Skip to main content Accessibility help
×
Home

Information:

  • Access

Actions:

      • Send article to Kindle

        To send this article to your Kindle, first ensure no-reply@cambridge.org is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about sending to your Kindle. Find out more about sending to your Kindle.

        Note you can select to send to either the @free.kindle.com or @kindle.com variations. ‘@free.kindle.com’ emails are free but can only be sent to your device when it is connected to wi-fi. ‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

        Find out more about the Kindle Personal Document Service.

        Violence Exposure and Ethnic Identification: Evidence from Kashmir
        Available formats
        ×

        Send article to Dropbox

        To send this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Dropbox.

        Violence Exposure and Ethnic Identification: Evidence from Kashmir
        Available formats
        ×

        Send article to Google Drive

        To send this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Google Drive.

        Violence Exposure and Ethnic Identification: Evidence from Kashmir
        Available formats
        ×
Export citation

Abstract

This article studies the conditions that lead peripheral minorities to identify with the state, their ethnic group, or neighboring countries. We contribute to research on separatism and irredentism by examining how violence, psychological distance, and national status determine identification. The analysis uses data from a novel experiment that randomized videos of actual violence in a large, representative survey of the Kashmir Valley region in the Indian state of Jammu and Kashmir, an enduring site of separatist and irredentist conflict. We find that a strong regional identity is a counterweight to irredentism, but violent repression by the state can push members of the minority to identify with an irredentist neighbor. Violence increases perceived distance from the nation and reduces national identification. There is suggestive evidence that these effects are concentrated among individuals with attributes that otherwise predict higher levels of identification with the state. Information about integrative institutions and increased national status brought about by economic growth is insufficient to induce national identification in a context where psychological distance from the nation is large.

Footnotes

We are especially grateful to Lauren Pinson for inputs on research design and Aijaz Wani, who helped facilitate the training at Kashmir University, provided comments on the survey instrument, and was our local partner during implementation. Chris Blattman, Thad Dunning, Guy Grossman, Dipin Kaur, Devesh Kapur, Dorothy Kronick, Cathy Lee, Shamika Ravi, Ashutosh Varshney, and Steven Wilkinson provided helpful comments on design, analysis, and implementation. Iba Marwein, Atulesh Shukla, and Bhartendu Trivedi of Morsel Pvt. Ltd. managed the survey in challenging conditions. We thank enumerators and survey respondents for participating in the study. Nair acknowledges funding from the South Asia Studies Council, MacMillan Center, and Georg Walter Leitner Political Economy Program at Yale. Sambanis acknowledges funding from Yale's MacMillan Center. This study received IRB approval under Yale Human Subjects Committee Protocol #1507016145. All views and errors are the authors’.

Why do peripheral minorities sometimes identify with their country and sometimes with their ethno-regional group? When will their identification with a neighboring country be stronger than their national identity? Weak national identification in peripheral regions can contribute to separatism and irredentism. Reclaiming lost parts of the homeland has motivated enduring conflicts between China and Taiwan, North and South Korea, Greece and Turkey, Armenia and Azerbaijan, Serbia and Croatia, Somalia and Ethiopia, and in other cases. Irredentism has been a feature of major wars, including Germany's and Hungary's land grabs prior to and during World War II. Russia's annexation of Crimea in 2014 was a reminder that irredentism is as important today as it is understudied. Irredentism takes different forms, including indirect intervention by neighboring states who use their co-ethnics across the border to destabilize regional competitors by sowing conflict. Studies have shown that the presence of cross-border kin groups is a risk factor for civil war.1 Therefore, understanding the conditions under which governments can induce the loyalty of peripheral minority populations is critical for the study of separatism, irredentism, and nation building.

Affective factors—the salience of ethnic versus national identification—are important determinants of these conflict processes. There is no consensus on the causes of separatism and irredentism, but most studies focus less on grievances and more on factors that affect the feasibility of rebellion.2 Peripheral areas can be at greater risk for separatism because of interference by co-ethnic populations across the border, because greater physical distance from the center might impede the penetration of these areas by state institutions, or because these areas are more likely to be dominated by territorially concentrated ethnic minorities. These factors could also impact affective determinants of separatism because they contribute to a sense of cultural distance between center and periphery. The greater that distance, the lower the national identification of peripheral populations and the more likely it is for ethnic conflict to emerge.3 Regional adversaries of the state can benefit from that distance by supporting their co-ethnics or co-religionists across the border to fuel ethnic conflict.4 In response, states might react by using violent repression of ethnic minorities who make claims for self-determination when these claims serve the interests of regional adversaries. Is violent repression an effective strategy to suppress challenges to state authority by peripheral groups, or is it more likely to weaken those groups' loyalty to the state and the nation?

We contribute to the understanding of peripheral ethnic conflicts by studying the determinants of ethnic versus national identification in the Kashmir Valley in the Indian state of Jammu and Kashmir, one of the most well-known yet understudied cases of irredentist and separatist conflict (we do not cover the Hindu-dominated Jammu, Buddhist-dominated Ladakh, or Pakistan-administered Kashmir regions).5 Kashmir has been at the center of an enduring rivalry between India and Pakistan since the late 1940s. Insurgency and government crackdown has left over 60,000 people dead. Data gathering and rigorous academic scholarship have been rare. We explore the relationship between violence and social identification in this ongoing separatist conflict using novel survey experimental methods that allow us to go some way in resolving the selection issues that mar prior observational studies on the link between violence and ethnic identification.

While the literature assumes that ethnic violence reifies ethnic cleavages, no previous study has measured the effect of violence exposure on support for irredentism in a realistic setting during an active conflict. We provide such an analysis using data from face-to-face interviews with a large sample of Kashmiris while randomizing exposure to videos of actual ethnic violence and police repression of protests. This allows us to study how violence exposure affects the way Kashmiris feel about the Indian state and how they respond to Pakistani irredentism. Drawing inspiration from social identity theories of intergroup conflict, we also test whether integrative institutions reduce social distance between Kashmiris and the Indian state and if material or psychological benefits that arise from being part of a high-status nation with a growing economy can mitigate the effects of large psychological distance.6

A large historical, case-based literature exists about the Kashmir conflict,7 but there have been few theoretically motivated quantitative studies that aim to understand the determinants and dynamics of ethnic and national identification at the micro level. The existing literature is equivocal on whether a strong ethno-regional identity weakens the national identity.8 Previous empirical studies suggest that a tradeoff between ethnic and national identity exists under some conditions,9 but there are several open questions. Can the Indian state use policies to weaken Kashmiri ethno-regional identity as a strategy to reduce the intensity of separatist conflict in the region? Are integrationist policies—recruiting Kashmiris in the armed forces—likely to reduce the psychological distance Kashmiris feel toward India? Can India's growing international status make Kashmiris more likely to identify as Indian? Are Pakistan's overtures more likely to be accepted if India uses repressive tactics? We address all these questions, which are central to the specifics of the Indo-Pakistani conflict, but also inform the broader literature on ethno-nationalism and separatism.

For the most part, the literature on separatism and irredentism has adopted a cross-national analytical frame. We provide new micro-level data from a crucial but hitherto understudied case of separatism and irredentism, while utilizing a theory-driven approach to measuring national and ethnic identification. Our research design helps to overcome issues of selection by randomizing exposure to videos targeting psychological mechanisms that are theorized to be important drivers of social identification. These videos are widely available and depict actual violence. To our knowledge, similar interventions have not formed the basis of other large-scale survey experiments in developing country contexts, which typically use researcher-written vignettes that may appear artificial.

We find that there is a trade-off between the salience of ethno-regional and national identities. The Kashmiri identity is unassailable and diminishes identification with India. Repression of protests and the use of violence by the Indian state fuels the conflict by further diminishing national identification. Our key argument is that, conditional on having a strong ethno-regional identity which is itself the function of prior levels of conflict, violent crackdowns by the state will generate resentment, pushing regional minority group members to identify less with their nation. This creates a window for irredentist states, which can benefit by the anti-nationalist sentiment in the region.

We show that Kashmiris’ identification with Pakistan grows substantially with violence exposure, and that ethnic identification rises from its already high base. Individuals with attributes that are more likely to predict higher levels of national identification (having more education, or belonging to less separatist geographic regions, for instance) resemble their co-ethnics after treatment. In the absence of violent repression, however, a strong regional identity acts as a counterweight to irredentism (and identification with Pakistan is even weaker than with India). We find that regional identification is not easily supplanted by the national identity of the neighboring state.

This and several other findings from the Kashmir region point the way to the management of separatism and irredentism in border regions. Reducing incidents of violent repression is likely to weaken the identification of minorities with co-ethnics across the border, thereby reducing the plausibility of a strategy of irredentism, which would diminish the risk of interstate conflict. Yet, exposure to violence also causes study participants to express a lower willingness to engage in peaceful protests or endorse violence against the central government, illustrating the (potentially short-term) benefits of coercion from the state's perspective. We find that the psychological distance that is created by active conflict in Kashmir cannot be overcome by the perceived benefits of India's growing economic power or by efforts to integrate Kashmiri youth in state institutions. Cultivating national identification in the midst of active ethnic conflict might not be possible. Our research suggests that when perceived ethnic distance is too high, partly as a result of experiences of ethnic violence, material and institutional considerations will not be strong enough to induce national identification. This result can inform the strategies through which states can try to regain the loyalty of peripheral populations, and it also has implications for the literature on counter insurgency (COIN) and nation-building interventions. Winning hearts and minds via the provision of institutions and material benefits might not be possible in the context of active conflict that creates psychological distance between the state and minority populations.

Related Literature and Theory

In his classic study of ethnic conflict, Horowitz tells us that the relative prestige of two countries competing over control of minorities in border regions helps determine whether or not the minority will support irredentism.10 He gives the example of the Sudeten Germans, who abandoned ethnically conciliatory parties and became staunch supporters of Hitler in the mid-1930s when Nazi Germany became a major power. Horowitz is one of the few political scientists to consider how inter-group status comparisons affect separatism. To date, there is no quantitative/empirical assessment of this claim, but Sambanis and Shayo focus attention on status comparisons in a formal model of ethnic conflict that is compatible with Horowitz's constructivist insights.11 We draw on these studies to design our empirical investigation on the determinants of social identification.

