EMBEDDED MOBILIZATION Nonstate Service Provision as Electoral Strategy in India By TARIQTHACHIL* I. Introduction HOW do elite parties win over poor voters in developing democracies without alienating their elite core constituencies? Can the private provision of social services by religious parties and their affiliates prove an effective technique for winning votes? This study sits at the nexus of these two crucial concerns for scholars of comparative politics. To address these questions jointly, I examine how social service provision enabled a party of the privileged to win over poor electorates, despite the latter's historical opposition to the former's political ideology. Specifically, I analyze how the upper-caste Bharatiya Janata Party (bjp) has relied on the services provided by its grassroots affiliates in the Hindu nationalist Sangh Parivar (family of organizations) to make unexpected inroads among lower-caste voters in India.1 Despite growing interest in the potential electoral impact of private service provision by religious parties, there is a paucity of systematic empirical research on the subject. In particular, no study has yet examined how services impact individual political preferences in developing democracies. Yet without such disaggregated empirics, it is impossible * I would like to acknowledge the helpful suggestions for improving this work from Ronald Herring, Kenneth Roberts, Christopher Way, Amaney Jamal, Askhay Mangla, Irfan Nooruddin, Daniel Slater, Susan Stokes, Emmanuel Teitelbaum, Steven Wilkinson, and three anonymous reviewers. I am also grateful to feedback from presentations at Brown, Chicago, Cornell, and Yale. In Chhattisgarh, Bhupendra Sahu provided exceptional research assistance. Most importantly, I would like to thank all the informants whose willingness to share their time and opinions made this work possible. The research was made possible by a grant from the Social Science Research Council. All errors remain mine alone. Please direct correspondence to tariq.thachil@yale.edu. 1 This analysis is restricted to those services provided by nonstate actors from their own private resource base; it excludes provision funded by public monies. World Politics 63, no. 3 (July 2011), 434-69 Copyright © 2011 Trustees of Princeton University doi: 10.1017/S0043887111000116 EMBEDDED MOBILIZATION 435 to offer convincing answers to the most pressing questions about how to theorize service provision as an electoral tactic. Can a welfare-based strategy actually provide the basis for demographic expansion in a party's support base? If so, what is the nature of the ties between parties and voters forged by such provision? And is this tactic best conceptualized as a purely material exchange of services for votes, or does provision influence a broader community of voters? A study of Hindu nationalist strategy within India is particularly well suited for shedding light on such fundamental questions about service as an electoral tactic in poor democracies. Toward that end, this analysis uses a combination of original survey data and extensive qualitative fieldwork to provide a microlevel study of this politics of provision.2 This empirical work generates valuable theoretical insights for building a conceptualization of nonstate service provision as an electoral tactic. First, the costs and benefits of this strategy emerge as particularly well tailored to the capabilities and needs of politically conservative actors embedded within wider social movements. The material appeal of services provides a basis for reaching out to the poor, while eschewing the more fundamental compromises with elite interests entailed in implementing pro-poor policies or moving patronage flows outside of traditional elite circles. Service provision can thus be a particularly effective way for parties like the bjp to manage the delicate balancing act of simultaneously appealing to poor communities and appeasing rich ones. Second, a microlevel study of the implementation of this tactic in India yields specific insights for the broader study of the relationship between political parties and social movements.3 The availability of movement partners is clearly essential for realizing the potential of service to create unlikely alliances. Such a central role for nonelectoral organizational affiliates in consolidating voter support, particularly among the poor, has traditionally been understood as the purview of leftist political formations. Previous work on party-voter linkages in Western Europe and Latin America in particular has emphasized the importance of dense networks of organizations such as labor unions and peasant associations in politicizing and mobilizing previously marginalized communities for leftist or indigenous parties.4 2 Harik 1996 provides a rare example of a survey conducted to examine the basis of support for a religious party (Hezbollah in Lebanon) but does not look at service provision specifically. 3 Notable exceptions include Van Cott 2005; Van Cott 2007; and Goldstone 2003. See also Mc-Adam and Tarrow 2010. 4 Bartolini and Mair 1990; Roberts 2002; Van Cott 2005. 436 WORLD POLITICS However there are several distinctions worth noting between the two sets of actors. Unions and peasant confederations typically seek to mobilize constituencies for whom their electoral partners have clear programmatic appeal. In the case of socially conservative parties, by contrast, service-providing organizations must overcome the limited attractiveness of their electoral affiliates for poor voters. This fundamental difference has significant operational repercussions, most notably, that nonelectoral partners of parties like the bjp must be far more circumspect about revealing their ideological affiliations. In fact, Hindu nationalist service providers must take great pains to present themselves as "apolitical," in order to gain access to communities distrustful of their political motivations. Winning electoral support from such a position is a difficult task and requires a range of novel mobilizing strategies. Due to the delicate nature of the work of service organizations, the ties created by them are necessarily more tenuous than the encapsulating linkages created by the overlapping efforts of unions and peasant confederations on behalf of parties with clear pro-poor agendas.5 Moreover, in the latter case, the "bonding" linkages created within marginalized constituencies can prove even more durable when buttressed by highly salient class and/or ethnic cleavages.6 In contrast, my work shows that the "bridging" work that service affiliates seek to perform across social divides for elite parties proves far more successful where poor communities are not highly politicized in either their class or ethnic identities.7 Outlining the local dynamics of provision in India also challenges the conceptual separation of nonelectoral activists seeking to protect the ideological "purity" of the movement from electoral pragma-tists seeking to maximize votes at any ideological cost.8 In the case of Hindu nationalism, services provided by "nonelectoral" activists were hardly uninformed by electoral considerations, nor were BJP personnel removed from ideological constraints. Instead, electoral actors and movement activists found their shared interests converging around the recruitment of lower castes. Indeed, the case of Hindu nationalism revealed the irony that a service strategy produced electoral returns be- 5 Roberts 2002. 6 Roberts 2002, 8. 71 borrow this terminology of course from Putnam 2000 and the well-known literature on social capital, where "bonding" social capital refers to ties that link members within a homogenous community (based on class, race, ethnicity, religion, and so on) and is contrasted with social capital that "bridges" these groups. For an application of this distinction in India, see Varshney 2002. 8 For a general attempt to address the separation of the study of political parties from that of social movements, see Goldstone 2003. For an India-specific example, see Desai 2007. EMBEDDED MOBILIZATION 437 cause it was implemented by nonelectoral activists who seemed to be removed from the "dirty river" of party politics, despite their clearly political motivations. A final point I seek to make concerns the nature of the support that is forged through service provision. I argue that it would be theoretically inadequate to classify the linkages built by a service-based electoral strategy as simply another form of clientelism. First, to be elector-ally successful, services need to be provided consistently, usually over multiple electoral cycles, and they cannot conform to the quid pro quo logic understood as so central to clientelist politics.9 Further, due in part to the recent proliferation of impressive studies of clientelist politics, the relationships between parties and voters are increasingly seen as the product of episodic material transactions. Accordingly, a great deal of scholarly attention has been devoted to ways in which budgets are composed, policies are targeted, and conditional transfers are dispensed, particularly around elections.10 While crucially important, such analyses have tended to diminish the wider influence of quotidian politics and nonmaterial mechanisms on electoral behavior. To the extent that such studies have focused on grassroots organization, scholars have tended to focus solely on the ability of such groups to gather information for parties in order to monitor whether a quid pro quo protocol is being upheld.11 Yet this study of service activists in India reveals a host of additional ways in which grassroots organizations can influence electoral preferences. Specifically, I argue that the service embeds providers within communities, enabling activists to engage in a host of nonmonitoring mobilizing activities, from recruiting candidates and workers to influencing voters through rumor and suggestion. In doing so, I seek to illustrate the benefits of sociologically informed analyses of the crucial everyday interactions between parties and voters, and not only the periodic material transactions between them. In the sections that follow, I first briefly outline the failed efforts of the bjp with lower castes in India and the reasons why a service strategy offered an attractive alternative for the party. I then use qualitative fieldwork, including extensive interviews with villagers, activists, and party functionaries, to outline the workings of a service-based electoral 9 See Kitschelt and Wilkinson 2007; Stokes 2007. 