Residential segregation in urban India and persistence of caste

  • Blog Post Date 01 July, 2020
  • Articles
  • Print Page
Author Image

Naveen Bharathi

University of Pennsylvania and Harvard University


Author Image

Deepak Malghan

Indian Institute of Management Bangalore


Author Image

Sumit Mishra

Institute for Financial Management and Research


Author Image

Andaleeb Rahman

Cornell University


B.R. Ambedkar had exhorted lower-caste people to move towards cities to defy localism and benefit from the virtues of cosmopolitanism that urbanisation might provide. Using 2011 enumeration block-level Census data for five major cities in India - Bengaluru, Chennai, Delhi, Kolkata, and Mumbai - this article finds that not only are Indian cities highly segregated, but population size seems to have no association with the extent of segregation. In fact, the largest cities are some of the most segregated.

Across urban India, housing discrimination based on caste or religious identity is commonplace (Thorat et al. 2015). How widespread is such discrimination that is often attributed to apparently benign 'cultural preferences'? Anecdotal evidence about neighbourhood segregation in urban India suggests that modernisation and urbanisation have not been able to make a dent in one of the constitutive features of India’s traditional caste society the complete segregation of residential space. However, there is little systematic country-wide data on actual patterns of neighbourhood-scale residential segregation in India. Are bigger cities likely to be less segregated than smaller towns? Are more diverse cities and towns less segregated?

Evidence from around the world has shown how residential segregation leads to a widening of social distance between groups, thereby reinforcing historical hierarchies and social prejudices. In the US, segregation of African Americans and other ethnic minority in cities has generated a rich vein of scholarship, which argues that racial segregation is inimical to the development process as it alienates the marginalised communities  socially and economically (Cutler and Glaeser 1997). More recently, Chetty et al. (2014) have shown that high segregation lowers the chances of social and economic mobility.

"Segregation of spaces" (Ghurye 1969) is central to caste-based social and economic marginalisation of vast swathes of India. In many parts of India, village habitations continue to be segregated as caste-based hamlets, with ‘lower’ caste groups occupying the spatial periphery. This spatial hierarchy has been central to the maintenance of social hierarchy, and also to regulation of differential access to public goods, such as drinking water (Mukherjee 1968). It was against such discriminatory practices of the social life in villages, that B.R. Ambedkar, the founding father of Indian Constitution, advocated greater migration to urban areas for the marginalised caste groups, Dalits. The promise of migration to urban areas has a major presumption inherent in it  the anonymity provided by the city shall mute the historical baggage of caste identity, replacing it with ‘class’ distinctions instead (Beteille 1997). Caste, therefore, would cease to control the spatial organisation of spaces as it does in a village (Swallow 1982). Prominent public intellectuals continue to advocate urbanisation as a panacea for caste-based spatial segregation as it is believed that ‘caste is losing, and will continue to lose, its strength’ as India urbanises (Prasad 2010). Is the urbanisation experience of modern India consistent with these prescriptions? Is caste-based spatial segregation decreasing as India rapidly urbanises?

Limitations of using wards as the spatial measure of ‘neighbourhood’ while studying segregation

Between the decennial Census counts of 2001 and 2011, Dalit population in urban India has increased by 40%.1 How have these historically marginalised and formerly ‘untouchable’ groups assimilated in Indian’s burgeoning urban centres? To answer this question, we need a fine-grained neighbourhood-scale analysis of spatial segregation patters  the distribution of social groups across urban neighbourhoods. However, until recently neighbourhood-scale data were not available for a systematic analysis of residential segregation patterns in urbanising India. Historically, the Census of India reports caste information as three broad aggregate categories  Scheduled Caste (SC), Scheduled Tribe (ST), and the residual Others (OTH), at the ward level. Size of a ward in urban areas is sufficiently large, and average population sizes vary across them. For example, population size of a ward could be between 1,500-6,000 in smaller towns, and 30,000-200,000 people in the larger metropolitan cities (Prasad 2006). Ward, therefore, is not the most useful spatial unit of analysis to study segregation. Yet, most of the segregation studies in India  limited by availability of administrative data at finer spatial resolutions  rely on this coarse data (Dupont 2004, Sidhwani 2015, Vithayathil and Singh 2012).

Measuring neighbourhood-scale segregation in urban India

We explicate this spatial resolution problem in our recent research (Bharathi et al. 2019), and show why wards may not be the most useful spatial unit of analysis. One needs to go to finer geographic scales, such as a Census enumeration block, which we advocate as a better proxy for a neighbourhood. An enumeration block (EB), on an average, contains around 100-125 households with a total population of 650-700.2 The EB is therefore a more realistic approximation of what constitutes a ‘neighbourhood’.