With Horowitz, we assume that separatism reveals weak national identification. This might seem obvious, but it is a position not explored systematically in a literature that tends to dismiss affective explanations for ethnic conflict in favor of purely instrumental ones. Consistent with the Common In-group Identity Model (CIIM) in social psychology,12 we assume that a tradeoff exists between nested subordinate and superordinate identities. The CIIM is consistent with classic views of nationalism and nation building13 according to which national attachment diminishes the risk of domestic interethnic conflict. But how malleable is national (and ethnic) identification, particularly during active conflict?14 Does exposure to ethnic violence harden identities to such an extent that integrationist and other state policies designed to increase national identification are doomed to fail?

Few studies have addressed this question. Grossman, Manekin, and Miodownik study the effects of violence exposure and find that combat experience in the Israeli army radicalizes attitudes, reducing the space for political compromise in the Arab-Palestinian conflict.15 This is consistent with our expectations. But the broader literature is equivocal on whether or not violence exposure makes people less pro-social and increases out-group prejudice. Some argue that violence exposure during wartime increases pro-sociality by increasing altruism16 and promoting civic engagement.17 Yet others find the opposite—violence exposure decreases trust18 and increases in-group bias.19 None of these studies explore if explicitly integrationist policies or the perception of greater material or psychological benefits from rising national status reduces any negative effects of violence exposure. Our study is designed to address this question empirically in the context of the Kashmir conflict.

Our expectation is that ethnic violence makes ethnic identities more salient because it draws attention to the incompatibilities between the state and an ethnic minority and it heightens the significance of ethnic attributes that differentiate the minority from the rest of the nation.20 There is no empirical evidence that we are aware of to contradict the claim that ethnic violence increases perceived distance from the nation. However, social identification is a complex process and the effects of violence on social distance might be offset by other considerations, such as material benefits that stem from being part of the nation or other instrumental motives.21

If Horowitz's reasoning is correct, then as a country's international status rises relative to its competitors, the intensity of domestic ethnic conflict should decline as more people will identify nationally for both psychological and material reasons. As national status increases, more citizens will identify nationally because they will derive greater self-esteem from being part of a rising global power.22 Similarly, the country's growing economic or military power should induce national identification because of the perception that more material benefits will accrue to its citizens by virtue of being part of the nation. The state can also solidify national identification by integrating minority groups in state institutions, such as the national army.23

Integration with state institutions can provide minorities with ways to address their grievances and should diminish the feeling of political exclusion that can fuel conflict.24 Military integration also provides recruits and their families with material rewards (jobs) while reducing their perceived distance from the nation: a Kashmiri serving in the Indian national army must feel closer to the rest of the nation as his sense of citizenship is reinforced by his profession. In addition, observing one's coethnics being represented in national institutions should reduce the psychological distance between one's ethnic and national identities. These arguments, which inform our empirical design, are supported by a wealth of experimental evidence in social psychology on the determinants of identification with a more inclusive super ordinate group. Shifting identification from subordinate to super ordinate groups has been shown to reduce in-group bias and out-group prejudice in the Common In-Group Identity Model (CIIM).25

Membership in a social group could induce individuals to identify with that group.26 To identify with a group means to care about the group's relative social standing (status) and about the payoffs of in-group members relative to an out-group. Controlling for “distance” between the individual and the rest of the group, social identification with a group is more likely the higher the group's status because belonging to high-status groups provides individuals with more utility. These key insights of social identity theory form the basis for our empirical inquiry.

To sum up the theoretical argument thus far, the main expectation is that any factor that increases a country's national status and diminishes perceived distance from the nation will induce citizens of that country to identify nationally at the expense of competing social (ethnic, religious, regional) identities.

We focus on how ethnic violence affects social identification and weigh its effects against other factors (status, interest) that should affect national identification. Theoretical models have established how exposure to ethnic violence increases perceived social distance27 and prolonged ethnic violence can reify ethnic identities, making it difficult for policy interventions to switch identification from ethnic to national. Prior literature has focused extensively on the question of ethnic distance as a determinant of irredentism and secession,28 but the effects of distance could be offset by material inducements to identify with the nation.29 There is now consensus that nationalist sentiment is not unwavering and that it can be affected by several variables, from the content of mass schooling systems,30 to economic modernization,31 colonial legacies,32 public goods provision,33 and external security threats.34 These mechanisms are all examples of “material and institutional” determinants of social identification.35 The main insight here is that identification is not static—it changes over time in response to social conditions. But how malleable is identity in the context of an active conflict? Can integrative institutions—such as minority recruitment in the military—and a growing economy offset ethnic distance between an ethnic minority and the nation?

No study to date has established empirically the relative weight of distance-related determinants of national versus ethnic identification versus material or status-related determinants. Our expectation is that, in an active conflict (an important scope condition for our study), the population will discount promises of material rewards and institutional integration will appear shallow, overshadowed by the weight of ongoing violence. If an ethnic minority does not see itself as part of the nation, then the nation's rising status relative to regional competitors will be seen as a threat rather than as a source of greater self-esteem. And integration of part of the minority in state institutions will be seen as co-opting and weakening the group rather than reducing its distance from the nation. Thus, our expectation is that, while status gains and institutional integration can be effective in some contexts, violence exposure will have a stronger effect on identification.

To summarize, we draw on social identity theory, the CIIM, and the applications of those theories to political science36 to test the following five hypotheses:

H1

Strong ethno-regional identification will reduce national identification.

H2

Inter ethnic violence will reduce national identification.

H3

Violent repression by the state will increase identification with irredentist states.

H4

Higher national status will not increase national identification substantially given high levels of ethnic distance.

H5

Exposure to integrative institutions will not increase national identification substantially given high levels of ethnic distance.

Finally, in assessing the effects of violence exposure on nationalism, we establish connections between the literature on ethnic/national identification and the literature on repression. We do so by exploring heterogeneous effects of violence exposure and consider whether violent repression is effective in discouraging protest behavior.

Context

The setting for our study is the Kashmir Valley region of the Indian state of Jammu and Kashmir, a region that is central to the enduring rivalry between India and Pakistan.37 India is a large multi ethnic state with a civic nationalist tradition. In some Indian regions, ethnic and national identities are not conflicting. In the state of Kerala (at the other, southern, extreme of India), for instance, a strong subnational identity coexists with strong national identification.38 Yet in other regions, as our evidence on Kashmir will show, ethnic conflict creates a trade-off between ethnic and national identities.

Kashmir presents an interesting case to test our hypotheses. Perceived social distance from the nation is high among Kashmiri Muslims because of high levels of prior conflict. Objective differences between Kashmiris and other Indians are made salient by Indian politics (incidents of Hindu-Muslim violence, electoral platforms that play on religious difference, the India-Pakistan conflict). Our qualitative fieldwork and interviews indicate that Kashmiris perceive themselves as a high-status group, and the theories of Horowitz and Sambanis and Shayo suggest that high-status groups are less likely to identify nationally.39 This perception of inter group status is based on culture as well as physical attributes such as height, facial features, and skin color, which are emphasized in the region to draw sharp boundaries between Kashmir and the rest of India. As such, many individuals express a willingness to pay significant economic and personal costs to avoid further integration. The intensity of ethnic identification in Kashmir combined with active conflict makes this a least-likely case for predictions regarding the positive effects of national status on national identification. This sets a useful benchmark for future studies in other areas of India, where there is likely a greater abundance of factors supporting nationalism.

The Kashmir conflict has its origins in the partition of the Indian subcontinent along religious lines at the end of British rule in 1947. Each of the numerous princely states that made up the subcontinent could join either India or Pakistan, or remain independent. In practice, geography and popular pressure meant that none were ultimately able to exercise the latter option.

Both Pakistan and India claimed the territory of Kashmir. Kashmir had a majority Muslim population with a sizable Hindu minority but was governed by a Hindu ruler who vacillated between joining India or Pakistan. After Pakistan supported the invasion of tribesmen from the northwest to force the issue, the king signed an “instrument of accession” to merge Kashmir with India. India subsequently sent troops to the area, and the first war between the two countries broke out in 1947–48, which ended with a UN-sponsored ceasefire. Sheikh Abdullah, the most popular leader in the Kashmir Valley, opposed Kashmir's joining Pakistan at the time of the partition and called Pakistan an aggressor.40

While Pakistan retained about 35 percent of the territory in the north and west, India secured most of the Sunni Muslim-dominated Kashmir valley, as well as the eastern and southern areas of Ladakh and Jammu, respectively. Ladakh is principally made up of Shia Muslims and Buddhists while Jammu has a majority Hindu and Sikh population (though one-third of the population is Muslim).41 Three more wars have been waged between India and Pakistan over the disputed territory (in 1965, peripherally in 1971, and in 1999), but the 1948 ceasefire borders (demarcated by the “line of control’’) have remained basically unchanged, even though both countries claim the territory in its entirety.

The central government quickly grew distrustful of Kashmiri leaders, fearing that they harbored secessionist aims. As a consequence, it imprisoned Abdullah for long stretches of time. Jammu and Kashmir formally became a state in India in 1957, after the Jammu and Kashmir Constituent Assembly approved a merger with India in early 1956. Under the terms of its incorporation Jammu and Kashmir was given special status and autonomy under Article 370 of the Indian constitution. Elections during the 1960s, 1970s, and 1980s were marred by widespread fraud that allowed the center, particularly under Indira Gandhi, to install puppet governments without widespread legitimacy. Two exceptions were the elections in 1977 and 1983, which by most accounts were not fraudulent.42 By the late 1980s this crisis of legitimacy led to growing disenchantment with the Indian government, and a concomitant rise in separatist and irredentist views. A flawed election in 1987 served as the spark that caused these grievances to explode into a secessionist insurgency.

The insurgency enjoyed substantial material support from Pakistan and was accompanied by the growth of irredentist Islamic parties that favored union with Pakistan, in a trend that paralleled the contemporaneous rise of political Islam elsewhere in the world.43 Nevertheless, while some militant groups preferred union with Pakistan, others resolutely sought independence, a division that was mirrored in the broader population.

