Do Internal Migrants Divide or Unite Across Ethnic Lines? Ethnographic and Experimental Evidence from Urban India ABSTRACT: Despite rapid internal migration to cities across the developing world, little is known about identity formation within poor urban migrant communities. When will ethnic differences socially and politically divide poor migrants in the city? And when will ascriptive differences be obscured by shared class identities? I study these crucial questions through six months of ethnographic fieldwork and a large-scale vignetteexperiment (N=3018) conducted among internal migrants in urban India. Intra-class ethnic divisions prove highly salient in interactions exclusively between poor migrants, but irrelevant in interactions that also include urban elites. I introduce a new mechanism to explain this variation: a solidarity effect produced by city elites who perceive and treat migrants of different ethnicities in shared class terms. Elite behavior stems from the concentration of poor migrants into ethnically heterogeneous, economically homogeneous urban worksites and settlements. Such concentration reverses the visibility advantage ethnicity enjoys over class in villages structured along ethnic lines.
Tariq Thachil Department of Political Science Yale University 115 Prospect Street, New Haven 06520
[email protected] (607)-229-3197
Keywords: Comparative Political Behavior, Urban Politics, Ethnic Politics, Democratization, South Asia.
In November of 2013, I spoke with Rajesh, one of the millions of poor laborers who migrate each year from villages in northern India to New Delhi to work on construction sites dotting the national capital’s ever-expanding urban periphery. I asked him what the biggest difference was between city and countryside, expecting him to mention the number of buildings, the hectic traffic, the price of food. Instead, he told me: My village is divided by caste. There is one section for Thakurs [upper castes], and one for Harijans [lower castes] like me. So at home, everyone around me is of my caste community. But in the city that is not that way. Here I am with people of all caste communities…Thakurs [upper castes], Harijans [lower castes], it is a zoo of castes. But they are all ‘labor log’ [laborers] like me.1 Rajesh’s observation points to a key implication of urban migration across the global south: the spatial concentration of poor migrants into residential settlements and worksites that are ethnic diverse, yet relatively economically homogenous.2 Such concentration forces poor migrants to engage in intra-class, inter-ethnic interactions far more frequently than in their home villages. As Weiner (1978: 26) noted in a seminal study of the politics of internal migration, such interactions have both ‘integrative and divisive potential.’ Increased inter-ethnic engagement in cities could either attenuate or reinforce the salience of ethnic differences among poor migrants. Yet four decades after this observation, the conditions under which either potential is realized remain poorly understood. Under what conditions will ethnic differences divide poor migrants in their multi-ethnic urban communities? And when will such differences be obscured by commonalities of shared class positions in the city? 1
Interview, Rani Nagar, New Delhi. 11/8/2014. All interviews conducted by author.
Names of chowks have been changed to protect respondent confidentiality. 2
Following Chandra (2012), I define ethnic groups as any subset of people for whom
descent-based attributes are necessary for membership.
This paper addresses these crucial questions through a study of poor migrants in urban India. To examine when and how ethnic differences divide migrants of comparable economic positions, I constructed a sample that was highly ethnically heterogeneous, yet highly occupationally homogenous. To do so, I sampled respondents from an original listing of over 200 casual labor spot markets where many poor migrants assemble to find work. I analyze this population through a novel blend of street ethnography and a large vignette-based survey experiment (N=3018). My study finds ethnic differences prove salient in interactions exclusively between poor migrant workers, such as selecting another migrant as a roommate or informal worksite leader. By contrast, ethnic differences are irrelevant for the same migrants in interactions that involve urban elites, such as securing employment, or voting in urban elections. I argue this variation is explained by an elite-induced solidarity effect, stemming from the uniform treatment poor migrants receive at the hands of the city’s privileged. Indiscriminate treatment from elites stems from the concentration of poor migrants of all ethnicities at shared urban worksites and settlements. Such concentration renders their shared class identity more visible to urban elites than their disparate ethnic ones. Cities thus reverse the ‘visibility advantage’ ethnicity has been argued to enjoy over class in Indian villages organized along ethnic lines (Chandra 2004). This paper make several theoretical and methodological contributions to literatures on ethnic and urban politics. First, it adds to our understanding of the sociopolitical consequences of internal urban migration in poor countries. Internal migrants are helping shift the demography of Asian and African countries. In West Africa, there are now 100 million more urban residents than in 1990, and countries like Ghana have reported majority urban populations for the first time in their histories (Nathan 2014). The percentage of Indonesians living in cities has steadily risen over the past two decades, and 2000-2010 marked the first decade in which India added more people to its cities than its villages. Yet internal migrants from these regions remain far less studied
2