10 There is obviously a large literature on this subject. For some prominent examples, see Calvo and Murillo 2004; Stokes 2005; Magaloni 2006; Kitschelt and Wilkinson 2007; and Diaz-Cayeros et al. 2010. 11 Stokes 2005. 438 world politics strategy at the grassroots level. Finally, I use an original survey conducted with lower-caste voters in central India to test whether this strategy actually succeeded in winning votes for the bjp, both among direct beneficiaries of services and among a broader community of poor voters. II. Poor Voter Linkage Strategies and the BJP's Dilemma The bjp's efforts to woo lower castes were seen as particularly daunting, given that the Brahminical ideology it espouses has largely been understood as one appealing to Hindu elites and not to those subjected to the daily humiliations of caste practice or to those whose spiritual traditions have been denigrated as improper or even uncivilized. Further, the party supported policies that largely appealed to upper castes, whose economic interests and preferences were the diametrical opposite of those of most poor voters.12 The ideology and policy platform of the party led observers of Indian politics to conclude that the bjp would never be able to expand its profile beyond upper-caste, middle-class voters.13 Indeed, several of the party's most famous setbacks, including in a series of state assembly elections in 1993 and in the 2004 national campaign, have been largely attributed to its failure in attracting lower-caste support.14 However, recent electoral evidence suggests several instances of counterintuitive support for the bjp from two of India's most marginalized communities: Dalits (former "untouchable" castes) and Adivasis (India's indigenous tribal populations). In the 2004 elections the party won more than 30 percent of the combined vote for Dalits and Adivasis, in seven major states.15 Such results mark a significant achievement for any party in India's fragmented polity but were particularly notable for the bjp, given the party's status as an upper-caste formation. Moreover the gains were recent: in 1996 the bjp enjoyed this level of success in only a single state (Gujarat). These recent inroads marked a relatively unnoticed, but fundamental puzzle in Indian politics: why were many of the country's most subjugated citizens voting for the party of its most privileged? 12 Chhibber 1999. 13 Indeed the BJP came to be referred to as the "Brahmin-Bania" (two major upper-caste categories) party of India; Graham 1990. 14 Brass 1993; Datar 2004. 15 The states were Assam, Chhattisgarh, Gujarat, Jharkhand, Karnataka, Madhya Pradesh, and Rajasthan. Data from the 1996 and 2004 National Election Study, conducted by the Center for the Study of Developing Societies, New Delhi (weighted averages with blank responses excluded). EMBEDDED MOBILIZATION 439 The literature on political parties has emphasized three ways of appealing to the poor electorally: programmatic, clientelist, and identity-based strategies.16 Perhaps the most explicit way for parties to recruit the poor is by forging what Stokes terms "programmatic redistributive linkages," whereby parties provide broad benefits that serve the material interests of particular classes of voters.17 The classic example of such an approach toward poor voters is provided by social democratic parties in continental Europe whose success with working classes was predicated on pursuing policies beneficial to these populations.18 In India the longtime dominance of communist parties in their stronghold states of West Bengal and Kerala has also been understood as in large part the product of their implementation of redistributive policies, most importantly in land reform.19 Programmatic strategies are most often contrasted with clientelist techniques, which are typically defined as the selective dispensing of excludable benefits to individuals in return for their support.20 Clien-telism can be channeled through the circulation of public sector benefits (patronage) or private resources of a particular party. Earlier accounts also emphasized the sociopolitical asymmetries between patrons and clients as a key component in the success of clientelism.21 In India, too, early forms of patronage politics as practiced by the dominant Congress Party during the 1960s and 1970s were described as resting on the incorporation of local elites who used their influence to deliver the votes of those poor communities under their sway.22 More recent accounts of clientelistic techniques, while certainly not denying the inherent asymmetries between patrons and clients, have tended to emphasize their transactional nature. These analyses describe individualized exchanges of material benefits for votes flowing between networks of parties, brokers, and voters.23 The literature on India has also reflected this trend, with accounts of newer caste-based parties emphasizing the creation of circuits of patronage within caste communities that challenge those de- 16 I draw here on the typology developed by Kitschelt 2000. However, the category of "identity-based" linkages differs somewhat from Kitschelt's category of "charismatic" linkages. While both emphasize the importance of symbolic gestures, the latter is far more concerned with the deployment of such politics by specific individuals to garner a personalistic following, rather than by political formations seeking to politicize certain group identities. 17 Stokes 2007. 18 Bonoli and Powell 2004; Mair, Müller, and Plasser 2004. 19 See Kohli 1987; Herring 1988; Desai 2007. 20 See Kitschelt and Wilkinson 2007; Stokes 2007. 21 Scott 1972. 22 Weiner 1967; Kochanek 1976; Kothari 1964. 23 For example, Calvo and Murillo 2004; Stokes 2005; Magaloni 2006. 440 WORLD POLITICS pendent on intercaste hierarchy. According to Chandra, parties like the Bahujan Samaj Party (bsp) consciously filled their candidate lists with lower-caste personnel, which in turn signaled to coethnics that benefits would be more likely to flow through their caste communities than had been the case under the earlier Congress system.24 Yet both programmatic and clientelist techniques of courting the poor present certain obstacles for elite parties like the bjp. Programmatic redistribution directly compromises the material benefits of elite core constituencies. A party following such a strategy therefore risks losing its primary base of support, especially if core voter loyalty is regarded not as fixed but as subject to attenuation without sustained benefit flows.25 Certainly the bjp's policy platform has reflected this more fluid view of the party's own core supporters by not deviating from prioritizing their interests in officially articulated policy. In fact the most dramatic change in the party's policy positions—its shift from critic to enthusiastic enabler of liberalizing economic reforms—came once party leaders realized that these reforms had broad backing from their elite supporters.26 If the bjp's elite base made programmatic policies an unlikely source for electoral expansion, other strategies proved equally difficult. Replicating the Congress's system required sustained incumbent access to patronage, which even the Congress itself was increasingly unable to achieve. Moreover, the decline of one-party hegemony in India had created a more fragmented polity, with caste serving as a primary building block for politics.27 In such a world the bjp was at a distinct disadvantage in wooing lower castes. Mirroring the BSP strategy of incorporating lower-caste elites met with stiff resistance within the party. The bjp strategist at the helm of efforts to implement a variant of this tactic (termed "social engineering") confessed that the since such an approach required displacement of upper castes from candidate lists it resulted in a massive elite backlash prompting its discontinuation.28 The third option, identity-based mobilization, was certainly the bjp's preferred strategy as a religious nationalist formation. Much ink has been spilled on the party's use of sectarian (or, in South Asian parlance, "communal") mobilizations in the late 1980s and early 1990s to emphasize interreligious cleavages over intra-Hindu caste differences. If successful, agitation-based politics might have allowed the bjp to 24 Chandra 2004. 25 I borrow this insight about the unfixed quality of core voter loyalty from Diaz-Cayeros, Mag-aloni, and Estevez 2010. 26 Chhibber 1997; Lakha 2007; Palshikar 2003. 27 See Chhibber 1999; Yadav 1999. 28 Author interviewwith K. N. Govindacharya, New Delhi, May 15, 2008. EMBEDDED MOBILIZATION 441 expand its base without posing any material threat to elite interests. Unfortunately for the party the agitations organized in the late 1980s and early 1990s clearly failed, by its own account, to resonate with poor voters.29 In the 1993 state assembly elections, widely understood as a referendum on the bjp's involvement in the demolition of a centuries-old mosque in the town of Ayodhya, the party suffered numerous defeats. Moreover, the losses were to parties with largely lower-caste bases of support, prompting scholars to argue that Hindu nationalism had reached its electoral saturation point: it would never be a party of all Hindus.30 Yet despite these pessimistic prognoses about the limits of its demographic appeal, the bjp did manage by 2004 to make inroads with Dalit and Adivasi voters in several states. Given the inadequacy of conventional linkage strategies, what tactic helped the bjp woo these least likely of supporters? This article argues that social services provided by grassroots affiliates of the bjp were crucial in developing unexpected pockets of lower-caste support for the party. Even in the relatively modest literature examining welfare provision by religious movements, there is no clear consensus on how such provision is targeted. Some authors have argued that religious welfare has been deployed to develop vertical ties between elite leadership and potential poor recruits, while others see it operating by and for the middle classes.31 In their careful spatial study of welfare efforts by Hezbollah and the Future Movement in Lebanon, Cammett and Issar argue that the recipients of service are determined by the political motivations of providers.32 When the latter are driven by electoral concerns, social services function as "bricks and mortar cli-entelism" targeting out-group voters (defined in Lebanon in sectarian terms). By contrast, organizations committed to ideological mobilization rather than the electoral process will more narrowly channel benefits to mobilize in-group cadres.33 At first blush, this framework raises the question of how to classify the communities the bjp sought to mobilize. From the perspective of religious identities, the bjp viewed lower castes as falling within its own "in-group" of Hindus.34 Yet when viewing these electorates through the 29 See Hansen 1999, 226. 