By ASHUTOSH VARSHNEY
(This article was presented at the annual meeting of the American Political Science Association, Boston, 1998, and various versions were also presented in seminars and symposiums at the following universities and other institutions: Chicago, Columbia, Harvard, Illinois (Urbana-Champaign), Michigan (Ann Arbor), MIT, Notre Dame,Oxford,Texas (Austin),Toronto, Uppsala, Yale, the World Bank, and the Ford Foundation (Delhi).Work on the larger project of which this article is part was supported by the SSRC-MacArthur Program in International Security, the U.S. Institute of Peace, the Samuel P. Huntington Fund of Harvard University, and the Ford Foundation. For comments, criticisms, and suggestions, I would like to thank Hans Blomkvist, Kanchan Chandra, Partha Chatterjee, Pradeep Chhibber, Elise Giuliano, Donald Horowitz, Gary King, Atul Kohli, David Laitin, Scott Mainwaring, Anthony Marx, James Morrow, Bhikhu Parekh, Elizabeth Perry, Robert Putnam, Sanjay Reddy, Susanne Rudolph, Jack Snyder, James Scott, Manoj Srivasatava, Alfred Stepan, Steven van Evera, the late Myron Weiner, Yogendra Yadav, Crawford Young, and three anonymous reviewers of this journal).
Much scholarly work has been done on the topics of civil society and ethnic conflict, but no systematic attempt has yet been made to connect the two . The conclusions of my recent, India-based project,2 supplemented by non-Indian materials, suggest that the links between civil society and ethnic conflict are crying out for serious attention. Does civic engagement between different ethnic communities also serve to contain ethnic conflict? Does interethnic engagement differ from intraethnic engagement from the perspective of ethnic conflict? What role do civic organizations play in times of ethnic tensions and why? These questions are not simply of academic relevance, and they have yet to be systematically researched. Given that violence marks many multiethnic societies, our research may well have great practical meaning if we can sort out some key relationships.
This article argues that there is an integral link between the structure of civic life in a multiethnic society, on the one hand, and the presence or absence of ethnic violence, on the other. To illustrate these links, two interconnected arguments are made. First, interethnic and intraethnic networks of civic engagement play very different roles in ethnic conflict. Because they build bridges and manage tensions, interethnic networks are agents of peace, but if communities are organized only along intraethnic lines and the interconnections with other communities are very weak or even nonexistent, then ethnic violence is quite likely. The specific conditions under which this argument may not hold will be theoretically specified toward the end. Their empirical relevance can be ascertained only with further research.
Second, civic networks, both intraethnic and interethnic, can also be broken down into two other types: organized and quotidian. This distinction is based on whether civic interaction is formal or not. I call the first associational forms of engagement and the second everyday forms of engagement. Business associations, professional organizations, reading clubs, film clubs, sports clubs, NGOs, trade unions, and cadre-based political parties are examples of the former. Everyday forms of engagement consist of simple, routine interactions of life, such as whether families from different communities visit each other, eat together regularly, jointly participate in festivals, and allow their children to play together in the neighborhood. Both forms of engagement, if robust, promote peace: contrariwise, their absence or weakness opens up space for ethnic violence. Of the two, however, the associational forms turn out to be sturdier than everyday engagement, especially when confronted with attempts by politicians to polarize people along ethnic lines. Vigorous associational life, if interethnic, acts as a serious constraint on politicians, even when ethnic polarization is in their political interest. The more the associational networks cut across ethnic boundaries, the harder it is for politicians to polarize communities.
The article also briefly considers how interethnic civic organizations developed in India. Much of India’s associational civic structure was put in place in the 1920s, a transformative moment during the freedom movement against the British, when a new form of politics emerged under the leadership of Mahatma Gandhi. After 1920 the movement had two aims: political independence from the British and social transformation of India. Gandhi argued that independence would be empty unless India’s social evils were addressed, drawing attention to Hindu-Muslim unity, the abolition of untouchability, self-reliance, women’s uplift, tribal uplift, labor welfare, prohibition, and so on. The associational structure of India before Gandhi had been minimal, but the Gandhian shift in the national movement laid the foundations of India’s associational civic order. In the process, between the 1920s and
1940s, a host of new organizations came into being. Historical reasoning, therefore, requires that we draw a distinction between proximate and underlying causation. The role of intercommunal civic networks has been crucial for peace at a proximate level.Taking the long view, however, the causal factor was a transformative shift in national politics. Once put in place by the national movement, the civic structures took on a life and logic of their own, constraining the behavior of politicians in the short to medium run.
The article is organized as follows. The first section clarifies three key terms whose meanings are not self-evident: ethnicity, ethnic conflict, and civil society. The second section deals with the puzzle that led me to discover the relevance of civil society for ethnic conflict. The third section summarizes how the puzzle was resolved and presents the arguments that can link ethnic conflict and civil society. The fourth section presents empirical evidence in support of the arguments made. The fifth section considers causation and endogeneity. The final section summarizes the implications of the project for studies of civil society but also suggests a possible set of conditions under which the basic argument about interethnic and intraethnic engagement is unlikely to apply.
I. CLARIFYING CONCEPTS AND TERMS
The terms “ethnic,” “ethnic conflict,” and “civil society” mean different things to different people.To preempt misunderstanding, one needs to specify the meaning one is using. There are two distinct ways in which the term “ethnic” is interpreted. In the narrower construal of the term, “ethnic” groups mean “racial” or “linguistic” groups. This is the sense in which the term is widely understood in popular discourse, both in India and elsewhere. For example, for politics and conflict based on religious groupings, Indian scholars, bureaucrats, and politicians since the time of the British have used the term “communal,” not “ethnic,” reserving the latter term primarily for linguistically or racially distinct groups.
There is, however, a second, broader definition. As Horowitz argues, all conflicts based on ascriptive group identities—race, language, religion, tribe, or caste—can be called ethnic.3 In this umbrella usage, ethnic conflicts range from (1) the Protestant-Catholic conflict in Northern Ireland and the Hindu-Muslim conflict in India to (2) black-white conflict in the United States and South Africa, (3) the Tamil-Sinhala conflict in Sri Lanka, and (4) Shia-Sunni troubles in Pakistan. In the narrower construction of term, (1) is religious, (2) is racial, (3) is linguistic-cum-religious, and (4) is sectarian. In the past the term “ethnic” would often be reserved for the second and, at best, third conflicts but would not be extended to the first and fourth.
Proponents of the broader usage reject such distinctions, arguing that the form ethnic conflict takes—religious, linguistic, racial, tribal—does not seem to alter its intensity, duration, or relative intractability. Their emphasis in on the ascriptive and cultural core of the conflict, and they distinguish it primarily from the largely nonascriptive and economic core of class conflict. Ethnic conflict may indeed have an economic basis, but that is not its core feature. Irrespective of internal class differentiation, race, language, sect, or religion tends to define the politics of an ethnic group. Contrariwise, class conflict tends on the whole to be economic, but if the class into which one is born is also the class in which one is trapped until death—and this is true for large numbers of people—then class conflict takes on ascriptive overtones. Following Horowitz, it is now well understood that the latter characteristics apply not to ethnic systems in general but to ranked ethnic systems, such as America during the period of slavery, South Africa during apartheid, and India’s caste system. Ranked ethnic systems merge ethnicity and class; unranked ethnic systems do not.
The larger meaning, one might add, is also increasingly becoming the standard meaning in the social sciences, even if that is not yet true of politics and activism. I will use the term “ethnic” in this broader sense. In other words, I may distinguish between communal (that is, religious) and linguistic categories, but I will not differentiate between those that are communal and ethnic. Ethnicity is simply the set to which religion, race, language, and sect belong as subsets in this definition. Does “ethnic conflict,” our second term, have a uniquely acceptable meaning? On the whole, the existing literature has failed to distinguish between ethnic violence and ethnic conflict. Such conflation is unhelpful. In any ethnically plural society that allows free expression of political demands, some ethnic conflict is more or less inevitable, but it may not necessarily lead to violence.4 When there are different ethnic groups that are free to organize, there are likely to be conflicts over resources, identity, patronage, and policies.
The real issue is whether ethnic conflict is violent or waged via the polity’s institutionalized channels. If ethnic protest takes an institutionalized form—in parliaments, in assemblies, in bureaucratic corridors, and as nonviolent mobilization on the streets—it is conflict but not violence. Such institutionalized conflict must be distinguished from a situation in which protest takes violent forms, rioting breaks out on the streets, and in its most extreme form civil war ensues or pogroms are initiated against some ethnic groups with the complicity of state authorities. Given how different these outcomes are, explanations for institutionalized conflict may not be the same as those for ethnic riots, on the one hand, and for pogroms and civil wars, on the other. Ethnic peace should, for all practical purposes, be conceptualized as an institutionalized channeling and resolution of ethnic demands and conflicts: as an absence of violence, not as an absence of conflict. The world would arguably be a happier place if we could eliminate ethnic and national conflicts from our midst, but such a postethnic, postnational era does not seem to be in the offing in the near term. Indeed, many postmodern conflicts, even in richer societies, are taking ethnic forms on grounds of authenticity of living styles and distinctiveness of expression.5
Though highly popular and much revived in recent years, the concept of civil society also needs to be subjected to critical scrutiny. According to conventional notions in the social sciences, “civil society” refers to that space which (1) exists between the family, on the one hand, and the state, on the other, (2) makes interconnections between individuals or families possible, and (3) is independent of the state. Many though not all of the existing definitions also suggest two more requirements: that the civic space be organized in associations that attend to the cultural, social, economic, and political needs of the citizens; and that the associations be modern and voluntaristic, not ascriptive. According to the first requirement, trade unions would be part of civil society, but informal neighborhood associations would not.