Figure 1. Variation of SC+ST population within wards in Bengaluru

Consider Figure 1, where we provide a visual representation of the EB-level population shares of SC/STs within various wards of Bengaluru.3 It is not difficult to observe that, EBs within a particular ward, are markedly heterogeneous in terms of their caste composition. There are many EBs (finer lines) with very few SC/STs. At the same time, there are clusters of substantial SC/ST population within a ward (dotted borders), which is largely inhabited by the OTH category people. The point we are trying to impress upon here is the following: when population clustering takes place at a micro-level, communities might be highly segregated even within a ward, which is diverse in terms of caste composition. Spatial segregation in urban areas, therefore, should be studied at finer geographic scales, say a street.

Using EBs as our unit of analysis, we calculate caste-based segregation for the five major Indian metropolitan cities of India  Chennai, Delhi, Kolkata, Mumbai, and Bengaluru. This is the first ever attempt to study caste-based residential segregation in Indian cities using the finest available spatial scale, the EBs.

Segregation across five major Indian metropolitan cities

We use a simple 'dissimilarity index’ to measure residential segregation. The dissimilarity metric captures the degree of ‘evenness’ of a given geographical unit compared to a more aggregated one, and represents the proportion of population that has to be moved to achieve perfect evenness. With Indian national Census data, the dissimilarity index measures how population shares (SC, ST, and OTH) at the EB or ward level are different from the larger spatial aggregates, at ward or city level, respectively.The index, D, varies between 0 and 1, with the zero indicating complete integration of the groups, while 1 represents the case of extreme segregations. In Table 1, we report the segregation metric for the Indian metropolitan cities.

Table 1. Patterns of segregation: Dissimilarity index


D (Ward-City)

D (Block-City)

Median D (Block-Ward)

Pop. Wtd. Mean D (Block-Ward)


























Note: ‘Pop. Wtd. Mean’ refers to population-weighted mean (average).

The dissimilarity index, as computed traditionally by Vithayathil and Singh (2012), at the ward-city level is reported in column 1 to benchmark the metric at the finer scale of EB. The second column reports the dissimilarity index computed at the block-city level  a measure of how the caste composition of EBs in a city are different from that of the city as a whole. The numbers in the second column are substantially larger than the first one. The comparison between these two columns illustrates how ignoring intra-ward segregation amounts to neglect of a significant portion of segregation in a city.

The last two columns provide a direct measure of intra-ward segregation, where for each of the wards, we computed how caste compositions vary across EBs within it. Again, it is apparent that there is substantial heterogeneity within the wards regardless of the population weights used. Ranking of these cities in terms of segregation, however, remains unchanged, regardless of the metrics employed. Among the five cities, Kolkata is the most segregated, while Bengaluru is the least. There could be several reasons for why these cities have different levels of segregation, and its potential impact, which remains an open question of further research.

Now that we know how segregated each of the EBs and wards in a city are, it is instructive to ask if there is a pattern to it. Is there a relationship between the size of a ward and the level of segregation? Do larger wards exhibit greater segregation? We, however, do not find such a relationship to hold. In Figure 2, we plot this association for each of the cities, and no clear pattern emerges.

Figure 2. Relationship between ward size and block-ward dissimilarity index

Promise of urbanisation: A look at India’s 147 largest cities

Continuing along this line of inquiry, we have extended our study to include all 147 cities in India that have a population of at least 0.3 million residents, as per the Census of 2011. We rank these 147 cities, based upon their population size, and look at the association between population size and the degree of segregation.5 We assume that larger cities are the ones that exhibit greater economic dynamism, and are more cosmopolitan and egalitarian in nature. As a result, they would be less discriminatory spaces, as the modernisation argument would suggest. On the contrary, we find that there is no association between segregation and city size for these 147 cities (Figure 3). The six metropolitan cities, including Hyderabad, are outliers, but in terms of greater segregation.

Micro-segregation, even in diverse wards and cities

Are more diverse wards less segregated across these cities? This is an important question because greater diversity of the larger spatial aggregate implies greater chances of group interactions, and a possible sense of community feeling. We, however, do not find this hypothesis to hold. Greater diversity, be it across these 147 cities, or the wards that they contain, has no association with the level of segregation (Figure 4).6 This highlights the extent of micro-segregation, even within diverse cities and wards, which is often invisible to us, and what we term as ‘fractal urbanism’ (Bharathi et al. 2020).

Figure 3: City size and segregation

Figure 4. Diversity and segregation

What are the implications?