As the insurgency grew during the 1990s, India's control over large swathes of the Valley declined precipitously. The central government dismissed the state government, imposed President's rule, and gave the armed forces a largely free hand to quell the insurgency. The insurgency and the state repression that followed produced enormous violence with tens of thousands of casualties (as well as an exodus of a previously significant Kashmiri Pandit Hindu minority). Although insurgent attacks and state repression remained quite common during the 1990s, the Indian government had contained the insurgency sufficiently to hold free and fair elections in 1996 and 1998 with relatively high turnout.

People living in Indian-administered Kashmir became unwilling or unable to bear the massive human and economic costs of the insurgency, and militants were increasingly drawn from Pakistan's side of the border over time. In an attempt at stifling India's successful counter insurgency, Pakistan supported an incursion into the strategically important peaks around the Kargil district in 1999. The ensuing war left hundreds of soldiers on both sides dead. The prospect of a nuclear confrontation led the United States to broker a ceasefire that led to a return to the status quo ante. Relatively free and fair elections have been held since then and cross-border incursions and the intensity of the insurgency declined until 2016, when violence again surged.

As our survey shows, large proportions of the population of the Kashmir Valley state a preference for Kashmir's independence. A much smaller minority seeks union with Pakistan. While irredentist sentiment was at its peak in the late 1980s and early 1990s,44 it declined during the first decade of the twenty-first century because of Pakistan's struggles under military rule, slowing economic growth, pressure from the United States, and the rise of violent extremism within Pakistan's borders. Irredentist and secessionist views vary considerably even within the Valley and over time. This variation allows a unique opportunity to study the micro-dynamics of ethnic and national identification.

Study Design

Our design uses randomized interventions in the context of a face-to-face household survey to examine the effects of violence, status, and psychological distance on social identification. We use condensed YouTube videos that have been viewed thousands of times, or ask respondents to interact with maps that have direct real-world analogues. Such interventions have only recently become feasible at such a large scale in developing country settings with the advent of phone- and tablet-based survey methods. They allow researchers to present respondents with realistic content rather than relying on contrived vignettes or other constructed interventions that are the standard practice in survey experiments. The videos have both a priming component (they raise the salience of particular issues and identities) and an informational component, providing facts to respondents that they may not have been familiar with.45

Randomization

We used four different treatments, sometimes in combination. Table 1 summarizes the experimental design. We describe the treatments here and include screen shots of each of the treatments in the appendix.

  1. 1. exposure to violence: Respondents assigned to this condition were shown a video of protesters and a violent crackdown. The video begins with a large group of protesters who are shouting the slogan “go back” in an urban setting before troops attempt to disperse the group using tear gas and ammunition. Fleeing protesters are chased by troops on foot and in armored cars. Some protesters continue to throw stones and the police and military respond by firing in the air and racing through the streets to break up the gathering. Our discussions with local researchers and nonacademics suggested that the film both realistically portrayed scenes that are common in the region and might be something that is shared by people using phones. After viewing the video, respondents were asked whether they had personally witnessed such events and were asked to share their opinions in an open-ended format. To our knowledge, this is the first study to experimentally manipulate exposure to violence to study its effects on polarization. Our goal is to understand whether state-led repression increases psychological distance from the nation; whether it increases identification with irredentist neighboring states; and whether it makes people more or less willing to participate in protests and endorse violence against the government.

    Table 1. Study design and implementation

  2. 2. economic growth and national status: This treatment is a brief newsreel from a major India-wide news channel. The announcer describes the World Bank's Global Outlook Report, which predicts that India is poised to overtake China as the fastest growing major economy in the world. The announcer notes (in Hindi) that “your country will move ahead of China within two years” because of hopes raised by economic reforms that have been undertaken by the new government. She further states that there is talk of economic growth that is poised to reach all corners of the country and economy as India's economy grows at 7 percent or higher while China's slows down to grow at 6.9 percent or lower. Rapid economic growth in India is the source of considerable pride and many news outlets extensively cover India's changing status as an “emerging giant” and “world beater.” The explicit comparison with China is particularly important because it conveys messages of increasing international status.46

  3. 3. integrative institutions: The political science literature on power sharing has focused on constitutional arrangements, electoral rules, and national-level institutions as solutions to the problem of political order in “divided” societies.47 Yet, most people do not interact with politicians as much as they do with the military, police, and the bureaucracy. Accordingly, we examine whether national institutions that integrate disaffected ethnic minorities in the military can reduce ethnic identification.48 In recent years, increasing numbers of Kashmiri youth have begun joining the national-level bureaucracy, police forces, and even the Indian Army (something that may have been unthinkable during the period of widespread insurgency). This trend is the subject of debate, pitting people who support such moves in an environment of high unemployment against those who view it as a betrayal of the struggle for independence. To test the importance of the psychological effects of having more integrated national institutions, we expose a random subset of respondents to a video about Kashmiri youths joining the Indian Army. The video shows the passing-out parade of 179 newly commissioned soldiers in the Indian Army's Jammu and Kashmir Light Infantry regiment. The video mentions that several inductees are from insurgency-hit areas, and then features a commanding officer who says that the army's recruitment drives meet great demand (a good sign in his view), before switching to an interview with a new inductee who says that it has been his ambition to join the army from an early age.

  4. 4. common geography: The final intervention presented subjects with a map of India with the location of four major Indian cities, including the capital, highlighted. Respondents were then asked to match a list of cities to the locations on the map. Maps play a central role in theories of nationalism49 and the treatment allows us to prime national identity in a relatively unobtrusive way. Viewing the map with neighboring countries shaded should induce participants to think of themselves as Indian, as should the exercise of matching city names to the map. We use different versions of the map to generate three sub-treatments. In two versions, the map shows the state's boundaries as claimed by India, with one additionally showing a small Indian flag located in the top right-hand corner of the map to make the shared national identity more cognitively salient. A third version of the treatment additionally shows the Line of Control between India and Pakistan in addition to the boundaries as claimed by India. Figures A2 to A5 in the appendix include photographs from the video interventions and pictures of the maps used to raise the salience of a common national identity.50

To examine the sensitivity of any increase in national identification induced through these mechanisms, we subsequently cross treatment groups 2, 3, and 4 with violence exposure or a control condition. After randomization, respondents were asked a series of questions designed to measure social identification along a number of dimensions.

Measuring Identity

Cross-national studies of nationalism have for the most part relied on off-the-shelf measures of self-reported attitudes culled from large repeated surveys. The World Values Survey (WVS) and AfroBarometer surveys are a main source51 that include Likert-scale scores of national pride. These surveys claim to offer representative views of the population but conflict regions and mobilized minorities are often under represented.

The standard WVS measures of identity (e.g., “how proud are you of your nationality?”) produce almost uniformly high values for national identification with very little movement even in countries in the midst of ethnic or sectarian civil war. We provide a richer set of measures of identity than the previous literature, utilizing some questions that are similar to WVS, but with some modifications.

  1. 1. rank ordering of social identities: Subjects are asked to rank their Indian identity, Kashmiri identity, religious identity, and occupational identity in order of importance. We reverse-code this variable to take values from 1 to 4, with higher values indicating greater national identification.

  2. 2. self-identification: Respondents were asked whether they saw themselves as “Kashmiri only,” “Indian and Kashmiri,” or “Indian” only. This variable is coded to take a value of 1 if respondents identified themselves as “Indian” or “Indian and Kashmiri” and 0 otherwise.

  3. 3. kashmir's political status: We probed respondents’ political opinions as a measure of identification by asking them whether they thought (1) Kashmir should stay with India, maintaining the current status quo; (2) remain with India but be given more autonomy; (3) become independent after the two parts on either side of the border are merged; (4) or become a part of Pakistan. We code the variable to take a value of 1 if respondents selected option (1) or (2) and 0 otherwise.

  4. 4. india-pakistan preference: Respondents’ political attitudes were also measured by whether they thought Kashmir should remain part of India (1) or be merged with Pakistan (0) if independence were not an option. Using this variable, we intend to measure the extent to which exposure to violence makes Kashmiris more open to Pakistan's irredentism.52

  5. 5. quasi-behavioral measure of identification: To better understand whether stated identification correlates with in-group bias as measured by behavior, enumerators told respondents that INR 30,000 (about USD 450, a sizeable sum in the local context) would be donated to nongovernmental organizations providing help to the poor, medical aid, or education assistance based on the responses of people participating in the survey. Respondents were asked what percentage of the bonus (in increments of 10%) they would like to donate to NGOs providing assistance to people living in Kashmir and NGOs providing assistance to people all over India. Subjects were told that they could donate the full amount to either NGO or some percentage to both and that one person's name would be randomly chosen by computer. After giving respondents a moment to make their decision, the enumerators then handed over the tablet to respondents to allow them to make their choice confidentially, using clearly numbered allocations with small coin images denoting the amount that would be donated to NGOs.53 We measure national identification by the percentage of the bonus that is donated to the all-India NGO.

  6. 6. identification index i: We combine the separate measures into a single index of national identification using inverse covariance weighting.54 Given the non response to the India-Pakistan preference question we exclude this variable from the index.55

  7. 7. identification index ii: Again we compute an index of identification, but in this version include the India-Pakistan preference question.

  8. 8. participation in peaceful protests against the indian government: Interviewers read the following statement: “When the government threatens our rights, it sometimes becomes necessary to protest against these actions.” Then respondents were asked how likely they are to participate in peaceful protests against the Indian government in the future. Respondents were asked to rate this likelihood on a scale of 1 to 5, with this likelihood symbolized by circles of increasing size. Enumerators handed the device over to respondents, who made their choice, which activated a new screen.

  9. 9. endorsement of violence against the indian government: Immediately following the question about peaceful protests, respondents were asked, “and sometimes, the government's actions are so wrong that some believe that violent protest becomes necessary. Do you think violence against the Indian government is sometimes justified?” Again, subjects selected circles of increasing size, and we code this endorsement on a continuous scale from 1 to 5.

Prior to randomization, subjects were asked a short battery of demographic questions that sought information about age, gender, marital status, religion, household size, news consumption, occupation, income, and wealth.