Following the second requirement, philately clubs and parent-teacher associations would be civic, but a black church or an association of Jews active on behalf of Israel would not. Should we agree with the latter two proposals? Can nonassociational space also be called civic or part of civil society? Must associations, to constitute part of civil society, be of a “modern” kind—voluntaristic and crosscutting, rather than ascriptive and based on ethnic affiliation? The modernist origins of civil society are originally attributed to Hegel’s nineteenth-century theoretical formulations.6 In recent years, however, it has often been suggested that the revival of a modernist notion of civil society derives from debates in Eastern Europe and the English translation of Habermas’s Structural Transformation of the Public Sphere.7 Because the concept of civil society has been so important to the field of political philosophy, it is mostly political philosophers who have explored it in recent times.8 In comparison, the analytic work on civil society in the more empirical fields of the social sciences has not been as voluminous,9 though the need for it should be quite clear.
Only by systematic empirical investigation of the associational and nonassociational forms of civic life can we determine whether the functions and forms attributed to civil society in the normative literature exist as more than simply theoretical propositions. As an illustration of the modernist biases of the conventional definitions of civil society, consider the theoretical arguments of Ernest Gellner, whose writings on civil society have been plentiful as well as influential. “Modularity,” argues Gellner, “makes civil society,” whereas “segmentalism” defines a traditional society.10 By modularity, he means the ability to transcend traditional or ascriptive occupations and associations.
Given a multipurpose, secular, and modern education and given also the objective availability of plentiful as well as changing professional opportunities in posttraditional times, modern man can move from one occupation to another, one place to another, one association to another. In contrast, traditional man’s occupation and place were determined by birth. A carpenter in traditional society, whether he liked it or not, would be a carpenter, and all his kinsmen would be carpenters. He would also not generally be involved in associations; and if he were, the association would most likely be an ascriptive guild of carpenters. An agrarian society, argues Gellner, might be able to avoid the tyranny of the state. That is because the power of the state could not reach all segments of a traditional society, given the decentralized nature of production structure, the low level of communication technology, and the relatively self-sufficient character of each segment. But that does not mean that such a society would be “civil,” for instead of the “tyranny of state,” it would experience the “tyranny of cousins.” Civil society, concludes Gellner, is not only modern but also based on strictly voluntary associations between the state and family, not on ethnic or religious considerations.11 Such claims can be empirically challenged. First, a remarkably large number of studies show that ethnic and religious associations combine ascription and choice. Not all Christians have to be members of a church in a given town, nor all blacks members of a black church. Moreover, it has also been widely documented that ethnic associations can perform many “modern” functions, such as participating in democratic politics, setting up funds to encourage members of the ethnic group to enter newer professions, and facilitating migration of ethnic kinsmen into modern occupations and modern education.12
A similar objection can be raised with respect to the requirement that associations be formal. In much of the developing world, especially in the countryside and small towns, formal associations do not exist. That does not mean, however, that civic interconnections or activities are absent. If what is crucial to the notion of civil society is that families and individuals connect with other families and individuals— beyond their homes—and talk about matters of public relevance without the interference of the state, then it seems far too rigid to insist that this must take place only in “modern” associations. Empirically speaking, whether such engagement takes place in associations or in the traditional sites of social get-togethers depends on the degree of urbanization and economic development, as well as on the nature of the political system. Cities tend to have formal associations; villages make do with informal sites and meetings. Further, political systems may specify which groups may have access to formal civic spaces and establish organizations and which ones may not.
Nineteenth-century Europe provided the propertied classes with access to a whole range of political and institutional instruments of interest articulation; trade unions for workers were slower to arrive. Some of the spirit of these remarks is conveyed in the commentary generated by Habermas’s distinction between the “lifeworld”and “system” in The Structural Transformation of the Public Sphere. In its original formulation, the distinction indicated a radical rupture between the significance of everyday interaction and that of interaction made possible by institutions and organizations. The latter, according to Habermas, was associated with a modern public sphere. Everyday interaction made life, but organized interaction made history.13 The new history of popular struggles launched by those not formally admitted to the public sphere in much of nineteenth-century Europe and America—women, peasants, workers, minorities—suggests the limited utility of the original Habermas distinction.14 Indeed, in his more recent positions, Habermas has all but dropped his earlier, radical distinction.15 Streetcorner activity can now be viewed as a serious civic form if more organized and institutional civic sites are not available—whether generally or to some particular groups.
The point, of course, is not that formal associations do not matter. One of the arguments of this paper is that they do. But at least in the social and cultural settings that are different from those of Europe and North America, if not more generally, the purposes of activity rather than the forms of organization should be the critical test of civic life. Tradition is not necessarily equal to a tyranny of cousins, and capitalist modernity does not always make civic interaction possible. At best, such dualities are ideal types or based on normatively preferred visions. Empirically speaking, tradition often permits challenging the cousins when existing norms of reciprocity and ethics are violated.16 Similarly, even capitalist modernity may be highly unsocial and atomizing, if people in America stay home and watch soap operas on TV, instead of joining PTAs and other civic organizations.17 Both informal group activities and ascriptive associations should be considered part of civil society so long as they connect individuals, build trust, encourage reciprocity, and facilitate the exchange of views on matters of public concern—economic, political, cultural, and social. That they may have very different consequences for conflict or peace is an entirely different matter. The latter is an argument about what type of civil society is better for governance and peace, not whether civil society per se is endowed with benign possibilities.
II.WHY CIVIL SOCIETY?
A PUZZLING FEATURE OF ETHNIC CONFLICT
Civil society is a new variable for the study of ethnic conflict. How it emerged as a causal factor in research on Hindu-Muslim relations, therefore, requires a brief explanation of what my project sets out to do and why. Sooner or later, scholars of ethnic conflict are struck by a puzzling empirical regularity—that despite ethnic diversity, some places (regions, nations, towns, villages) manage to remain peaceful, whereas others experience enduring patterns of violence. Similarly, some societies with an impressive record of ethnic peace suddenly explode in ways that surprise the observer and very often the scholar as well. Variations across time and space on the whole constitute an unresolved puzzle in the field of ethnicity and nationalism.
How does one account for such variations? The standard research strategy, with some exceptions,18 has been to seek the commonalities across the many cases of violence. Although this approach will continue to enlighten us, it can give us only the building blocks of a theory, not a full-blown theory of ethnic conflict. The logic underlying this proposition is simple, often misunderstood, and worth restating.19 Suppose on the basis of commonalities we find that interethnic economic rivalry (a), polarized party politics (b), and segregated neighborhoods (c) explain ethnic violence (X). But can we be sure that our judgments are right? What if (a), (b), and (c) also exist in peaceful cases (Y)? In that case, violence is caused by the intensity of (a), (b), and (c) in X; or there is an underlying factor or contextual element that makes (a), (b), and (c) conflictual in one case but not in the other; or there is yet another factor (d) that differentiates peace from violence. It will, however, be a factor that we did not discover precisely because peaceful cases were not studied with the conflictual ones.
This argument for the necessity of studying variance leads to another important methodological question: at what level must variance itself be studied? Should the unit of analysis be nations, states, regions, towns, or villages? What methodologists call a large-N analysis can help us identify the spatial and temporal trends in violence and allow us to choose the appropriate level for analyzing variance. The project, therefore, considered all reported Hindu-Muslim riots in the country between 1950 and 1995.20 For purposes of identifying larger trends, two results were crucial.
First, villages constitute a remarkably small portion of communal rioting. Between 1950 and 1995 rural India, where a majority of Indians still live, accounted for a mere 3.6 percent of the deaths in communal violence. Hindu-Muslim violence turns out to be primarily an urban phenomenon. Second, within urban India, too, Hindu-Muslim riots are highly locally concentrated. Eight cities21 account for a hugely disproportionate share of communal violence in the country: nearly 46 percent of all deaths in Hindu-Muslim violence (Table 1, column 4).
As a group, however, these eight cities represent a mere 18 percent of India’s urban population (and about 5 percent of the country’s total population, both urban and rural). Put otherwise, 82 percent of the urban population has not been “riot prone.” Given such high local concentrations in urban India, the large-N analysis clearly establishes town/city as the unit of analysis. India’s Hindu-Muslim violence is city specific, not state specific, with state (and national) politics providing the context within which the local mechanisms linked with violence are activated. To understand the causes of communal violence, we must investigate these local mechanisms.
Following this reasoning, the project selected six cities—three from the list of eight riot-prone cities and three peaceful ones—and arranged them in three pairs. Thus, each pair had a city where communal violence is endemic and a city where it is rare or entirely absent.To ensure against comparing apples and oranges, roughly similar Hindu-Muslim percentages in the city populations constituted the minimum control in each pair. The first pair—Aligarh and Calicut—was based on population percentages only. The second pair—Hyderabad and Lucknow— added two controls to population percentages, one of previous Muslim rule and a second of reasonable cultural similarities. The third pair—Ahmedabad and Surat—was the most tightly controlled. The first two pairs came from the North and South. The third came from the same state of Gujarat, sharing history, language, and culture but not endemic communal violence. All of these cities, at this time, have a population of above half a million, and the biggest, Hyderabad, is a metropolis of over four million people.