30 Brass 1993,258. 31 Zubaida 1993; Clark 2004. 32 Cammett and Issar 2010. 33 This question is embedded within a larger debate on whether parties primarily direct benefits to their core supporters as rewards for their unstinting support (Cox and McCubbins 1986; Nitchter 2008) or to new "swing" voters, whom they wish to woo; Dixit and Londregan 1996. 34 It bears noting that while Hindu nationalists may view lower castes as inherently Hindu, many members of these latter communities would resist such classification; see Illiah 1996. 442 WORLD POLITICS prism of the caste-based ethnic identities preeminent in Indian politics, the party clearly did not see Dalits and Adivasis as part of it core support base. Thus, classifying Dalits and Adivasis as part of a "Hindu in-group" would be conceptually indefensible given the realties of identity politics in contemporary India. Viewed in its proper context, the targeting of lower castes by the bjp therefore fits Cammett and Issar's argument of electorally motivated actors in a politically competitive environment targeting ethnic out-group voters. My study thus seeks to build on this prior work by extending the analysis down to the unexamined level of individual voters. Specifically, how do service organizations exert electoral influence over the areas they serve, and is that influence restricted to direct beneficiaries or to a wider community? In answering these questions, however, this study raises questions about the conceptual framework offered by these prior analyses. First, in India the actual service providers themselves were not electoral prag-matists but were activist ideologues working for nonelectoral wings of the movement.35 The interaction of electoral and nonelectoral arms within the same movement confounds our ability to draw conceptual distinctions between electoral pragmatists targeting out-group voters and nonelectoral ideologues mobilizing core bases. While Hindu nationalist activists certainly saw their goals as broader than simply winning elections, they also clearly found ways for their presence to help the bjp electorally. Ideologues can have electoral interests, just as electoral actors seek mobilizational resources, and it was this confluence of interests that underpinned politically motivated service provision in India. Further, services were electorally potent precisely because they were furnished by nonelectoral activists with time horizons lengthened by their ideological commitments. Second, ascertaining when parties use services to target in-group or out-group voters adopts an either/or focus similar to that of studies examining whether electoral actors prioritize distributing benefits to reward their loyal "core" supporters or to make inroads among ambivalent swing voters.36 However, as critiques of this latter literature have suggested, such a focus serves to divert attention from examining how parties might develop strategies to appeal to both core and noncore electorates simultaneously.37 A central contention of this article is that what made a service strategy attractive for a party like the bjp was precisely 35Jaffrelotl993. 36 Cox and McCubbins 1986; Dixit and Londregan 1996; Calvo and Murillo 2004; Stokes 2005. 37 Diaz-Cayeros, Estevez , and Magaloni 2010. embedded mobilization 443 its ability to appeal to noncore lower castes without threatening elite interests. To argue that vote-maximizing actors will always prioritize targeting out-group or swing electorates implicitly assumes that these actors view the support of their core constituencies as unconditional. Yet such an assumption is simply unfounded in the bjp's case, where the party's efforts to woo lower castes were clearly constrained by worries about alienating its upper-caste base. A service approach was therefore pursued precisely because it entailed neither the economic concessions of a programmatic redistributive strategy nor the political compromise of a "social engineering" tactic.38 Finally I argue that even when used to win votes, service provision should not be narrowly classified as simply a variant of clientelist strategy. There can be no doubt that a major part of the appeal of services for poor communities is material, particularly in areas where basic health and educational services remain woefully inadequate, as they do still in many parts of India. Yet to be successful, service activists had to provide services over multiple electoral cycles without linking provision to the quid pro quo protocol that underpins clientelist exchange. Further, to provide services reliably, activists must literally embed themselves within communities. This embedded quality generated several nonmaterial mechanisms through which activists could affect political choices, including even those of many voters not directly benefiting from their services. Providers exploit their formally nonpartisan status and the high regard accruing from their dedicated provision to garner credibility and influence within their local communities. This standing is subsequently used to collect local information helpful for formulating local canvassing strategy, to recruit new candidates and party workers for the bjp, and even to subtly shape voter opinions through rumor and suggestion around elections. Viewing service as simply material exchange would therefore sorely limit our ability to comprehend the full range and complex interplay of these multiple mechanisms. III. Setting the Context: Hindu Nationalism in Chhattisgarh The remainder of this article uses empirical evidence from central India to examine and test the electoral impact of a service-based electoral strategy on lower-caste support for the bjp. The research was conducted in three rounds of fieldwork between 2007 and 2010 in Chhattisgarh, 38 For an example of this rhetoric, see internal Sangh circular as quoted in Mathur 2008, 125. 444 world politics a recent entry into India's list of states. Previously the state's territory was considered part of Madhya Pradesh, India's largest in terms of area and one of the most populous. However, on November 1,2000, sixteen administrative districts of Madhya Pradesh were formally declared to comprise the state of Chhattisgarh. It was to be the first of three new Indian states, along with Jharkhand and Uttarakhand, formed during a two-week period. While national attention to this fledgling state is primarily focused on the activities of dense networks of the Maoist Nax-alite movement in the southern part of the state, more conventional party politics in Chhattisgarh has been relatively understudied. Nonetheless, the state remains a particularly promising one in which to study the bjp's attempts at social expansion among lower-caste communities more closely. Adivasis (known as Scheduled Tribes in the lexicon of the Indian government) comprise 31.8 percent of Chhattisgarh's population, one of the highest concentrations of "tribals" in the country, which along with the state's 11.6 percent Dalit (Scheduled Caste) population gives marginalized voters a substantial electoral presence. However, the bjp faced an uphill battle in wooing both communities in Chhattisgarh, as the Congress Party had established effective dominance in the region through its typical strategy of incorporating local leaders and their networks of patronage into its party structure.39 Yet in recent elections the bjp has fared extremely well in Chhattisgarh, winning large legislative majorities in several state and national elections between 2003 and 2009. What made the party's recent electoral successes in the state surprising was that they were fueled by remarkable inroads with lower-caste voters. According to a survey conducted by the Center for the Study of Developing Society (csds) in New Delhi, the bjp managed to win an impressive 36 percent of the Adivasi vote in the 2003 Chhattisgarh assembly elections, matching the vote share of the Congress. This strong performance was replicated in the 2004 national election, the 2008 assembly election, and then again in the 2009 national election, in which the bjp actually outstripped the Congress with Adivasis, a previously unthinkable feat.40 What role did social services play in facilitating this unlikely rise? In Chhattisgarh, the main service arm of the Hindu nationalist Sangh Parivar is the Vanvasi Kalyan Ashram (Association for the Welfare of "Tribals," vka), which targets Adivasi areas, with a secondary role played by another wing, Seva Bharati (Service to India). The vka was 39 Berthet 2008. 40 Data from the 2009 nes. 5 2 E o o ^; it CTl "D o "D o — 5 q. embedded mobilization 445 F founded in 1952, in the northern Chhattisgarhi district of Jashpur, and its roots in the state are inextricably linked to the long history of Christian missionary activity in the region.41 Yet despite this long history in | Chhattisgarh, the VKAs growth from a marginal to a mass organization g- has been a relatively contemporary phenomenon (see Figures 1 and 2). 2 Did this relatively late expansion also underpin the bjp's recent im-| proved performance with the state's marginalized electorates? § While past scholarship has separately examined the vka's activities, 3 no analysis has rigorously analyzed the impact of Hindu nationalist | service wings on electoral outcomes.42 To address this question, I relied | on a mixed-methods research strategy. First, I observed the techniques 3 these service wings deployed to win votes, using qualitative fieldwork 3? consisting of extensive interviews with Sangh activists and local politi-| cians, semistructured interviews with over eighty households in Jash-jj pur district, and observation of the service chapters themselves. This b research revealed specific mechanisms through which the work of these If nonelectoral networks can potentially benefit the bjp. And then to as-o- sess whether these mechanisms are actually affecting the bjp's electoral p, performance with poor voters, I conducted an original survey of Dalit z and Adivasi respondents across four Chhattisgarhi districts selected to 5 vary on the key independent variable of service network strength. I 3 present my empirical findings from this study in the next two sections. C 0 1 IV. The Local Politics of Service Provision: *i Qualitative Evidence from Central India 03 ^-- u o £ The information in this section is drawn largely from repeated inter- | actions with bjp and Sangh personnel and ordinary villagers during "S 2007-8.43 During that period I observed the activities of the vka in m Chhattisgarh in order to gain a sharper sense of exactly how the orga- nization operated. The first point to note is that while the material appeal of the services the vka offers is quite clear, it is not predicated on their physical quality. With a couple of exceptions the bulk of services offered at the village level are extremely rudimentary. The vka's famous Ekal Vidyalayas (one-teacher schools), which constitute the majority of their schools in the state, operate at close to the dollar-a-day budget advertised in the organization's publicity materials. Costs are kept low by u o ~ o E ^ S & 41 See Sundar 2006 - 00 I g 42 Yadav 2004; Froerer 2006; Berthet 2008; Pandey 2008. 