Caveats

As with other survey and field experiments, researcher control over treatment assignment comes with trade-offs. Any real experience, including experiences of violence (being personally injured or losing a family member) will likely have deeper, more multifaceted effects on social identification than any experimental treatment that can be manipulated by researchers in the field or the lab. Differences in the form or intensity of violence could generate different reactions. Moreover, our results cannot speak to the persistence of any measured effects, or how they are likely to be affected when respondents are exposed to competing or complementary information. Because the interventions correspond to stimuli and information that respondents might have encountered in their own lives, we can only study the effect of an additional or marginal increment of these phenomena on respondents’ identification.

Despite these important caveats, our empirical strategy allows us to mitigate otherwise intractable problems of causal identification in conflict settings. The realism of our treatments compares favorably to those used in many similar studies. Governments’ efforts to control social media content through regulation and lengthy bans (“e-curfews”), suggest that they take the dissemination of such content seriously. As the video of Kashmiri youths joining the army suggests, they also view the production of such content as a potentially important means of shaping preferences. Since we want to explore the psychological mechanisms underlying social identification, the lack of a real economic or material benefit or cost is an advantage. A conservative interpretation of our study is that it sheds light on the short-term psychological determinants of identification. It accomplishes this goal by drawing on data from a large, representative sample of a region of significant policy and scholarly interest. To the extent possible, we rely later on secondary observational and qualitative literature to buttress our core findings and to demonstrate that they also correspond to long-term trends and have applicability beyond the survey setting.

Another set of caveats concerns research in conflict settings. Each element of the study design was approved by a university institutional review board. It was designed and implemented in partnership with academics at Kashmir University, and utilized questions and media that have been used previously in this setting or correspond to content that respondents would encounter in their daily lives.

The violence portrayed in the videos is not greater than what individuals would encounter by watching the news in the region or elsewhere in the world. While the video shows violent images, no individuals are shown being killed or maimed, and the video was edited to reflect a believable simulacrum of actual events without unduly risking trauma to individuals. Unlike a number of field experiments in developing countries, our study does not alter political institutions or economic relations in the community, and we do not create winners and losers. From a policy perspective we show that while coercion may reduce the incidence of protests against the state, it also might create the longer-term conditions that encourage protest by strengthening separatist and irredentist sentiment.

Enumerators were encouraged to conduct interviews with only the respondent present (if possible) and respondents could decline to answer any question. While the low level of education precluded respondents from taking the survey themselves, they recorded responses to particularly sensitive questions (such as the endorsement of protests) using intuitive symbols on the phones used to record their responses (which automatically skipped to the next screen). Their responses were confidential and all identifying information was kept separate from survey responses. The data were transmitted to a secure server on a daily basis, after which it was not retained on the devices. A considerable amount of effort was put into recruiting qualified enumerators, training them extensively, and supervising and auditing their work. While the enumerators are from the same region as the respondents, potentially biasing our findings through interviewer effects, these biases (and non response rates) would likely have been larger if the interviewers were from outside Kashmir.

Implementation and Sample Characteristics

Our sampling strategy was designed to maximize the representativeness of our subject pool with respect to the population of adults over the age of eighteen in the Kashmir Valley; we did not expand the scope of the study to Jammu or Ladakh for logistical reasons. Accordingly, we sampled half the state legislative assembly constituencies in proportion to population size in all ten districts of the Valley. Subsequently, we selected polling booths at random and used electoral rolls to identify a sampling frame. Because most government services require voter registration, almost all individuals are registered on electoral rolls (even if many polling booths report near-zero turnout for national and state-level elections). After generating a random number and interval between voter numbers, enumerators attempted to contact respondents. If they failed to locate a respondent, they simply selected the next person defined by the interval and attempted to contact that individual. This produced a known, verifiable sampling frame.

After a week-long training, a survey team consisting of twenty-three enumerators (eighteen female) and four managers (two female), all with at least a BA degree and most with a master's degree, completed the surveying in November–December 2015.56 Interviews typically lasted between thirty and forty minutes.

Our sample consists of 2,522 people (49% female, 51% male). Basic demographic information collected includes education level, age, income, type of occupation, and a number of other variables. While education levels are low in this population (36.4% of the subjects have no formal education), this is not atypical compared to the rest of India. Median monthly household income is between 5,000 and 10,000 Indian Rupees (USD 75 to USD 150) (though, as with other household surveys in India, non-cash income and systematic underreporting of incomes mean that this may be an underestimate).

A large share (35%) of the sample is composed of unemployed women, most of them in their late twenties to early forties. Laborers, students, agriculture workers, and small business owners are the most heavily represented groups. About 7 percent work in state government. Very few (0.4%) work for the central government and an equally small percentage are professionals (doctors, engineers, lawyers). The population is primarily Sunni Muslim (93% of our sample) with just over 5 percent of our sample being Shia Muslim. Hindus are no longer well-represented in the region (they make up just 0.2% of the sample).

Interview subjects were remarkably open about their views. Refusal rates were very low despite the sensitive nature of many of the questions. When asked their opinions about the status of Kashmir, only 5.7 percent of subjects refused to answer and 65 percent reported pro-independence attitudes. The one question with relatively high refusal rates (31%) concerns preferences over India versus Pakistan (i.e., strong proponents of Kashmiri independence are reluctant to support either India or Pakistan). People are distrustful of the state (63% report that they never trust the government referring to the state government; 89% report distrusting the central government of India); and they are even more distrustful of the media (89%). Over 90 percent report distrust toward the Indian army and negative attitudes extend to the local police (82%). These negative views seem to be supported by a widely shared perception of increasing human rights violations (64% of the sample think these have increased over the previous two years, compared to 22% who think they have decreased). Moreover, these violations are taking place in a climate of decreasing insurgent violence. Survey subjects for the most part (55%) share the view that militancy has decreased. This suggests a shared perception of unwarranted repression by the state. About (38%) of respondents claim to have personally witnessed violence by the police or army and even fewer (7%) report having been injured by violence.

The vast majority of respondents rank their regional or religious identities as most important to them and their national identity comes last. The privileging of parochial identities at the expense of the national one is reflected in clear evidence of in-group bias in our quasi-behavioral measure of identification: on average, 75 percent of respondents’ donations go to the Kashmir-only NGO and only 25 percent to the all-India NGO.

It is against this backdrop of intense ethno-regionalism that we apply the experimental interventions described earlier.

Violence Exposure and Ethnic Identification

Our analysis begins by focusing on the effects of exposure to violence on ethnic and national identification, first in an observational setting using self-reported violence exposure and then using the experimental data.

We code a dichotomous variable that takes a value of 1 (and 0 otherwise) if a respondent answered in the affirmative to any of the following statements: personally witnessed violence against others by the police or military; been personally injured in the conflict; had home or business been damaged in the conflict; been taken into custody or detained by police/army; someone in the immediate or extended family has been injured in the conflict or died or disappeared in mysterious circumstances. According to this measure, 57 percent of the sample has been exposed to violence. For each of the outcome measures, we then estimate three separate OLS linear probability models of the form:

(1)$$f\,(Y_i) = \beta _0 + \beta _1ReportedViolenceExposure_i + \gamma \Theta _i + {\rm \epsilon} $$
(2)$$f\,(Y_i) = \beta _0 + \beta _1ReportedViolenceExposure_i + \gamma \Theta _i + \pi X_i + {\rm \epsilon} $$
(3)$$f\,(Y_i) = \beta _0 + \beta _1ReportedViolenceExposure_i + \gamma \Theta _i + \pi X_i + \mu _i + {\rm \epsilon} $$

Where i indexes individual respondents and β 1 is the principal quantity of interest we estimate by regressing an outcome of interest Y on a dummy variable indicating reported violence exposure. The outcome variables have been coded to indicate greater national identification or a greater willingness to participate in or higher endorsement of protests when they take higher values. We utilize data from the full sample with Θi representing a vector of controls for random assignment to the five principal experimental conditions (with the control group being the reference category). X i is a vector of individual-level covariates including age, gender, education, religion, household head, wealth, and government occupation. μ i are constituency fixed effects and ε is the error term. We calculate heteroskedasticity-robust (Huber-White) standard errors for all regressions. Figure 1 summarizes the results graphically using a coefficient plot of the estimated “effect” of violence exposure on national identification. The appendix has the corresponding regression tables.

Figure 1. Coefficient plot for correlation between self-reported exposure to violence (1/0) and national identification

The main pattern that emerges from Figure 1 is that self-reported violence exposure is negatively correlated with national identification across the range of different survey responses and when these measures are combined into an index of national identification. In addition, those respondents who report being exposed to violence in the past are also more likely to endorse peaceful and violent protests against the Indian government.

There are obvious reasons to be skeptical of these correlations. Exposure to violence is itself a function of weak national identification. Individuals who identify most strongly with their ethnic group may be more likely to support separatism or irredentism and may also be most willing to pay personal costs in the form of state coercion to further these political goals. Simply controlling for individual or geographic characteristics that predict both violence exposure and identification is unlikely to produce credible estimates using cross-sectional data (and risks post-treatment bias). Unobservable variables may account for any measured correlation between violence exposure and identification. Individuals who identify ethnically may be more liable to report exposure to violence to justify their political preference or for other instrumental reasons. Exposure to violence at the level of a geographic aggregate need not affect all individuals and will also produce sorting according to preferences and identification. Panel data might help to mitigate some problems of reverse causality and selection, but even then unobservable time-varying confounding variables may lead to bad inferences.

Our experimental design helps to resolve some of these problems by randomizing exposure to a real-world analogue of state coercion. It also allows estimation using simple difference-in-means tests or corresponding linear probability models with minimal functional form assumptions. For each of the outcome measures detailed below we estimate OLS and linear probability models of the form:

(4)$$f\,(Y_i) = \beta _0 + \beta _1ViolenceExposure_i + {\rm \epsilon} $$

where i indexes individual respondents. An outcome of interest Y is regressed on a dummy variable indicating if a respondent is assigned to a video of protests and state-led crackdowns (with the control group serving as the reference category). Table 3 summarizes the results.

Table 2. Sample characteristics

Table 3. Average treatment effects: exposure to violence

Notes: **p < .05; ***p < .01 (two-tailed); robust standard errors in parentheses.