Why was similarity in demographic proportions chosen as the minimum control in each pair? Both in India’s popular political discourse and in theories about Muslim political behavior, the size of the community is considered to be highly significant. Many politicians, especially those belonging to the Hindu nationalist Bharatiya Janata Party (BJP), who have often subscribed to the idea of “Muslim disloyalty” to India, have argued that the demographic distribution of Muslims makes them critical to the electoral outcomes. Muslims constitute more than 20 percent of the electorate in 197 out of a total of 545 parliamentary constituencies in India. In a first-past-the-post system, where 30 percent of the vote is often enough to win a seat in multiparty contests, these percentages make the Muslims electorally highly significant.22 The higher the numbers of Muslims in a given constituency, argue politicians of the BJP, the greater the inclination of mainstream political parties to pander to their sectional/communal demands and the lower the incentive, therefore, for Muslims to build bridges to Hindus. Thus, according to this argument, appeasement of Muslims, based on their large numbers in a democracy, is the cause of communal conflict and violence in India.23
That Muslim demography has political consequences is, however, not an argument confined to the Hindu nationalist BJP. Leading Muslim politicians also make a demographic claim, but with the causation reversed. The higher the numbers of Muslims in a city or town, they argue, the greater the political threat felt by the leaders of the Hindu community, who react with hostility to legitimate Muslim anxieties about politics and identity. An unjustified, even self-serving opposition on the part of Hindu leaders, they argue, is the source of communal hostilities.24 Both extremes of the political spectrum thus rely heavily on demography for their explanations.
These popular arguments are shared by social scientists as well, although their reasoning is different. It has been argued, for example, that when a city/constituency has a Muslim majority or plurality, Muslims typically prefer Muslims-only confessional parties over centrist intercommunal parties.25 Muslims support centrist parties when their share of the population/electorate is small in a town/constituency. Smaller numbers make it rational to seek the security of a large and powerful mainstream party. Can one find cases—cities or constituencies—where similar demographic distributions lead to very different forms of political behavior? Selecting from a larger sample of such cases, the project did precisely that. As described above, it compared three pairs of cities where a rough similarity in demographic proportions coexists with variance in political outcomes—peace or violence.
III. RESOLVING THE PUZZLE: THE ROLE OF CIVIL SOCIETY
The preexisting local networks of civic engagement between the two communities stand out as the single most important proximate explanation for the difference between peace and violence.Where such networks of engagement exist, tensions and conflicts are regulated and managed; where they are missing, communal identities lead to endemic and ghastly violence. As already stated, these networks can in turn be broken down into two parts: associational forms of engagement and everyday forms of engagement. Both forms of engagement, if intercommunal, promote peace, but the capacity of the associational forms to withstand national-level “exogenous shocks”—such as India’s partition in 1947 or the demolition of the Baburi mosque in December 1992 in the presence of more than three hundred thousand Hindu militants—is substantially higher.
What are the mechanisms that link civic networks and ethnic conflict? And why is associational engagement a sturdier bulwark of peace than everyday engagement? One can identify two mechanisms that connect civil society and ethnic conflict. First, by promoting communication between members of different religious communities, civic networks often make neighborhood-level peace possible. Routine engagement allows people to come together and form organizations in times of tension. Such organizations, though only temporary, turned out to be highly significant. Called peace committees and consisting of members of both communities, they policed neighborhoods, killed rumors, provided information to the local administration, and facilitated communication between communities in times of tension. Such neighborhood organizations were difficult to form in cities where everyday interaction did not cross religious lines, or where Hindus and Muslims lived in highly segregated neighborhoods. Sustained prior interaction or cordiality facilitated the emergence of appropriate, crisis-managing organizations.
The second mechanism also allows us to sort out why associational forms of engagement are sturdier than everyday forms in dealing with ethnic tensions. If vibrant organizations serving the economic, cultural, and social needs of the two communities exist, the support for communal peace tends not only to be strong but also to be more solidly expressed. Everyday forms of engagement may make associational forms possible, but associations can often serve interests that are not the object of quotidian interactions. Intercommunal business organizations survive because they connect the business interests of many Hindus with those of Muslims, not because of neighborhood warmth between Hindu and Muslim families. Though valuable in itself, the latter does not necessarily constitute the bedrock for strong civic organizations.
That this is so is, at one level, a profound paradox. After all, we know that at the village level in India, face-to-face, everyday engagement is the norm, and formal associations are virtually nonexistent.26 Yet rural India, which was home to about 80 percent of India’s population in the early 1950s and still contains two-thirds of the country, has not been the primary site of communal violence. By contrast, even though associational life flourishes in cities, urban India, containing about onethird of India’s population today and only 20 percent in the early 1950s, accounts for the overwhelming majority of deaths in communal violence between 1950 and 1995.
Why should this be so? Figure 1 presents a formal resolution of the paradox. It depicts the relationship between size and civic links diagrammatically, holding the level of civic engagement constant. Moving from circle 1 to 4, we can see why associational engagement is necessary in cities if we answer the following question: how many links will have to be made if we wish to connect each individual with every other individual as we move from villages to cities? Let N represent the number of persons in a village or city, and K the number of links that must be made if everybody is to be connected with everyone else.
The four circles in the diagram increase the size of the local setting. In circle 1, our diagrammatic representation for a small village, there are only two individuals (N = 2); to connect them, we need only one link (K = 1). In circle 2 there are three individuals (N = 3); we need at least three links (K = 3) to connect them all. This circle can represent a small town. Circle 3, in which we have four individuals (N = 4), can be called a city.To connect one with all, we need six links (K = 6). In circle 4, our diagrammatic substitute for a metropolis, there are five individuals (N =5), and at least ten links (K = 10) are needed to connect each of them to everybody else. Thus, for a given level of civic density (in this case, each person connected to everyone else), K rises faster than N. This whole relationship can be written as K = N (N–1)/2.
This formula essentially means that as we move from villages to towns and from towns to cities, we need many more links to connect people than the increase in population. Cities tend naturally to be less interconnected; some degree of anonymity is inevitable. We can now understand what associations do, when villagelike intimacy is no longer possible. Since each association can represent a lot of people, organizations end up reducing N in cities and making a lower K viable. That is why everyday engagement may be effective in villages, with smaller Ns, but not in cities, with higher Ns. Therefore, to maintain the same level of civic engagement in cities as in villages, one needs associations, rather than informal and everyday interaction. The explanation above is deductive. It explains why everyday engagement has very different meanings in rural and urban settings, but it still does not tell us how exactly associations prevent or mitigate communal conflict when they do. That is an empirical question, to which we now turn.
Organized civic networks, when intercommunal, not only do a better job of withstanding the exogenous communal shocks—like partitions, civil wars, and desecration of holy places; they also constrain local politicians in their strategic behavior. Politicians who seek to polarize Hindu and Muslims for the sake of electoral advantage can tear at the fabric of everyday engagement through the organized might of criminals and gangs. All violent cities in the project showed evidence of a nexus of politicians and criminals.27 Organized gangs readily disturbed neighborhood peace, often causing migration from communally heterogeneous to communally homogenous neighborhoods, as people moved away in search of physical safety. Without the involvement of organized gangs, large-scale rioting and tens and hundreds of killings are most unlikely, and without the protection afforded by politicians, such criminals cannot escape the clutches of law. Brass has rightly called this arrangement an institutionalized riot system.28 In peaceful cities, however, an institutionalized peace system exists. Countervailing forces are created when organizations such as trade unions, associations of businessmen, traders, teachers, doctors and lawyers, and at least some cadre-based political parties29 (different from the ones that have an interest in communal polarization) are communally integrated.
Organizations that would lose from a communal split fight for their turf, alerting not only their members but also the public at large to the dangers of communal violence. Local administrations are far more effective in such circumstances. Civic organizations, for all practical purposes, become the ears and arms of the administration. A synergy emerges between the local wings of the state and local civic organizations, making it easier to police the emerging situation and preventing it from degenerating into riots and killings. Unlike violent cities, where rumors and skirmishes, often strategically planted and spread, quickly escalate into riots, the relationships of synergy in peaceful cities nip rumors, small clashes, and tensions in the bud. In the end, polarizing politicians either do not succeed or eventually give up trying to provoke and engineer communal violence. Figure 2 represents the argument diagrammatically.
This argument, it should be clarified, is probabilistic, not lawlike. It indicates the odds but should not be taken to mean that there can be no exceptions to the generalization. Indeed, pending further empirical investigation, lawlike generalizations about ethnic violence may not be possible at all. Upsetting the probabilities, for example, a state bent on inciting ethnic pogroms and deploying its military may indeed succeed in creating a veritable ethnic hell. My argument, therefore, would be more applicable to riots than to pogroms or civil wars. A theory of civil wars or pogroms would have to be analytically distinguished from one that deals with the more common form of ethnic violence: riots. Indeed, perhaps the best way to understand the relationship between civic life and political shocks is via an analogy from meteorology. If the civic edifice is interethnic and associational, there is a good chance it can absorb ethnic earthquakes that register quite high on the Richter scale (a partition, a desecration of a holy place); if it is interethnic and quotidian, earthquakes of smaller intensity can bring the edifice down (defeat of an ethnic political party in elections, police brutality in a particular city) ; but if engagement is only intraethnic, not interethnic, small tremors (unconfirmed rumors, victories and defeats in sports) can unleash torrents of violence. A multiethnic society with few connections across ethnic boundaries is very vulnerable to ethnic disorders and violence.