within comparative politics than transnational south-north immigrants, despite outnumbering the latter by a significant margin (UNESCO 2013). Within India, political scientists have importantly studied how internal migrant influxes can activate nativist mobilizations (Weiner 1978, Bhavnani and Lacina 2014, Gaikwad and Nellis 2014). However they have rarely examined processes of conflict and cooperation among migrants themselves. This paper shifts attention to such processes by providing a rigorous test of the conditions under which ethnic differences do and do not generate intra-class divisions among poor migrant communities. My research design includes (to the best of my knowledge) the first survey experiment conducted among internal migrants in the developing world. The experiment explores intra-migrant relations across important and precisely defined issue areas- including labor market competition, housing, informal leader selection, and voting- that are typically studied separately. Studying these issues collectively reveals situational variation in ethnicity’s salience, which in turn suggests a novel pathway of migrant identity formation. Existing scholarship assumes enforced proximity within cities will prompt poor migrants to bridge ethnic divides by generating mutual affection through increased inter-ethnic contact. My findings suggest instead that such proximity reduces ethnic divisions more indirectly, by affecting how urban elites perceive these migrants. Second, this study contributes to the rich constructivist literature on ethnic politics. In revealing that poor migrants divide along ethnic lines in certain urban interactions but completely ignore ascriptive differences in others, my findings align with the constructivist view that ethnic identity salience can shift, sometimes rapidly. Yet scholars of Asia and Africa have primarily used this insight to analyze how the salience of one ethnic identity, say tribe or caste, can divide individuals who are co-ethnics along a second dimension, such as language group or religion (Laitin 1986, Chandra 2005, Posner 2005). My study shows how this framework can be extended to study the
3
conditions under which ethnic identities succeed or fail in dividing members of the same class. Further, I show the ‘visibility advantage’ ethnicity is argued to enjoy over class in the countryside cannot be assumed to extend to urban environments. To the extent that this advantage has centrally informed ethnicity’s role in structuring patronage politics, my findings suggest a hitherto unexplored mechanism through which urbanization might erode the logic of ethnic voting. Finally, I deploy a unique methodological strategy of adapting conversations from sustained street ethnography to construct experimental vignettes grounded in the words and experiences of my informants. This strategy of developing ‘ethnographic vignetteexperiments’ is important particularly given rising concerns that survey experiments are often designed to focus on a researcher’s theoretical goals at the expense of remaining intelligible and meaningful to respondents (Morton and Williams 2010). Additionally my worksite-based strategy for sampling poor mobile migrants, who often lack official addresses in the city, can easily be extended to other settings.
ETHNICITY AND CLASS WITHIN POOR MIGRANT COMMUNITIES The central purpose of this paper is to understand the conditions under which ethnic differences divide poor migrants, and when such differences are superseded by commonalities of class and occupation. Since Marx (1848), scholars have studied such issues within the context of wealthy Western democracies, but equivalent analyses within poorer nations remain rare. In large part, this neglect is a function of the rural demography of sub-Saharan African and South Asian countries, where most contemporary scholarship on ethnic politics has been based (e.g. Wantchekon 2003, Chandra 2004, Miguel 2004, Posner 2005, Dunning and Nilekani 2013, Ichino and Nathan 2013, Chauchard 2014). Villages in these regions rarely contain ethnically diverse populations of the same class position. In sub-Saharan Africa, geographically bounded ethnic divisions ensure relatively homogenous village communities (McCauley
4
2014). In South Asia, villages are internally diverse, but ethnic differences are ‘ranked’ and correlate strongly with economic status (Horowitz 1985, Herring and Agarwala 2006). Yet studies of cities as different as Nairobi, Accra, Lusaka, Bangalore, and Jaipur increasingly paint a different picture.3 Sub-Saharan African cities concentrate poor migrants from regionally distinct ethnic groups into multi-ethnic slums and informal marketplaces (Marx et. al 2014, Nathan 2014, Resnick 2012). Indian cities bring together poor migrants of different castes in diverse squatter settlements, casual labor spot markets, and street vendor bazaars (Agarwala 2013, Auerbach 2013, Krishna 2013, Sethi 2011).4 Migrants within these communities obviously have far greater chances to engage in intra-class, inter-ethnic interactions. To assess the conditions under which ethnic differences do and do not divide poor urban migrants in these interactions, we must first specify major theoretical mechanisms through which they might do so. Existing scholarship suggests two major mechanisms through which ethnicity socially divides poor migrants: by inhibiting intra-class cooperation, or by exacerbating intra-class 3
This is not to deny that some level of ethnic sorting occurs in South Asian and African
cites, but this tends to be dwarfed by economic sorting. Practically everyone living in an Indian urban slum is poor, while practically no one living within a middle-class urban neighborhood is (Auerbach 2013, Heller 2014). Similarly Akoh (2007:121) notes that the rapid growth of African cities, especially the ‘chaotic expansion’ of informal poor settlements renders ethnic segregation ‘more of a myth than reality’. 4
For example, recent slum census data collected from the Indian government for a major
slum resettlement program (Rajeev Awaas Yojana) show the vast majority of slums in north Indian cities of Jaipur, Bhopal, and Lucknow are highly diverse across caste and religious lines.