(1) Rank Ordering of Social Identities: Rank of Indian identity (1 to 4) with higher values indicating greater national identification.

(2) Self-Identification: Respondent sees herself as “Indian Only” or “Indian and Kashmiri” (1) or “Kashmiri Only” (0).

(3) Kashmir's Political Status: Kashmir should remain with India (1) rather than becoming independent or joining Pakistan (0).

(4) India-Pakistan Preference: If independence is not an option, should Kashmir remain with India (1) or join Pakistan (0).

(5) Behavioral Measure of Identification: Allocation to all-India NGO (%) rather than NGO working only in Kashmir.

(6) Index of National Identification I: Excluding India-Pakistan preference.

(7) Index of National Identification II: Including India-Pakistan preference.

(8) Participation in Peaceful Protests: Stated willingness to participate in peaceful protests against the Indian government (1 to 5).

(9) Support for Violence: Endorsement of violence against the Indian government (1 to 5).

When respondents were asked to voice their opinions (in an open-ended format) about the video, large numbers noted that the security forces were using excessive force and that the army and police presence should be reduced. A significant fraction also reported that the events portrayed in the video were common occurrences. Indeed, 47.4 percent of respondents said they had seen such events. Of the remainder, 46.8 percent of respondents reported that they had not seen such events, while 5.8 percent refused to answer the question.

Table 3 shows that exposure to violence generally caused consistently lower levels of national identification among respondents. When respondents are asked to rank their religious, ethnic, national and occupational identities in order of how important they are to them, the national identity is ranked last (1 in our coding scheme) by 78 percent of control group respondents. A further 16 percent of respondents say that this identity is second-to-last in importance. The mean national identification falls from 1.30 in the control condition to 1.24 in the violence condition. Similarly, the proportion of individuals who see themselves as either Indian or Indian and Kashmiri declines from 23 percent57 to 19 percent, the average percentage of a donation that respondents prefer to send to an Indian NGO (over the Kashmiri) falls from 25.6 percent to 22.8 percent and the proportion of respondents who support a policy whereby Kashmir remains a part of India (rather than becoming independent or joining with Pakistan) declines from 29 percent to 25 percent.

These results are not significant at conventional thresholds, but point in the same direction and column 6 shows a statistically significant decrease in the index of national identification that combines these various measures. It is remarkable that state-led repression further reduces identification with the nation even when this identity is already extraordinarily weak.

Moreover, these effects are still stronger where identities are in greater flux and not as liable to floor effects—in particular in response to questions about whether respondents identify more strongly with India or Pakistan. If independence is not an option, the percentage of subjects preferring India over Pakistan declines by nine percentage points from 61 percent to 52 percent. Thus, violent repression by the state causes greater identification with an irredentist neighbor. Along with the treatment effects on the combined indexes (columns 6 and 7), these results suggest that national identification declines with exposure to state violence.58

If state-led repression serves to strengthen parochial and non-national identities, then why do governments turn to violence in the first place? Column 8 suggests an answer. While national identification declines when study subjects are exposed to the violence treatment, they are also significantly less likely on average to say they would participate in peaceful protests. In short, the government may not lose the loyalty of an already hostile population; but by using repression it might deter moderates from collective action, even as the seeds of future grievances are sown.59

What mechanisms might account for this finding? Respondents may update their beliefs about the likelihood of being the victim of state violence themselves. Indeed, a rational fear of consequences might explain respondents' declining willingness to participate in protests.60 Yet, if such purely instrumental motives alone are at work, it is not clear why study subjects don't express higher levels of identification with the nation when this could reduce the probability of becoming the target of state coercion. An alternative mechanism draws on the psychological theories of national and ethnic identification that we set forth previously. These theories postulate that individuals do not solely choose identities with reference to material considerations but also derive self-esteem from membership in particular groups. Exposure to state-led violence could increase the psychological distance between ethnic and national identities and thereby increase ethnic identification and support for irredentism, which over the longer term may increase the incidence of separatist violence, even if state coercion reduces protests in the short term.61

Heterogeneous Effects of Violence Exposure

The overall impact of violence exposure is to reduce national identification and increase irredentism and ethnic identification, but an important question is whether these effects are concentrated among particular subgroups of the population. Column 3 of Table 4 examines heterogeneous effects by measuring the effect of the violence exposure intervention in interaction with respondent characteristics measured prior to randomization. The dependent variable is the index of national identification. The estimated quantities are descriptive rather than causal (the variables that define the subgroups are not randomly assigned) and, as with other analyses of effect heterogeneity, there is a risk that the findings are attributable to sampling variability.

Table 4. Heterogeneous treatment effects of violence exposure on national identification index

Notes: **p < .05, ***p < .01 (two-tailed); robust standard errors in parentheses.

Keeping these important caveats in mind, the data suggest that violence exposure reduces national identification by more among those with greater education and among respondents from constituencies with above-median (control group) national identification.

Exposing respondents to state-led repression makes national identifiers statistically indistinguishable from their co-ethnics. While the proportion of less-educated respondents who see themselves as Indian or Indian and Kashmiri remains unchanged (19%) when these study subjects are exposed to violence, the equivalent proportion declines from 32 percent to 17 percent among those with at least a secondary school education. High-education respondents ranked their Indian identity lower after being exposed to violence and reduced the amount they donated to the NGO working all over India by one-third, reducing their allocation from 31.4 percent to 19.9 percent. Among the more educated subgroup, the proportion who believe that Kashmir should remain a part of India (rather than becoming independent or joining Pakistan) declines from 27 percent to 21 percent. When faced with a binary choice of Kashmir joining Pakistan or remaining part of India, support for remaining with India falls by more than a quarter—from 59 percent to 43 percent, compared to a smaller fall (seven percentage points) among the less educated. Similarly, we find somewhat larger treatment effects among people from above-median national-identifying constituencies. Treatment heterogeneity may be because of a “floor effect” since baseline national identification among the less educated is already very low (though we observe similarly heterogeneous effects for the endorsement of protests in Figure A8 and Table A13 in the appendix and there is more variation in responses to the protest endorsement questions). This result also likely reflects the fact that state coercion has the potential to polarize moderates even if they themselves are not the direct targets of this violence.62

In summary, we find that exposure to state-led repression causes national identification and willingness to participate in or endorse actions against the central state to decline, particularly among those who generally identify more strongly with the nation.

Status, Integrative Institutions, and Common Geography

Next, we turn to the effects of national status and integrative institutions on identification. We asked study participants to view a video describing India's economic growth and rising status in the world, watch a video of Kashmiri youths joining the Indian army, and point out Indian cities on a map.63 Half the respondents assigned to each of these treatment conditions were subsequently assigned to watch the video of state crackdowns on protesters. We estimate OLS regressions with the following form:

(5)$$\eqalign{& f\,(Y_i) = \beta _0 + \beta _1EconomicGrowth_i + \beta _2IntegrativeInstitutions_i + \cr & \quad \beta _3CommonGeography_i + \nu _i + {\rm \epsilon}} $$

Once again, the unit of analysis is the individual i and we examine differences in means and proportions across treatment and control conditions using the same set of outcome variables specified previously. ν i is a fixed effect for the additional (orthogonal) randomization of violence exposure.64 Table 5 summarizes a straightforward empirical result: national identification does not significantly increase (and endorsement of actions against the Indian government does not substantially decrease) in response to any of the treatments.

Table 5. Average treatment effects: economic growth, integrative institutions, and common geography

Notes: **p < .05; ***p < .01 (two-tailed); robust standard errors in parentheses.

(1) Rank Ordering of Social Identities: Rank of Indian identity (1 to 4) with higher values indicating greater national identification.

(2) Self-Identification: Respondent sees herself as “Indian Only” or “Indian and Kashmiri” (1) or “Kashmiri Only” (0).

(3) Kashmir's Political Status: Kashmir should remain with India (1) rather than becoming independent or joining Pakistan (0).

(4) India-Pakistan Preference: If independence is not an option, should Kashmir remain with India (1) or join Pakistan (0).

(5) Behavioral Measure of Identification: Allocation to all-India NGO (%) rather than NGO working only in Kashmir.

(6) Index of National Identification I: Excluding India-Pakistan preference.

(7) Index of National Identification II: Including India-Pakistan preference.

(8) Participation in Peaceful Protests: Stated willingness to participate in peaceful protests against the Indian government (1 to 5).

(9) Support for Violence: Endorsement of violence against the Indian government (1 to 5).

The setting, while important from a scholarly and policy perspective, is also one where ethnic identities are deeply entrenched and, conversely, where national identification is weak. This is shown by the fact that over two-thirds of the sample supports independence for the region and three-quarters consider themselves Kashmiri only (rather than Indian or Indian and Kashmiri). Thus, ethnic distance is so large that it may take a huge change in national status or in material incentives to induce individuals to identify nationally. Kashmir could be considered a least likely case for the theory that integrative institutions or high national status would induce national identification.

Respondents’ views on the news segment on India's accelerating growth are telling. Only about one-third of respondents felt that India's economic growth would benefit them; the remainder either noted that the growth would not benefit them personally or else provided an irrelevant answer. This sense among respondents that they would not personally benefit from growth was also reflected in respondents' belief that other Kashmiris would not be made better off.65 Nor did a majority think that India's rapid economic growth would change how it is seen around the world.66 We interpret this as consistent with a pre-existing large psychological distance from the nation because of a history of conflict and mutual suspicion. Kashmiris who do not see themselves as part of the nation will not expect to benefit from India's economic power and are less likely to accept (or state) that India's international status is rising.

The news segment on economic growth is readily comprehensible to Urdu speakers (Urdu is understood by almost all individuals living in Kashmir) but some of the spoken words and all the written words (which are restated by the anchor) are in Hindi, which differs from Urdu in script and some vocabulary. The language of the video and its overtly nationalistic tone may have caused some respondents to identify ethnically. Nevertheless, while the treatment may have had these effects, it also corresponds to how individuals might actually be exposed to this content in their lives (rather than in a form that is contrived by the researchers as is the case in many vignette-based experiments).67 The video's much greater correspondence with reality than is true of many survey experiments has the cost that we cannot identify the precise mechanisms at work, but it does suggest that any increase in national identification that could be generated by information that India is rising in economic status is not sufficient to displace the weight of other factors and considerations that increase “ethnic” distance from the nation.

