Ethnic Parties and Public Spending

Abstract

Social scientists largely see ethnic politics as inhibiting public goods provisioning within developing democracies. Such parties are thought to uniformly rely on distributing excludable benefits to co-ethnics, rather than on providing public goods to all. We argue that ethnic parties can vary substantially in how they mobilize support and behave in office. Much of this variation depends on the breadth of the identity they activate. Although “narrow” ethnic parties do indeed entrench patronage politics, the rise of more “encompassing” ethnic parties can actually improve levels of voter autonomy, expand the effective size of winning coalitions, and increase spending on broadly available public goods. We develop and test this argument with evidence from the Indian states, including a nationally representative survey of 20,000 Indian voters and a panel data set of 15 major states over four decades.

Keywords

corruption and patronage democratization and regime change political parties political economy politics of growth/development

Introduction

Does the rise of ethnic politics inhibit the provision of public goods within developing democracies? Social scientists have demonstrated how shared preferences, visions of a “linked fate” (Dawson, 1995), and dense social networks enable coethnics to build mutual trust (Fershtman & Gneezy, 2001) and to punish free riding within their communities (Besley, Coate, & Loury, 1993). By contrast, scholars anticipate that ethnically heterogeneous societies will vehemently disagree over which public goods the government should provide (e.g., Alesina, Baqir, & Easterly, 1999; Lieberman & McClendon, 2013). Yet in examining the role of ethnicity within democratic politics, scholars increasingly distinguish between the social fact of ethnic diversity and the electoral salience of ethnic divisions (Posner, 2006). Although ethnic diversity exists in many countries, interethnic disputes over the spoils of government are most likely to arise in developing countries. In such contexts, information constraints and weakly defined policy platforms are argued to induce parties to rely on visible and stable ethnic markers to help solve the standard commitment problems that plague discretionary electoral exchanges (Chandra, 2004). In such situations ethnicity facilitates politicians trading excludable benefits for votes, thereby dampening the prospects for more inclusive distributive strategies.

Although valuable, these studies overlook an important dimension of ethnic politics that can mediate its deleterious effects on public goods provisioning. Although all ethnic parties originate within the confines of ascriptive categories that have the potential to facilitate patronage, the communities that they mobilize frequently lack access to political power. This type of political exclusion is especially high in “ranked” ethnic systems (Horowitz, 1985, p. 22), where ethnic divisions largely overlap with those of class. In such contexts, ethnic parties often challenge existing social hierarchies by directly mobilizing marginalized ethnic groups. Such efforts disrupt the prior encapsulation of these voters by traditional political elites, thereby helping to increase the former’s political autonomy.

This “autonomy effect” of ethnic parties has received comparatively little attention in the broader literature on ethnic parties, but importantly shapes how these new political actors distribute public goods. In this article, we argue that the increasing autonomy of marginalized voters mobilized by ethnic parties can make it difficult for politicians to continue relying on a small number of elites to deliver an electoral victory. Through their “de-encapsulation” of voters, ethnic parties expand the effective size of local winning coalitions (the percentage of votes required to stay in power). This expansion raises the incentives for politicians to spend on public goods, given the increasing returns to scale of such a strategy (Kitschelt & Wilkinson, 2007; Stokes, Dunning, Nazareno, & Brusco, 2013).

When do ethnic parties follow conventional expectations in entrenching patronage systems, and when do they exert the salutary expansion of winning coalitions via an autonomy effect? We argue the key lies in the inclusiveness of the ethnic identity mobilized by a given party (Miguel, 2004; Singh, 2011). When ethnic parties appeal to voters along more encompassing identities shared by a higher proportion of all voters, they both challenge voter encapsulation by traditional elites and look to mobilize an expansive ethnic core. The emergence of such “encompassing ethnic parties” (EEPs) boosts the size of local winning coalitions and consequently spending on public goods and services. By contrast, parties mobilizing narrower ethnic bases face fewer incentives to follow inclusive distributive strategies with increasing returns to scale. Instead, such “narrow ethnic parties” (NEPs) find it more appealing to replicate prior patronage politics within their restricted ethnic base. Supporters of NEPs thus find themselves re-encapsulated within pyramids of patronage, albeit ones now headed by co-ethnics.

We test our argument with evidence from India. In addition to its intrinsic importance as the world’s most populous democracy, India has served as a paradigmatic case for illustrating how ethnic parties undermine the prospect of widespread public goods provisioning. Over the past three decades, India’s party system has radically transformed from an arena dominated by a “catch-all party” (the Indian National Congress) into a “patronage democracy” structured along ethnic lines (Chandra, 2004; Chhibber, 1999).

Yet it is apparent to even a casual observer of Indian politics that the proliferation of ethnic parties across Indian states has not been accompanied by uniform patterns of public spending. To make sense of such variation, we first examine how ethnic parties mobilized along ascriptive identities of varying breadth, largely due to differences in the social processes through which these parties emerged. We follow this historical discussion by leveraging contemporary survey data of a nationally representative sample of 20,000 voters to verify our conceptual distinction between EEPs and NEPs. We also establish significant differences in how these parties affect voter autonomy—the key mechanism underpinning our theory. Finally, we analyze state-level panel data on public goods expenditures over a 40-year period to examine if subnational variation in EEP and NEP success corresponds with variation in expenditures on public goods. We find the success of EEPs correlates with increased spending on broadly accessible social services and decreased levels of poverty. The rise of NEPs instead corresponds with the continuation of a patronage-based system.

One key advantage of our subnational research design is that it helps control for the potentially confounding effect of key political institutions like electoral rules and levels of fiscal decentralization. Earlier studies show how these institutions affect both the structure of ethnic politics (Posner, 2006), and the nature of electoral strategies and public policies (Carey & Shugart, 1995; Iverson & Soskice, 2006). Our analysis demonstrates how the inclusiveness of ethnic parties can vary greatly within a common institutional setting, highlighting how the sociological origins of such parties condition their impact on voter autonomy and public spending.

Ethnic Parties, Voter Autonomy, and Public Spending

In this section, we present our argument for how the political mobilization of ethnic groups affects public spending. For the purposes of our analysis, the term ethnic group refers to any subset of people for whom ascriptive attributes are necessary for membership. Following scholarly consensus, we assume that individuals may “nominally” hold multiple ethnic identities by simple virtue of being born into particular descent-based categories. In India, voters are simultaneously arrayed across language groups, religion, caste groups, and subcastes (jati). We do not privilege any one of these categories as “ethnic” over the others (Chandra, 2012). Rather, our analysis examines how social and historical forces determine which of these nominal categories become activated, and how this differential activation of categories affects public spending.

Political parties play a major role in activating particular ethnic identities (Chandra, 2004; Chhibber, 1999). Following Chandra (2004), we view ethnic parties on the basis of their political appeals, as those which “champion the particular interests of one ethnic category,” while excluding others. Not all ethnic parties within a polity mobilize voters along the same dimension of ethnicity. In complex societies like India’s, we may see some ethnic parties mobilize on the basis of religion, others on the basis of caste, and still others on the basis of language. However, each of these parties practice exclusion: caste-based ethnic parties appeal to one caste group, excluding other castes. Language-based ethnic parties appeal to one language group, excluding other language groups. Thus, the latter are ethnic parties even if a given language group encompasses several castes. Although all ethnic parties practice exclusion, the key argument we develop in this section is that the extent of that exclusion can vary in ways that existing literature often ignores.