5
competitive tensions. Theories of ethnic voting imply ascriptive differences might politically divide poor migrants by generating variable preferences for formal political leaders who control patronage flows (Chandra 2004, Posner 2005). Recent works suggest this logic might also extend to the informal leaders and brokers poor migrants select from amongst themselves (Stokes et. al 2013, Auerbach 2013). Below, I distill each of these four pathways into falsifiable hypotheses for subsequent testing. While no list is exhaustive, these four hypotheses collectively constitute a theoretically motivated, multidimensional, and precise test of my central question. Cooperation Classical modernization theorists saw urban migration as uprooting villagers from parochial agrarian societies, thereby diminishing the salience of ethnic identities (Lipset 1959, Gellner 1964, Apter 1965). Gluckman (1960:57) famously argued that ‘every African is detribalized as soon as he leaves his tribal areas’. Such ‘detribalization’ was partly compelled by the fact that cities enabled poor migrants to interact with noncoethnics of the same class, helping them realize similarities stemming from shared class positions, and inducing them to cooperate with one another to overcome shared challenges produced by their common dislocation. Social contact theorists confirmed the importance of such interactions, finding intergroup biases were most frequently overcome in conditions of equal economic status and frequent interaction across social groups. Such conditions typify ethnically diverse, economically homogenous poor migrant communities (Allport 1954, Pettigrew and Tropp 2006). Instrumental theories of ethnic politics challenged the optimistic logic of these arguments, noting that cooperation within multi-ethnic contexts often takes place along ethnic lines (Fearon and Laitin 1996, Habyarimana et. al 2007). Co-ethnic individuals are often embedded within dense social networks, enabling them to build mutual trust
6
(Fershtman and Gneezy 2001), and more effectively punish free riding within their group (Besley, Coate and Loury 1993). Consequently, ethnic differences might divide poor urban migrants by inhibiting the likelihood of intra-class cooperation in shared urban spaces: H1: Ethnic differences decrease the likelihood of migrants engaging in intra-class cooperation. Competition Critics of modernization theory provide a second pathway through which ethnic differences might socially divide poor urban migrants: by exacerbating animosities produced by intra-class competition. Precisely because poor migrants encountered equally poor non-coethnics in the city, fierce competition for scarce jobs in African cities was often filtered through ethnic lenses (Abernethy 1967, Bates 1983). As Bates (1983: 160) notes, a major consequence of urban competition ‘is to create…a sense of threat and disadvantage…and perceive it in ethnic terms [emphasis added]’. Similarly, Hindus migrating to Indian towns were argued to perceive competitive threats more negatively if they came from Muslims (and vice versa), ensuring ‘class identities could never be distilled’ out of religious divides (Chakrabarty 1989: 218, Rudolph and Rudolph 1993). A key observation from these studies was that poor migrants faced competitive pressures from co-ethnics and non-coethnics alike, but perceived the latter as more threatening. This logic was supported by psychological studies showing people respond more strongly to identical negative threats from out-groups compared to those from their own in-groups (Riek et. al 2006, Paolini et. al 2010). Ethnographic studies of immigrants in North America arrived at similar conclusions. For example, Waters (1999: 71) finds poor West Indians living in New York are more likely to forgive harmful competitive acts by co-ethnics compared to African-Americans of the same class.
7
H2: Ethnic differences increase the likelihood of migrants perceiving intra-class competitive acts as threats. Formal Politics Intra-class political divisions are usually conceptualized as divergent preferences for tax and transfer policies among citizens of the same class (Shayo 2009, Huber 2013). However, such frameworks may be less valid for developing democracies characterized by non-programmatic distributive politics. In such contexts, voters choose the candidates most likely to reward them with patronage when in office. Ethnicity often structures such choices, because the visibility and stickiness of ethnic markers, especially in comparison to class, bestow informational advantages crucial to solving politician-voter commitment problems inherent to clientelist pacts (Posner 2005). This form of ethnic clientelism is anticipated to be especially effective among poor populations (Kitschelt and Wilkinson 2007). Consequently, theories of ethnic voting imply poor migrants will privilege their diverse ethnic identities at the polls, diminishing their prospects of forming inter-ethnic, class-based electoral blocs. While intuitive, this general logic has not been systematically scrutinized among migrants or indeed other segments of the population. Existing studies test whether ethnic identities are electorally salient, but not whether such salience produces intra-class variation in electoral preferences. Yet this distinction is important, especially in ‘ranked’ ethnic systems like India’s where overlapping economic and ethnic differences make it difficult to ascertain if ethnic voting is actually class voting in disguise (Huber and Suryanarayana 2013). By removing the correlation between ethnicity and class, multiethnic, homogenously poor migrant communities help tease out such distinctions. In sum, if poor migrants of the same class systematically favor co-ethnics when evaluating politicians in urban elections, ethnic differences will create political divisions among them.
8
H3: Ethnic differences divide migrant preferences for formal leaders in city elections. Informal Politics Studies of how ethnicity shapes political behavior overwhelmingly focus on formal electoral preferences. Yet the political lives of poor migrants are hardly reducible to the narrow domain of elections. Most poor internal migrants in the global south work within the informal economy, and live in quasi-legal squatter settlements. Consequently, activities within these informal arenas are often more substantively important than the formal electoral arena studied by political scientists (Bremen 1996). Importantly for my analysis, prior scholarship documents the emergence of informal political figures within migrant communities. Studies of distributive politics increasingly recognize the importance of such brokers (Stokes et. al 2013), who range from Indian ‘fixers’ securing valuable goods for squatter settlements (Auerbach 2013) to Mexican street market leaders who control the allocation of selling spaces to vendors (Cross 1998). Given the importance of these resource flows, the logic of ethnic patronage could easily extend into the selection of these leaders. If poor migrants are systematically disinclined to support non-coethnic urban informal leaders, ascriptive differences could inform intra-class divisions in this key political decision in the city: H4: Ethnic differences divide migrant preferences for informal leaders within their communities.