Qualitative data and recent studies in other contexts provide some confidence that these effects are not solely a function of the treatments, but also comport with broader trends and literature. The outbreak of Kashmir's insurgency in 1989, and its intensification during the 1990s and early 2000s, corresponds to a period of rapid economic growth and rising international status for India (See Table 6 and Figure 2). The null result in the experimental context thus reflects a general phenomenon.

Figure 2. Plot of GDP growth India and insurgency-related casualties in Kashmir over time

Table 6. GDP growth in India and insurgency-related casualties in Kashmir by decade

Notes: Data on GDP growth are from the World Bank 2017. Data on casualties are from South Asia Terrorism Portal 2017.

Unemployment is high and the state's economic development has been damaged by conflict, but according to the National Sample Survey (2011–12) Jammu and Kashmir has the ninth lowest rate of poverty among India's twenty-nine states (10.4% compared to a national average of 21.9%). And while economic growth in recent years has trailed the national mean, state GDP growth was still high.68 While Kashmir accounts for about 1 percent of India's population, it received as much as 10 percent of central government transfers in the 2000–2016 period.69 Numerous discussions with individuals in the region suggest that many people see the demand for political independence as a product of cultural and psychological distance rather than economic grievance. Indeed, many are aware that deep poverty is far more endemic in other regions compared to Kashmir.

The results from the other interventions also suggest low levels of national integration in our sample. When respondents were shown the video of youths joining the army, 53 percent of them said they felt that the youths had made the wrong decision, though the remainder stated that they thought that they were right to enlist (some participants pointed to high youth unemployment as a justification for this view). The inability of the military's integrative effort to elicit greater national identification in the survey context echoes the fact that irredentist and separatist sentiment has endured (even as the scope of militancy has waned) despite intense government efforts to win hearts and minds through infrastructure investment, army “Goodwill Schools,” healthcare facilities, vocational training, youth exchange programs, and a variety of other interventions.

The effectiveness of the Indian army's strategy to win “hearts and minds” through its “Operation Sadbhavana” has not yet been subjected to quantitative scrutiny. But based on qualitative data and interviews Nabi and Ye and Chakrabarti come to the conclusion that the effort has been unsuccessful because citizens’ political preferences and identities do not change (even if they derive material benefits from these efforts) in an environment in which violence and coercion remain widespread.70 Indeed, Nabi and Ye note that during large-scale protests in 2010, “many Sadbhavana projects became targets … as they were seen as symbols of the Indian military occupation. Hoardings which carried the message ‘Jawan and awam Aman hay Mukam’ (soldiers and people, together can bring peace) were the first targets.”71

A growing literature that examines the impact of development programs on conflict suggests that development spending can increase conflict,72 or has mixed effects and does not affect security.73 The effectiveness of such interventions is likely to be highly dependent on local context. Dasgupta, Gawande, and Kapur find that the roll-out of India's National Rural Employment Guarantee Scheme generated a large decrease in Maoist conflict violence.74 To the extent that the Maoist insurgency in India is motivated principally by economic grievances, the apparent failure of development programs and integrative efforts in the Kashmir Valley suggests that high levels of ethnic identification are likely to be an important feature of the local context that determines the success or failure of “hearts and minds” programs.75

The “null” results from our common geography intervention indicate that symbols of a common national identity (the flag; a map of India) are insufficient to bridge the gap dividing the Kashmiris from the rest of India. Three-quarters of the respondents were unable (or did not want) to identify a single city on the map of India that was shown to the respondents assigned to the “Common Geography” group. Low levels of education could account for these results.76 The map treatment was not delivered in the form of a video, unlike the other three treatments related to violence, economic growth, and integrative institutions, so comparing the effect magnitudes of the map treatment to the others is difficult given its subtlety and minimal informational component. Overall, our results suggest that altering the low salience of a common national identity through psychological mechanisms in an environment of violent conflict that reinforces ethnic identification is likely to be a difficult task.

Conclusion

We have examined how mechanisms of violence, status, and distance shape social identification in the context of an active conflict. In studying these mechanisms, we have used new data from the Kashmir Valley, a central case of separatist and irredentist conflict. Our research design was novel in its use of condensed videos to study social identification and after randomly assigning respondents to watch videos that are easily and widely available, we quantified identification using an array of measures. A number of results emerge from this analysis. We find that ethnic and national identities in Kashmir are competing and not complementary and that strong ethno-regional identification results in severe in-group bias. At the same time, a strong ethnic identity also acts as a counterweight to irredentism, and identification with Pakistan is even lower than it is with India. We also find that state-sponsored repression reduces national identification, particularly where identities are in greater flux—in this case identification with India or Pakistan can change significantly as a result of violence exposure. These effects seem to be more concentrated among those who otherwise identify more strongly with the nation. Identification in this context—a hard case where psychological distance to the national identity is large—is not sensitive to increases in national status or interventions designed to reduce psychological distance by integrating Kashmiris into national institutions.

The most robust result in cross-country studies of civil war is the negative correlation between per capita income or income growth77 and civil war risk. This correlation has led many analysts to argue that the answer to civil war is economic development. There are likely several complex mechanisms underlying the correlation between development levels and civil war and the existing literature has not yet identified those mechanisms. To the extent that social identification is one of the factors fueling violent separatism, our study suggests that development strategies based on increasing rates of economic growth will not be sufficient to end the conflict. Similarly, our study speaks to the design and timing of peace-building strategies based on investments in integrative institutions. Such investments are unlikely to be successful unless threats to security are first addressed. Ongoing violence generates enmity and suspicion that diminish the potential of integrative policies.

The practical implications of our findings are sobering. They suggest that national identification in the Kashmir Valley is both extremely low and unlikely to increase in the short run, even in an era of high growth and rising international prestige for India. Integrative institution building and economic development could be winning strategies in many contexts, but our study suggests that they cannot buy the target population's loyalty when social distance is very large. A hypothesis emerging from our study is that nation-building strategies in divided places will succeed only once the population's desire for security is satisfied and violence ends. Moreover, the availability of an “exit” option can complicate nation building in peripheral regions such as Kashmir, where there are interferences by cross-border kin groups and neighboring states with irredentist designs. Low levels of national identification are likely to result in continued demands for political independence that will spill over into street-level protests (and perhaps continued militancy) and are in turn likely to be checked through state-led coercion, thus ensuring that places like Kashmir remain caught in a trap of high ethnic identification and conflict.

Supplementary Material

Supplementary material for this article is available at <https://doi.org/10.1017/S0020818318000498>.

1. Cederman, Girardin, and Gleditsch 2009.

2. Fearon and Laitin 2003.

3. The term nation is used here to refer to all individuals and groups living in a given country. A premise underlying our analysis is that the national identity is open and available to any citizen of the country. Whether or not citizens identify with the nation (country) varies over time as a function of variables we consider in our theory, including violence. Violence among co-nationals changes the meaning of the nation.

4. Mylonas 2013.

5. We use the term region to refer to subnational regions, peripheral, border, or administrative regions (e.g., Kashmir, Punjab, Kerala), or a larger regions (e.g., South Asia) depending on the context. The key idea is that ethnic identification can be defined by territorial boundaries.

6. Horowitz 1985.

7. See Bose 2009; Schofield 2000; Staniland 2014; Varshney 1991, and their references. See also results of public opinion polls in Braddock 2010; Irwin and CVoter 2008; Kumar et al. 2015.

8. Early studies of ethnic identification in the United States found that ethnic identities can co-exist with a strong national identity, though each becomes salient in different contexts. Waters, 1990. It is less clear whether ethnic and national identities can co-exist in states with ongoing ethnic violence or if ethnic—as opposed to civic—nationalism is inherently incompatible with this pluralism.

9. Staerklé et al. 2010 show evidence from twenty countries that majority subgroup members within a superordinate category such as a nation are more likely than minority subgroup members to identify with the nation because they believe the nation represents their subgroup's norms and values. Using survey data from sixteen Sub-Saharan African countries, Robinson 2014 shows that urbanization, education, and employment in the modern economy strengthen national identities at the expense of ethnic ones.

10. Horowitz 1985, 286.

11. Sambanis and Shayo 2013.

12. Gaertner and Dovidio 2000.

13. Hechter 2000.

14. Previous studies have shown that ethnic distance and polarization increase with conflict intensity. For example, Bhavnani and Miodownik 2009 show that the salience of ethnic—as opposed to national—identity is greater in Nigerian states in the Delta region that are rich in resources and experience high levels of ethnic conflict.

15. Grossman, Manekin, and Miodownik 2015.

16. Voors et al. 2012.

17. Bellows and Miguel 2009; Blattman 2009. However, these effects are likely restricted to the ingroup and greater pro-sociality toward one's in-group might actually translate into behaviors that further marginalize outgroups (e.g., voting for ethnic parties, joining associations that exclude minorities).

18. Besley and Reynal-Querol 2014; Cassar, Grosjean, and Whitt 2013; Nunn and Wantchekon 2011.

19. Shayo and Zussman 2011.

20. Sambanis and Shayo 2013. Ethnic violence by one's co-nationals increases ethnic distance from the nation and that, in turn, could reduce identification with the nation unless offset by other factors, such as material interests or inter-group status considerations.

21. Some studies of the phenomenon of “ethnic defection” posit that violence might actually weaken ethnic identification. For example, Staniland 2012 explains switching sides in ethnic war as the result of exposure to what he calls “fratricidal” violence. Such violence, however, is motivated by within-group elite competition. Our theoretical priors are that within-group (intra-ethnic) violence would divide the group along ideological or other grounds, which might diminish individuals’ attachment to the group. National identity might become stronger as a result.

22. Sambanis and Shayo 2013. The derivation of self-esteem from group membership is what makes identification with social groups important for individual behavior. Bettencourt et al., 2001.

23. Samii 2013. Samii's study, as well as most studies of different forms of power sharing, assess the effects of integrationist institutions in the context of a post war settlement. Other features of those peace settlements could interact with the effect of integrationist institutions, as can the overall balance of power between groups at the end of the conflict. We explore the effect of military integration in an active conflict, so our results might differ from previous studies.