43 All interviews and focus group discussions were conducted in Hindi and translated from Hindi q- ^ to English by the author. 446 WORLD POLITICS 14000h 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 Year Figure 1 Growth in Vanvasi Kalyan Ashram Projects (1981-2007) Source: Data collected by author from vka annual records, Jashpur, Chhattisgarh. employing teacher-activists willing to work for low pay (Rs. 800-1000 or roughly $18-$22 per month) and by holding class in either a home or a public space in the village. Similarly most of the health services offered by the organization are provided by activists with very simple training in first aid and basic hospice care. While there are trained doctors who are part of the VKA network, they tend to manage the handful of dispensaries and care centers the Ashram operates in Chhattisgarh. The majority of village activists have no formal medical background but are trained to treat the most common illnesses encountered in the villages, for example, by dispensing quinine during malaria season or mixing simple sugar solutions for children suffering from dysentery. The implication of the very basic standard of services offered by the VKA is that such services are appealing only in areas where even basic public provisioning is absent, and even then only to those voters who cannot pay to obtain these services privately. Adivasis in a state like Chhattisgarh, with its inadequate public infrastructure, meet both criteria.44 In most villages visited for this study, the closest primary health 44 Sharma 2004. EMBEDDED MOBILIZATION 447 60000 50000 3 40000 > 8 30000 bo >■ o 20000 £ 10000 n—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 Year Figure 2 Growth in Vanvasi Kalyan Ashram Villages Contacted (1981-2007) Source: Data collected by author from vka annual records, Jashpur, Chhattisgarh. care center was more than ten kilometers away, and thus the most ready supply of medicines often came from vka health activists living in the village or were provided by the organization's mobile dispensaries that travel in a rotation from village to village within a district. Even within the comparatively denser networks of public school infrastructure, rural areas experience severely inadequate coverage.45 In more than 70 percent of the villages visited for this study, in those areas where there were primary school buildings in the village, the teacher was either not present or not teaching during spot checks during school hours (in three cases it was revealed that they had been absent for at least the past week).46 To assess the comparative appeal of different providers, I interviewed more than eighty randomly selected households across five villages in the district of Jashpur where both vka and government facilities were 45 Sundar 2004. 46 This local pattern is in line with the results of broader studies of public infrastructure in India. A recent study ranked Chhattisgarh in the bottom quarter of Indian states in terms of teacher attendance, with almost one-third of public school teachers missing school every day and fewer than half actually doing any teaching in class; see Kremer et al. 2005. 448 WORLD POLITICS present. In these interviews, 34 percent of respondents rated the VKA schools as providing the best quality of education, more than twice as many as rated the government most highly (16 percent).47 Why did respondents prefer VKA services? Interview evidence suggests that ideological attachments play a minimal role in household decisions about education. Forty-six percent of interviewees said the primary reason they sent their children to school was for them to get a salaried job, while only 3 percent identified "learning cultural values" as the main impetus.48 What made the VKA rate favorably was the regularity and availability of service, rather than the cultural orientation of providers.49 Over 50 percent of respondents agreed that VKA activists worked "much harder" than their government counterparts (compared with only 9 percent who thought the reverse). Indeed, even among parents who enrolled their children in public schools, the largest number agreed that VKA teachers were more dedicated. Villagers complained of public school teachers who came from urban areas and refused to show up for work when allotted a remote rural posting, their dereliction protected by the well-known difficulties of removing a public sector employee from his or her post.50 By contrast, Sangh activists were identified as committed to residing in the villages in which they worked, ensuring that school would be held with far greater regularity. Thus, even the admittedly basic services of the VKA held real material appeal for Adivasi communities. However, viewing this strategy as exclusively an exchange of goods for votes would miss several important mechanisms linking provision to shifts in electoral preferences. Specifically, service work helped upper-caste activists gain access to initially distrustful communities, generated cadres of party workers and candidates for the bjp, and perhaps most crucially embedded networks of "apolitical" Hindu nationalists within poor constituencies. Activists then deployed their carefully constructed nonpartisan status to become far more effective electoral mobilizers than those within the formal party apparatus. The first point to be noted is that service work enabled upper-caste activists to interact regularly with villagers, a necessary prerequisite to changing their opinion of the Sangh and the BJP as elitist political formations. Prior activism by the Sangh, based less on service and more on 47 Private fee-charging institutions were preferred by 19 percent of respondents, while 13 percent favored Christian missionary education. 48 Author conducted household interviews, Jashpur, August 2010. 49 Focus group discussion, Daurdooba village, Jashpur district, March 13, 2008. 50 Author interview with Pharan, resident, Gutakiya village, August 1, 2010. EMBEDDED MOBILIZATION 449 efforts at ideological mobilization, had met with vocal resistance from these same local communities. Following vka activists in their forays into new villages revealed they had learned from these past experiences. In their initial interactions with villagers, service workers consistently downplayed any polarizing ideological language in favor of portraying themselves as neutral service providers. The head doctor in the Jashpur unit of the vka explained this strategy in some detail: They [Christian missionaries in the area] worked out a great medium to gain new followers: through service. See education and healthcare are the first thing a person needs for their life to be lived. If anyone is sick, and if someone treats them and makes them better, then they will forever have their ear. If someone is illiterate, and someone teaches them to read, then again they will always listen to their teacher and have their sympathy. That is why we work in health and education. See I told you about the dispensary [in Jashpur] and the medical vans [which go around villages to dispense basic medical care]. Then we have hostels and schools in the village. Through these mediums you can go to anyone at anytime. After we established ourselves, then we can begin work with Dharma Jagran [spiritual awakening], and get people to come sit with us, to learn bhajans [prayers/religious chants] to come to mandir [temple], etc. Slowly, slowly, this bhav [feeling] is brought into tribals.51 While vka activists were initially reticent about their strategies and motivations, such frank admissions became increasingly common after repeated meetings. One example came from a trained doctor who has worked for the organization across India for many years: In the beginning, when we go to hold medical camps, people are suspicious as to what the motive is, because those who came to "help" before [referring here to colonial administrators] would come because they were really interested in their labor. Also they have heard that we don't approve of practices like the drinking of liquor or the worshipping of trees. There is distrust at first, and will continue until you actually go to them and work amongst them. And at first we don't tell them anything about how to pray We just talk to them about health and education. Then gradually over time we begin talking to them about Hindu culture.52 The depoliticized discourse of "service" thus provides a key entry point through which the Sangh can access communities distrustful of their political agendas.53 These activities articulate a different vision of Hindu 51 Author interview with Brigendra Singh, vka doctor, Jashpur, April 1, 2008. 52 Autnor interview with Pankaj Bhatia, Jashpur, April 2,2008. The vka medical unit traveling in a van with basic medical supplies runs these medical "camps." They are held for one day in each village, with a team doing a rotation among ten to fifty villages; they offer free check-ups and basic medical care. 53 Baviskar 2005, 5108, makes a similar point in a study of the growth of Hindu nationalism in Madhya Pradesh. 