IV. EVIDENCE THAT CIVIL SOCIETY MATTERS
To establish causality, a modified technique of process tracing was applied to each pair of cities. The technique of process tracing works back from the outcome—peace or violence—step by step, looking to identify what led to what. It can be shown that process tracing, as applied to one case, may not conclusively establish causality.30 The argument about the desirability of variance, summed up in Section II, is also applicable to why case-based process tracing can give us history but not necessarily causality. Therefore, a modification was applied to the technique. In each pair, we looked for similar stimuli that led to different outcomes in the two cities and then identified the mechanisms by which the same trigger produced divergent outcomes. Civil society emerged as a causal factor from such comparisons. If we had studied only violent cities, where interconnections between Hindus and Muslims were minimal or absent in the first place, we would not have discovered what intercommunal civic links can do. A controlled comparison based on variance can thus turn process tracing into a method for establishing causality.
SIMILAR PROVOCATIONS, DIFFERENT RESPONSES
The process outlined above was applied to all three pairs in the project. Civic links between the two communities, combined with the use of such links by local administrations, kept tensions from escalating into riots. To explain how this sequence was established, let me concentrate only on the first pair of cities. Presenting all cities together in a stylized fashion will not give a good sense of the process involved. The first pair consists of Aligarh and Calicut. The former is a riotprone city in the North Indian state of Uttar Pradesh (UP), and the latter a peaceful city in the South Indian state of Kerala. Calicut has not had a single riot in this century; Aligarh figures in the list of the eight most riot-prone cities (Table 1, column 4). Both cities are roughly 36–38 percent Muslim, with the remaining population overwhelmingly Hindu.31
Between 1989 and 1992, when the Hindu nationalist agitation to destroy the Baburi mosque in Ayodhya (hereafter referred to as the Ayodhya agitation) led to unprecedented violence in much of India, both cities experienced rumors, tensions, and small clashes. But the final outcomes were very different. In Calicut the local administration was able to maintain law and order. Unfounded rumors circulated in the city that pigs had been thrown into mosques. Similarly, there were rumors that Muslims had attacked the famous Guruvayur temple, a site greatly venerated by Hindus in the state. Such rumors often led to riots in several cities in India and frequently did so in Aligarh. In Calicut the peace committees and the press helped the administration quash the rumors. The storm of the Ayodhya agitation, the biggest since India’s partition and one that left hundreds dead in several cities, skirted Calicut and left it unharmed.
By contrast, blinded by a Hindu nationalist fervor during the Ayodhya agitation, the city of Aligarh plunged into horrendous violence. Unlike Calicut’s newspapers, which neutralized rumors after investigating them, Aligarh’s local newspapers printed inflammatory falsehoods. Two of the largest-circulation Hindi newspapers wrote in lurid detail of Muslim nurses, doctors, and staff of the Aligarh Muslim University (AMU) hospital killing Hindu patients in cold blood.32 Some Hindus were indeed killed outside the university campus,33 but nobody was murdered in the AMU hospital.34 The rumors were believed, however. And gangs of Hindu criminals went on a killing spree. Some of them stopped a train just outside the city, dragged Muslims out, and brutally murdered them. The press underreported their acts of killing.
Although these newspapers were later reprimanded for unprofessional behavior by the Press Council, the damage had already been done. Gruesome violence rocked the city for several days, leading to nearly seventy deaths and many more injuries. As in the past, Aligarh’s local mechanisms of peace were remarkably inadequate to the task of dealing with an exogenous shock—in this case, the Ayodhya agitation. The criminals who engaged in killings could not be brought to justice. Not only were they protected by politicians, but they also had remarkable journalistic connections—Muslim criminals with the Urdu press and Hindu thugs with the Hindi press. Effective peace committees could not be formed at the city level in Aligarh, for it was difficult even to get the Hindu nationalists and Muslim politicians together. Rumors were often started and then exploited by political organizations. Instead of investigating rumors, the press simply printed them with abandon. Contrast the situation with Calicut. Two points were common to all accounts given by administrators of Calicut between 1989 and 1992 (as well as those posted there since the mid-1980s) about how the peace was kept. First, politicians of all parties helped establish peace in the city, instead of polarizing communities, as in Aligarh. Second, city-level peace committees were critical to management of tensions.35 They provided information to the administration, became a forum at which all were welcome to speak out and express their anger, gave a sense of participation to local actors, and provided links all the way down to the neighborhood level, where, in addition, citizens formed smaller peace committees.
By contrast, those peace committees that do emerge from below in Aligarh have often tended to be intrareligious, not interreligious. They are organized at the neighborhood level to protect coreligionists from a possible attack by other communities and do not facilitate communication with those other communities. Rather, they simply increase the perception of risk and harden the attitudes of those who participate in them. The members of these committees take turns policing their community. The process forms a very different kind of consciousness from what there would be if the committees were interreligious, since by definition intrareligious committees are based not on interreligious trust but rather on a lack of such trust. Moving within one’s own community, hearing rumors that no one can verify or disprove, staying up in the middle of the night for weeks together, collecting firearms and other small weapons to ensure that retaliation is swift in the event of attack—these activities of intrareligious committees fuel and reflect a communal consciousness, not a consciousness that builds bridges.
THE VARIETY OF CIVIC NETWORKS
Why did the two cities respond so differently? Why did politicians of all kinds cooperate in Calicut but not in Aligarh? Most of all, why did even those politicians of Calicut who stood to benefit from Hindu-Muslim polarization, like the Hindu nationalists of the BJP, avoid working to inflame communal passions and instead cooperate in peacemaking efforts? The BJP leader of Calicut admits that Hindu-Muslim polarization would be in his party’s political interest because it would lead larger numbers of Hindus to vote for the BJP instead of noncommunal parties, as most do currently. But he is also convinced that it would not be wise for his party to systematically initiate the polarizing process, because it might then be blamed for undermining the local peace. If, however, the radical Islamic groups were to launch a violent campaign, it would doubtless benefit the party and the BJP would be happy to respond in kind.36
To understand why the BJP is unwilling to engage in polarizing activities in Calicut, one needs to survey the texture of civic life there. Hindu-Muslim civic integration runs so deep in Calicut (and, many would argue, in the state as a whole) that polarization is a highly risky strategy. If a party can be clearly linked to activities destroying the decades-long Hindu-Muslim peace, there is a good chance it will be punished by the electorate. The reverse is true in Aligarh, where the utter weakness of crosscutting links opens up space for communal politicians to play havoc. Consider first the quotidian forms of citizen engagement in the two cities. According to survey results, nearly 83 percent of Hindus and Muslims in Calicut often eat together in social settings; only 54 per cent in Aligarh do.37 About 90 percent of Hindu and Muslim families in Calicut report that their children play together; in Aligarh a mere 42 percent report that to be the case.
Close to 84 percent of Hindu and Muslims in the Calicut survey visit each other regularly; in Aligarh only 60 percent do, and not often at that. The Hindus and Muslims of Calicut simply socialize more often and enjoy it much of the time, whereas Hindu-Muslim interactions in Aligarh are comparatively thin. Aligarh’s statistics on all of these interactions would be much lower if we had concentrated only on the violent neighborhoods. We see from the few peaceful but integrated neighborhoods that politics has not destroyed civic interaction in all parts of the town, as some of the neighborhoods have managed to keep their distance from the hegemonic political trends elsewhere in the town. It should be noted, however, that an overwhelming proportion of respondents over the age of sixty reported that their neighborhoods in Aligarh had been much more integrated in the 1930s and 1940s.38 But in the 1930s, as politicians started using thugs to spread violence, migration began to communally homogenous localities. Neighborhood-level intimacy was simply unable to withstand the depredations of the emerging politiciancriminal nexus.
What about the associational forms of engagement? Much like Tocqueville’s America, Calicut is a place of “joiners.” Associations of all kinds—business, labor, professional, social, theater, film, sports, art, reading—abound. From the ubiquitous traders associations, to the Lions and Rotary Clubs found in almost all towns in India, to the otherwise rare reading clubs, the head-loaders (porters) association, the rickshaw-pullers association, and even something like an art-lovers association—citizens of Calicut excel in joining clubs and associations. Religiously based organizations also exist, as they do in Aligarh; what is distinctive is the extent of interreligious interaction in nondenominational organizations.39 Consider the economic life of Calicut, which is based primarily on merchandise trade.