The standard view of ethnic parties anticipates their success to uniformly dampen the prospects for inclusive public spending. Drawing on the framework developed by Bueno de Mesquita, Smith, Siverson, and Morrow (2003), Chhibber and Nooruddin (2004) argue the incentives for public goods provisioning are a function of the size of the minimum winning coalition (the number of votes required to win an election) relative to the size of the “selectorate” (those who participate in the selection of leaders). In plurality systems, a proliferation of parties typically decreases the number of votes any single party needs to cross a winning threshold. This reduction enables parties to appeal to narrower electorates, including specific ethnic communities. Based on evidence from India, Chhibber and Nooruddin argue parties in more fragmented states face heightened incentives to provide excludable benefits to co-ethnics, instead of public goods to the entire electorate.

The argument that party fragmentation shrinks the official size of winning coalitions is correct. At the same time, the proliferation of political parties does not equally attenuate the effective size of winning coalitions, even within plurality systems. Bueno de Mesquita et al. (2003) clarify this distinction, noting the size of a winning coalition is not determined simply by the institutional features of a party system but by its social characteristics. For example, elite-dominated systems often produce what the authors term “small, restrictive winning coalitions” even under broad procedural franchise. They cite the example of elections in Kenya during the 1980s, when this coalition often comprised of a single elite individual—A chief electoral officer capable of completely distorting the actual outcome in favor of the ruling party.

In other instances, economic elites can dominate restrictive winning coalitions. Baland and Robinson (2007) describe small winning coalitions of landlords in Chile, who “systematically controlled rural voting until the late 1950s” by “giving their laborers the ballot of the party of their choice” (p. 130). Hierarchical organizations such as labor unions also often deliver en bloc votes, thereby rendering the effective size of a winning coalition much smaller than the official vote share of a victorious party (Roberts, 2002). In each case, the individual autonomy of voters crucially influences whether the official winning coalition reflects its effective, or actual, size.

The emergence of ranked ethnic systems, like caste in India, can similarly distort the autonomy of voters, sustain incentives for vertical incorporation, and produce “small restrictive winning coalitions.” In such contexts, political parties can target benefits to key “high-rank” elites, whose position at the top of vertical social hierarchies enables them to control and deliver inter-ethnic electoral majorities. In these systems, new ethnic parties often emerge by appealing to “low-rank” ethnic groups historically excluded from the top echelons of power. These parties thus disrupt the prior encapsulation of non-elite voters by traditional elites, and hold the potential for increasing levels of voter autonomy within the political system. Thus, if the entry of ethnic parties leads to a reduction in the official size of this coalition in a plurality system, it can potentially expand the effective winning coalition size of a polity via this autonomy effect.

Whether ethnic parties follow through on this potential depends on how they target their supporters, which in turn hinges on whether they mobilize encompassing or narrow ethnic identities. When ethnic parties mobilize encompassing identities that most non-elite voters share, they challenge prior elitist political networks and look to mobilize a broad ethnic core. This combination ensures the rise of such “encompassing ethnic parties (EEPs)” expands the effective size of the local minimum winning coalitions. To mobilize this broad base, EEPs find it more efficient to spend on broadly accessible public goods than narrowly targeted patronage.

However, ethnic parties can also emerge along identity categories shared by far narrower slices of the electorate. The smaller target constituencies of such “narrow ethnic parties” (NEPs) incentivize their leaders to simply replicate narrow patronage-based strategies within their restricted ethnic cores. Such replication means non-elite voters move from being encapsulated by patronage pyramids headed by small winning coalitions of traditional social elites to being encapsulated by those headed by low-rank co-ethnic elites. In this scenario, the entry of ethnic parties does little to improve voter autonomy, expand the effective size of winning coalitions, or increase incentives for parties to engage in inclusive public spending.

Thus, although all ethnic parties practice exclusionary politics, the extent of that exclusion varies greatly between EEPs and NEPs. How can we empirically operationalize such differences? Ethnic parties vary in their ethnic exclusiveness along two key dimensions. The first is the relative size of a party’s target ethnic group. For example, the United Malays National Organization (UNMO) in Malaysia, an ethnic party targeting Malays, is more encompassing than the Malaysian Chinese Association, which targets ethnic Chinese. This is simply because Malays constitute more than 50% of the Malaysian electorate, whereas ethnic Chinese constitute only 25%. For reasons we explore in the next section, some parties mobilize ethnic identities that demarcate narrow slices of the electorate while others mobilize identities shared by many or even most voters. Second, ethnic parties differ in the extent to which they rely on the support of their target ethnic group. An ethnic party that draws 80% of its support from its core ethnic base is less encompassing than one that draws on its core ethnic base for only 40% of its vote share.¹

To measure ethnic exclusion in a precise and replicable way, we draw on Huber’s (2012) ethnic concentration index (ECI) for individual ethnic parties, a measure that taps into both attributes discussed above. As we subsequently discuss, ECI can be calculated for any dimension of ethnicity in any country, and subsequently to distinguish EEPs and NEPs in that polity. Yet which dimension of ethnicity is most relevant for distinguishing NEPs and EEPs will require contextual justification and case-specific knowledge. India’s dizzying array of ethnic identities makes it an unusually challenging case to analyze. We anticipate our methodological approach will actually be more straightforward to implement in less ethnically complex societies with fewer ethnic identity dimensions to adjudicate between.

The Rise of EEPs and NEPs in India

Why do some ethnic parties mobilize more inclusive ethnic identities than others? Our subnational research design allows us to control for important explanations of such variation, such as the nature of electoral institutions. We focus instead on the sociological reasons for why EEPs or NEPs emerged in particular Indian states, detailing two different pathways of ethnic party formation. NEPs were self-consciously engineered by opportunistic elites as vehicles for entering into electoral politics, and concentrated their efforts among the narrow ethnic groups from which their particular leaders hailed. By contrast, EEPs grew out of prior non-electoral collective action, and thus did not have active choice over which ethnic identity their supporters were mobilized along. Because broader social forces determined the ethnic identities that EEPs mobilized, these identities were much broader than what a purely Downsian actor might have selected. Once formed, both EEPs and NEPs have been limited in their ability to shift the ethnic identity along which they mobilize support. Consequently, initial differences between EEPs and NEPs have persisted, with lasting implications for the politics they pursue.

Low-Caste Movements and the Rise of EEPs

The Indian National Congress dominated the first two decades following independence from British colonial rule in 1947 (Corbridge & Harriss, 2000). The Congress benefitted from its status as the party of independence, but equally from its establishment of an effective electoral machine. An influential scholarship has documented how the Congress “system” relied on upper-caste intermediaries such as local landlords to exercise their clout over local communities at election time (Kothari, 1964; Weiner, 1967). As Jaffrelot (2003) notes, the Congress incorporated “upper-caste notables,” who delivered the votes of non-elite voters within their sphere of social and economic influence (p. 66). Such “vertical mobilization” of non-elites by traditional notables ensured Indian democracy resembled Bueno de Mesquita et al.’s (2003) “restrictive winning coalition,” with the Congress focused on channeling patronage to local elites rather than providing public goods to all citizens (see Rudolph & Rudolph, 1967). As the Congress faced little electoral opposition, non-elite voters had few alternatives to this elite-dominated vertical incorporation.

However, following Nehru’s death in 1964, the Congress’ dominance began to crumble. This shift was partly produced by opportunistic politicians within the Congress who felt thwarted by the narrow apex of the party’s “big tent,” and sensed better opportunities outside its fold. The Congress leadership attempted to pass anti-defection laws to limit the autonomy of its regional leaders, which only added fuel to the fire (Nikolenyi, 2009). Yet it would be a mistake to read the process of fragmentation as one produced solely by the frustrations and opportunism of individual elites. Most early attempts to form new parties in India took place while the Congress still enjoyed comfortable majorities in state legislatures. As Ziegfeld (2012) notes, incentives against forming such parties were high, as they had little chance of being part of a governing coalition.

