31 | Findings from the 1921 Census on Caste and its Afterlives

By Nishitha Mandava, Gaurav Kalyani & Shivakumar Jolad

Published on: 13 Aug 2026

Introduction

By the 1921 census, the colonial state had now two decades of experience in caste enumeration. It recognised that caste was one of the most defining features of Indian society and continued to affect one’s life in every aspect, including marriage, education and so on. 

At the same time, it acknowledged that caste posed significant administrative and statistical difficulties. Despite these difficulties, the caste census of 1921 would assume great political significance with it being invoked during the deliberations of the Simon Commission and the Second Round Table Conference (RTC). 

People belonging to a variety of religions and castes utilised it to voice competing vociferous claims for their political representation.

The inclusion of caste in the census had become a subject of unrelenting debate. The Census Commissioner, J.T. Marten noted that there had been repeated proposals to omit the question on caste and discontinue the classification of the population by caste and tribe. Similar suggestions had been made during the 1901 census, the argument being that the distribution of castes changed only gradually and therefore did not require enumeration every ten years.

The issue resurfaced in 1920 when a resolution introduced in the Legislative Council criticised the caste enquiry for recognising and thereby perpetuating caste distinctions through official action. Critics also argued that the returns were unreliable because individuals from lower castes frequently claimed membership in higher-status groups for social mobility, rendering the statistics inaccurate.

Marten rejected these claims by distinguishing between the social consequences of caste and its value as an administrative category. Whatever one’s views on caste as a social institution, he emphasised, it remained impossible to conduct a meaningful discussion of India’s population and its society without taking caste into cognisance. In the report, he described caste as ‘the foundation of the Indian social fabric’ and ‘the best guide to the changes in the various social strata in the Indian society.’ (Marten, 1924).

He also pointed out that unlike Western societies where wealth, education and occupation increasingly shaped social differentiation, Indian society continued to be organised primarily through religion and caste. Since caste remained an all-encompassing feature of an individual’s official and social identity, he concluded that it needed to be taken into account during the census exercise.

Marten clarifies that enumerating caste did not mean defending or preserving the institution. Responding to the criticism that the census reinforced caste identities, the Commissioner cited the Mysore official Thyagarajaiyar, who observed that although caste appeared less visible among urban and reformist circles, it continued to organise everyday social life for the majority of the population (Marten, 1924). 

The 1921 census shifted its emphasis from constructing an ideal ossified classification of the caste system to managing the practical problems of recording it accurately. In doing so, the report reveals both the colonial state’s continued dependence on caste as a category of governance and its growing recognition of the difficulties involved in translating a complex social institution into statistical data. In this essay, we discuss some of the key findings from the 1921 census.

Obtaining Reliable Returns

Having justified the continuation of the caste enquiry, the report acknowledges the various troubles in obtaining reliable returns. Interestingly, it observes that public interest in caste improved rather than weakened the accuracy of the census, since special efforts had been made in 1901 and 1911 to prepare standardised lists of caste names and eliminate ambiguous terminology (Marten, 1924).

Enumerators conducting field-work were instructed to avoid vague descriptions, and instead record recognised caste names, while previous census records and the ethnographic survey were used to interpret doubtful entries. The report acknowledged that minor errors arising from copying, classification and local ignorance were inevitable but considered them relatively insignificant in the larger exercise (Marten, 1924).

The Commissioner identified a more critical source of error–deliberate misrepresentation of caste. Because one’s caste status wielded much more power than their age, occupation or birthplace in determining their socio-economic position in the society, many respondents believed that the census could influence or even fix their social status. This perception encouraged numerous communities, particularly those seeking upward mobility, to claim higher caste identities during enumeration.

For instance, in Punjab the creation of the ‘agricultural tribe’ category by the Punjab Alienation of Land Act of 1901, membership to it was almost indispensable for any purchase or sale of land. This provoked caste groups to manipulate and make claims to be considered agricultural and such efforts to manipulate official classifications continued to be documented as late as the 1931 census. Such lobbying efforts are evident from the fact that the groups counted as agricultural tribes kept increasing sharply (Cassan, 2010).