The city, with a population of about seven hundred thousand in 1995, was dominated by merchants and traders.40 About one hundred thousand people depended partially or wholly on trade, and, although exact numbers are not available, estimates indicate that the city had between ten and twelve thousand traders.41 It was the rare trader who did not join a trade association. These associations—ranging from organizations of traders who deal in food/grains to those who deal in bullion—were, in turn, members of the Federation of Traders Associations (Vyapari Vyavasayi Ekopana Samithi). In 1995 as many as eleven out of twenty-six trade associations registered with the federation had Hindu, Muslim, (and Christian) office holders: if the president of the association was from one community, the general secretary was from one of the others.42 These associations refuse to align with any particular political parties in electoral contests:
“We don’t want to enter politics because our unity will be broken. We have debates in our association, so conflicts, if any, get resolved.” Moreover, the depth of engagement was such that many transactions were concluded without any formal contracts. “Our relationships with Muslim businessmen are entirely based on trust. Payments as large as 10 to 15 lakhs ($30,000–$35,000) are sometimes due. We send bills, but there are no promissory notes valid in the courts of law. Payments come in thirty days.We work through brokers. There is no breach of trust.”43
Aligarh also has a traders association (Vyapar Mandal). In the late 1980s it had about six thousand members. In the 1970s it had even acquired a fair number of Muslim members, who emerged on the business map after the Gulf migration. The association, however, began to engage in infighting over whether it should support and work for a political party, the argument being that supporting a party favorable to traders would benefit all of them. In the 1980s the association finally split into two bodies: a “secular” organization and a “nonsecular” one, with the nonsecular faction joining the BJP and the Muslims turning to the “secular” faction.44
Unlike trade-based Calicut, Aligarh also has a significant industrial sector and is among the largest producers of locks in India. The lock manufacturing is mostly small scale. Moreover, different units specialize in different parts of the manufacturing process. Yet Aligarh has not developed an economic symbiosis between Hindus and Muslims. It is impossible to estimate the number of people working in Aligarh’s lock industry, as no surveys have been conducted.45 We know from ethnographic work, however, that the workers come from both Muslim and Hindu communities, as do the firm owners.We also know that there is virtually no intercommunal dependence. The informal credit market, normally dominated by Hindu lenders (mahajan), was the only Hindu-run economic activity on which some Muslim manufacturers used traditionally to depend. Over the last few decades rotating credit societies have emerged.46 But these are intra-Muslim societies that build trust within communities, not across them.47
If the businessmen are not integrated, what about the workers? Since they numerically constitute a larger proportion of the city than the businessmen, interreligious links formed in trade unions could, in principle, more than make up for an absence of such links among the businessmen. But trade unions hardly exist in Aligarh. Decrepit offices of the local branches of national trade unions, with no staff and little data, greet researchers who study labor activities. By contrast, trade unions thrive in Calicut. The largest unions are linked to two major national trade-union federations: CITU, which is associated with the Communist Party (Marxist), and INTUC, whose political patron is the Congress Party.48 Both of these unions are intercommunal. Calicut does have a political party of the Muslims, the Muslim League, which regularly wins general elections. It also sponsors a trade union, the STU. The STU, however, is neither as large as the local units of CITU or INTUC nor as vibrant. It is the weakest and smallest of the three. Muslim workers by and large vote in assembly elections for the Muslim League, but they tend typically to join INTUC or CITU for protection of their labor rights.
The Marxist and atheistic character of CITU does not stop them from joining CITU’s unions, if they think that CITU will do a better job of fighting for their rights and wages. In the process, they come in contact with Hindu workers, intercommunal links are formed, and a Hindu-Muslim division of the workforce does not take place. A most unlikely site for unionization—the “head loaders” or porters—is worth mentioning, for it shows the associational abilities and success of Calicut workers. Distributed over hundreds and thousands of shops and small business units, porters in Indian bazaars are rarely unionized. But they are in Calicut (and in Kerala). In 1995 there were nearly ten thousand head loaders in Calicut—about 60 percent Hindu and 40 percent Muslim. Most were part of INTUC and CITU trade unions. There are head loaders in the bazaar in Aligarh, but there they have no associations.
A final and highly distinctive aspect of associational life in Calicut concerns its social and educational activities. The city has had an array of film clubs, popular theater, and science societies. There is nothing unusual about film clubs—they are popular throughout South India. But societies interested in bringing theater and science to the masses are rather uncommon. Even more uncommon have been reading clubs. The literacy rate in Kerala today is the highest in India. “Reading rooms,” a unique Kerala institution, accompanied Kerala’s remarkable rise in literacy and formed deep social networks between the 1930s and the 1950s. Young people from most communities would get together several times every week to read newspapers and cultural and political books. The fascinating story of the birth of reading clubs has recently been told by Menon:
Between 1901 [and] 1931, the rise in the numbers of literate was phenomenal. The growing numbers of schools and the rise in literacy found expressions in the numbers of reading rooms that were established both in the countryside and in the towns. . . . One of the novelties in the organization of reading rooms was the [communitarian] drinking of tea, as one person read the newspapers and the others listened. . . . Tea and coffee lubricated discussions on the veracity of the news and of political questions, and a new culture emerged around the reading rooms. It was premised upon sobriety and knowledge rather than drunken companionship transcending consciousness which characterised the toddy shops.
The importance of tea and coffee lay in the fact that they were recently introduced beverages and did not fit into any taboos regarding what could be shared between castes. Tea shops and reading rooms all over Malabar provided [a] common place for people to meet and to drink together regardless of caste [and community]. . . . The reading rooms emerged as central to both formal attempts at organization by the left wing of the Congress as well as local initiatives.49
The cumulative outcome of the reading-room movement is worth noting. In our Calicut sample, as many as 95 percent of Hindus and Muslims reported reading newspapers—a statistic that is likely to be even higher than in most cities of the richer countries of the world. Calicut today, with a population of over seven hundred thousand, has twenty newspapers and magazines!50 By contrast, while most Hindus in the Aligarh sample read newspapers, less than 30 percent of Muslims did so. Information often travels in the Muslim community by word of mouth. As links with the Hindu community are nonexistent, it takes only a few people to spread nasty rumors and make them stick.
To sum up, the civic lives of the two cities are worlds apart. So many Muslims and Hindus are interlocked in associational and neighborhood relationships in Calicut that peace committees during periods of tension are simply an extension of the preexisting local networks of engagement.51 A considerable reservoir of social trust is formed out of the associational and everyday interactions between Muslims and Hindus. Routine familiarity facilitates communication between the two communities; rumors are squelched through better communication; and all of this helps the local administration keep peace. In Aligarh, however, the average Hindu and Muslim do not meet in those civic settings—economic, social, educational—where mutual trust can be forged.
Lacking the support of such networks, even competent police and civil administrators look on helplessly, as riots unfold. The other pairs in the project experienced similar processes. The different outcomes, however, resided neither in the absence of religious identities nor in the presence of tensions, provocative rumors, and small clashes. Decisive, rather, was the presence of the intercommunal networks of engagement. Intracommunal networks, by contrast, did not contain, or stop, violence.
V. ENDOGENEITY AND THE UNDERLYING CAUSATION
Before we accept the argument about civic engagement, two more questions must be explored. First, how can one be sure that the causation did not flow in the other direction? Did communal violence destroy the Hindu-Muslim civic networks in riot-prone towns, or did the presence of such networks prevent violence from occurring? Might we not have a case of endogeneity here? Second, process tracing can at best establish short-run causality. Is the underlying causation different from proximate causation? Are there historical forces that explain the vitality or absence of civic networks? What emerges if we turn the independent variable of the short-run analysis—civic networks—into a variable to be explained historically?
The city of Surat, the third historically peaceful city in the project, helps us address the problem of endogeneity and establishes the shortrun primacy of civic networks. In Surat (Gujarat) a nasty riot occurred after the destruction of the Ayodhya mosque, the first such riot in nearly seventy years. An overwhelming proportion of violence, however, was confined to the slums; all 192 deaths took place in the shantytowns. The old city, by contrast, witnessed some arson and looting but no deaths. Subjected to the same stimuli, the preexisting social networks accounted for the variance within the city. Surat has experienced an industrial boom in the last twenty years, becoming the small-industry capital of India.
Among cities of more than a million people, Surat has registered one of the highest population growth rates since 1980. Migrants from within and from outside the state have poured into the city and settled in the shantytowns. Working in small industrial units and unprotected by the labor laws of the Factory Act, most of these migrants work exceptionally long hours, returning to the slums and shantytowns only to sleep and eat. There are few institutionalized settings for building civic ties. When the mosque came down in Ayodhya in December 1992, the slums were the site of awful brutality and violence. In the old city, however, peace committees were quickly formed. The business associations of Surat, whose members live primarily in the old city, are especially integrated. These Hindus and Muslims, who had lived side by side for years and had participated in the old city’s business and social life, were able to come together to lower tensions. They set up neighborhood watch committees and deployed their own resources and organizations in checking rumors and communicating with the dministration. As a result, the local administration was more effective in the old city than in the industrial shantytowns, where civic networks were entirely missing and criminals were free to commit acts of savagery and violence.
What about the long-run causation? Have the Hindu-Muslim civic networks always been robust in peaceful towns, directing their Hindu-Muslim politics and making it possible for them to withstand exogenous shocks? Historical research conducted in the cities demonstrates that civic networks—quotidian and associational—determined the outcome in the short to medium run, but in the long run intercommunal networks were politically constructed. The 1920s were a transformative moment in the nation’s politics because it was then that mass politics emerged in India under the leadership of Mahatma Gandhi. Politics before Gandhi had been highly elitist, with the Congress Party a lawyers’ club that made its constitutional arguments for more rights with the British in the Queen’s English.