RESEARCH DESIGN Selecting a Study Population Poor urban migrants in India are a predictably vast, heterogeneous population that is rapidly expanding, and contributing to the rising share of Indians who live in cities
9
(Figure 1).5 Accordingly, my research required precisely selecting a study sample from this massive population to fit my empirical and theoretical purposes. --- FIGURE 1 ABOUT HERE--Empirically, I wanted to ensure I focused on a substantively important subsample of poor urban migrants. Most such migrants in India are defined by two traits: informality and circularity. Well over 90% of Indian migrants work in the informal sector, and largely observe circular patterns of migration (NSS 2007). Further, most of these circular migrants are poor men seeking employment in urban areas.6 By contrast, most permanent migrants in India are rural women moving between villages after marriage (Census of India 2001, UNESCO 2013). In addition to being theoretically appropriate, circular Indian migrants constitute a substantively important population, recently estimated at 100 million, or roughly the population of Mexico (Deshingkar and Akter 2009). Focusing on circular migrants can prompt important objections. One is that such migrants are not truly ‘urban’, working in the city during agricultural lean seasons, while spending most of their time in the countryside. Yet intergenerational declines in landholdings for the rural poor in India have forced longer absences. Original survey evidence found most poor circular migrants stay in the city for between 9-11 months, returning only for family emergencies or to celebrate weddings and religious holidays. A second concern is that circularity and informality are merely ‘transitional’ stages to formal employment and a permanent shift to the city. Yet a considerable literature on India refutes such linear assumptions, which are often derived from Western experiences (Rogaly 1999, Gidwani and Sivaramakrishnan 2003). Instead, circularity and informality 5
Internal migration is estimated to account for about 1/3 of urban population growth in
India. Other factors include natural urban population growth and reclassification of large villages (UNESCO 2013). 6
72% of circular male migration is to cities (NSS 2007, p. 95). 10
often persist across several generations, and have proven to be ‘the overwhelming and enduring reality of Indian urban economies, both past and present’ (Gooptu 2001: 2). While my empirical interests compelled a focus on circular migrants, my theoretical interests required a subsection of these migrants who were ethnically heterogeneous, economically homogenous, and had opportunities to regularly interact at shared worksites and residences in the city. Such a sample would afford the best leverage on assessing how ethnic differences might divide migrants of the same class in their urban interactions. I therefore focused on urban circular migrants working in the construction sector, which is the majority employer of such migrants in India (Figure 2). Finally, I focused geographically on two north Indian cities in the so-called ‘Hindi belt’, home to seven of the top ten inter-state, rural-urban migration streams (IIHS 2011). ---FIGURE 2 ABOUT HERE--Having defined this study population, the biggest logistical challenge was locating respondents. Given their high mobility and lack of fixed urban addresses, sampling my respondents at their places of residence was not a viable strategy. In order to address this problem, I accessed informants at the informal labor ‘spot markets’ that are ubiquitous across Indian cities. Men assemble at these labor chowks [markets] in the hopes of finding temporary employment at construction sites that pepper India’s expanding urban landscape. Most work available at the chowk is via short-term informal oral contracts, typically lasting a single day. Sampling at these markets greatly simplified the task of locating informants, but also entailed challenges that I detail later in the paper and in Supplement C. Vignette-Experiment Design My empirical strategy centered on a vignette-based survey experiment administered among a large sample of over 3000 migrant laborers. To my knowledge, this is the first survey experiment implemented among internal urban migrants in the
11
global south. Respondents were presented with four vignettes, each designed to test one of the hypotheses discussed above. I will describe how each vignette was constructed in the next section. I first focus on describing the experimental design and implementation protocol. Each vignette is centered on a hypothetical migrant or political candidate, whose identity is manipulated by changing two pieces of information about him: his name, and the state from which he has come.7 I used four different names, which provided respondents with information about three nested ethnic attributes. The first is the migrant’s (or candidate’s) narrow sub-caste (jati). Jati refers to endogamous subcastes that denote traditional occupations, are highly localized, and number in the hundreds. Hindu hierarchies aggregated jatis of a similar socio-economic status into larger caste groupings or varnas. Colonial and post-colonial state practice reinforced the salience of varnas by using these categories for social policy purposes (most famously for affirmative action quotas). Finally, each name indicated the migrant/candidate’s religious identity to be Hindu or Muslim (roughly 95% of Indian citizens belong to one of these two faiths). Figure 3 depicts the nested structure of these ethnic categories. ---INSERT FIGURE 3 ABOUT HERE--Each hypothetical migrant was also assigned a state of origin, either Uttar Pradesh or Bihar (see Table 1). This decision was facilitated by the fact that these two large states accounted for over 85% of the migrants we found at the surveyed marketplaces in my study cities of Delhi and Lucknow. While I do not focus on region-of-origin results in this paper, it bears noting that results presented here are for migrant respondents who are
7
Female seasonal migrants constitute a relatively small proportion of chowk workers. I
therefore utilized a male survey sample.