This turned the census into a site for political negotiation, where communities attempted to shape the official record in ways that reflected their aspirations for socio-economic advancement.

Caste Composition Among Hindus

The 1901 census occupies a foundational status in India’s census history as an exhaustive list of castes was prepared for the purpose of the ethnographic survey. In 1911, only certain castes were recorded, excluding those that fell below a certain percentage of the population. In the current census, further reductions were made. And even in the castes recorded, it does not give a faithful picture of the actual strength of the castes.

Marten (1924) qualifies that the interest in the 1921 census lies in,

The 1921 census recorded a total of 217 million population belonging to the Hindu religion and they were further classified in the following manner:

Table 1. Estimated composition of the Hindu population, India 1921

Here two features warrant our immediate attention. First, the Depressed Classes, which the report concedes, ‘has no final definition nor is it certain exactly whom it covers.’ (Marten, 1924).

Unlike caste, which was recorded directly from individual census returns, the category of the Depressed Classes had to be assembled from several sources. They relied on lists already prepared by the Education Department, where these communities had been identified for the purposes of educational assistance, together with estimates provided by provincial census superintendents.

The term Depressed Classes was already prevalent by this time, for instance, in 1920 B.R. Ambedkar had organised the All India Conference of the Depressed Classes in Nagpur (Ishita Banerjee-Dube, 2014). A detailed note on the evolution of the usage of this term is provided further below in the essay.

Not every province cooperated equally in census enumeration of these classes. The United Provinces, for example, declined to prepare such an estimate because, as Marten put it, ‘extreme delicacy of official sentiment shrank from facing the task.’ This remark suggests that colonial officials themselves recognised the political sensitivity involved in officially identifying and listing communities considered ‘untouchable.’

Despite these uncertainties, Marten estimated that approximately 53 million people belonged to the Depressed Classes, while suggesting that the actual figure was probably between 55 and 60 million because many communities had likely been omitted. He cautioned that these numbers should be treated as conservative estimates rather than precise statistics (Marten, 1924).

Marten described the Depressed Classes as communities ‘all of whom are considered impure,’ meaning that the defining criterion was not economic deprivation but ritual status within the Hindu caste order (Marten, 1924). In other words, the census understood the Depressed Classes primarily through the institution of untouchability rather than through poverty or occupation.

He adds that the disabilities experienced by these communities were much more prevalent in southern India, where, in his view, orthodox Hinduism had produced a more intense and elaborate system of social differentiation than in northern India. At the same time, he noted a paradoxical development: it was also in southern India that the Depressed Classes had begun to develop ‘class consciousness and class organisation.’ (Marten, 1924).

Christian missionary activity, government educational initiatives, and political reforms had helped these communities organise collectively, while constitutional reforms had begun to provide them with representation in provincial legislative assemblies.

The second were the hill and tribes people which the report describes them in the following manner: ‘the country opens out, these people leave their native forests and hills for the attraction of the plains, where they settle down and adopt the habits, language, and occupation of their Hindu or Hinduised neighbours.’

Since only about 9.5 million had returned to a tribal religion, ‘between 6 and 7 millions of more or less recently Hinduised tribes’ already sat inside the Hindu community (Marten, 1924).

Enumerating ‘Respectable’ People

In 1885, the British passed a legislation which limited the powers of the zamindari classes in Bengal, making it difficult for them to collect rents. One of the classes to be directly hit by this was the ‘bhadralok’ of Bengal–an exclusive group of upper-caste Hindus who made money from collecting rents. Their rent incomes decreased drastically and directly affected their status in society.

It was also around the same historical moment that English education began to attain prominence, especially as an avenue for a government job. The popularity of western education among the bhadralok reflected a new way of mobilisation under the colonial masters (Roy, 2015).

