process, 65, 114, 146, 180. )

An ongoing and robust political dialogue with Pakistan might help mitigate the pressures of the incident and raise the costs of entering into a state of crisis. A leader or government invested in bilateral dialogue and conflict resolution might attempt to firewall relations from the vicissitudes of inevitable friction or even provocations by spoilers. This on its own does not necessarily impute normative value on such dialogue, but its potentially mitigating effects on a crisis atmosphere should be factored into cost/benefit assessments of dialogue.128Ashley J. Tellis, Are Peace Talks Worth a Damn? (Washington, D.C.: Carnegie Endowment for International Peace, 2017).

Though complexity and duration are slightly positively correlated with each other (with a correlation coefficient of .55), the presence of these variables is more likely to draw domestic audience attention and galvanize public concern. Such public engagement can generate pressure on the government to treat a provocation seriously due to the humiliation of the incident, fear induced by government lapses in protecting its citizens, or anger due to the transgressive nature of the provocation. Thus, a violent act alone or fatalities may not be enough to cross red lines, but if the provocation employs certain terror-inducing qualities and lasts long enough to attract substantial attention, it might generate pressure for the government to treat it as a crisis, even if the government has no ability to respond.


The contribution of this essay is two-fold. First, in addition to summarizing the mainstream strategic studies literature defining and detailing crises, the essay has introduced a novel approach to the agency of a state in “selecting” into a crisis. Crisis onset then is not simply a function of specific conditions, exact variables, or distinct thresholds. Instead, it is a subjective position clearly influenced by political considerations of context or leadership. Second, the essay helps to distill and theoretically ground a set of hypotheses trying to explain the conditions under which a crisis emerges.

The results are extremely preliminary at this point and ought to be taken with several grains of salt, but they suggest some support for the political rather than the structural or material correlates of crisis onset. 129A broader study is underway to expand upon these preliminary findings. This study utilizes a larger dataset of provocations and codes for a range of additional independent variables or properties of a provocation. To the extent that variables of abnormality – duration and complexity – have an impact on crisis onset, states can start to think about ways to counter them. Aside from the discussion of deterrence or compellence by punishment, these variables may suggest that deterrence by denial as well as by resilience, might be useful strategies to consider.

From a policy standpoint, effective denial and greater resilience would demand investing in personnel skills and organizational tools to raise the barriers to entry for nonstate actors and to rapidly mitigate the costs if they did manage to conduct an attack. This would place a premium on better intelligence to pre-empt or quickly disrupt complex attacks that require substantial national security reforms to enable higher quality analysis, communication, and organizational efficacy.

The essay also helps to set an agenda for future research on South Asian crises. The same set of structural, material, and political lenses might be directed to examine India-China or Afghanistan-Pakistan crises. Further study can also begin to tease out more precise statistical relationships between the variables identified in this essay, ideally with a larger dataset, but also to evaluate additional factors in crisis onset not yet considered. Other hypotheses that have been suggested for consideration include the gruesomeness of an attack, international attention paid to it, and other contextual features such as the domestic political timing (proximity to elections), temporal proximity to planned bilateral meetings, and the state of conventional deterrence (based on the range of military options and concepts evolving within India and/or Pakistan). Finally, rich and detailed insider accounts starting to emerge as well as interviews can help scholars trace decision-making in crisis episodes and tease out which variables are ultimately most salient in the minds of leaders.130Dulat, Kashmir; Menon, Choices; Gokhale, Securing India the Modi Way; Shyam Saran, How India Sees the World: Kautilya to the 21st Century (New Delhi: Juggernaut Books, 2017); and Khurshid Mahmud Kasuri, Neither a Hawk nor a Dove: An Insider’s Account of Pakistan’s Foreign Relations (Oxford: Oxford University Press, 2015).

The authors especially wish to thank Arzan Tarapore, as well as Daniel Altman, Phil Haun, Feroz Khan, Vipin Narang, Rich Nielson, Roger Petersen, Stephen Tankel, Rachel Tecott, Chris Twoomey, Joshua White, along with participants in seminars hosted by the U.S. Naval Postgraduate School and the Massachusetts Institute of Technology Security Studies Program.

Sameer Lalwani is a Senior Associate and Co-Director of the South Asia Program at the Stimson Center where he researches deterrence, strategic competition, and counter-insurgency. He is also an Adjunct Professor at the George Washington University (GWU) and was previously a Stanton Nuclear Security Postdoctoral Fellow at the RAND Corporation. Lalwani completed his Ph.D. from the Massachusetts Institute of Technology’s Department of Political Science where he was an affiliate of its Security Studies Program. He has conducted extensive fieldwork in India, Pakistan, Sri Lanka, and the British Archives. His work has been published by Security Studies, the Journal of Strategic Studies, Small Wars & Insurgencies, Foreign Affairs, Foreign Policy, CTC Sentinel, The New York Times, RAND, the Cato Institute, Sage, and Oxford University Press. He has had scholarly affiliations with GWU’s Institute for Security and Conflict Studies, the Institute for Defence Studies and Analyses, Lahore University of Management Sciences, New America, the Tobin Project, the Smith Richardson Foundation, and the Center for New American Security.

Hannah Haegeland is a Research Analyst in the Stimson Center's South Asia Program working on nuclear security, crisis escalation and management, and regional politics. Haegeland’s analysis has been featured in Foreign Affairs, Foreign Policy, War on the Rocks, Defense One, The National Interest, Arms Control Wonk, The Daily O, and The Diplomat. She originally joined Stimson as a Scoville Peace Fellow in 2015. Previously she was a Boren Fellow at the Indian trust no man's land and a Fulbright Scholar in Nepal. Haegeland completed her M.A. in South Asian Studies from the Jackson School of International Studies, University of Washington. She has been a Gorton International Policy Center Global Leaders Fellow, Conlon Fellow, and four-time winner of the U.S. Department of Education's Foreign Language and Area Studies Fellowship for Urdu and Hindi. Haegeland earned a B.A. in history and English literature from Concordia College, Moorhead.