450 WORLD POLITICS nationalism to poor voters: one of a movement interested not solely in preserving upper-caste traditions but also in attending to the basic welfare of the poor. In addition to enabling bjp partners to gain access to poor voters, service activities have generated several organizational benefits for the party, a point local party leaders repeatedly stressed. First, service chapters have yielded a small, but influential cadre of Adivasi politicians for the party from among early generations of recruits, many of whom have now assumed positions of prominence in the bjp state unit: Look, they [the vka and rss] are our party's main backbone, the soul of the party. They are the tree, we are the fruit. They may work directly, indirectly, but ultimately we are a seed that had been sown by them ... our current st [Scheduled Tribe/Adivasi] leadership have come from the grassroots, come up through the vka system and are trained by them. Our st leadership includes Vishnu Dev Sai (Member of Parliament, Raigarh), Nand Kumar Sai (mp Sarguja), Ganesh Ram Bhagat (mp Jashpur) all affiliated with the vka.54 The fact that these politicians come up through a tradition of grassroots activism with the Sangh makes them far more popular with their local constituencies than earlier token candidates handpicked because of their perceived quiescence by threatened upper castes in Delhi and Raipur (the state capital). Local bjp leaders in the state have noted how the work of their non-electoral partners, specifically the VKA, has helped the party embed itself within previously hostile communities. Nand Kumar Sai, the first Adivasi state-level president for the bjp (in Madhya Pradesh) and current MP from Sarguja parliamentary constituency in Chhattisgarh, observed: See the Sangh does do a lot of work—you are right—then there is the Vanvasi Kalyan Ashram which has worked in health and education, where no one else has gone ... all of this work in different spheres has really helped the bjp establish a presence with these sections. The party has managed to consolidate through this extensive organization.55 In addition to this core of political leaders picked from among early recruits, the activities of the vka have provided the bjp with a cadre of youthful party workers from among its later recruits. In the thirty villages visited for this study, the average age of the bjp's primary karyakar-tas [party workers] was 27.5, over ten years younger than the Congress 54 Author interview with Satyanand Rathiya, BJP state cabinet minister (independent charge), Raipur, March 31, 2008. 55 Author interview with Nand Kumar Sai, New Delhi, April 28, 2008. o id EMBEDDED MOBILIZATION 451 average of 38 years. Moreover the majority of the party's workers have prior affiliations with the vka network, mostly as students. The salience of these workers in door-to-door campaigning is particularly heightened in the context of Indian elections in remote Adivasi-dominated areas, where exposure to conventional mass media is typically low. Even rival Congress Party candidates acknowledge the profound impact these cadres have had: "All the work for the bjp is done by the Sangh organizations only! The vka takes on the name of development of tribals but really they are all working for the bjp full time as well. The day they stop working in this region, the bjp will be finished."56 However, given the weakness of Congress's own network in the state (indeed, the party's state president confessed that there was "no grassroots organization to speak of") and its lack of nonelectoral movement | partners, the party has been able to do little to counter the organiza- jj tional punch the Sangh provides its rival.57 b While the benefits of having stocks of personnel from which to !• build candidate lists and party worker rosters are invaluable, perhaps the most valuable assets are the teacher-activists who remain outside p, the bjp's formal apparatus: 03 5 Our teachers are not respected at first, because people are not sure about what 2! they are trying to do in the village. But over time they gain high social status in 6 the village. vka teachers are working with children, and even today teachers in § tribal society command much respect, which is helpful for the Sangh.58 a) J! Once service activists successfully insert themselves into the fabric of g village communities, they serve as a key preexisting cadre for the bjp to o tap into during elections. The impact of this network increased com- mensurably in the 1990s, as the number of villages in the vka's ambit grew, spreading the word about the organization and establishing a reputation for its activists. Many of these activists were at first reticent about admitting to their role at election time. However, over time they began to talk more openly about the valuable data they were able to gather and share with the local bjp unit. vka nirakshak (inspector) Dhananjai Kumar revealed u o ™ o -o <- 57 56 Author interview with Sarwan Bhagat, former Congress mla candidate, Jashpur, March 28, 2008. 57 Author interview with Chandradas Mahant, Raipur, April 5, 2008. Indeed as Manor 2005 has noted, the Bjp's formal organizational strength is often overstated. In Chhattisgarh it has been the | g networks the party has inherited from its movement partners, not the party's own apparatus that have 5 8 provided it with a decisive edge. £. P: 58 Author interview with Birbal, vka worker, vka hostel, Raipur, March 22, 2008. £ ° E o o ^; it CTl "D o "D o — 5 q. R & 452 WORLD POLITICS to me an excellent example of such information.59 His assignment, in addition to monitoring the performance of chapters within a thirty-village radius, included maintaining a detailed log of the caste composition of households in this area to help the bjp make local decisions about candidate selection and canvassing strategies. In addition to organizational information, the depoliticized discourse surrounding service work allows Sangh activists to appear nonpartisan and thereby enhance their leverage at election time. This self-professed removal from the "dirty" sphere of party politics is particularly useful for activists tasked with such delicate jobs as spreading rumors about rival competitors from the Congress: In addition to doing social work, the vka has managed to do other work for the bjp such as telling people the truth about how [Congress Party leader and former chief minister] Ajit Jogi is trying to pass [himself] off as a real Adivasi while he was not one, as well as reassuring them that bjp candidates are friends of the vka and will not engage in corruption.60 Ajit Jogi, a former Congress chief minister in the state, was a victim of this efficient rumor-spreading network. Jogi was accused of falsely certifying himself as a member of a Scheduled Tribe (Adivasi), a move that allowed him to contest seats reserved for ST candidates, as well as gain popularity among the state's important Adivasi electorates. As the informant quoted above noted, VKA activists worked hard prior to the 2003 elections to spread this rumor and met with great success in casting doubt on the validity of Jogi's candidacy precisely because the rumor came from outside the formal political arena.61 The most delicate task of all is providing political "advice" around election time, while appearing nonpartisan. In the buildup to elections vka workers are assigned to mobilize support for the bjp in the villages surrounding those in which they reside. In these villages, activists organize baithaks (meetings) at which they attempt to exert influence without explicitly lobbying residents: See whenever I go for political work, if I'm holding a baithak, I start the meeting the same way that I do for normal meetings (not around elections), with a song 59 Author interview with Dhanajai Kumar, April 24, 2008. 60 Author interview with Birbal, vka worker, vka hostel, Raipur, March 22, 2008. 61 The controversy made headlines following the 2001 ruling by the National Scheduled Tribe and Scheduled Caste Commission, which was headed by BJP leader, Dilip Singh Bhuria, that Jogi was in fact a member of the Satnami (Dalit/Scheduled Caste) community, and not an Adivasi. The Bilaspur High Court overturned this decision in 2006, enabling Jogi to serve once more as a state legislator from an Adivasi constituency. EMBEDDED MOBILIZATION 453 or some light discussion to set a good mood. And I tell people at the meeting that all I want to do is talk about what is best for our Adivasi community. Then I tell them that I know about the personality, about the behavior of the candidates. I tell them look I am not from any political party, and so I am not in this to win an election—I just want us to choose the best person we can, that I can tell you who a good person is.62 Many activists work in this vein, trying to maintain an air of nonpar-tisanship by advocating particular candidates on the basis of their ability to serve Adivasis, rather than their adherence to Hindu nationalist views: So for example villagers might ask us, which candidate do you like? Now first we always ask them—well who are you thinking of supporting? If they are thinking of the bjp, then we don't say much—why put more pressure \_dabav\} But if they say Congress, then we ask—have you considered all the issues when making this decision? Then we samjhao [explain] why they should reconsider. If after hearing us they say we still want to vote for Congress, we never put pressure \zabar-dasti nahin karte]—we don't want them to think we are working for a party.63 The idea is to avoid being aggressive in trying to influence voters but still sway them toward the bjp candidates {apne taraph jhukana hai). Clearly this requires striking a delicate balance, and it remains to be seen whether maintaining a position of neutrality while advocating candidates of the bjp can remain a viable strategy indefinitely. However it is unlikely that such a task could have been accomplished for any length of time were it not managed by activists perceived to be removed from party politics and dedicated to providing much-needed services to villagers. Close study of the Sangh Parivar in Chhattisgarh thus reveals the myriad mechanisms through which social service provision might impact electoral politics. While the material benefits such work offers are clearly valued by underserved and impoverished communities, the influence of a service approach is not reducible to notions of material exchange. Instead, my research indicates that service providers were also able to recruit candidates and party workers for the bjp and collect useful demographic information about local communities. At the same time, these providers capitalized on their "apolitical" status as service providers to become extremely influential in swaying local electoral decisions themselves. While qualitative research is well suited to uncover the workings of these providers, it is less useful for testing the effect of 62 Author interview with G. Pradhan, Beharana village, July 24, 2010. 63 Author interview with Madhru Ram Bhagat, Gutukiya, July 22, 2010. 