The 1921 census includes an estimate of the bhadralok (‘respectable people’) of Bengal as a distinct social category. The report associates it primarily with three upper-caste Hindu communities, Brahmins, Kayasthas and Baidyas, whose combined population was estimated at nearly three million. The census therefore treated the bhadralok as an educated middle class that was rooted in particular caste groups (Marten, 1924).

The bhadralok were concentrated in Calcutta, the adjoining districts of western Bengal, and parts of eastern Bengal such as Chittagong. It linked this distribution closely to literacy, especially English literacy, arguing that districts with larger bhadralok populations also supplied a disproportionate share of clerks and educated government employees.

The report suggested that the geographical distribution of this class also corresponded with the centres of earlier political movements, implying that education, colonial employment, and political activity were closely intertwined.

The most interesting part of the discussion, however, is the attempt to expound these regional differences through historical conquest. Quoting the Bengal Census Superintendent, Mr. Thompson, the report suggested that variations in the proportion of bhadralok Hindus reflected the history of Muslim expansion in Bengal.

It claimed that high concentrations of upper-caste Hindus in districts such as Bakarganj and Chittagong resulted from migration, as members of the ‘better class Hindus’ supposedly sought refuge in more remote areas during periods of Muslim political dominance (Marten, 1924).

Conversely, the relatively high proportions of bhadralok in Burdwan and Bankura reflected the fact that these districts had not been fully conquered by the Mughals in the same way as eastern and central Bengal.

The report also sought to explain the comparatively low proportion of bhadralok in northern Bengal. It argued that this could not simply be attributed to Mughal rule, since Muslim political authority had first become established around Gaur (near Malda), while northeastern Bengal was incorporated much later. 

Instead, Thompson speculated that these areas had remained outside the control of earlier Aryan settlers for longer than central and southern Bengal. Thus, differences in the modern distribution of educated upper-caste Hindus were explained through a combination of medieval Muslim rule and much older theories of Aryan settlement.

Communities beyond Hindus

The report notes that caste-like structures can be observed in religions beyond Hinduism and Sikhism. It intended to identify analogous forms of social stratification across religious communities, while also documenting the ways in which communal identities were changing under modern political conditions.

The Sikhs are described as ‘mostly recruited from the Hindu castes.’ This historical origin created a dilemma for census officials. Although many Sikhs continued to identify with particular castes, an increasing number rejected caste as incompatible with Sikh religious identity.

Following a resolution in the Legislative Assembly, enumerators were instructed not to insist on recording caste if Sikhs refused to state it. The results varied sharply by region. In Punjab, where caste identities remained widely acknowledged, only about 2% of Sikhs declined to record a caste. In the North-West Frontier Province, however, more than half, 54%, were enumerated without any caste designation (Marten, 1924).

Marten interpreted this as evidence of a growing collective Sikh identity, suggesting that ‘corporate feeling’ or communal affiliation was temporarily superseding narrower caste loyalties. But the report also cautioned that caste had by no means disappeared from Sikh society, as roughly one-third of the entire Jat population of Punjab identified as Sikh. This illustrated that caste continued to shape Sikh social composition even as some Sikhs rejected its official enumeration.

The report adopted a similarly nuanced approach toward Indian Muslims. It began by affirming the orthodox Islamic principle that ‘Islam recognises no caste distinctions.’ However, it immediately qualified this statement by cautioning that social practice differed considerably across regions. 

In the north-west, where the Muslim population was believed to have a stronger foreign ancestry, caste-like restrictions were said to be largely absent. Elsewhere, particularly in regions with large populations descended from Hindu converts, the census observed that ‘the influence of Hinduism has powerfully affected Muhammadan custom.’ (Marten, 1924).

Muslims were described as being divided into broad status groups such as Pathans, Moghuls, Saiyids and Sheikhs, alongside occupational communities including Boras, Khojas, Memons and Julahas. Although these groups lacked the formal religious sanction of Hindu caste, the report observed that they nevertheless practised endogamy. For example, ‘a Sheikh will not marry a Kulu’ in Bengal. 