What proved crucial in enabling these new parties to find niches of support were the shifts wrought by broader social mobilizations challenging caste hierarchies. The “gradual erosion of authority” of dominant castes over marginal ones was first noted in ethnographies of Indian villages (Lerche, 1999; Mendelsohn, 2002). In southern India, initial challenges to traditional authority came early, some even pre-dating national independence. In Kerala, Sree Narayana Guru organized agitations among members of the Ezhava (toddy tapper) caste community demanding entry into temples previously reserved for upper castes. Backward caste associations in the neighboring state of Tamil Nadu, such as the Nadar Mahajana Sangam (NMS), fought social discrimination, and provided communal facilities to their membership and alms to poorer members. However, most of these early “horizontal mobilizations” (Rudolph & Rudolph, 1967) focused on advancing a “politics of dignity” rather than winning elections. Consequently, the Congress’ electoral machine remained relatively unchallenged.

Yet many of these social movements produced a legacy of mass mobilization that regional politicians would later draw on to forge their electoral bases (Jaffrelot, 2003). Social reformers sought to unite marginalized citizens by activating ethnic identities shared by all, such as region, and de-emphasizing narrower ethnic identities such as caste and subcaste (jati) that divided non-elite communities. In Kerala, the “untouchable caste” Pulaya associations allied with “backward caste” Ezhavas to successfully pressure local upper castes under the banner of a shared Malayali linguistic identity (Franke & Chasin, 1989).² In Tamil Nadu, E. V. Ramaswamy (a.k.a. Periyar) quit the Indian National Congress in 1925, complaining that the party served only Brahmin elites. Periyar devoted himself to directing a “self-respect movement” that attempted to unite non-Brahmin castes under the shared regional identity of “Adi Dravidians” (original inhabitants of the South). This ethnoregional identity was defined in opposition to Brahmin elites, who were depicted as descendants of “Aryan” outsiders hailing from the Hindi-speaking north.

It is no coincidence that the strongest early electoral challenges to Congress hegemony came from parties in areas where such anti-elite regionalism was most pronounced. Ziegfeld (2012) notes that “despite incentives to the contrary, these parties remained resolutely regionalist in their outlook, and regional in their geographic scope, quite likely because ideological considerations overrode political interest” (pp. 17-18). Periyar and his followers explicitly drew on the social reform organizations they had built when founding an independent political party (the Dravida Kazhagham, later the Dravida Munnetra Kazhagham [DMK]) in Tamil Nadu. Similarly, ethnic parties in Andhra Pradesh depended heavily on the legacy of prior regional mobilizations, including the dramatic fast unto death of Potti Sreeramulu demanding a state for Telugu speakers. Under the leadership of Nandamuri Taraka Rama Rao, the Telugu Desam Party (TDP) looked to unite a large swathe of non-elite voters under a Telugu ethnolinguistic identity. Such unity was achieved in explicit opposition to the elites in control of the Congress party. For example, when elected to office, the TDP abolished the posts of village officers traditionally held by upper castes within the Congress and increased quotas for non-elites in local government institutions (see Vakil, 1990; Vaugier-Chatterjee, 2009). This phenomenon of regional political mobilization repeated itself throughout India in subsequent years. For example, the Asom Gana Parishad in the eastern state of Assam emerged out of efforts to mobilize an Assamese-speaking majority, partly to protest the dominance of a Bengali-speaking minority in the state’s administrative services.

Party Fragmentation and the Rise of NEPs

Thus, EEPs in India emerged in arenas where prior non-electoral collective action sought to unify non-elite communities under a shared ethnic banner against traditional elites. A different type of ethnic party was formed after the erosion of the Congress’ national dominance that EEPs helped bring about. Forming ethnic parties in the post-Congress era “did not require ideological motivation since [these] parties were no longer a losing proposition” (Ziegfeld, 2012, pp. 17-18). Thus, opportunistic leaders from lower caste communities had an opportunity to form relatively narrow vehicles through which to fulfill their personal ambitions.³

These office-seeking elites had little interest in activating broad ethnic identities like region or language shared by an electorally inefficient supermajority of voters in a state. Instead, they looked to concentrate on mobilizing members of the narrow caste groupings and even narrower endogamous subcastes (jatis) from which they came. Take the case of the Bahujan Samaj Party (BSP) in the northern state of Uttar Pradesh. A prototypical NEP, the BSP was founded in 1984 by Kanshi Ram, a member of one of India’s Dalit (former “untouchable” castes). Ram self-consciously sought to focus his initial appeals among his Dalit co-ethnics. Even within this aggregate caste category, the BSP has specifically concentrated its personnel and support among members of the Chamar jati, from which Ram himself hailed.

Other NEPs follow a similar pattern. The rise of the Samajwadi Party (Uttar Pradesh) and Rashtriya Janata Dal (Bihar) in the 1990s was disproportionately fueled by support from the Yadav jati from which their respective leaderships came (Robin, 2009). The Indian National Lok Dal in the northern state of Haryana is a family-run party, which depended on the Jat farming jati. Although primarily concentrated in the north, NEPs are also present in other parts of the country. The Janata Dal (Secular) in the southern state of Karnataka is dominated by members from the Vokkaliga jati of the party’s most influential leader, Deve Gowda.

Thus, the ethnic bases of NEPs were informed by narrow electoral considerations, not broader social change sought by non-electoral collective action (Varshney, 2012).⁴ The restricted ethnic bases of NEPs prompted their leaders to replicate the Congress’ patronage strategies within their core constituencies (Chandra, 2004; Jaffrelot & Kumar, 2009). Note that our theory does not rely on an intrinsic difference between EEP and NEP leaders. We do not posit that EEP leaders were altruistic while NEP leaders were greedy. Instead, instrumental leaders of both types of parties were affected by the conditions of their emergence. NEP leaders were free to activate the specific ethnicities serving their electoral needs, whereas EEP leaders found themselves “subsumed by” prior ethnic mobilizations that “could not be compromised or accommodated” (Rudolph & Rudolph, 1960, pp. 12-13).

The sticky legacies of these origins help explain the seemingly inefficient electoral strategies of ethnic parties. Downsian frameworks anticipate that NEPs and EEPs should converge in mobilizing ethnic electorates whose numbers most closely approximate a winning plurality. NEPs should seek to expand their ethnic bases, while EEPs should try to narrow them from the inclusive supermajorities constructed by their non-electoral movement antecedents. Yet nationally representative election survey data from 1996, 1999, and 2004 reveal no evidence of a systematic broadening of the ethnic base of NEP parties or narrowing of EEPs.⁵ Such difficulties of reinvention are also underscored by the experiences of prominent NEPs, which have struggled to transcend the ethnic divides they helped politicize.⁶

Ethnic Parties and Voter Autonomy

The preceding section provided an overview of the differential rise of EEPs and NEPs within India’s postcolonial era. In this section, we detail our protocol for empirically identifying EEPs and NEPs, and classify ethnic parties in India according to this typology. We then use survey data to test voter-level implications of differences in how these parties win support.