Gandhi seized control of the movement in 1920 and quietly revolutionized it by arguing that the British were unlikely to give independence to India until the Indian masses were involved in the national movement. Gandhi talked of two intertwined battles of independence (swaraj), one against an external adversary, the colonial power, and another against an internal enemy, India’s social evils. He was interested not only in political independence from the British but also in the social transformation of India, arguing that the former could not be meaningful without the latter. He first concentrated on three social objectives:
Hindu-Muslim unity, abolition of untouchability, swadeshi (buy Indian, wear Indian, think Indian).To these were later added other projects of social transformation: women’s welfare, tribal welfare, labor welfare, prohibition, and so on. In the process millions of his followers created a large number of organizations between the 1920s and the 1940s. Before Gandhi the civic structure of India had been quotidian. After the Gandhian moment in the national movement it became associational. The biggest organization, of course, was the Congress Party, which led the movement politically and developed cadres all over India during the 1920s.52 The argument about social reconstruction also created a second set of organizations, the voluntary agencies. The Congress Party was primarily political, and organizations that dealt with education, women’s issues, the welfare of the tribals and “untouchables,” selfreliance, and the homespun movement were immediately concerned with their social projects.
The civic order that emerged was not identical in different places. The movement had greater success in putting together Hindu-Muslim unity in towns where a Hindu-Muslim cleavage had not already emerged in local politics. India’s towns had been having elections for local governments since the 1880s. If local politics emphasized some other cleavages—for example, caste cleavage among the Hindus or Shia-Sunni divisions among the Muslims—then the Congress Party and Gandhian social workers found it easier to bring Hindus and Muslims together in the local civic life. If, however, Hindu-Muslim differences were the dominant axis of local politics, the national movement could not build integrated organizations with the same success.
Though originally a child of politics, these organizations, once firmly in place, acquired relative autonomy from politics. Depending on how integrated or communal they were, they began to create very different pressures in politics. To sum up, the role of intercommunal civic networks has been crucial for peace at a proximate level. In a historical sense, however, a space for them was created by forms of mass politics that emerged all over India in the 1920s.
For problems of endogeneity, this reasoning suggests a twofold conclusion. If a historical perspective is applied, it turns out that a transformative ideological shift in national politics, seeking to address social evils and to reorient the fight for independence, was the cause of a systematic organizational effort. In the short to medium run, however, the organizational civic order, instituted by the national movement, became a constraint on the behavior of politicians. Given the thrust of the national movement, the civic constraint on politics was especially serious if building or destroying bridges between Hindus and Muslims was the object of politicians’ strategies.
VI. CONCLUDING OBSERVATIONS
Are the conclusions of this paper India specific or have they resonance elsewhere? Two sets of concluding observations—one on civil society and one on ethnic conflict—are in order. Putnam has used the term “social capital” for civic networks.53 My use of the term “networks of engagement” differs from Putnam’s in two ways. First, my focus is on interethnic and intraethnic civic ties, not civic ties per se. Communal and ethnic organizations, focusing on a single religious or ethnic group only, can be shown to generate a great deal of trust among their members. If they are plentiful, such organizations, by Putnam’s definition, can endow a place with a high degree of social capital. However, in my materials, these organizations not only are often incapable of preventing Hindu-Muslim riots but are also associated with the escalation of communal violence.
What matters for ethnic violence is not whether ethnic life or social capital exists but whether social and civic ties cut across ethnic groups. Stated differently, trust based on interethnic, not intraethnic, networks is critical. Second, while civic engagement in Putnam’s work rightly includes both formal and informal interactions between individuals and families, the difference between the two forms should also be noted. For ethnic peace, everyday engagement between ethnic groups may be better than no interaction at all, but it is also qualitatively different from the more formal, organized engagement. Everyday interethnic engagement may be enough to maintain peace on a small scale (villages or small towns), but it is no substitute for interethnic associations in larger settings (cities and metropolises). Size reduces the efficacy of informal interactions, privileging formal associations.54
My findings also have implications for the literature on ethnic conflict. Although disaggregated statistics on local or regional dispersions of ethnic violence have not been systematically collected for many countries, it should first be noted that the data that we do have—for example, for the United States or Northern Ireland55—show roughly the same larger pattern that exists in India. On the whole, ethnic violence tends to be highly concentrated locally or regionally, not spread evenly geographically across the length and breadth of the country. A countrywide breakdown of ethnic relations, more characteristic of civil wars, is rare: we tend to form exaggerated impressions of ethnic violence, partly because violence and not the quiet continuation of routine life is what attracts the attention of media. It is more common to have pockets of violence coexisting with large stretches of peace.
If we systematically investigate the links between civil society and ethnic conflict, there is a good chance we can get a good theory that can explain these local or regional variations.The reason for this intuition is quite simple. Though networks of communities can be built nationally, internationally, and, in this electronic era, also “virtually,” the fact remains that most people experience civic or community life locally. Business associations or trade unions may well be confederated across local units and business or labor leaders may also have national arenas of operation, but most of the time most businessmen and workers who are members of such organizations experience associational life locally.
The type and depth of these local networks—whether they bring ethnic communities together or pull them apart, whether the interactions between communities are associational or informal—are the variables that have the potential for explaining the observable patterns of ethnic violence and peace. Though such research has not been done, some potentially powerful indications are available. A few existing studies of post-1969 Catholic-Protestant violence in Northern Ireland have dealt with intranational variance in violence. John Darby, for example, has studied three local communities in Greater Belfast—Kileen/Banduff, the Upper Ashbourne Estates, and Dunville.56 All three communities have mixed populations, but the first two have seen a lot of violence since the late 1960s, whereas the third has been quiet. Darby found that churches, schools, and political parties were segregated in all three communities, but Dunville had some distinctive features not shared by the other two.
In contrast to the segregated voluntary groups in the first two communities, Dunville had mixed rotary and lions clubs, soccer clubs, and bowling clubs, as well as clubs for cricket, athletics, boxing, field hockey, swimming, table tennis, and golf. There was also a vigorous and mixed single parents club. These results are quite consistent with my Indian findings. Studies of racial violence in the U.S. are also of interest, but in a different and potentially highly challenging way. There is—to the best of my knowledge—no good theory emerging from these studies that can explain city-level variance in racial violence in the 1960s. Why were Newark (New Jersey), Detroit (Michigan), Los Angeles (California), which together accounted for a very large proportion of all deaths in the 1960s riots, so violent? And why did Southern cities, though politically engaged, not have riots?57 The studies show that economic inequalities between African Americans and white Americans neither explained the timing nor the location of riots, but no firm alternative explanations have been provided. Lieberson and Silverman’s work comes reasonably close to what I am arguing for India: they emphasize local integration, especially African American participation in the local government structures.58 But to my knowledge no scholar has investigated whether civic associations—labor unions, churches, PTAs, and so on—were on the whole racially better integrated in the peaceful cities.59
If they were not—and here lies the innovative potential of American race relations in a comparative sense—we might need an initial distinction in our theory between (1) multiethnic societies that have a history of segregated civic sites (unions, churches, schools, business associations, and so on)—for example, the United States and South Africa—and (2) multiethnic societies where ethnic groups have led an intermixed civic life—for instance, India and Sri Lanka. Interracial or intercommunal civic engagement may be a key vehicle of peace in the latter, but, given the relative absence of common black-white civic sites in countries like the United States, there may not have been any space historically for interracial associational engagement, leading to puzzles about the precise nature of mechanisms that led to peace in a different historical and social setting. If we think about the above distinction further, it may actually be more accurate to say that groups, not societies as a whole, have a history of segregation. In India, where political parties, unions, business associations, film clubs, and voluntary agencies are by and large ethnically quite mixed, segregation has marked relations between the Scheduled Castes, who were “untouchable” for centuries, and the “upper castes.”
Historically, there have been no civic or associational sites where the upper castes and the former untouchables could come together. Similarly, Protestants, Catholics, and Jews could eventually find common civic sites in the U.S., but blacks and whites on the whole could not.60 “Self-policing,” a mechanism of peace proposed recently by Fearon and Laitin, may well be relevant to such segregated settings.61 In the terminology developed in this article, it means intraethnic, or intracommunal, policing. If exercised by elders, by an ethnic association, or by civic organizations such as black churches, intraethnic policing may lead to the same result as interethnic engagement does in India. Crosscountry research must take such alternative possibilities seriously. Much remains to be learned.
APPENDIX: RESEARCH MATERIALS
THE NATIONAL LEVEL
The large-N analysis of Hindu-Muslim riots is primarily based on a reading of the daily Times of India, covering a span of forty-six years (1950–95). In case of doubts, reports appearing in other journals were checked, but the Times of India was chosen as the primary source because it is the only newspaper that (1) covers the entire period (1950–95); (2) had a truly national coverage of Hindu-Muslim violence; and (3) had, unlike some other newspapers, often refused to run the potentially most inflammatory stories in this period about communal violence without double-checking them. The Times of India did have some problems of bias, but these problems were resolvable.
Moreover, the newspaper was read interpretively, not literally.Newspaper reports sometimes do not carefully distinguish between intrareligious violence, on the one hand, and interreligious violence, on the other. At other times communal riots are simply presented as a clash between two communities. And the term “communal,” even if applied correctly, can represent Christian-Hindu clashes (as in the Northeast), or Christian-Muslim violence (as in Kerala), or Hindu-Sikh clashes (as in Punjab). An interpretive reading of the reports was thus necessary, based on a detailed understanding of the variety of religious groups, festivals, and contentious issues found in different parts of India. Unless the labeling of the riot in the newspaper was supported by the description of the symbols and issues involved, to which an interpretive reading was applied, a communal riot was not coded as a Hindu-Muslim riot. This data base was put together in collaboration with Steven I. Wilkinson of Duke University.