12
also from these two states.8 ---INSERT TABLE 1 ABOUT HERE— As Table 1 shows, a respondent could receive any one of eight different names in a particular vignette. The randomization was done to ensure no respondent was ever presented with the same name more than once across the four vignettes. Because this was a paper-based survey, the vignette-questions were asked in four fixed places within the instrument. However, the order of the vignettes was randomized across these four spots to minimize biases from question order effects. Each name created either an ethnic match or mismatch between the respondent and the hypothetical migrant along a given dimension of ethnicity. The salience of that ethnic dimension was determined by comparing the pooled mean response values for all experimental subgroups in which the respondent and hypothetical migrant matched along a particular identity, to those of all subgroups where they did not.9 Note that due to the nested nature of the identities, all respondents with the same jati necessarily shared the same caste and religion, but not vice versa. Similarly, all respondents of the same caste shared the same religion (Hinduism), but not vice versa. Survey Implementation Survey experiments of the form described above offer important advantages, notably helping reduce social desirability biases by providing unobtrusive ways to detect ethnic preferences. The experimental design also permits simultaneous assessment of the relative salience of several different dimensions of ethnicity (jati, caste, religion, and 8
Due to random sampling at labor markets, some selected informants came from other
states. These respondents were interviewed but their responses were excluded from this paper’s analysis. 9
For recent examples of a similar strategy within India see Dunning and Nilekani 2013.
13
region). However, such experiments frequently raise concerns of external validity. To help address these concerns, I replicated the full ethnographic and experimental study in two Indian cities. The first was New Delhi, the national capital, and one of three Indian megacities with a population over 10 million. This site was selected as the largest recipient of migrant construction laborers in the country (UNESCO 2013). The second city was Lucknow, one of India’s fifty ‘medium-sized’ cities (whose populations range from 1 to 10 million), and capital of Uttar Pradesh, India’s most populous state. A study of two cities is still limited in its scope. The significant costs and logistics of combining ethnography and large-scale surveys of a mobile migrant population prevented further replication. Still, repeating the study across one large and one mediumsized city considerably improves the scope of my findings, given that megacities and medium-size cities collectively account for the majority of India’s urban population, and its urban circular migrant population (additional details in Supplement Part C). To further improve the validity of my findings, I made substantial efforts to interview a representative sample of migrant construction workers within each city. This effort included a zone-wise enumeration of informal labor markets in Delhi and Lucknow. A team of RAs worked in conjunction with Departments of Labor in both cities, as well as with local NGOs, builders and contractors. This massive mapping exercise took place over a combined four months, and enumerated over 200 labor markets. It is difficult to claim this list is exhaustive, given the rapidity with which new markets arise and old ones disappear. Yet these efforts almost certainly included all major labor markets, and are the most comprehensive efforts to map informal labor markets in India. Within each city, the survey was administered at a random sample of labor markets selected from this list. Figure 4a and 4b shows these sites were well spread out across both cities. In total, 3018 migrant workers were interviewed across 74 labor markets (1818 in Delhi, 1200 in Lucknow). A further 350 workers were interviewed
14
during pilot surveys. Extensive details about the mapping exercise, sampling strategy and survey protocol are included in Supplement Part C. ---FIGURES 4A AND 4B ABOUT HERE--It is worth highlighting here that unusually stringent efforts were made to ensure the survey protocol was followed (detailed in Supplement Part C). I led intensive training sessions for all enumerators, including rehearsals at out of sample chowks. Teams included an external monitor who was independent of the survey firm. I reviewed monitor evaluations daily, and accompanied enumerators for the entire Delhi survey, and 3/4 of the Lucknow survey, rotating between the three teams in a random and unannounced pattern. All in all, I observed 470 of the 3018 completed surveys (15.5%). Vignette Construction Using Street Ethnography A second concern with the validity of vignette-based survey experiments is whether they are actually intelligible and meaningful to respondents, let alone generalizable beyond them (Morton and Williams 2010). Such concerns are especially acute for studies centered on poor and vulnerable respondents. Here, I relied on a novel approach: the use of extensive ‘street-ethnography’ at labor markets to identify the reallife scenarios that best captured each hypothesis (Shah 2014). I then adapted quotes from my fieldwork to word the specific vignettes in the language and experience of my informants. To access the widest possible range of migrant experiences, I conducted ethnographic fieldwork at marketplaces that varied across a number of key dimensions: geographical location, market age, surrounding neighborhood, and types of construction work typically found. I visited each of these marketplaces in alternation from 7 a.m. to noon (its operational hours). I conducted a total of 86 such daily visits between September and November of 2013 in Delhi, and in February in Lucknow. During these visits, I took part in numerous informal conversations with seasonal migrant workers over
15
tea, biscuits, and unfiltered cigarettes. I visited the rooms, shacks, and shelters in which they slept, and the scorching construction sites at which they worked. In two instances, I traveled with workers back to their home villages. Such repeated interactions were essential in gaining the trust of poor migrants who are often harassed by local authorities and residents. I now turn to discussing how this fieldwork helped me operationalize H1H4. 1. Cooperation Vignette The most important cooperative decision my ethnographic informants said they made in the city was deciding whom to share accommodation with. At first glance, sharing housing seems a poor measure of cooperative endeavors, given it is usually a highly private and sensitive decision. Yet such choices have a different flavor in the tough world of the chowks, where privacy and discretion are often unaffordable luxuries. Poverty, high city rents, and the need to live close to the chowk constrained the options of my informants, forcing them to share cheap rented rooms with each other.10 These rooms, which I often visited, are typically shared by between 2 and 6 workers. One of the first tasks for migrants arriving in the city is finding such shared accommodation. In some instances, they arrange to stay with family or friends from their home villages. However migrants more often find roommates at the chowk itself, and frequently live with other workers they had only known for a few days.11 In addition to being substantively important, sharing housing is a theoretically appropriate arena in which to investigate the impact of ethnic differences on intra-class 10
The author survey (2014) found the majority of migrants lived within 6 kilometers of
the chowk at which they came to find work, and lived in rented rooms. 11
50% of respondents to the author survey (2014) said they had no help from anyone in
their home village in finding accommodation before arriving in the city.