Classifying EEPs and NEPs

For our empirical analysis, we needed to classify NEPs and EEPs within the Indian party system. As a first step, we made a list of all the parties who placed either first or second within-state assembly elections in the 15 major Indian states from 1967 to 2007.⁷ Our intuition here was that to influence public spending patterns, the party had to be either in office or the primary opposition party in the state legislature.⁸ This exercise generated a list of 31 parties (see Table S1.1 in the supplement for the full list). Our next step was to isolate those parties that could be credibly classified as “ethnic.” To do so, we first consulted a range of secondary sources on each of the 31 parties within our analysis. As mentioned earlier, we define ethnic parties on the basis of their appeals to the particular interests of one ethnic category to the exclusion of others. Thirteen of the 31 parties in our sample fulfilled this criterion.⁹

Our third and final step was distinguishing NEPs from EEPs within the pool of ethnic parties. Conceptually, we specified this distinction in terms of the breadth of the ethnic electorate each type of party seeks to mobilize. To measure this, we adapt an ECI developed by Huber (2012). The index is calculated according to a formula:

E C I_{k} = \sqrt{1 / 2 \sum_{g = 1}^{G} {(V_{k}^{g} - S_{g})}^{2}}

where $V_{k}^{g}$ is the proportion of party k’s support from group g. S_g is the size of a given group, and G is the number of groups produced by a given dimension of ethnicity.

Note that while our distinction between ethnic and non-ethnic parties is based on our adjudication of their electoral appeals (which requires contextual knowledge), our distinction between EEPs and NEPs is based on ECI, which can be calculated for any dimension of ethnicity in any country. For example, if we define ECI on the basis of language, the index varies as a function of the percentage of a party’s vote coming from speakers of its target language group and the size of that language group relative to speakers of all G languages. These ECI scores can then be used to classify parties as encompassing or narrow.

In our analysis of India, we focus on ECI scores calculated along the ethnic identity of jati or subcaste. We chose jati because it is the narrowest ethnic identity available to Indian political parties, and hence captures the most fine-grained variation in support bases of ethnic parties. High jati-based ECI scores indicated a party targets supporters based on the narrowest ethnic category available on the Indian menu. Low jati-based scores indicated the party drew support from a wide range of jatis, as it mobilized a broader ethnic identity encompassing multiple jatis. As these latter actors still mobilized on the basis of an ascriptive identity such as language or region, they are still ethnic parties according to our framework. Our focus on jatis is validated by recent analyses finding it to be the most appropriate dimension for capturing variation in the ethnic bases of Indian political parties (Huber & Suryanarayan, 2014).

To calculate individual indices for each party, we utilized electoral data from a representative nation survey of more than 20,000 Indian voters conducted in 2004.¹⁰ This survey, the National Election Study (NES), is widely seen as India’s most reliable post-poll electoral survey. We classified parties as EEPs or NEPs based on a cutoff value of 0.1 in this jati-wise ECI score. We used this particular threshold because we felt it provided a classification that was both empirically grounded in the survey data but also conceptually grounded in historical differences between ethnic parties in India. The five EEP parties this threshold yields are broadly understood to mobilize along broad ethnolinguistic and ethnoreligious dimensions. The eight NEPs meanwhile are more commonly understood to have specific jati strongholds.

However, to ensure our results were not driven by a particular classificatory threshold, we performed our analysis using a purely computational threshold (the sample mean ECI). Each of our key findings was robust to using this alternative categorization (Tables S2.4 and S3.4 in the supplement). The results were also robust to using a continuous variable based on the ECI values instead of a dichotomous classification (Table S2.5). The use of computational or continuous measures enables this analysis to be easily exported to other country contexts.

Table 1 lists each of these 13 ethnic parties, along with their primary ethnic base, and ECI scores. The mean index value for NEPs was .195, nearly 3 times that for EEPs (.074).¹¹ The second column of the table notes that the ethnic bases of the eight NEPs tend to be focused within narrow subcastes communities (jatis). For example, the survey found more than 50% of the support for the BSP in Uttar Pradesh came from the Chamar subcaste, a formerly “untouchable” jati that comprised roughly 12% of the population. Similarly, 35% of the vote for the Janata Dal (S) in Karnataka comes from the Vokkaliga jati, which makes up only 14% of the state’s population. By contrast, EEPs, like the DMK in the southern state of Tamil Nadu or TDP in Andhra Pradesh, tend to target larger linguistic or regional ethnic communities. Accordingly, no single jati accounted for more than 11% of these parties’ overall vote shares. For additional information on each party, see section S.2 of the supplement.

Table 1.

Classifying Encompassing and Narrow Ethnic Parties.

	Ethnic dimension	States	ECI
Encompassing ethnic parties
1. Asom Gana Parishad	Linguistic (Assamese)	Assam	.081
2. All India Dravida Munnetra Kazhagham	Linguistic (Tamil speakers)	Tamil Nadu	.063
3. Dravida Munnetra Kazhagham	Linguistic (Tamil speakers)	Tamil Nadu	.078
4. Shiromani Akali Dal	Linguistic/religious (Punjabi Sikh)	Punjab	.093
5. Telugu Desam	Linguistic (Telugu speakers)	Andhra Pradesh	.056
Mean ECI for EEPs			.074
Narrow ethnic parties
6. Bahujan Samaj Party	Dalits (untouchable castes, specifically Chamar jati)	Haryana, Madhya Pradesh, Punjab, UP, Uttarakhand, Chhattisgarh	.304
7. Haryana Vikas Party (active until 2004)	Dominant caste Jat jati Hindus	Haryana	.209
8. Indian National Lok Dal	Dominant caste Jat jati Hindus	Haryana	.208
9. Janata Dal Secular	Dominant caste Vokkaliga jati	Karnataka	.193
10. Janata Dal (United)	OBC castes (specifically Kurmi and Koeri jati)	Bihar	.115
11. Rashtriya Janata Dal	OBC castes (specifically Yadav jati) and Muslims	Bihar	.132
12. Samajwadi Party	OBC castes (specifically Yadav jati) and Muslims	UP, Uttarakhand	.236
13. Shiv Sena	Maratha and Kunbi jati communities	Maharashtra	.161
Mean ECI for NEPs			.195

ECI calculated using Voter Survey Data from 2004 National Election Study (N = 20,390). ECI = ethnic concentration index; EEP = encompassing ethnic parties; UP = Uttar Pradesh; OBC = Other Backward Classes; NEP = narrow ethnic parties.

Testing the Mechanism: Voter Autonomy

We have argued NEPs and EEPs dislodge non-elite voters from hierarchical networks dominated by traditional elites. Yet given their narrow bases, NEPs largely re-encapsulate their supporters within narrow patronage networks dominated by co-ethnic elites. Consequently, we anticipate NEP supporters are less likely to enjoy electoral autonomy and be more likely to express views reflecting their captivity within jati-based vote blocs. By contrast, the expansive cores of EEPs incentivize them to eschew hierarchical strategies in favor of inclusive public spending tactics with increasing returns of scale. Consequently, we anticipate supporters of EEPs to enjoy a higher autonomy at the polls.

We test these expectations using several measures of autonomy derived from the 2004 NES survey. First, we analyzed how strongly respondents agreed with the view that they should vote in the same manner as other members of their jati.¹² Our theory anticipates supporters of NEPs should be more likely to agree with such an obligation, whereas EEP supporters should not.

The key explanatory variables for this analysis are binary measures identifying partisan supporters of EEPs and NEPs.¹³ We prefer this partisanship measure because it sets a higher threshold for identifying core supporters of a party than support in a single election. However, each of our results is robust to using vote choice in the 2004 election as the basis for inclusion (see supplement Table S2.3).

In addition to the key variable of partisanship, we also include a range of control variables that might influence the voter autonomy from caste networks. First we include an indicator of the cultural salience of caste for a voter, which might arguably influence the political salience he or she attaches to this identity. Our measure of ethnic parochialism is support for banning inter-caste marriage. We also include measures of income and education, which we anticipate will positively correlate with voter autonomy. Finally, we control for age, gender, religion, whether a respondent was from marginalized lower caste or tribal communities, and whether they lived in an urban location. All else equal, we expect older, male, urban, and upper-caste voters to enjoy greater political autonomy.