THE CITY LEVEL
Apart from documentary and archival research for the cities, interviews were conducted at two levels—elite and cross-sectional. I conducted 20–25 interviews at the elite (political, bureaucratic, religious, educational) level in each city. To survey the cross-section in a methodologically defensible way, a stratified sample of 100–140 households was also drawn for each city. To reach the poor, literacy was used as a principle basis for stratification. Illiteracy is a good proxy for poverty, deprivation, and “subalternity” in India. Five or six neighborhoods were then selected in each city—two Hindu dominated (one violence prone, the other peaceful), two Muslim dominated (one violence prone and a second peaceful), and finally, one or two “mixed” neighborhoods. (In peaceful cities neighborhoods where tensions recently surfaced replaced the violence-prone category.) Respondents were selected on the basis of literacy. If 50 percent of Muslims and 30 percent of Hindus in the neighborhood were illiterate, the neighborhood sample of twenty interviewees (ten Hindu, ten Muslim) included five illiterate Muslims and three illiterate Hindus. This procedure was repeated in all neighborhoods. A team of two research assistants—one Hindu, one Muslim—was trained in each city, yielding a research team of twelve in six cities. To ensure candor, Muslim respondents were interviewed by Muslim research assistants, and Hindu respondents by Hindus. We thus got about 700 cross-section interviews in six cities and nearly 125 interviews with elites. The survey was used for two different purposes. The first was to study Hindu and Muslim attitudes toward politics, administration, police, religion, and history and in particular to identify the everyday forms of engagement between the two communities in neighborhoods.
The second purpose was to respond to some standard criticisms of social science research on ethnic conflicts. Unlike works on the functioning of legislatures, executives, and bureaucracies, which typically deal with institutionalized forms of elite politics, research on communalism, ethnicity, and nationalism tends to be part of mass politics and runs up against a by now popular criticism, made especially by postmodernist critics. The complaint is that even while talking about the masses, our sources on communalism and nationalism end up being highly elitist, or “official.”We consult government reports.We conduct interviews with a select group of political leaders, educational and religious elites, bureaucrats, and police officers. And we read newspapers,
which at least in poor countries, do not necessarily represent the opinions of the poor masses. In particular, the use of “official records” on communal violence has been vehemently criticized. It is argued that official records are unreliable, especially when the state, both colonial and postcolonial, may itself be involved, at least partially, as an instigator of communal divisions and/or violence. To deal with these objections and to collect “unofficial transcripts,” I turned my survey, first of all, into a way of collecting brief oral histories on specified questions. This was an unconventional use of survey methods because those who sample rarely collect oral histories and those who collect oral histories rarely sample their respondents. Forty percent of the sample (280 respondents out of 700 separate households in six cities) was sixty plus in age: the aim was to retrieve local memories of Hindu-Muslim realtions in the 1930s and 1940s. Accounts of Hindu- Muslim relations in the 1930s and 1940s are plentiful for the national or provincial level but not for the town level. Since the city was the unit of analysis in the project, local materials for the 1930s and 1940s had to be created—in part, orally.
Moreover, by using literacy as a stratification principle, I also used the survey to collect the so-called “subaltern narratives.” Illiteracy, as already argued, is a good proxy for subalternity in India: those who are illiterate also tend to be very poor. And since such a large part of India is illiterate, a sample stratified according to literacy allowed me access to a large number of illiterate people, especially in the Muslim community, who rarely get interviewed by the newspapers or researchers in a systematic way. To put it differently, through a sampling technique, I sought to “hear the voices of the subaltern.” The subaltern voices are typically heard by those who focus on one town or one village. If one wants to reach the subaltern in a multitown project, stratified sampling— sampling roughly the same proportion of the subaltern as in the population—was by far the most methodologically defensible way of being representative.To make subaltern voices as authentic as possible, considerable prior rapport was struck to create a nonthreatening situation for the interviewees.
The survey data have been converted into statistics only where such conversion is more meaningful: for example, in assessing the degree and nature of everyday engagement between communities or attitudes toward history, personal laws, and administration.Wherever textual summaries were more useful, statistics have not been used.
To sum up, the following kinds of research materials were used: (1) archival research for historical periods on which historians have not yet written and oral records cannot be created; (2) documentary research for contemporary issues; (3) purposive and focused interviews with the elite in all six cities; (4) stratified survey research for the cross section, including the illiterate poor; and (5) a reading of each day’s Times of India between 1950 and 1995 to figure out the long-run and large-N distribution of communal violence over forty-six years.
1 This is not to say that community life within ethnic groups has not been studied as part of civil society. A striking recent example, though not the only one, is Michael Walzer, What It Means to Be American? (New York:Marsilio, 1992). The view that ethnic (or religious) community life can be called civic is, of course, contested by many. The debate is summarized in Section I.
2 Ashutosh Varshney, Ethnic Conflict and Civic Life:Hindus and Muslims in India (New Haven: Yale University Press, forthcoming).
3 For an analysis of why, on the basis of a myth of common ancestry, ethnicity can take so many forms (language, race, religion, dress, diction), see Donald Horowitz, Ethnic Groups in Conflict (Berkeley: University of California Press, 1985), 41–54. One might add that this definition, though by now widely accepted, is not without problems. If all ascriptive divisions can be the basis of ethnicity, can the landed gentry or women’s groups be called ethnic? So long as we equate ascriptive identities with ethnic identities, there is no good answer to such questions.
4 Indeed, such conflict may be inherent in all pluralistic political systems, authoritarian or democratic. Compared with authoritarian systems, a democratic polity is simply more likely to witness an open expression of such conflicts. Authoritarian polities may lock disaffected ethnic groups into long periods of political silence, giving the appearance of a well-governed society, but a coercive containment of such conflicts also runs the risk of an eventual and accumulated outburst when an authoritarian system begins to liberalize or lose its legitimacy.
5 Ronald Inglehart, Modernization and Postmodernization (Princeton: Princeton University Press,1997).
6 For the early history of the idea, see Adam Seligman, The Idea of Civil Society (Princeton: Princeton University Press, 1992).
7 Jurgen Habermas, The Structural Transformation of the Public Sphere: An Inquiry into the Category of Bourgeois Society, trans. Thomas Burger and Frederic Lawrence (Cambridge: MIT Press, 1989). For a debate built around the publication of the English translation, see Craig Calhoun, ed., Habermas and the Public Sphere (Cambridge: MIT Press, 1994).
8 See also Charles Taylor, “Modes of Civil Society,” Public Culture 4 (Fall 1990); Michael Walzer, “The Idea of Civil Society,” Dissent 38 (Spring 1991); and Jean Cohen and Andrew Arato, Civil Society and Political Theory (Cambridge: MIT Press, 1992).
9 The debate generated by Putnam’s work is finally leading to empirically based scholarship. See, inter alia, Sheri Berman, “Civil Society and the Collapse of the Weimar Republic,” World Politics 49 (April 1997); the special issue on social capital of American Behavioral Scientist 39 (April 1997); and Deepa Narayan and Michael Woolcock, “Social Capital: Implications for Development Theory, Research and Policy,” World Bank Research Observer 15 (August 2000).
10 Ernest Gellner, “The Importance of Being Modular,” in John Hall, ed., Civil Society: Theory,History, Comparison (Cambridge: Blackwell, 1995). This article is a good summary of a large number of Gellner’s writings on civil society, written in both the reflective and the activist mode. Many of these writings, including some polemical essays, have been put together in Gellner, Conditions of Liberty: Civil Society and Its Rivals (New York: Penguin Press, 1994).
11 For a similar argument, see Edward Shils, The Virtue of Civility (Indianapolis, Ind.: Liberty Fund, 1997).
12 For pioneering work on the modernist uses of ethnicity, see Lloyd Rudolph and Susanne Rudolph, The Modernity of Tradition (Chicago: University of Chicago Press, 1967); and Myron Weiner, Sons of the Soil (Princeton: Princeton University Press, 1978).
13 See the brief but thoughtful discussion in Harry Boyte, “The Pragmatic Ends of Popular Politics,” in Calhoun (fn. 7).
14 Starting with E. P. Thompson, The Making of the English Working Class (Harmondworth: Penguin Press, 1968), many such historical works by now exist. For a quick review of how they relate to Habermas, see Mary Ryan, “Gender and Public Access: Women’s Politics in Nineteenth Century America,” and Geoff Elly, “Nations, Publics, and Political Cultures: Placing Habermas in the Nineteenth Century,” both in Calhoun (fn. 7).
15 Jurgen Habermas, “Further Reflections on the Public Sphere,” in Calhoun (fn. 7).
16 James Scott, Moral Economy of the Peasant (New Haven: Yale University Press, 1978).
17 For an argument along these lines, see Robert Putnam, “The Strange Disappearance of Civic America,” American Prospect 8 (Winter 1996); and idem, “Bowling Alone,” Journal of Democracy 6 ( January 1995).
18 Among the towering exceptions are Horowitz (fn. 3);Weiner (fn. 12); and W. Crawford Young, The Politics of Cultural Pluralism (Madison: University of Wisconsin Press, 1976).
19 For an elaborate argument for the need for variance in social science research, see Gary King, Robert Keohane, and Sydney Verba, Designing Social Inquiry (Princeton: Princeton University Press, 1993). For why discovering commonalities may matter even in a world where variance exists, see Ronald Rogowski, Symposium on Designing Social Inquiry, American Political Science Review 89 ( June 1995).