454 world politics their work on voting more broadly. Are the various mechanisms outlined above influential enough to affect the bjp's pattern of success with lower castes? Are participants in the activities of these organizations really more likely to vote for the bjp at the polls? Are nonparticipants influenced enough by this tactic to ever vote for the bjp? To assess these questions, I designed and conducted an individual voter survey in four districts across Chhattisgarh in 2008. The analysis is presented in the next section. V. Does Service Win Votes? Testing the Electoral Impact of Provision The survey for this study was administered among 360 Dalit and Adi-vasi voters in 24 villages in the districts of Bastar, Jashpurnagar (Jash-pur), Mahasamund, and Raigarh.64 While the villages and the respondents were randomly selected within districts, the districts themselves were selected for their marked variance on the key independent variable of interest: the organizational strength of the Sangh. Thus, although Hindu nationalists have a formidable service apparatus across Chhattisgarh, such strength is not uniformly distributed. The Sangh does not provide district-level data for its projects across India, but I was able to obtain figures for the number of service projects being run by the state unit of Vanvasi Kalyan Ashram and Seva Bharati (the two main service organizations in the state) in each of the four districts in which the survey was conducted. I then divided this number of projects by the district's combined Dalit and Adivasi population to get a per capita measure of density of provision (reported in Table 1). Examining voting preferences across districts within a single state affords a great deal of comparative control, allowing the effects of Sangh service organizations on bjp support levels to be more clearly revealed. All four districts are comparable in their levels of party competition, land area, literacy rates, and ethnolinguistic profile; all also have sizable Dalit and Adivasi communities and are primarily rural with less than 10 percent of the population living in urban areas. Further, none of the four districts had experienced significant sectarian rioting along either religious or caste lines.65 Of the Dalit and Adivasi respondents 64 For details on the survey instrument, sampling procedure, and other material, see the online appendix for this article, at http://pantheon.yale.edu/~tt282. 65 Indeed, the only district that experienced any rioting was Raigarh, one of the districts in which the BJP was not successful. Thus the idea that the party could use rioting to garner votes appears not to have empirical support in the specific case of Dalit and Adivasi voters. Riot data are from Varshney-Wilkinson Dataset on Hindu-Muslim Violence in India, 1950-1995, 2004. EMBEDDED MOBILIZATION 455 Table 1 Voting Patterns across Districts in Chhattisgarh Number of Index of Sangh Service Sangh Strength Percentage District Projects per Capita Voting for the BJP Jashpur 545 1.07 59.55 Bastar 512 .619 67.78 Raigarh 20 .096 28.89 Mahasamund 26 .059 19.77 Sources: vka headquarters, Jashpur, Seva Bharati office, Raipur, Chhattisgarh, and author-conducted survey, Chhattisgarh (2008). surveyed, half were in districts with "high organizational strength" (Bastar and Jashpur) and half were in districts with "low organizational strength" (Mahasamund and Raigarh). My argument anticipates that districts with denser networks of service organizations should have higher levels of support for the party than comparable districts without them.66 The results of the survey presented in Table 1 do indeed demonstrate startling differentials in bjp support between the high-strength districts and the low-strength districts. Such aggregate differences are telling: the neighboring districts of Jashpur and Raigarh have few demographic and political differences yet display markedly disparate lower-caste voting patterns.67 To try to determine whether service wins votes, however, it is necessary to examine data at the individual level. The argument presented here would expect participants in the activities of service organizations to be more likely than nonparticipants to support the bjp. Second, service's impact need not be restricted to participants, as qualitative field-work uncovered several ways in which the vka might help change the opinions of those members of the wider community who hold service 66 Congress politicians such as state leader Ajit Jogi have cried foul over the party's reversal in Bastar in the 2003 election, arguing that the result was due to improper electoral procedures. Jogi maintains that the bjp, which was in power in New Delhi at the time, forcibly deployed the Central Reserve Police Force (crpf) to the region under the guise of "keeping elections fair," when in reality they stuffed ballot boxes for the bjp in these areas. Kumar 2004 notes, however, that such allegations flew on both sides of the aisle, as some of his sources from Bastar informed him that the 1998 elections had not been fair, as "it was alleged that many booths were non-functional and district officials had gifted those seats to the then Congress Chief Minister Digvijay Singh on a platter." As a result, it is difficult to state conclusively that election irregularities hurt one party's performance systematically more than others. 67 The VKAs share in the four districts are 398 projects in Jashpur, 317 projects in Bastar, 13 projects in Raigarh, and 8 projects in Mahasamund. Sewa Bharati runs the remaining projects in each district. 456 world politics organizations in high regard. Whether opinions of service wings are consequential enough to influence electoral decisions is therefore an additional matter for empirical verification. To test these arguments I therefore present the following falsifiable hypotheses: HI. Lower-caste voters who participate in the activities of Sangh social service organizations will be more likely to vote for the bjp than those who are not participants. H2. Among nonparticipants, those who have a good opinion of service activities will be more likely to support the bjp than those who do not. The analysis worked with a tripartite conceptualization of voters: those who identified as members of Sangh organizations, those who identified as participants in their activities, and those who were neither of the above (henceforth, nonparticipants). Someone receiving treatment from one of the vka's mobile dispensaries during a "medical camp" in a given village would be a participant but not necessarily a formal member who attends weekly meetings or even works for the vka. Having disaggregated respondents into these three categories, we can then examine how effective and extensive each type of linkage was in terms of helping the bjp at the polls. Turning to the survey results again, we see that the probability of voting for the bjp varied significantly between nonparticipants and the other two categories. The percentages in each column in Table 2 provide an initial measure of the efficacy of each linkage type, while the number of respondents reported in parentheses indicates its extensive-ness. Not surprisingly, membership in a Sangh social service organization guarantees support for the bjp at the polls, but the results for non-member participants are striking. Nine in ten voters who have simply participated in activities organized by Sangh affiliates voted for the bjp in the last state elections. While more rigorous testing is obviously required, the results seem to provide preliminary evidence for the efficacy of the Sangh strategy in affecting electoral outcomes. Table 2 also reveals that while the efficacy of participation remained robust across different types of districts, the extent of this participation differed substantially. In Jashpur and Bastar, the districts in which the service network was strong, Sangh organizations have encompassed more than four times as many respondents as in the low-strength districts of Mahasmudh and Raigarh. In these latter regions, organizational presence appears not to have been pervasive enough to yield a significant electoral base of voters. CT1 O 0) £ ? E o o ^; £: CTl max values) Change in Predicted Probability of Variable Name Voting for BJP (Percentage Points) Participation (0-1) 55.82 SanghOpinion (0-3) 57.10 Age (18-85) -66.47 aPredicted values for model 4 were obtained using simulations conducted with the Clarify software package while keeping other variables at their mean values. variable that retained an impact in the fixed-effects specification (model 4) was Religiosity, although it did not appear to confound the impact of the participation, opinion, and age variables. In order to ascertain the specific substantive effects of participation in and opinion of Sangh service activities, I estimated the change in the predicted probability of voting for the bjp as each causal variable varied from its observed minimum to its observed maximum value (using the results of model 4). It is clear from Table 4 that participation in Sangh service activities has a profound impact on lower-caste voting decisions. Dalit and Adivasi participants were 55.82 percentage points more likely to vote for the bjp than their nonparticipant counterparts. Age also had a large negative effect, suggesting that the bjp has made significant progress in recruiting new voters within these electorates, with a stunning 65 percent of voters between the ages of eighteen and twenty-nine voting for the party. Taken in conjunction with the earlier evidence on youth cadres being cultivated by service networks, the data suggest a strong generational effect of this strategy. Isolating the substantive effect of the variable SanghOpinion for the sample of nonparticipant voters is of particular importance for assessing whether the influence of Sangh service organizations extends beyond those they directly envelop into their fold. Figure 3 charts the impact of opinion of Sangh work on the likelihood of supporting the bjp for nonparticipants specifically, once again using predicted probabilities from model 4. The figure shows that, on average, a nonparticipant with the most positive opinion of Sangh social services is twice as likely to support the bjp as one with the least favorable opinions, holding other variables at their mean values. This impact of the opinion of Sangh service wings on nonparticipant vote choices is significant, because it indicates that activists can exert a considerable political influence even embedded mobilization 461 Figure 3 Effect of Opinion of Sangh Work on Voting (Nonparticipants) when respondents do not benefit directly from their activities. Such evidence resonates with the qualitative evidence presented earlier in suggesting that a service approach resists narrow classification as one based purely on individualized material exchange. VI. Robustness Checks: Exact Matching and Vote Switching This section presents some tests undertaken to verify the robustness of the key finding from the previous section of participation