Table 2 presents the results of our analysis. We use an ordered logistic model for our ordinal dependent variable, and all models include robust standard errors clustered by parliamentary constituencies. Our primary specifications confirm our major expectations. EEP supporters reported significantly less compulsion to vote in the same manner as their co-ethnics, whereas NEP supporters were significantly more likely to feel such compulsions. These results were statistically significant at the .001 level and were robust to narrowing the sample to include only non-elite “backward caste” voters (Model 2) as well as the inclusion of constituency-level fixed effects (Model 3). The final specification is particularly exacting and finds that these systematic differences between NEP and EEP partisans persist even when comparing respondents within the same electoral constituency.

Table 2.

Ethnic Parties and Voter Autonomy.

	Dependent variable: “One should vote with one’s caste community”
Variables	(1) All voters	(2) “Backward caste” voters	(3) All voters (fixed effects)
Encompassing ethnic parties partisans	−0.381***(0.106)	−0.556***(0.145)	−0.203**(0.070)
Narrow ethnic parties partisans	0.360***(0.102)	0.495***(0.156)	0.299***(0.095)
Ethnic parochialism	0.081*** (0.010)	0.092***(0.013)	0.072***(0.010)
Lower caste or tribal	0.155***(0.046)		0.074**(0.038)
Income	−0.025*(0.013)	−0.000(0.027)	−0.076***(0.013)
Education	−0.171***(0.013)	−0.195***(0.018)	−0.159***(0.012)
Age	−0.004***(0.001)	−0.005***(0.001)	−0.002**(0.001)
Female	0.140***(0.035)	0.082*(0.048)	0.215**(0.033)
Hindu	−0.079(0.067)	0.022(0.080)	−0.184(0.062)
Urban	−0.077*(0.041)	0.006(0.064)	−0.139***(0.048)
Latent variable Cut Point 1	−1.089(0.125)	−0.937(0.159)	−1.625(0.123)
Cut Point 2	−0.488(0.124)	−0.297(0.161)	−0.944(0.119)
Cut Point 3	0.036(0.131)	0.222(0.157)	−0.347(0.116)
Cut Point 4	0.727(0.124)	0.952(0.156)	0.439(0.117)
Log likelihood	−29,886.026	−11,798.159	−28,064.746
Number of constituencies	392	389	392
Observations	20,204	7,931	20,204

Ordered logistic regressions. Dependent variable is an ordered variable measuring the degree to which respondents agree with the statement that they must vote with their caste. Reported latent variable cut points were all tested and found to be statistically significant. Robust standard errors clustered by parliamentary constituency in parentheses.

p < .1, **p < .05, ***p < .01.

These effects are also substantively meaningful. Figure 1 shows the simulated probability of EEP partisans strongly disagreeing with the need to vote with their caste group is 42%, compared with just 27% for NEP partisans. By contrast, the probability of NEP partisans strongly agreeing with this statement is nearly 31%, compared with only 18% for EEP partisans. The results for control variables were largely along expected lines, with higher voter autonomy positively correlating with education and negatively correlating with ethnic parochialism, being lower caste, or female.

Figure 1.

Vote with caste community?

The differences between EEP and NEP partisans reported in Table 2 are also robust to two alternative measures of the dependent variable. The first is a binary variable identifying respondents who said they made their electoral decision independent of any external influence. The second identifies respondents who said their caste community leader was the biggest influence on their vote choice. The detailed results are presented in the supplement (Table S2.2), but the main findings of this analysis are that NEP partisans were significantly more likely to identify co-ethnic caste leaders as the single biggest influence on their vote choice. EEP partisans were meanwhile significantly more likely to report voting independent of any external influence. Figure 2 shows the probability of EEP partisans voting independently is roughly 15 percentage points higher than for NEP partisans at any given level of education. In fact, the difference in autonomy between NEP and EEP partisans is considerable, roughly the equivalent to going from having no education to a high school diploma.¹⁴ These analyses are also robust to alternative categorizations of ethnic parties.¹⁵

Figure 2.

Voter autonomy, ethnic partisanship, and education.

Thus, we find considerable evidence of the autonomy mechanism underpinning our theory, across a range of measures and model specifications. EEP partisans exhibit greater individual autonomy, whereas NEP partisans are encapsulated by jati elites and compelled to vote with their jati.

Ethnic Parties and Public Spending

So far, we have distinguished between two types of ethnic parties (EEPs and NEPs) and have demonstrated that voters affiliated with EEPs are more autonomous than voters affiliated with NEPs. The next step of our theory anticipates the downstream consequences of this difference on public spending. We expect that because EEPs expand the effective size of local winning coalitions via their salutary impact on voter autonomy, their success should correlate with higher expenditures on broadly accessible public goods and less spending on more narrowly targeted goods and services. Conversely, the success of NEPs does not improve voter autonomy and therefore should be associated with less spending on broadly accessible goods and services and more spending on narrowly targeted goods and services.

To test this argument, we analyze patterns of public spending and poverty in 15 major states, which together comprise approximately 95% of India’s population in a time-series cross-section framework for the period 1960 through 2007.¹⁶ We choose the state as the unit of analysis because state governments in India exercise significant control over the dispensation of social and economic services that most immediately affect local communities. We analyze how patterns in public spending in these states vary in relation to the nature of party systems, and in particular to the success of socially EEPs or NEPs. The remainder of this section describes our measures and methods of analysis in greater detail.

Dependent Variables: Patterns of Public Spending and Poverty

To analyze how EEP dominance affects spending on inclusive public goods and services, we look at two categories of the budget. First, we follow earlier studies in using the amount that states budget for development projects as a general proxy for overall commitment to the provision of public goods and services (Chhibber & Nooruddin, 2004; Saez & Sinha, 2009). While we acknowledge that development spending can be used clientelistically in some circumstances (Hicken, 2011), development expenditures have at least higher ex-ante potential to boost public goods provisioning than spending on explicitly non-developmental budget categories, like interest payments on the debt or administrative services, because unlike the latter they are explicitly intended to improve overall levels of economic welfare. This measure provides a good initial test of our theory in that privileging the appropriate budget category is a necessary, if not sufficient, step for improving the supply of public goods.

Second, we analyze the effects of EEP dominance on administrative expenditures, a subcategory of non-development expenditures. Administrative expenditures are those related to public administration (courts, police, tax collection, etc.), plus public pensions and debt payments. A large percentage of spending in this category is comprised of payments of salaries and pensions. As public sector jobs are frequently allocated as rewards to political supporters rather than on the basis of merit, administrative expenditures are frequently used as a source of political patronage (Chhibber & Nooruddin, 2004).

If our argument is correct, then the increased dominance of EEPs should correlate with more development spending as a whole; but since administrative services are designed to benefit a narrower segment of society (namely public servants), we would expect EEP dominance to correlate with lower levels of administrative spending. NEPs may spend more on administrative services, however, because such expenditures are more in line with the standard practice of winning votes through narrowly targeted patronage.

Previous studies have examined development and administrative spending solely in terms of the percentage of total budgetary expenditures. In our analysis, we also look at levels of spending in per capita terms. Looking at levels of spending enables us to address the concern that differences in the percentage-based variables are reflective of parties simply shifting money between budget categories, as opposed to an increased commitment to improving aggregate infrastructure and services. We can be more confident of the commitment of EEPs to boosting overall levels of provisioning if, in addition to devoting a higher percentage of the budget to development spending, EEPs also spend more money per capita on development-related services.