16
cooperation among migrants. While villages remain highly segregated on the basis of both caste and religion, my informants agreed there were abundant opportunities and fewer constraints to living with non-coethnic workers in the city: See here [in Delhi] people of all castes can stay in the same room. Village sections are separated by caste. And so people will say, I can’t drink water from here, or touch someone from there. But we are all here in the city because of the majburi [compulsions] of dal-roti [hunger/poverty]. And you meet people of all communities at the chowk.12 Yet other interviews suggested ethnic differences could still inhibit the solidary potential of such intra-class cooperative opportunities: It is best to live with people from your caste. Because the thing is if you share a room, someone can always make off with all of your things while you are working on the site. So you need someone you can trust…If 2-4 men come from my caste community, then they become my bandhas [friends].13 Another respondent expressed similar preferences, worrying that members outside his caste group might be financially irresponsible, ‘leaving me responsible for all of the rent.’14 Thus migrants can live with members of other ethnicities in the city, but do they? To assess this question across my survey sample, I designed the following vignette: One day I was speaking with a laborer at a chowk like this one. He told me about another laborer whose name was [NAME]. He is a member of the [HINDU CASTE/MUSLIM] community from [STATE]. The two were not good friends but
12
Interview, Rani Nagar, 11/10/2013.
13
Interview, Tilak Nagar, 9/25/2013.
14
Interview, Tilak Nagar, 11/11/2013.
17
used to meet at the chowk everyday. One day [NAME] asked my friend if he could share his room with him, and said they would share paying the rent.15 Next, each respondent was asked: Q1. Imagine you are looking for someone to share a rented room in the city with. If you met a laborer like [NAME] from [STATE] everyday at the chowk, would you let him share your room with you? H1 anticipates the proportion of positive answers to Q1 will be lower when the hypothetical migrant and respondent are non-coethnics than when they are coethnics. 2. Competition Vignette The most obvious source of competitive tensions among my ethnographic informants stemmed from the oversupply of workers at each chowk, which pressurizes migrants to undercut one another when negotiating wages from prospective employers: A lot of people here are such motherfuckers. Let us say you are a thekedar [contractor] who drives up to the chowk on a motorcycle. I am talking with you to try and fix a job. I am talking to you fixing the rate. While I am talking someone else comes and puts their leg up onto the motorcycle, and goes away with you, agreeing to work for less. Such people are the reason our dehadi [daily wage rate] remains so low.16 Such rate-katna [rate-cutting] is thus seen as a primary reason for wage rates remaining low, preventing migrants from accumulating the savings for which they leave their villages. 15
The vignette provides explicit mention of the hypothetical migrant’s jati to improve
comprehension, consistent with similar prior experiments in India (Dunning and Nilekani 2013, Bossuroy and Selway 2011). 16
Field notes, Rani Nagar, 9/22/2013. 18
Such competitive dynamics clearly have divisive potential. More importantly for my analysis, some informants suggested competitive animosities were exacerbated by ethnic differences. One lower caste worker in Delhi explained: It is one thing when we have to go for less money- that is usually just on days we can’t find work, but upper castes?...They [upper castes] are kamchors (lazy), because they have never had to work hard before coming to the city. In the village they are landlords so they can take it easy. So here they know they will go to the site and give such little effort [at the construction site] so they don’t mind always taking less money and cutting our rates.17 Note the informant acknowledges that migrants of all caste engage in rate-cutting, but perceives this act as a systematic threat when committed by an ethnic outgroup (upper castes), and an occasional understandable compulsion when committed by an ingroup. Other informants disparaged rate-cutting outgroup migrants as lacking intelligence, which made them willing to accept lower wages to the detriment of the informant.18 Qualitative evidence thus suggests perceptions of labor market threats are conditioned by ethnic difference. Are migrants systematically more likely to perceive rate-cutting as an economic threat if committed by a non-coethnic laborer? To test this possibility, I presented survey respondents with an experimental vignette, adapted from one of the quotes above: Once a laborer was standing at a chowk like this one and trying to fix a job with an employer, who he told his daily wage rate. As he was doing so, another laborer approached the employer from the other side. This second laborer’s name was [NAME] and he is a member of the [HINDU CASTE/MUSLIM] community from [STATE]. He told the employer he would do the work for less money. 17
Interview, Tilak Nagar, 10/29/2013.