24. Cederman, Wimmer, and Min 2010.

25. See Dovidio et al. 2010 for a review.

26. For evidence, see Brewer 1979.

27. Sambanis and Shayo 2013.

28. Cederman, Wimmer, and Min 2010; Saideman and Ayres 2000.

29. See Hechter 2000; Sambanis and Milanovic 2014; Spolaore 2008.

30. Darden and Grzymala-Busse 2006.

31. Gellner 1983; Horowitz 1985.

32. See Davidson 1992; Herbst 1990; Kedourie 1993.

33. See Hechter 2013; Miguel 2004; Robinson 2014.

34. See Alesina and Spolaore 2005; Sambanis, Skaperdas, and Wohlforth 2015.

35. Some of these mechanisms capture long-run influences on identification (e.g., colonial legacies; socialization via the schooling system) while others reflect macro-level conditions that do not change in our study (e.g., economic modernization and external security threats).

36. Horowitz 1985; Sambanis and Shayo 2013.

37. The material for this section comes from Saideman 2005 and the references therein, but many of the central facts are widely known; see Varshney 1991 for another political history of the conflict.

38. Singh 2011.

39. Horowitz 1985; Sambanis and Shayo 2013.

40. Varshney 1991.

41. Irredentist and separatist sentiment, as well as insurgency-related violence, are concentrated in the Kashmir Valley rather than Ladakh or Jammu.

42. Varshney 1991.

43. Saideman 2005, 213.

44. Saideman 2005, 213.

45. A recent field experiment in Nigeria found that short clips of actors reporting corruption that were spliced into a feature film caused citizens to report corruption at higher rates, suggesting that such video interventions can shape attitudes and behavior in unfavorable contexts. Blair, Littman, and Paluck 2017.

46. In an attempt to study the priming of material interests and instrumental motives that might be activated by the message of higher economic growth, we ask all respondents if they believe that higher growth will benefit them personally and if it will benefit other people in the state, before asking a randomly chosen subset of half of the respondents whether higher growth will “change how the country is seen around the world.” Our analysis marginalizes over these subtreatments.

47. Horowitz 1985; Lijphart 1977.

48. Previous studies of military integration's effects focused on the soldiers themselves. Samii 2013 uses data from a natural experiment in Burundi to show that quota-based integration of the military after civil war decreased prejudicial behavior among soldiers without increasing the salience of ethnic identities. We focus on the effects on the broader society, rather than on integrated soldiers. Another important difference from Samii's study is that we focus on an ongoing conflict, which provides a different perspective on institutional integration.

49. Anderson 1983.

50. We anticipated that psychological distance from the nation could be reduced by raising the salience of the shared national identity between Kashmiris and other Indians. We expected this effect to be larger in versions of the map where the flag was shown in the background. All three versions of the map treatment asked respondents to name the four cities on the map, with the expectation that this exercise would highlight the common geographic/national identity shared by all Indians and would increase identification with the nation. The versions of the maps with the Lines of Control were intended to allow us to measure if any increase in national identification because the treatment is moderated by a reinforcement of the idea that this territory is disputed.

51. Eifert, Miguel, and Posner 2010; Robinson 2014; Shayo 2009.

52. Unlike other questions, which respondents answer freely, about a third of respondents refuse to answer this question. Survey respondents with a strong preference for Kashmir becoming independent refused to say whether they would prefer being part of India or Pakistan (in fact, saying that Kashmir should become independent is the best predictor of refusing to answer this question). Response rates are not significantly different among respondents who view the video compared to those in the control condition.

53. Enumerators made choices for respondents who were illiterate. The full amount had to be donated to target in-group preferences without having to deal with inter personal differences in general altruism.

54. Anderson 2008.

55. The substantive and statistical significance of the results remain unchanged if we additionally exclude the question on respondents’ preferences on Kashmir's political status or use alternative methods of computing the indices such as principal components or a standardized index using the method of Kling, Liebman, and Katz 2007.

56. A letter of introduction from the University of Kashmir helped our enumerator teams gain access.

57. Reassuringly, a similar proportion to that found in another recent survey in the Kashmir Valley by Kumar, Palshikar, Sardesai et al. (2015, 70).

58. By most accounts, support for Pakistan was historically low. Varshney (1991, 1005) quotes the US Ambassador in Delhi who, in a letter to the State Department in the 1950s, reported that Sheikh Abdullah, the greatest Kashmiri leader, would have preferred Kashmir to stay in India rather than be annexed by Pakistan, if independence was not an option.

59. The diminished willingness to participate in protest need not imply that individuals identify less with Kashmir. Protest behavior depends both on social identification and material payoffs, which are shaped partly by the costs of those actions. State repression increases the costs of protests. If individuals believe that protests are unlikely to succeed in changing the behavior of the Indian state then their assessments of the material benefits of protest relative to its costs in the face of Indian repression should discourage them from protesting.

60. See Young 2017 for recent experimental findings that indicate that fear diminishes participation in dissent in authoritarian settings.

61. Respondents’ willingness to endorse violent conflict also falls, though the effect is not significant. States may be primarily concerned with ethnic identification to the extent that it correlates with individuals’ willingness to take costly short-term actions in defense of group identities, which may explain the strategic use of coercion.

62. Table A13 in the appendix shows that the results with respect to education are robust to a multiple-outcomes correction using the Benjamini-Hochberg method. Higher levels of identification with India among the educated may be because of a variety of factors. We observed that more educated individuals were more likely to travel to and study in other parts of India, consume news and media with a broader geographic scope, primarily speak Urdu or Hindi rather than Kashmiri, and seek upward mobility and career opportunities with government institutions and regions outside the state. These are the more cosmopolitan members of our sample who may also be more open to integrating with India. Higher national-identification constituencies also tend to be located in the economic and political center of the state rather than in the periphery (though there are exceptions).

63. Each of the treatments had small variations, but we marginalize over these to keep the presentation simple. We do not find that national identification is different across treatment and control conditions and there are no meaningful differences when analyzing the sub treatments separately.

64. The goal of this additional randomization was to study if any increase in national identification produced by the interventions is sensitive to opposing effects from violence exposure. The effect of violence exposure on national identification is again negative, though not significant. Table 7 in the appendix has results for each treatment combination separately. Tests for an interaction between the crossed treatments fail to reject the null hypothesis of no interactive effect.

65. We do not have comparable data on perceptions of inter regional inequality in the control group so we cannot test if treatment effects are different among people whose priors are that Indian growth does not benefit Kashmir.

66. The nationalistic tone of the message delivered by the news anchor and the favorable comparisons to China are consistent with our premise that rising economic power is an element of high national status. The fact that most of our respondents do not accept that India's rising economic power will change how it is seen around the world is evidence that a large ex ante psychological distance colors their interpretation of the news report.

67. The concepts of economic power and national status always have “subtext” that might have different meaning for majority versus minority groups. For many Kashmiris, there will always be a nationalistic undertone to a message about Indian growth and our video is designed to capture that. Extensions of our design could consider varying features of the video message to manipulate the subtext that is communicated to the respondent. For example, if the videos can be produced internally, the identity and features of the news anchor could be varied as well as the name of the news organization along with background information that could provide cues influencing viewer's ethnic versus national identification.

68. Bhandari and Chakraborty 2014.

69. Raghavan 2014.

70. Nabi and Ye 2015; Chakrabarti 2015.

71. Nabi and Ye 2015, 63.

72. Crost, Felter, and Johnston 2014; Nunn and Qian 2014.

73. Beath, Christia, and Enikolopov 2016.

74. Dasgupta, Gawande, and Kapur 2017.

75. Beath, Christia, and Enikolopov 2016 argue that proximity to a supply of foreign insurgents based in Pakistan is an important source of heterogeneous effects on violence across villages that received development aid in Afghanistan. The organizational support that insurgents in the Kashmir Valley receive from Pakistan is undoubtedly crucial to explaining the insurgency's persistence. Yet, such organizational support likely generates returns from Pakistan's perspective because of already low national identification and separatist and irredentist sentiment in the region.

76. A related finding from a survey published in the newspaper The Hindu shows that 14 percent of rural youth across India cannot identify India on a map and 36 percent cannot name the Indian capital. (Vikas Pathak, “36 Percent Rural Youth Can't Name India's Capital, Finds Survey,” The Hindu, 16 January 2018, retrieved from <http://www.thehindu.com/news/national/36-rural-youth-cant-name-indias-capital-finds-survey/article22451734.ece>.) This highlights the importance of mass-level education for nationalism to work: nationalist symbols such as the flag or the map, or the name of major cities, must be common knowledge for them to generate a sense of belonging to the same nation.