One final concern is that development spending—both in percentage and per capita terms—may not tell us about whether benefits reach intended recipients. Perhaps politicians narrowly capture development spending for themselves, with few benefits for the common voter. Thus, any observed correlation between EEP success and development spending might perversely measure increased elite capture of resources. To address this possibility, we specify a model that directly tests the impact of ethnic party success on the well-being of disadvantaged voters. Specifically, we measure the impact of EEP (and NEP) electoral success on the head count ratio, or the percentage of the population below the poverty line. The head count ratio is an appropriate outcome to look at because it provides the broadest measure of economic opportunities, and because in contrast to literacy, infant mortality, life expectancy, or other general measures of social and human development, it is available annually throughout our panel.¹⁷

Explanatory Variables

The two primary independent variables in our analysis are the percentage of seats in the Vidhan Sabhas (state assemblies) won by members of NEPs and EEPs respectively. Following the decline of the Indian National Congress, ethnic parties became dominant in more than half of India’s 15 major states. EEPs primarily succeeded in four states—Andhra Pradesh, Assam, Punjab, and Tamil Nadu (see Table 1). Caste-based NEPs primarily succeeded in an additional five states (Bihar, Haryana, Karnataka, Maharashtra, and Uttar Pradesh).¹⁸ Other states experienced the continued dominance of the Congress party and/or the rise of the Bharatiya Janata Party (e.g., Madhya Pradesh, Gujarat, and Rajasthan) and the rise of left parties (e.g., Kerala and West Bengal). The baseline category against which our EEP and NEP variables are measured is therefore primarily composed of these three national political formations.¹⁹

Modeling Strategy

In our analysis, we look at changes in spending on development and administrative services. We expect that in states where the Congress is superseded by EEPs, state budgets will devote more resources toward development spending and less toward administrative spending. Conversely, we expect that the rise of NEPs will not herald any such increasing commitment to development-related goods and services from the era of Congress dominance but may be associated instead with even higher levels of spending on administrative expenditures.

Note that our argument predicts how the rise of inclusive and exclusive parties affects patterns of spending within states. To model these dynamics, we analyze the relationship between public spending and party systems for the period 1967 through 2007 using ordinary least squares regressions of the form,

y_{s t} = β_{0} + μ_{1} e_{s t} + μ_{2} n_{s t} + {ξχ}_{s t} + α_{s} + δ_{t} + ε_{s t},

where $y_{s t}$ is a measure of public spending, $β_{0}$ is a constant, $e_{s t}$ is the percentage of seats held by EEPs in the state assembly in a given state-year, $n_{s t}$ is the percentage of seats held NEPs, and $χ_{s t}$ are a series of controls.²⁰ The models also include fixed effects for states $(α_{s})$ and years $(δ_{t})$ , and were estimated with robust standard errors clustered by state.

Fixed state effects enable us to model the within-state effects of EEP dominance on patterns of public spending and to ensure that the revealed impact of our explanatory variable is not a reflection of systematic unobserved heterogeneity between states. Consequently, fixed state effects enable us to rule out any counter-explanations for patterns in public spending that pertain to stable underlying differences between Indian states, such as regional culture, informal institutions existing in one or a handful of states, or the presence of a regionalist social movement. Fixed year effects help control for unobserved trends that might influence the outcomes in the analysis, or economic and institutional shocks common to all states in a given year. The inclusion of fixed year effects also helps us to address potential endogeneity concerns. Although theoretically, it is difficult to imagine public spending creating ethnic parties, it is possible that the rise of ethnic parties and patterns of public spending are endogenous to some third variable that might be trending through the panels. The inclusion of fixed year effects helps address this concern by controlling for unobserved trends and exogenous shocks.

Control Variables

Our specifications control for a variety of political factors that prior analyses identify as potentially influencing patterns of public spending. First, we include a standard measure of party fragmentation—the “effective number of parties” developed by Laakso and Taagepera (1979)—to test whether an increase in the number of relevant parties, irrespective of their social foundations, influences patterns of government expenditures (Chhibber & Nooruddin, 2004).²¹ We also include a dummy variable for election years to control for the influence of electoral cycles on spending (Khemani, 2004). In addition, we control for the competitiveness of elections by including a measure of winning margins (the percentage of votes won by the leading party state assembly minus the percentage of votes won by the runner-up). We also include a dummy for coalition governments, to account for the possibility that the uncertainty faced by such governments prompt them to increase spending on public goods to shore up support (Saez & Sinha, 2009).

Finally, we control for macroeconomic factors that might shape patterns of expenditure. As government spending is likely to be influenced by overall levels of wealth, we control for levels of development by including the log per capita state domestic product (SDP). We also include total central grants as a percentage of state revenues to control for any changes in state expenditures arising from a boost from central government monies.²²

Results

The results of our analysis, presented in Table 3, provide a compelling confirmation of our argument regarding the differential impact of EEPs and NEPs on patterns of public spending. The models demonstrate EEP success is positively associated with development spending and that there are significant differences between the degree to which EEPs and NEPs emphasize spending on public goods and services.²³

Table 3.

Ethnic Parties and Public Spending (1967-2007).

Variables	(1) Development expenditures (% of budget)	(2) Log development expenditures per capita	(3) Administrative services (% of budget)	(4) Log administrative services per capita	(5) Head count ratio
Encompassing ethnic parties	0.0308** (0.0127)	0.00103** (0.000424)	−0.0160** (0.00669)	−0.000390 (0.000791)	−0.0283* (0.0152)
Narrow ethnic parties	−0.00214 (0.0360)	−0.00178 (0.00133)	0.0184* (0.00989)	0.000712 (0.00106)	0.0448 (0.0473)
Effective parties	−0.000574 (0.297)	−0.00293 (0.00664)	0.0376 (0.0701)	0.00474 (0.00888)	0.896*** (0.234)
Election dummy	−0.413 (0.445)	−0.00959 (0.0121)	0.313 (0.192)	0.0181 (0.0116)	−0.495* (0.257)
Winning difference	−0.0306 (0.0475)	0.000388 (0.000878)	0.0132 (0.0135)	0.00243* (0.00133)	−0.0212 (0.0363)
Coalition government	−0.252 (1.115)	0.0133 (0.0193)	0.131 (0.288)	0.0316 (0.0381)	−0.518 (0.886)
Grants from center	30.70** (10.43)	1.082*** (0.223)	−3.895** (1.490)	0.441*** (0.136)	6.418 (7.984)
Log real per capita SDP	5.175 (4.331)	0.246*** (0.0646)	1.024 (1.418)	0.177 (0.140)	−11.66** (4.026)
Constant	13.68 (37.53)	3.104*** (0.553)	5.325 (12.12)	2.197* (1.189)	140.8*** (34.31)
Observations	540	540	534	534	454
R ²	.755	.964	.793	.944	.953

All regressions include fixed state and year effects. Robust clustered standard errors in parentheses. Development and administrative expenditures are taken as a percentage of total state government expenditures. See text for details on other variables in the analysis.

p < .1. **p < .05. ***p < .01.