effects on lower-caste support for the bjp. The first problem to address was the nonrandom assignment of "treatment," that is, the participation in Sangh service activities. The problem with such nonrandom assignment is well known: it is impossible to assume that the treatment is independent of observed and unobserved baseline variables. If participants systematically differ from nonparticipants, assessing the influence of participation on voting poses a host of difficulties for causal inference. For a number of logistical and ethical reasons, not all research, including the present study, is amenable to experimental techniques of randomizing assignment. In such situations, matching analysis has become an increasingly popular option for ensuring sufficient overlap 462 WORLD POLITICS between the treatment and the control units along certain key observed covariates, in order to better isolate the effect of the treatment itself. In essence, matching can be used to prune the existing sample to create pools of treatment and control observations that most closely resemble each other. Table 5 shows the difference in means for the treatment group (participants) and the control group (nonpartici-pants). While most of the differences were not significantly different from zero, there were still underlying differences in the two populations. To account for these differences, I used the Matchlt software library to match treatment and control samples exactly across all the four potential confounders included in the prior models.72 Naturally, the strict criteria of exact matches ensure that the distance between the two groupings is minimized (indeed, it is zero).73 Since matching analysis requires complete data sets, I first used the Amelia software package in R to impute the missing values in the local survey data.74 As a result, I was working with five imputed data sets and consequently with five matched data sets.75 With exact matching, a simple difference of means test between the treatment and control group is sufficient to assess the impact of the treatment variable. The second column of Table 6 shows that for each of the five matched data sets, participants were at least 43.5 percentage points more likely to support the BJP than exactly matched nonpar-ticipants. The substantive results are thus robust to the exact matching preprocessing. While matching is clearly no panacea and cannot address differences on unobserved criteria, the fact that we can obtain consistent results after accounting for such theoretically important potential confounders lends confidence to the prior findings. However, matching cannot address potential issues of endogeneity between voting and participation in service wing activities. Specifically, could participants have some prior preference for the bjp that draws them into the Sangh's service network? Could participants mostly be longtime supporters of the bjp, something not picked up with a cross-sectional survey? Qualitative evidence from this study goes against such a supposition in several respects. First, aggregate electoral evi- 72 Ho et al. 2009. 73 The main drawback of this technique is that its stringency often produces too few matches to proceed with further analysis. However, since the control variables were all discrete measures, sufficient matches were generated for subsequent analysis. 74 See Honaker, King, and Blackwell 2009; King, Joseph, and Scheve 2001. 75 The number of units in each of the matched data sets were 110 (39 treatment units), 120 (39), 115 (38), 114 (39), and 124 (42). For an example of multiple imputation followed by matching, see Simmons and Hopkins 2005. EMBEDDED MOBILIZATION 463 Table 5 Comparing Participants and Nonparticipants (Pre-Match) p-value Variable Name Participants Nonparticipants Means Difference (two-tailed) Communalism 3.08 3.11 -.03 .63 Income 2.25 2.47 -.22 .28 Religiosity 4.94 4.49 .45 .01 CasteCom .415 .502 -.087 .252 Table 6 Effects of Participation on Voting and on Vote Switching (PosT-match)a (1998-2003) Difference ofMeans for Difference ofMeans for Matched BJPvoted (2003) SwitchedtoBJP (1998-2003) Date Set Number (Participants-Nonparticipants) (Participants- Nonparticipants) 1 .470** .282*** 2 479*** .317*** 3 .482*** .315*** 4 .476*** .288*** 5 .435*** .261*** "significant at the .01 level, ***significant at the .001 level. Participants and nonparticipants matched exactly on Religiosity, Communalism, Castelnßuence, and Income variables using Matchlt software package. Matching conducted on five imputed data sets (imputation conducted using Amelia software package). dence suggests the bjp had never previously enjoyed levels of support equivalent to what the survey indicates it has now attained in districts where these networks are strong. Indeed, the vka's aggregate expansion (early 1990s) appears to have preceded the bjp's aggregate gains in the state (late 1990s and early 2000s). Further, while both bjp and Sangh leaders outlined the many ways in which the party benefited from the efforts of its affiliates, none articulated any mechanisms by which the bjp funnels its supporters into service organizations; nor did I see evidence of such activity. However, to probe this issue further at the individual level, the survey asked respondents how they voted in the prior, 1998 state assembly elections. A variable SwitchedtoBJP was then created from these responses, coded 1 for those voters who shifted to supporting the bjp between 1998 and 2003. The third column of Table 6 shows that participants were 28.2 percent more likely than exactly matched nonparticipants 464 world politics to have switched to voting for the bjp. This additional result provides fairly robust evidence that service participants were not simply longtime supporters of the bjp. VII. Conclusion The purpose of this study was to construct an empirically informed conceptualization of service provision as an electoral strategy based on the first individual-level analysis of how nonstate service provision can win votes in a developing democracy. Original survey data revealed that service was successful in mobilizing support for a traditionally elite-based religious nationalist party even among its historically least likely supporters. Lower-caste participants in service activities were far more likely to support the bjp than nonparticipants, even when accounting for important potential confounding factors and prior voting preferences. These individual-level effects also aggregated into electoral shifts, with the party performing dramatically better among poor voters in districts with dense service networks than in highly comparable districts without them. The survey also revealed that service's impact extended beyond direct beneficiaries, with opinions of service activities exerting considerable influence on the preferences of nonparticipant voters. As noted at the outset, these empirical findings have important theoretical implications for how we conceptualize service provision as an electoral tactic. The fact that services affect more than those who benefit from them indicates the inadequacy of viewing the tactic as one based purely on individualized material exchange. While clearly the services offered by the Sangh have great material appeal in areas lacking basic public infrastructure, this study outlines how provision yields political dividends through a variety of additional mechanisms. Qualitative fieldwork revealed how service functions to provide the bjp with popular political candidates, youthful party workers, and influential teacher-activists, all of whom help the party electorally. Furthermore, the painstaking nature of provision, coupled with the ideological commitments of providers, meant that a service approach did not follow the quid pro quo protocol of machine politics.76 Taken together, these insights point to the limitations of conceptualizing service as merely another form of clientelism and suggest instead that politically informed provision should be autonomously conceptualized as an electoral tactic, rather than forced to conform to existing categories. 76 Medina and Stokes 2002; Diaz-Cayeros, Magaloni, and Estevez 2010. EMBEDDED MOBILIZATION 465 The local working of service also draws attention to dimensions increasingly neglected in analyses of the relationships between parties and ordinary voters. Most current scholarship has focused on the episodic, individualized, and transactional nature of party-voter relations, particularly around elections. Service reveals, by contrast, the importance of quotidian social interaction. Its remarkable ability in enabling the bjp to win over communities historically hostile to the party is dependent at least in part on how provision both necessitates everyday contact and provides a depoliticized framework through which activists can interact with ordinary voters. Service thus embeds parties within communities in ways that selective conditional cash transfers or public sector jobs and contracts do not. Indeed, a major yet largely neglected asset of service, particularly in arenas dominated by clientelist politics, is that it frees parties to continue channeling conventional patronage to core supporters.77 At the same time, however, providing social services entails massive organizational costs. This particular mix of costs and benefits provides some insight into the perceptible affinity between a service-based approach and conservative religious political formations. Most of these formations, not only in South Asia but also in the Middle East and Southeast Asia, typically begin with small pockets of urban middle-class support but then need to expand demographically to remain effective.78 However, these groups also typically have extensive organizational resources and personnel with deep ideological commitments, making them perfect candidates for implementing a service-based approach to expansion. Of course, a service strategy cannot succeed in all political contexts. In other work, I examine the conditions affecting this approach's electoral success more systematically, but two factors are worth brief mention.79 The appeal of services in Chhattisgarh was clearly enhanced by the inadequacies of local public infrastructure. But equally important was the fact that poor communities had not been actively mobilized electorally along either class or caste lines. In sharp contrast to sectarian riots, the tactic for which Hindu nationalists have received greatest attention, service as an electoral strategy is less likely to succeed in (or succeed in producing) a sharply polarized environment. Had the poor been horizontally politicized in central India as lower castes or indeed as working classes or marginal peasants, an already difficult task may have proved insurmountable. 