77. See Collier and Hoeffler 2004; Miguel 2004.

References

Alesina, Alberto, and Spolaore, Enrico. 2005. The Size of Nations. Cambridge, MA: MIT Press.
Anderson, Benedict. 1983. Imagined Communities: Reections on the Origin and Spread of Nationalism. New York: Verso.
Anderson, Michael L. 2008. Multiple Inference and Gender Differences in the Effects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects. Journal of the American statistical Association 103 (484):1481–95.
Beath, Andrew, Christia, Fotini, and Enikolopov, Ruben. 2016. Winning Hearts and Minds Through Development: Evidence from a Field Experiment in Afghanistan. Working Paper.
Bellows, John, and Miguel, Edward. 2009. War and Local Collective Action in Sierra Leone. Journal of Public Economics 93 (11):1144–57.
Besley, Timothy, and Reynal-Querol, Marta. 2014. The Legacy of Historical Conflict: Evidence from Africa. American Political Science Review 108 (2):319–36.
Bettencourt, B., Charlton, Kelly, Dorr, Nancy, and Hume, Deborah L.. 2001. Status Differences and In-Group Bias: A Meta-Analytic Examination of the Effects of Status Stability, Status Legitimacy, and Group Permeability. Psychological Bulletin 127 (4):520–42.
Bhandari, Laveesh, and Chakraborty, Minakshi. 2014. Spatial Poverty in Jammu and Kashmir. Live Mint, 24 February. Available at <https://www.livemint.com/Politics/qIhZeetYN29chTZPq1s63N/Spatial-poverty-in-Jammu-and-Kashmir.html>.
Bhavnani, Ravi, and Miodownik, Dan. 2009. Ethnic Polarization, Ethnic Salience, and Civil War. Journal of Conict Resolution 53 (1):3049.
Blair, Graeme, Littman, Rebecca, and Paluck, Elizabeth Levy. 2017. Motivating the Adoption of New Community-minded Behaviors: An Empirical Test in Nigeria. Working Paper.
Blattman, Christopher. 2009. From Violence to Voting: War and Political Participation in Uganda. American Political Science Review 103 (2):231–47.
Bose, Sumantra. 2009. Kashmir: Roots of Conflict, Paths to Peace. Cambridge, MA: Harvard University Press.
Braddock, Robert W. 2010. Kashmir: Paths to Peace. Technical Report, King's College London and Chatham House. Available at <http://www.operationspaix.net/DATA/DOCUMENT/4747~v~Kashmir___Paths_to_Peace.pdf>.
Brewer, Marilynn B. 1979. In-group Bias in the Minimal Intergroup Situation: A Cognitive-Motivational Analysis. Psychological Bulletin 86 (2):307–204.
Cassar, Alessandra, Grosjean, Pauline, and Whitt, Sam. 2013. Legacies of Violence: Trust and Market Development. Journal of Economic Growth 18 (3):285318.
Cederman, Lars-Erik, Girardin, Luc, and Gleditsch, Kristian Skrede. 2009. Ethnonationalist Triads: Assessing the Influence of Kin Groups on Civil Wars. World Politics 61 (3):403–37.
Cederman, Lars-Erik, Wimmer, Andreas, and Min, Brian. 2010. Why Do Ethnic Groups Rebel? New Data and Analysis. World Politics 62 (1):87119.
Chakrabarti, Kaustav Dhar. 2015. Sadbhavana and the Paradox of “Winning Hearts and Minds.” Economic and Political Weekly 50 (24):2123.
Collier, Paul, and Hoeffler, Anke. 2004. Greed and Grievance in Civil War. Oxford Economic Papers 56 (4):563–95.
Crost, Benjamin, Felter, Joseph, and Johnston, Patrick. 2014. Aid under Fire: Development Projects and Civil Conflict. The American Economic Review 104 (6):1833–56.
Darden, Keith, and Grzymala-Busse, Anna. 2006. The Great Divide: Literacy, Nationalism, and the Communist Collapse. World Politics 59 (1):83115.
Dasgupta, Aditya, Gawande, Kishore, and Kapur, Devesh. 2017. (When) Do Antipoverty Programs Reduce Violence? India's Rural Employment Guarantee and Maoist Conflict. International Organization 71 (3):128.
Davidson, Basil. 1992. The Black Man's Burden. Africa and the Curse of the Nation-State. New York: Three Rivers Press.
Dovidio, John F., Gaertner, Samuel L., Schnabel, N., Saguy, Tamar, and Johnson, James. 2010. Recategorization and Prosocial Behavior. In The Psychology of Prosocial Behavior: Group Processes, Intergroup Relations, and Helping, edited by Stürmer, Stefan and Snyder, Mark, 289309. Hoboken, NJ: Wiley-Blackwell.
Eifert, Benn, Miguel, Edward, and Posner, Daniel N. 2010. Political Competition and Ethnic Identification in Africa. American Journal of Political Science 54 (2):494510.
Fearon, James D., and Laitin, David D.. 2003. Ethnicity, Insurgency, and Civil War. American Political Science Review 97 (1):7590.
Gaertner, Samuel L., and Dovidio, John F.. 2000. Reducing Intergroup Bias: The Common Ingroup Identity Model. Philadelphia, PA: Psychology Press.
Gellner, Ernest. 1983. Nations and Nationalism. Ithaca, NY: Cornell University Press.
Grossman, Guy, Manekin, Devorah, and Miodownik, Dan. 2015. The Political Legacies of Combat: Attitudes Toward War and Peace Among Israeli Ex-Combatants. International Organization 69 (4):9811009.
Hechter, Michael. 2000. Containing Nationalism. New York: Oxford University Press.
Hechter, Michael. 2013. Alien Rule. New York: Cambridge University Press.
Herbst, Jeffrey. 1990. War and the State in Africa. International Security 14 (4):117–39.
Hlavac, Marek. 2014. Stargazer: LaTeX/HTML Code and ASCII Text for Well-formatted Regression and Summary Statistics Tables. Harvard University, Cambridge, USA. R package version 5.1.
Horowitz, Donald L. 1985. Ethnic Groups in Conflict. Berkeley: University of California Press.
Irwin, Colin, and CVoter. 2008. Peace in Kashmir: Myth and Reality. Technical Report. University of Liverpool and CVoter. Available at <http://www.peacepolls.org/peacepolls/documents/000295.pdf>.
Kedourie, Elie. 1993. Nationalism. Cambridge, MA: Wiley-Blackwell.
Kling, Jeffrey R., Liebman, Jeffrey B., and Katz, Lawrence F.. 2007. Experimental Analysis of Neighborhood Effects. Econometrica 75 (1):83119.
Kumar, Sanjay, Palshikar, Suhas, Sardesai, Shreyas, and Gupta, Pranav. 2015. Interpreting the Mandate in Jammu and Kashmir. Economic and Political Weekly 1 (19):6670.
Lijphart, Arend. 1977. Democracy in Plural Societies: A Comparative Exploration. New Haven, CT: Yale University Press.
Miguel, Edward. 2004. Tribe or Nation? Nation Building and Public Goods in Kenya Versus Tanzania. World Politics 56 (3):328–62.
Mylonas, Harris. 2013. The Politics of Nation-Building: Making Co-Nationals, Refugees, and Minorities. New York: Cambridge University Press.
Nabi, Peer Ghulam, and Ye, Jingzhong. 2015. Of Militarisation, Counter-insurgency and Land Grabs in Kashmir. Economic and Political Weekly 50 (46–47):5864.
Nunn, Nathan, and Qian, Nancy. 2014. US Food Aid and Civil Conflict. The American Economic Review 104 (6):1630–66.
Nunn, Nathan, and Wantchekon, Leonard. 2011. The Slave Trade and the Origins of Mistrust in Africa. The American Economic Review 101 (7):3221–52.
Raghavan, TCA Sharad. 2014. J&K gets 10% of Central funds with Only 1% of population. The Hindu, 24 July. Available at <http://www.thehindu.com/news/national/other-states/JampK-gets-10-of-Central-funds-with-only-1-of-population/article14506264.ece>.
Robinson, Amanda Lea. 2014. National Versus Ethnic Identification in Africa: Modernization, Colonial Legacy, and the Origins of Territorial Nationalism. World Politics 66 (4):709–46.
Saideman, Stephen M. 2005. At the Heart of the Conflict: Irredentism and Kashmir. In The India-Pakistan Conflict: An Enduring Rivalry, edited by Paul, T.V., 202–24. New York: Cambridge University Press.
Saideman, Stephen M., and Ayres, R. William. 2000. Determining the Causes of Irredentism: Logit Analyses of Minorities at Risk Data from the 1980s and 1990s. The Journal of Politics 62 (4):1126–44.
Sambanis, Nicholas, and Milanovic, Branko. 2014. Explaining Regional Autonomy Differences in Decentralized Countries. Comparative Political Studies 47 (13):1830–55.
Sambanis, Nicholas, and Shayo, Moses. 2013. Social Identification and Ethnic Conflict. American Political Science Review 107 (2):294325.
Sambanis, Nicholas, Skaperdas, Stergios, and Wohlforth, William C.. 2015. Nation-building Through War. American Political Science Review 109 (2):279–96.
Samii, Cyrus. 2013. Perils or Promise of Ethnic Integration? Evidence from a Hard Case in Burundi. American Political Science Review 107 (3):558–73.
Schofield, Victoria. 2000. Kashmir in Conflict: India, Pakistan and the Unending war. London: I.B. Tauris.
Shayo, Moses. 2009. A Model of Social Identity with an Application to Political Economy: Nation, Class, and Redistribution. American Political Science Review 103 (2):147–74.
Shayo, Moses, and Zussman, Asaf. 2011. Judicial Ingroup Bias in the Shadow of Terrorism. Quarterly Journal of Economics 126 (3):1447–84.
Singh, Prerna. 2011. We-ness and Welfare: A Longitudinal Analysis of Social Development in Kerala, India. World Development 39 (2):282–93.
South Asia Terrorism Portal. 2017. Fatalaties in Terrorist Violence 1988–2017 (Jammu and Kashmir). Available at <http://www.satp.org/satporgtp/countries/india/states/jandk/data_sheets/annual_casualties.htm>.
Spolaore, Enrico. 2008. Civil Conflict and Secessions. Economics of Governance 9 (1):4563.
Staerklé, Christian, Sidanius, Jim, Green, Eva G.T., and Molina, Ludwin E.. 2010. Ethnic Minority-Majority Asymmetry in National Attitudes Around the World: A Multilevel Analysis. Political Psychology 31 (4):491519.
Staniland, Paul. 2012. Between a Rock and a Hard Place: Insurgent Fratricide, Ethnic Defection, and the Rise of Pro-State Paramilitaries. Political Psychology 56 (1):1640.
Staniland, Paul. 2014. Networks of Rebellion: Explaining Insurgent Cohesion and Collapse. Ithaca, NY: Cornell University Press.
Varshney, Ashutosh. 1991. India, Pakistan, and Kashmir: Antinomies of Nationalism. Asian Survey 31 (11):9971019.
Voors, Maarten J., Nillesen, Eleonora E.M., Verwimp, Philip, Bulte, Erwin H., Lensink, Robert, and Van Soest, Daan P.. 2012. Violent Conflict and Behavior: A Field Experiment in Burundi. American Economic Review 102 (2):941–64.
Waters, Mary C. 1990. Ethnic Options: Choosing Identities in America. Berkeley: University of California Press.
World Bank. 2017. World Development Indicators: GDP Growth Annual (%). Available at <https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locations=IN>.
Young, Lauren E. 2017. The Psychology of Political Risk in Autocracy. Working Paper.