Columns (1) and (2) show the results for our analysis of development expenditures as a percentage of total expenditures and in log real per capita terms. In both models, the EEP coefficient is positive and statistically significant at the .05 level.²⁴ Columns (3) and (4) of Table 3 present the results of our analysis of administrative spending. EEP dominance in the state assembly predicts a reduction in spending on administrative services as a percentage of the total budget, whereas NEP dominance predicts an increase in such spending.²⁵ While we do not find a statistically significant relationship between EEP or NEP dominance and administrative spending in per capita terms, the signs of the coefficients are in the anticipated direction. Column (5) presents the results of our analysis of EEP dominance on poverty. In this analysis, we find confirmation of the expectation that EEP emphasis on public goods provisioning results in declining poverty. The EEP coefficient is significant at the .1 level and is in the predicted negative direction. This result suggests EEP success actually improves the fortunes of a broad swathe of disadvantaged voters and is not simply beneficial for party leaders and their coteries.²⁶

In addition to confirming the expected relationships between party type and public spending, our models suggest that EEPs and NEPs have a large substantive impact on spending patterns. Models (1) and (3) predict that a one standard deviation change in the dominance of EEPs corresponds to about a 0.25% increase in development spending as a percentage of the total budget each year, and a -0.04% decrease in administrative spending as a percentage of the total budget each year. These changes may sound small but can prove to be substantial in absolute terms when one considers that most of the budget is fixed and only a very small percentage can be devoted to new social services in a given state/year (Khemani, 2004).

To better demonstrate the substantive impact of ethnic parties on overall levels of spending, Figure 3 graphs predictions of levels of spending (in 1994 rupees per capita) and poverty at various levels of EEP and NEP dominance.²⁷ The graphs illustrate how increasing EEP dominance results in progressively higher levels of per capita development spending, lower levels of administrative spending, and lower levels of poverty, whereas NEP dominance produces just the opposite effects on these variables.

Figure 3.

EEP dominance, public spending, and poverty.

To take just one scenario, assume two hypothetical Indian states, one in which 60% of the seats are held by EEPs and one in which 60% of the seats are held by NEPs. Our models predict that the EEP-dominated state will spend Rs. 604 per capita on development, Rs. 84 per capita on administrative services, and that 37.6% of the population will fall below the poverty line. The NEP-dominated state, conversely, will spend Rs. 519 per capita on development, Rs. 90 per capita on administrative services, and will have a poverty rate of 41.5%. In other words, the EEP-dominated state will spend 16% more on development spending in a given year, 5% less on administrative services in a given year, and will experience 9% less poverty than the NEP-dominated state.

Generally speaking, other standard explanations for a high level of commitment to social spending do not find the same level of confirmation in the results of our analysis. The effective number of parties measure positively predicts levels of poverty, but is not significant in any of the spending regressions. This supports our contention that it is the inclusive or exclusive nature of parties driving fragmentation, rather than fragmentation per se, which influences patterns of public spending. In most of the models, other measures of the competitiveness of the political system (winning margins and the coalition government indicator) also fail to reach standard levels of statistical significance. Nor is the election dummy significant in any of the models. The models do suggest that some, but not all, economic factors play a role in the allocation of public goods. Central government transfers positively predict all types of spending, and higher per capita state domestic product correlates with higher development spending.

Conclusion

Scholars have largely conceptualized ethnic parties as exclusionary political actors. Ethnic party success hinders the supply of public goods, it is thought, because leaders of such parties primarily use patronage targeted to their co-ethnics to forge support. The exclusive nature of ethnic politics has been viewed as so complementary to the discretionary logic of patronage that one influential account went so far as to say the two are “locked into a stranglehold, with the one reinforcing the other.”²⁸

Our study seeks to qualify this conventional wisdom. We draw attention to the fact that many ethnic parties emerged to represent the interests of politically marginal groups, especially in “ranked” ethnic societies. In such contexts, ethnic challenger parties can increase the autonomy of non-elite voters, by disrupting the ranked hierarchies used to encapsulate them within multi-ethnic coalitions headed by traditional elites.

This “autonomy effect” endows ethnic parties with the potential to have a salutary impact on public spending. Whether these actors realize their potential is a function of the type of identities they mobilize. In some cases, ethnic parties emerged from non-electoral collective action that mobilized voters under encompassing identities. These “encompassing ethnic parties” increased non-elite voter autonomy and looked to replace elite-centered winning coalitions with broadly defined ethnic ones. The rise of EEPs therefore expanded the effective size of local winning coalitions, thereby increasing incentives for governments to spend on broadly accessible goods and services, and reducing poverty.

Whereas EEPs were formed out of broad, bottom-up collective action, other ethnic parties emerged as the opportunistically constructed vehicles of office-seeking elites. These top-down formations tended to mobilize narrower ethnic constituencies from which their leaders hailed. The success of such “narrow ethnic parties” leads to restricted winning coalitions of “high-ranked” political elites being replaced by equally narrow coalitions of “low-ranked” elites. The concentrated support bases of these parties thus incentivized them to replicate prior vertical patronage strategies within their ethnic cores, rather than to increase spending on inclusive public services.

These claims have been formulated and tested within the Indian states, yet we believe they carry implications beyond the subcontinent. The “autonomy effect” of ethnic parties we outline has been largely neglected and may well resonate with political experiences outside India. We expect such parallels to be more likely in cases where ethnic parties have mobilized marginalized groups under a relatively broad ethnic identity. As in India, such parties often emerge out of broader social mobilizations demanding political change. For example, Madrid (2008) describes the Movimiento al Socialismo (MAS) led by Evo Morales as one such “inclusive” party seeking to mobilize and unite Bolivia’s marginalized indigenous communities.

Previous ethnic parties in Bolivia had exclusively focused on narrow linguistic communities such as Quechuas or Aymaras, which comprised a small percentage of voters. Like EEPs in India, the MAS sought to mobilize voters sharing a much broader ethnic identity—that of indigenous peoples who had been historically underrepresented in government. After coming to power on the back of this broad non-elite coalition, the MAS has implemented a “highly redistributive” agenda, which has included land reform, non-contributory pension schemes, and cash transfer programs (Madrid, 2005).²⁹

Another potentially intriguing case to consider is that of the UNMO in Malaysia, which has built a broad following among Bumiputeras [literally “sons of the soil”], the country’s large and internally diverse ethnic Malay population. The UNMO forged this support through promises to address the shared economic marginality of Malays relative to the country’s wealthier ethnic Chinese and Indian minorities. Our argument would suggest the nature of this encompassing ethnic support importantly informs the UNMO’s record of implementing broadly progressive social policies, which have included some of the highest rates of spending on basic social services in all of Asia (Thachil, 2009).

More broadly, our argument conceptually implies that a heightened electoral salience of ethnicity need not always indicate a more exclusionary form of politics. We provide a replicable methodology for identifying EEPs and NEPs, and measuring the effects of their success on public spending. However, the specific dimensions of ethnicity relevant for such analyses need to be context-sensitive and also account for the specific conditions out of which ethnic parties emerge. Such contextually sensitive analyses can better anticipate how the electoral rise of ethnic parties will influence not only specific configurations of public spending but also broader patterns of democratic politics.

Footnotes

Acknowledgements

The authors would like to thank Pradeep Chhibber, Ronald Herring, Irfan Nooruddin, Ashutosh Varshney, and participants at seminars at Juan March Foundation, and Yale University for their helpful comments.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors thank the George Washington University Center for International Business Education and Research (GWU-CIBER) for financial assistance enabling this research.

Notes

Author Biographies

Tariq Thachil is an Assistant Professor of Political Science at Yale University. He research focuses on political parties, ethnic politics, and urban migration, with a regional focus on South Asia. His first book Elite Parties, Poor Voters: How Social Services Win Votes in India was published by Cambridge University Press (Studies in Comparative Politics series) in 2014.