Our work provides two analytical insights. First, what should be the ideal spatial measure of a neighbourhood in segregation analyses? In the context of the US, studies on residential segregation typically rely on the use of a ‘census tract’ as the preferred spatial unit of analysis (Iceland and Weinberg 2002, Massey and Denton 1987). The same census tracts continue to be the spatial unit of analysis, over time, allowing researchers and planners to study the pattern of urbanisation and its evolution. If EBs, are to be considered an ideal neighborhood, it is imperative for the policymakers to release data at such finer levels in the subsequent rounds too. In addition, to understand how segregated spaces affect economic outcomes, it also becomes important to have such information on other development indicators.

Our second contribution is a rather sobering commentary on India’s urbanisation process and spatial discrimination - not only are cities highly segregated, but population size seems to have no association with the extent of segregation. In fact, the largest cities are more segregated. This finding challenges the bedrock normative premise of urbanisation  the dilution of caste boundaries in Indian cities. Ambedkar’s exhortation to the lower-caste people to move towards cities to defy localism and benefit from the virtues of cosmopolitanism as urbanisation might provide, still remains a mirage 70 years after the Indian Constitution came into being.


  1. This migration of Dalit groups has also been recorded in earlier census periods. Mahars of Maharashtra are a particularly good example of a former ‘untouchable’ group migrating in large numbers into cities like Bombay and Nagpur at the beginning of the 20th century in order to ‘escape’ caste hegemony (Rao 2009).
  2. The Census of India, released for the first time in 2015, data on caste below the level of a ward for urban areas.
  3. We have the geo-referenced EB-level data for the city of Bengaluru  the only such map for any Indian city, to the best of our knowledge.
  4. It is computed as , where Si=SC+ST population in ith block/ward, and ri represents the rest of the population (OTH) in ith block/ward. Similarly, S and R stand for total population of SC+ST and OTH respectively in the ward/city. We have had to combine the two distinct categories  SCs and STs  because according to the Census definitions, Delhi has no tribal population.
  5. Segregation, D, is the median ward-EB difference for all the 147 cities.
  6. We use ‘fractionalisation metric’, the most standard measure, to compute diversity. It ranges between 0 and 1, and represents the probability that two different individuals picked at random would belong to two different communities, here SC/ST and OTH.

Further Reading

  • Beteille, Andre (1997), "India’s middle-class", Internationale Politik, 52(3):17-20.
  • Bharathi, Naveen, Deepak Malghan and Andaleeb Rahman (2019), "Neighbourhood-scale Residential Segregation in Indian Metros", Economic and Political Weekly, 54(30):64-70.
  • Bharathi, Naveen, Deepak Malghan, Sumit Mishra and Andaleeb Rahman (2020), "Fractal Urbanism: City Size and Residential Segregation in India", SocArXiv Papers, 9 June 2020.
  • Chetty, Raj, Nathaniel Hendren, Patrick Kline and Emmanuel Saez (2014), "Where is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States", The Quarterly Journal of Economics, 129(4):1553-1623.
  • Cutler, David M and Edward L Glaeser (1997), "Are Ghettos Good or Bad?", Quarterly Journal of Economics, 112:827-872.
  • Dupont, Véronique (2004), "Socio-spatial differentiation and residential segregation in Delhi: a question of scale?", Geoforum, 35(2):157-175.
  • Ghurye, GS (1969), Caste and race in India, Popular Prakashan, Bombay.
  • Iceland, J and DH Weinberg (2002), ‘Racial and ethnic residential segregation in the United States 1980-2000’, Bureau of Census, August 2002.
  • Massey, Douglas S and Nancy A Denton (1987), "Trends in the residential segregation of Blacks, Hispanics, and Asians: 1970-1980", American Sociological Review, 802-825.
  • Mukherjee, R (1968), The Way of Humanism, Academic Books.
  • Prasad, RN (2006), Urban local self-government in India, Mittal Publications.
  • Prasad, CB (2010), ‘New Order’, Himal South Asian, 1 April 2010.
  • Rao, A (2009), The Caste Question: Dalits and the Politics of Modern India, University of California Press.
  • Sidhwani, Pranav (2015), "Spatial inequalities in big Indian cities", Economic and Political Weekly, 50(22):55.
  • Swallow, DA (1982), "Ashes and Powers: Myth, Rite and Miracle in an Indian God-Man’s Cult", Modern Asian Studies, 16(1):123-158.
  • Thorat, Sukhadeo, Anuradha Banerjee, Vinod K Mishra and Firdaus Rizvi (2015), "Urban Rental Housing Market", Economic and Political Weekly, 27:47-53.
  • Vithayathil, Trina and Gayatri Singh (2012), "Spaces of Discrimination", Economic and Political Weekly, 47(37):60-66.
No comments yet
Join the conversation
Captcha Captcha Reload

Comments will be held for moderation. Your contact information will not be made public.

Related content

Sign up to our newsletter