77 I explore this argument in more detail in a separate study, Thachil 2011. 78 Wickham 2002; Hamayotsu 2007. 79 Thachil 2009. 466 world politics Finally, it bears reiterating that the ideological commitments of non-electoral activists proved critical in motivating them to furnish services for little private material gain and in lengthening their time horizons beyond electoral cycles. This long-term outlook in turn ensured that provision would be sustained long enough to generate the multiple political benefits articulated in this article. Thus, while electoral compulsions and prior failures at expansion clearly fueled the proliferation of service provision by Hindu nationalists in the 1990s, its success was ironically predicated on being implemented by nonelectoral activists. This interplay of activists and politicians, itself a hallmark of so many religious nationalist movements, thus questions the separation of ideological goals from electoral interests and instead highlights their potential symbiosis in forging unlikely coalitions of rich and poor. References Achen, Christopher. 2005. "Let's Put Garbage-Can Regressions and Garbage-Can Probits Where They Belong." Conflict Management and Peace Science 22, no. 4 (January): 327-39. Baviskar, Amita. 2005. "Adivasi Encounters with Hindu Nationalism in mp." Economic and Political Weekly, November 26: 5105-111. Berthet, Samuel. 2008. "Chhattisgarh: Tribals, OBCs, Reformist Movements and Mainstream Politics." In Christophe Jaffrelot, ed., Rise ofPlebians: The Changing Face of the Indian Legislative Assemblies. Routledge: New Delhi. Bonoli, Giuliani, and Martin Powell. 2004. Social Democratic Party Policies in Contemporary Europe. London: Routledge. Brass, Paul. 1993. "The Rise of the bjp and the Future of Party Politics in Uttar Pradesh." In Harold Gould and Sumit Ganguly, eds., India Votes: Alliance Politics and the Ninth and Tenth General Elections. Boulder, Colo.: Westview Press. Calvo, Ernesto, and Victoria Murillo. 2004. "Who Delivers? Partisan Clients in the Argentine Electoral Market. "American Journal of Political Science 48, no. 4: 742-57. Cammett, Melani, and Sukriti Issar. 2010. "Bricks and Mortar Clientelism: Sectarianism and the Logics of Welfare Allocation in Lebanon." World Politics 62, no. 3 (July): 381-421. Chandra, Kanchan. 2004. Why Ethnic Parties Succeed. New York: Cambridge University Press. Chhibber, Pradeep. 1997. "Who Voted for the Bharatiya Janata Party." British Journal of Political Science 27, no. 4 (October): 631-39. -. 1999. Democracy without Associations. Ann Arbor: University of Michigan Press. Clark, Janine A. 2004. Islam, Charity and Activism. Bloomington: Indiana University Press. Cox, Gary, and Mathew D. McCubbins. 1986. "Electoral Politics as a Redistribu-tive Game." Journal of Politics 48 (May): 370-89. o 00 id E o o ^; it CTl Core Voter Model: Some Evidence from Mexico." Working Paper, April. Diaz-Cayeros, Alberto, Beatriz Magaloni, and Frederico Estevez. 2010. "The 2 Politics of Public Spending: The Logic of Vote-Buying." In "Strategies of j§ Vote-Buying: Poverty, Democracy, and Social Transfers in Mexico." Book > manuscript. jf Dixit, Avinash, and John Londregan. 1996. "The Determinants of Success of % Special Interests in Redistributive Politics." Journal of Politics 58 (November): | 1132-55. Froerer, Peggy. 2006. "Emphasizing Others: The Emergence of Hindu National-5 ism in a Central Indian Tribal Community."/oz/r?za/ ofthe Royal Anthropological £> Institute 12 (March): 39-59. n Goldstone, Jack. 2003. States, Parties, and Social Movements. Cambridge: Cam- 3 bridge University Press. % Graham, Bruce. 1990. Hindu Nationalism and Indian Politics: The Origins and Devel- S opment of the Bharatiya fana Sangh. Cambridge: Cambridge University Press. s Hansen, Thomas. 1999. The Saffron Wave. Princeton: Princeton University Press. \ Herring, Ronald. 1988. "Stealing Congress Thunder: The Rise to Power of a 3 Communist Movement in South India." In Kay Lawson and Peter Merkl, eds., *j When Parties Fail. Princeton: Princeton University Press. Ho, Daniel, Kosuke Imai, Gary King, and Elizabeth Stuart. 2009. "Matchit: 5 Nonparametric Preprocessing for Parametric Causal Inference, "journal of Sta- 1, tisticalStudies 10, no. 2. At www.jstatsoft.org. ^ Honaker, James, Gary King, and Matthew Blackwell. 2009. "Amelia II: A Pro- ° gram for Missing Data." Software documentation, February 19. s Uriah, Kancha. 1996. Why I Am Not a Hindu. Calcutta: Samya. * Jaffrelot, Christophe. 1993. The Hindu Nationalist Movement in India. New York: § Columbia University Press. s -. 2005. "Hindu Nationalism and the Social Welfare Strategy: Seva Bharati as an Education Agency." In Christophe Jaffrelot, ed., The Sangh Parivar: A Reader. New Delhi: Oxford University Press: 211-24. King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve. 2001. "Analyz-|r ing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation." American Political Science Review 95, no. 1 (March ): 49-69. Kitschelt, Herbert. 2000. "Linkages between Citizens and Politicians in Democratic Politics." Comparative Political Studies 33, no. 6-7 (August-September): 845-79. Kitschelt, Herbert, and Steven Wilkinson, eds. 2007. Patrons, Clients, and Policies: Patterns of Democratic Accountability and Political Competition. New York: Cambridge University Press. Kochanek, Stanley A. 1976. "Mrs. Gandhis Pyramid: The New Congress." In H. C. Hart, ed., Indira Gandhis India: A Political System Reappraised. Boulder, U) o u o — o x! <- E ^ S g Colo.: Westview Press 5 2 Kohli, Arul. 1987. The State and Poverty in India. Cambridge: Cambridge University Press. 468 WORLD POLITICS Kothari, Rajni. 1964. "The Congress 'System' in India." Asian Survey 4, no. 12 (December): 1161-73. Kremer, Michael, Nazmul Chaudhury, F. Halsey Rogers, Karthik Muralidharan, and Jeffrey Hammer. 2005. "Teacher Absence in India. "Journal ofthe European Economic Association 3, no. 2-3: 658-67. Kumar, Sunil. 2004. "Fruits of Megalomania." Seminar, issue 534. Lokniti Team. 2009. National Election Study. New Delhi. Magaloni, Beatriz. 2006. Voting for Autocracy: Hegemonic Party Survival and Its Demise in Mexico. New York: Cambridge University Press. Mair, Peter, Wolfgang Miiller, and Fritz Plasser. 2004. Political Parties and Electoral Change: Party Responses to Electoral Markets. London: Sage. Manor, James. 2005. "In Part a Myth: The bjp's Organizational Strength." In Katherine Adeney and Lawrence Saez, eds., Coalition Politics and Hindu Nationalism. London: Routledge. Mathur, Shubh. 2008. Everyday Hindu Nationalism. New Delhi: Three Essays Collective. McAdam, Doug, and Sidney Tarrow. 2010. "Ballots and Barricades: On the Reciprocal Relationship between Elections and Social Movements." Perspectives on Politics 8, no. 2 (June ): 529-42. Medina, Luis Fernando, and Susan Stokes. 2002. "Clientelism as Political Monopoly." Paper prepared for the annual meeting of the American Political Science Association, Boston, August 29-September 1. Nitchter, Simeon. 2008."Vote Buying or Turnout Buying? Machine Politics and the Secret Ballot." American Political Science Review 102, no. 1: 19-31. Palshikar, Suhas. 2003. "Politics of India's Middle Classes." In Imtiaz Ahmad and Helmut Reifield, eds., Middle Class Values in India and Western Europe. New Delhi: Social Science Press. Pandey, Saumya. 2008. "Communalism: Narratives in Chhattisgarh." Economic and Political Weekly, January 12. Putnam, Robert. 2000. Bowling Alone. New York: Simon and Schuster. Ray, James Lee. 2003. "Explaining Interstate Conflict and War: What Should Be Controlled For?" Conflict Management and Peace Science 20 (September): 1-31. -. 2005. "Constructing Multivariate Analyses (of Dangerous Dyads)." Conflict Management and Peace Science 22, no. 4 (January): 277-92. Roberts, Kenneth. 2002. "Social Inequalities without Class Cleavages in Latin America's Neoliberal Era." Studies in Comparative International Development 36, no. 2 (Winter): 3-33. Robinson, James, and Thierry Verdier. 2003. "The Political Economy of Clientelism." Manuscript, University of California, Berkeley. Scott, James. 1972. "Patron-Client Politics and Political Change in Southeast Asia." American Political Science Review 66, no. 1 (March): 91-113. Sharma, Shashank. 2004. "Adivasiyon par Kisne Kiya Jadu?" Deshbandhu,]armaxy 20. Simmons, Beth, and Daniel Hopkins. 2005. "The Constraining Power of International Treaties: Theory and Methods. "American Political Science Review 99, no. 4 (November): 623-31. Srinivas, M. N. 1956. "A Note on Sanskritization and Westernization." Far Eastern Quarterly 15, no. 4 (August): 481-96. EMBEDDED MOBILIZATION 469 Stokes, Susan. 2005. "Perverse Accountability: A Formal Model of Machine Politics with Evidence from Argentina." American Political Science Review 99, no. 3: 315-25. -. 2007. "Political Clientelism." In Carles Boix and Susan Stokes, eds., Handbook of Comparative Politics. Oxford: Oxford University Press: 604-28. Sundar, Nandini. 2004. "rss: Teaching to Hate." Economic and Political Weekly, April 17. -. 2006. "Adivasi vs. Vanvasi." In Satish Saberwal and Mushirul Hasan, eds., Assertive Religious Identities: India and Europe. Delhi: Manohar: 357-90. Thachil, Tariq. 2009. 'The Saffron Wave Meets the Silent Revolution: Why the Poor Vote for Hindu Nationalism in India." Ph.D. diss., Cornell University. -. 2011. "Mixing Ideology and Organization: How Hindu Nationalists Constructed an Unlikely Coalition in India." Working paper. Tomz, Michael, Jason Wittenberg, and Gary King. 2003. "Clarify: Interpreting and Presenting Statistical Results. "Journal of Statistical Software 8 (January). Van Cott, Donna. 2005. From Movements to Parties in Latin America. Cambridge: Cambridge University Press. -. 2009. Radical Democracy in the Andes. Cambridge: Cambridge University Press. Varshney, Ashutosh. 2002. Ethnic Conflict and Civic Life: Hindus and Muslims in India. New Haven: Yale University Press. Varshney-Wilkinson Dataset on Hindu-Muslim Violence in India, 1950-1995. 2004. Version 2, Ashutosh Varshney and Steven Wilkinson, October 8. Wickham, Carrie Rosefsky. 2002. Mobilizing Islam. New York: Columbia University Press. Weiner, Myron. 1967. Party Building in a New Nation: The Indian National Congress. Chicago: Chicago University Press. Yadav, Yogendra. 2004. "bjp's Consolidation Will Be Hard To Break in Chhat-tisgarh." The Hindu, March 21. Zaidi, Askari. 2003. "Battlefield of Chhattisgarh." Hindustan Times, November 23. Zubaida, Sami.1993. Islam, the People and the State. London: I. B. Tauris.