Emmanuel Teitelbaum is an Associate Professor of Political Science and International Affairs at The George Washington University. He is author of Mobilizing Restraint: Democracy and Industrial Conflict in Post-Reform South Asia (Cornell University Press, 2011), as well as a number of academic articles on contemporary Indian politics and the sources class conflict and class compromise in the developing world.

References

Alesina

Baqir

Easterly

(1999). Public goods and ethnic divisions. Quarterly Journal of Economics, 114, 1243-1284.

Baland

J. M.

Robinson

(2007). How does vote-buying shape the economy? In Schaffer

(Ed.), Elections for sale: The causes and consequences of vote-buying (pp. 123-141). Boulder, CO: Lynne Rienner.

Besley

Coate

Loury

(1993). The economics of rotating savings and credit associations. American Economic Review, 83, 792-810.

Bueno

Mesquita

Smith

Siverson

R. M.

Morrow

J. D.

(2003). The logic of political survival. Cambridge, MA: MIT Press.

Carey

Shugart

(1995). Incentives to cultivate a personal vote: A rank ordering of electoral formulas. Electoral Studies, 14, 417-439.

Chandra

(2004). Why ethnic parties succeed. Cambridge, UK: Cambridge University Press.

Chandra

(2007). Counting heads. In Kitschelt

Wilkinson

(Eds.), Patrons, clients, and policies (pp. 84-109). Cambridge, UK: Cambridge University Press.

Chandra

(2012). Constructivist theories of ethnic politics. Oxford, UK: Oxford University Press.

Chhibber

(1999). Democracy without associations: Transformation of the party systems and social cleavages in India. Ann Arbor: University of Michigan.

10.

Chhibber

Nooruddin

(2004). Do party systems count? The number of parties and government performance in the Indian states. Comparative Political Studies, 37, 152-187.

11.

Corbridge

Harriss

(2000). Reinventing India. Cambridge, UK: Polity.

12.

Dawson

(1995). Behind the Mule: Race and Class in African-American Politics. Princeton, NJ: Princeton University Press

13.

Fershtman

Gneezy

(2001). Discrimination in a segmented society: An experimental approach. The Quarterly Journal of Economics, 116, 351-377.

14.

Franke

R. W.

Chasin

B. H.

(1989). Kerala: Radical reform as development in an Indian state. Oakland, CA: Institute for Food and Development Policy.

15.

Habyarimana

Humphreys

Posner

Weinstein

(2007). Why does ethnic diversity undermine public goods provision? American Political Science Review, 101, 709-725.

16.

Hicken

(2011). Clientelism. Annual Review of Political Science, 14, 289-310.

17.

Horowitz

(1985). Ethnic groups in conflict. Berkeley: University of California.

18.

Huber

(2012). Measuring ethnic voting: Do proportional electoral laws politicize ethnicity? American Journal of Political Science, 56, 986-1001.

19.

Huber

Suryanarayan

(2014). Ethnic inequality, political mobilization and the ethnification of political parties across states in India. Unpublished manuscript, Columbia University, New York, NY.

20.

Ichino

Nathan

(2013). Crossing the line: Ethnic geography and voting in Ghana. American Political Science Review, 107, 344-361.

21.

Iverson

Soskice

(2006). Electoral institutions and the politics of coalitions. American Political Science Review, 100, 165-181.

22.

Jaffrelot

(2003). The silent revolution. New Delhi, India: Permanent Black.

23.

Jaffrelot

Kumar

(Eds.). (2009). The rise of the Plebeians? New Delhi, India: Routledge.

24.

Khemani

(2004). Political cycles in a developing economy: Effect of elections in the Indian states. Journal of Development Economics, 73, 125-154.

25.

Kitschelt

Wilkinson

(2007). Patrons, clients, and policies. Cambridge, UK: Cambridge University Press.

26.

Kothari

(1964). The congress system in India. Asian Survey, 4, 1161-1173.

27.

Laakso

Taagepera

(1979). Effective number of parties: A measure with application to Western Europe. Comparative Political Studies, 12, 3-27.

28.

Lawrence

Sinha

(2009). Political cycles, political institutions, and public expenditure in India, 1980-2000. British Journal of Political Science, 40, 91-113.

29.

Lerche

(1999). Politics of the poor: Agricultural labourers and political transformations in Uttar Pradesh. Journal of Peasant Studies, 26, 182-241.

30.

Lieberman

McClendon

(2013). The ethnicity-policy preference link in sub-Saharan Africa. Comparative Political Studies, 46, 574-602.

31.

Madrid

(2005). Indigenous voters and party system fragmentation in Latin America. Electoral Studies, 24, 689-707.

32.

Madrid

(2008). The rise of ethnopopulism in Latin America. World Politics, 60, 475-508.

33.

Mendelsohn

(2002). The transformation of authority in rural India. In Shah

(Ed.), Caste and democratic politics in India (pp. 167-208). New Delhi, India: Permanent Black.

34.

Miguel

(2004). Tribe or nation? Nation building and public goods in Kenya vs. Tanzania. World Politics, 56, 327-362.

35.

Nikolenyi

(2009). Minority governments in India: The puzzle of elusive majorities. London, England: Routledge.

36.

Posner

(2006). Institutions and ethnic politics in Africa. Cambridge, UK: Cambridge University Press.

37.

Roberts

(2002). Social inequalities without class cleavages in Latin America’s neoliberal era. Studies in Comparative International Development, 36(2), 3-33.

38.

Robin

(2009). Bihar: The new stronghold of OBC politics. In Jaffrelot

Kumar

(Eds.), The rise of the Plebeians (pp. 65-102). New Delhi, India: Routledge.

39.

Rudolph

S. H.

(1960). The political role of India’s caste associations. Pacific Affairs, 33(1), 5-22.

40.

Rudolph

S. H.

(1967). The modernity of tradition. Chicago, IL: University of Chicago Press.

41.

Singh

(2011). We-ness and welfare: A longitudinal analysis of social development in Kerala, India. World Development, 39, 282-293.

42.

Stokes

Dunning

Nazareno

Brusco

(2013). Brokers, voters, and clientelism. Cambridge, UK: Cambridge University Press.

43.

Thachil

(2009). Neoliberalism’s two faces in Asia. Comparative Politics, 41, 473-494.

44.

Thachil

(2014). Elite parties and poor voters. American Political Science Review, 108, 454-477.

45.

Vakil

F. D.

(1990). Patterns of electoral performance in Andhra Pradesh and Karnataka. In Sisson

Roy

(Eds.), Diversity and dominance in Indian politics (pp. 249-275). New Delhi, India: SAGE.

46.

Van Cott

(2005). From movements to parties in Latin America. Cambridge, UK: Cambridge University Press.

47.

Varshney

(2012). Two banks of the same river? Rising north-south economic divergence in India. In Chatterjee

Katznelson

(Eds.), Anxieties of democracy (pp. 225-256). Oxford, UK: Oxford University Press.

48.

Vaugier-Chatterjee

(2009). Two dominant castes in Andhra Pradesh. In Jaffrelot

Kumar

(Eds.), The rise of the Plebeians (pp. 277-309). New Delhi, India: Routledge.

49.

Weiner

(1967). Party building in a new nation. Chicago, IL: University of Chicago Press.

50.

Yadav

Kumar

(2007, May 17). UP polls: BSP’s poor, powerful voter. IBNLive. Retrieved from http://ibnlive.in.com/news/up-polls-bsps-poor-powerful-voter/40749-4-2.html

51.

Yashar

(2005). Contesting citizenship. Cambridge, UK: Cambridge University Press.

52.

Ziegfeld

(2012). Coalition government and party system change: Explaining the rise of regional political parties in India. Comparative Politics, 45(1), 69-87.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.53 MB