18
Interview, Rani Nagar 11/13/2014.
19
Following this vignette, respondents were asked the following question: Q2: Some workers say people like [NAME] from [STATE] make it difficult for laborers like you to get work and earn money in the city. Do you agree? By focusing on migrants in the same occupational sector, and in using ethnographic insights, I am thus able to assess the impact of a precise wage threat. As Malhotra et. al (2013) note, such precision is crucial, but often overlooked, in testing how labor market competition shapes individual attitudes. H2 anticipates the proportion of positive answers to Q2 will be higher when the hypothetical migrant and respondent are non-coethnics than when they are coethnics. 3. Formal Politics Vignette To test H3, I presented survey respondents with a hypothetical political candidate running in a state assembly election. The candidate was also assigned a political party, although I do not examine partisan effects here, and accounting for responses to particular party cues do not affect the results presented in this paper.19 As in the other vignettes, this candidate was given a name from the list in Table 1, indicating their jati, caste, and religion. Instead of varying the hypothetical candidate’s region of origin, this vignette varied whether he was contesting assembly elections from an urban constituency within the destination city, or a rural constituency within the respondent’s home state.20 Here, I 19
In the Delhi survey the party was fixed to be the Indian National Congress. In the
Lucknow survey, the party was randomized between the Bahujan Samaj Party and Samajwadi Party. The effects of these partisan cues will be examined in a separate analysis. 20
I chose the state assembly level rather than local elections in order to maximize the
comparability of the candidate in both rural and urban scenarios. At the local level
20
took advantage of the somewhat unique political lives of my informants, who had direct experience of both urban and rural elections. Roughly 20% had voted in city elections, and over 70% had observed or participated in election rallies and campaigns held in their destination cities.21 These dual experiences provided an opportunity to compare how ethnic differences shaped intra-class political preferences across rural and urban elections among a population considerably familiar with both settings. Respondents were read a short speech the candidate was said to have given during an election campaign rally. The speech was deliberately worded to approximate a clientelistic appeal. This decision was taken because the logic behind H3 regards the electoral salience of ethnicity as rooted in its efficacy in facilitating quid pro quo politics that typifies India’s ‘patronage-democracy’ (Chandra 2004). Voters are anticipated to be more responsive to clientelist appeals from their co-ethnics, because they anticipate the latter to channel benefits along relatively sticky and visible ethnic markers. To simulate the clientelist language used by actual Indian politicians, I utilized a blend of phrases used in speeches given by real-life candidates I observed in six villages have village councils while large cities like Delhi have municipalities. Presenting candidates from other states standing in city elections would compromise the accuracy of the vignettes. Efforts were made to minimize chances the candidate was from a constituency in which the respondent had lived. The rural constituency was selected from a district from which few migrants come (Bihar: Sheikhpura constituency, Uttar Pradesh: Mirzapur constituency). A total of 8 respondents in the sample came from Sheikhpura or Mirzapur, and were excluded from the analysis. The urban constituencies were ones from which none of the sampled chowks was located. 21
As one informant said: ‘At the rallies that happen here- who goes? The local Dillivalas
[Delhites] don’t go. They round up all of us laborers and take us instead.’ (Interview, Tilak Nagar, 11/17/2013).
21
constituencies during city elections held over the course of my qualitative fieldwork.22 The exact text given to respondents were as follows: Now I will tell you about a political candidate from [PARTY]. This candidate ran in an election in [RURAL constituency /URBAN constituency]. His name was [NAME] and he is from the [HINDU CASTE/MUSLIM] community. I am going to read a short speech he gave during an election rally. After that I will ask you about your opinions of this candidate. This candidate said: My name is [NAME] and I am contesting assembly elections in [URBAN/RURAL constituency]. I have come to ask for your support. If I am able to win the election, then I will work hard to deliver benefits like BPL23 ration cards for getting cheap food to those who helped me succeed. I will also try to help those who helped me find government jobs. I will always remember to help those who helped me succeed. This language of mutual assistance, of ‘helping those who help me’ is a well-worn tactic used in Indian political speeches, especially in local campaign rallies in poorer areas my research was based in. When combined with a focus on providing excludable goods such as jobs and ration cards, such rhetoric closely resembles the scholarly definition of a clientelist appeal (Kitschelt and Wilkinson 2007). Politicians in India frequently make far
22
I observed rallies of Congress and BJP candidates in the following constituencies:
Badarpur, Chhattarpur, Mehrauli, Sadar Bazaar, Shahdara, Shakur Basti (November 1530, 2013). 23
BPL stands for Below Poverty Line ration cards, which grant access to a wide array of
subsidized foods.