20 The data set was put together in collaboration with Steven I.Wilkinson of Duke University.
21 The cities, as Table 1 (column 4) shows, are Ahmedabad, Bombay, Aligarh, Hyderabad, Meerut, Baroda, Calcutta, and Delhi. The last two are not normally viewed as riot prone. But because they have had so many small riots and had some large ones in the 1950s, they are unable to escape the list of worst cities in a long-run perspective (1950–95). In a 1970–95 time series, however, Calcutta is unlikely to figure and Delhi may also disappear.
22 Lloyd Rudolph and Susanne Rudolph, In Pursuit of Lakshmi (Chicago: University of Chicago Press, 1987), 196.
23 L. K. Advani, leader of the BJP, interviewed in Sunday, July 22, 1990.
24 Syed Shahabuddin, a prominent Muslim leader, has often made this argument in lectures, discussions, and political speeches.
25 Rudolph and Rudolph (fn. 22), 195.
26 M. N. Srinivas, Remembered Village (Berkeley: University of California Press, 1979).
27 These connections can be proven social scientifically, not legally. The latter requires establishing individual culpability, not obvious links between politicians and gangs as groups.
28 Paul Brass, Theft of an Idol (Princeton: Princeton University Press, 1997).
29 In a democratic system, political parties would be part of civil society, for not all of them may be linked to the state. In one-party systems, however, parties, even when cadre-based, tend to become appendages of the state, losing their civil society functions. India is a multiparty democracy.
30 For a debate on why process tracing will not easily establish causality, see APSA-CP (Winter 1997).
31 Calicut also has a small Christian population.
32 Aaj, December 10, 1990; Amar Ujala, December 11, 1990.
33 “For an Aligarh of Peace,” interview with District Magistrate A.K. Mishra, Frontline, December 22, 1990, 22–23.
34 Author interviews with AMU vice-chancellor M. Naseem Farooqui, Delhi, July 15, 1994; several AMU professors, August 1994; and local journalists, August 1994. For a thoughtful review of all such reports appearing in local Hindi newspapers, see Namita Singh, “Sampradayitka ka khabar ban jana nahin, kahbron ka sampradayik ban jaana khatarnak hai,” Vartaman Sahitya, September 1991.
35 Author interviews in Trivandrum with Amitabh Kant, district collector, Calicut, 1991–94, July 20, 1995; Shankar Reddy, police commissioner, Calicut, 1991–94, July 22, 1995; Siby Matthews, police commissioner, Calicut, 1988–91, July 21, 1995; K. Jayakumar, collector, Calicut, July 21, 1995; Rajeevan, police commissioner, Calicut, 1986–88, July 21, 1995. Politicians of the Muslim League and BJP confirmed their participation in peace committees. The political leaders interviewed were Dr.Muneer, Muslim League member of Legislative Assembly since 1991, July 23, 1995; K. Sreedharan Pillai, president BJP, Calicut district BJP committee, Calicut, July 25, 1995.
36 Author interview, Sreedharan Pillai, president, BJP district committee, Calicut, July 25, 1995.
37 Unless otherwise reported, the statistics here and below are from the survey conducted in Calicut and Aligarh. For the methodology, see the appendix.
38 Forty percent of the sample was older than sixty, which allowed us to gather recollections of the 1930s and 1940s.
39 It may be asked why people in Calicut join interreligious associations in such large numbers. Since violence and peace constitute the explanandum (the dependent variable) in this analysis and civic networks, the explanans (the independent variable), I only ask whether causality is correctly ascribed to civic networks or, alternatively, whether it constitutes a case of endogeneity. The question of why people join interreligious associations in Calicut but not in Aligarh is analytically different.To answer it requires a research design different from the one that investigates why violence or peace obtains in the two places, for the explanandum is violence in one case and associational membership in the other. That said, it is quite plausible to hypothesize that Calicut citizens have greater faith in the “rationallegal” functioning of the state, and therefore, instead of seeking to change the behavior of the state by capturing state power, they are confident they can exercise enough pressure on it through associations. It may also be that Calicut citizens identify less with caste and religion today than do the citizens of Aligarh, though historically there is no doubt that caste played an enormously important role in generating struggles for social justice there. For a recent account of the caste basis of such struggles, see Dileep Menon, Caste, Community and the Nation: Malabar, 1900–1948 (Cambridge: Cambridge University Press, 1995). Finally, integrated civic networks conceivably achieve much more than prevention of communal riots. They may, for example, be related to the better provision of social services in Calicut (and Kerala), but such outcomes are not the main object of analysis in this paper. Only communal violence is.
40 Calicut has no industry except tiles. It is small in size, with nine factories and about twenty-five hundred workers in all.
41 These numbers and the information below are based on extensive interviews with the president and general secretary of the Kerala Federation of Trade Associations (Kerala Vyapari Vyavasayi Ekopana Samithi hereafter Samithi). The Samithi is a powerful all-state body, based in all towns of Kerala. The Samithi keeps records and statistics and has a professionally run office. It is rare to find a traders association run so professionally in North India.
42 Data supplied by the Samithi, Calicut branch, July 1995.
43 Author interview with V. Ramakrishna Erady, wholesale rice dealer, Calicut, July 25, 1995.
44 Author interview with Mohammed Sufiyan, former president, Vyapar Mandal, Aligarh, August 1995.
45 It pays to underreport how much labor an industrial unit employs, for under Indian law the small informal sector does not have to pay pension and other benefits to its workers. Official statistics are thus entirely useless. Foucault’s concept of “popular illegality,” as one keen observer puts it, has caught the fascination of Aligarh’s lock manufacturers. Elizabeth A. Mann, Boundaries and Identities: Muslims, Work and Status in Aligarh (Delhi and Newbury Park: Sage Publications, 1992), 83.
46 Ibid., 101–2.
47 Ibid., 84–85.
48 Exact numbers of unionized members and their religious distribution are almost impossible to come by. Estimates based on the interviews are the best one can do. The description below is based on interviews with labor leaders in Calicut, especially a long and detailed interview with M. Sadiri Koya, state secretary, INTUC, August 4, 1993.
49 Menon (fn. 39), 145–49.
50 And the state of Kerala has “a library or a reading room within walking distance of every citizen.” K. A. Isaac, “Library Movement and Bibliographic Control in Kerala: An Overview” (Paper presented at the International Congress of Kerala Studies, Trivandrum, India, August 1994).
51 It may be suggested that this finding is close to being a tautology: a city is not riot prone because it is well integrated. This claim, however, would not be plausible for two reasons. First, a conventional explanation, which has long defined the common sense of the field, suggests that for peace, multiethnic societies require consociational arrangements. Consociationalism is an argument about segregation at the mass level and bargaining at the elite level, not integration at either level. My argument is very different. Second, religious fundamentalists have often fought violently to “purify” their communities of influences from other religions in society. Islamic fundamentalists have often sought to undermine Sufi Islam, which has traditionally combined the practice of Islam with the incorporation of neighboring influences. Communally integrated lives and belief systems have often been seen as a source of tension and conflict rather than peace. For the North American version of the debate, see H. D.Forbes, Ethnic Conflict (New Haven: Yale University Press, 1997).
52 It should, however, be pointed out that in Calicut and the neighboring areas, it is the left wing of the Congress Party, later splitting from the parent organization and becoming the Communist Party of India (CPI), that engaged in the most systematic association building.
53 Robert Putnam, Making Democracy Work: Civic Traditions in Italy (Princeton: Princeton University Press, 1993). It should be noted, however, that since writing Making Democracy Work, Putnam has introduced the notions of bridging and nonbridging civic networks. Putnam acknowledges the distinction further in Bowling Alone (New York: Simon and Schuster, 2000).
54 This reasoning also suggests a third way in which this research differs from Putnam’s Making Democracy Work. In Putnam’s formulation, the existence of social capital differentiates good governance from bad. The relationship between social capital and communal violence, however, yields a different formulation. If my argument is right, civic networks determine the presence or absence of riots, but they are politically constructed in the long run. Putnam’s study appears to emphasize the independent role of social capital in both the short run and the long run.
55 For the U.S., see Stanley Lieberson and Arnold Silverman, “The Precipitants and Underlying Conditions of Race Riots,” American Sociological Review 30 (December 1965); and for Northern Ireland, see Michael Poole, “Geographical Location of Political Violence in Northern Ireland,” in John Darby, Nicholas Dodge, and A. C. Hepburn, eds., Political Violence: Ireland in Comparative Perspective (Belfast: Appletree Press, 1990).
56 John Darby, Intimidation and the Control of Conflict in Northern Ireland (Dublin: Gill and MacMillan, 1986).
57 Donald Horowitz, “Racial Violence in the United States,” in Nathan Glazer and Ken Young, eds., Ethnic Pluralism and Public Policy (Lexington, Mass.: Lexington Books, 1983).
58 Lieberson and Silverman (fn. 55).
59 The Kerner Commission Report had an excellent chance to give us an explanation. It missed the chance because it studied the riot-afflicted cities only, not the peaceful ones.
60 Nathan Glazer, We Are All Multiculturalists Now (Cambridge: Harvard University Press, 1996).
61 James Fearon and David Laitin, “Explaining Interethnic Cooperation,” American Political Science Review 90 (December 1996).