22
more overtly quid pro quo speeches to poor constituents, including in the recent 2014 parliamentary elections.24 After hearing this vignette, respondents were asked: Q3: Would you vote for a candidate like [NAME] if they ran in a state assembly election in [RURAL constituency/URBAN constituency]? H3 anticipates the proportion of positive responses to Q3 will be lower for non-coethnics than for non-coethnics. 4. Informal Politics Vignette H4 focuses on how ethnic differences might fragment intra-class preferences for informal leaders from within migrant communities. To operationalize H4, I drew on conversations my ethnographic informants would frequently have about the need for greater collective unity: We come from all over India, but we are all mazdoors [workers] here. So why can’t we build ekta [unity] on that basis?...See take my five fingers- it you take each one separately, you can break them just like that- khatak [breaking noise]. That is what the thekedars [contractors] do to us. But if you keep them together as a fist, then you can break them?25 There was broad consensus on the best means for building such unity: an informal union that would regulate employment and fix wage rates. As one respondent told me, ‘If we have unity (ekta) among us, we can fix a minimum wage rate, and try to ensure payment. 24
For example, the deputy chief minister of the western state of Maharashtra warned
voters on the eve of polling he would “cut off their water supply” if they did not vote for his cousin: http://indianexpress.com/article/india/politics/vote-supriya-or-lose-waterthreatens-ajit/ 25
Interview, Rani Nagar, 9/18/2013. 23
But for that we need an organization.’26 Another respondent concurred: ‘See I think there should be a big register kept with all the laborers names. Then one man should be kept as the leader who will say that now it is your turn – that is how the taxi drivers do it, and do you see any of them starving?’27 Conversations with my informants also suggested ethnic differences could divide preferences for leaders of any such informal union: Laborer A: There should be a union. Laborer B: But then how will they choose who gets the jobs? Should the union leaders choose? Laborer A: Yes- they will have to choose. Laborer C: Then they will just choose their own men. Their own men from their own community…And [others] will get the job after waiting a year. But that man still has to pay rent every month or get beaten by the landlord right- so then what?28 This exchange suggests the distributive logic of ethnicity might extend to this more informal arena. This possibility was tested through the following vignette: Now I am going to tell you about a laborer I previously spoke to at a chowk like this one. After that I will ask your opinion about this laborer. His name is [NAME] and he is a member of the [HINDU CASTE/MUSLIM] community from [STATE]. He wanted to start an organization among laborers at his chowk. When a builder or contractor comes to this chowk to find workers, the organization headed by [NAME] from [STATE] would then decide which laborer went for which job, and make sure no one went for less than the standard rate. 26
Interview, Rani Nagar, 9/15/2013.
27
Interview, Rani Nagar, 9/30/2013.
28
Interview, Rani Nagar, 2/12/2014.
24
Following this vignette, respondents were asked the following question: Q4. Would you join an organization like this one if it were headed by a laborer like [NAME] from [STATE]? H4 anticipates the proportion of positive responses to Q4 will be lower for non-coethnic informal leaders than for co-ethnics. EXPERIMENTAL RESULTS The research design was successful in constructing an ethnically heterogeneous, but economically homogeneous sample of migrants. 75% of respondents earned wages between $2-4 per day on average, 77% of respondents did not have a secondary education, 75% had rural landholdings of less than one acre, and 74% did not have a household electric connection. At the same time, the sample was highly ethnically heterogeneous: 25% of respondents came from a Scheduled Caste (former ‘untouchable’ caste) background, 40% came from an intermediate caste (OBC) background, 20% from an upper caste background, and 15% were Muslims. Most significantly, given India’s ranked ethnic systems, there were no significant differences in the income of different ethnic groups. Figure 5 shows how this contrasts with villages in the migrant’s region of origin, where economic and ethnic statuses are highly correlated. Supplement Part A contains detailed summary statistics and tests showing randomization successfully ensured balance across baseline characteristics for the different experimental subgroups.29 I now turn to the experimental results, which are pooled across Delhi and Lucknow, as the main results presented here did not differ across the study cities. 29
These tests include difference of means tests for each of 16 baseline characteristics,
and F-tests on the joint significance of all treatment variables, from a regression of each baseline characteristic on treatment dummies and chowk fixed effects.
25
---FIGURE 5 ABOUT HERE--Social Vignettes The survey results show ethnic differences divide migrants socially by inhibiting intra-class cooperation rather than exacerbating animosities generated by intra-class competition. I fail to reject H2, as respondents are not more likely to perceive wagecutting acts as threats if committed by ethnic outgroups (left panel of Figure 6). Posttreatment questions found no support for potential mechanisms suggested by my ethnography: respondents did not rate wage-cutters from ethnic outgroups as systematically less hardworking or more selfish (and therefore less likely to save money to remit home) than ingroup wage-cutters (Results in Table S.3 in supplement). ---INSERT FIGURE 6 ABOUT HERE--By contrast, ethnic differences do significantly diminish the prospect of intra-class cooperation (right panel of Figure 6). Overall, I found respondents were roughly 12 percentage points less likely to share housing with another migrant from a different jati, roughly a 34% decrease. The effect of caste group difference (11.7 percentage points) and religious difference (11.8 percentage points) were similarly large and robust (p