Claims to higher status were often socially negotiated rather than historically verifiable. The census offered the telling observation that ‘a Sheikh is very apt to become a Saiyid when he achieves wealth and position,’ indicating that upward social mobility could be expressed through changes in communal identity (Marten, 1924). 

In contrast to the hereditary rigidity that the census associated with Hindu caste, Muslim status categories appeared more permeable.

The small domiciled European and Anglo-Indian were presented primarily through their occupational composition, with nearly two-thirds of adult males serving in the military, police or other branches of the Public Force. Because of this concentration in state service, the report regarded the European population as demographically ‘abnormal,’ noting its unusually low ratio of dependents to workers. 

Anglo-Indians, by contrast, were described as having a more typical family structure and a distinctive occupational profile. Nearly one-third of Anglo-Indian men worked in transport, especially the railways, while many others were employed as clerks and subordinate officials within the colonial administration (Marten, 1924). 

Rather than discussing them in terms of caste or social hierarchy, the census characterised these communities through their demographic structure and economic function.

Ambedkar on the 1921 Census 

The All-India Statutory Commission, also known as the Simon Commission, was appointed in 1927 to examine how the constitutional reforms introduced by the Government of India Act of 1919 were working and to recommend further constitutional changes. 

It was chaired by the British politician Sir John Simon and consisted of seven members, all of them British. The absence of any Indian member provoked widespread controversy towards the Simon Commission, leading almost all major political parties to boycott it. 

The arrival of the Commission’s members was met with country-wide strikes, black-flag demonstrations and resounding cries ‘Simon Go Back’ (Banerjee-Dube, 2014).

The leader of the Depressed Classes, B.R. Ambedkar presented evidence in front of this Commission to make a case for separate electorates for the Depressed Classes.

He proposed that ‘untouchables’ required separate electorates, that is, separate constituencies where they would be able to elect candidates from their castes to the legislatures. This is because universal adult suffrage was absent and instead only individuals with tax and property were allowed to vote which excluded much of the Depressed Classes (Banerjee-Dube, 2014). Hence, separate electorates would guarantee a representation of their voice. 

One of the pieces of evidence that Ambedkar analysed in front of the commission was the estimates of the Depressed Class made by the 1921 census. The Government of Bombay, drawing on the 1921 Census, had estimated their population in the Bombay Presidency at 1,478,390. 

In contrast, Ambedkar had given a figure of approximately 28 lakhs in his memorandum to the Commission. When questioned about this discrepancy, he explained that the figures used by the Bombay Government came from the Census Director’s calculations, whereas the figures in his memorandum were based on a different compilation of the population of the Depressed Classes (Ambedkar, 1928). 

He stressed that the census figures could not be regarded as exact because the tables from which they were derived did not include all the castes that he considered to belong to the Depressed Classes. Ambedkar specifically drew attention to the limitations of the Census tables, noting that the provincial figures were based on tables dealing with the ‘principal Indian castes.’ (Ambedkar, 1928).

According to him these calculations were not exhaustive. In Bombay, for instance, around ten castes that he regarded as undoubtedly belonging to the Depressed Classes were not included in the relevant tables. Hence, he urged that the smaller census figure should not be treated as the definitive population of the Depressed Classes.

When the Commission suggested that the actual number might lie somewhere between two and three million, Ambedkar accepted this as a possibility. This discussion directly connected to the question of representation. As the Commission observed, the larger figure would have implications for the extent of the special representation being demanded for the Depressed Classes.

The discussion also brought out the difficulty of defining who belonged to the ‘Depressed Classes.’ The Commission referred to the 1921 Census Director’s observation that the term had no final definition and asked whether it could include groups such as criminal tribes, hill tribes and other communities that stood outside the ordinary Hindu social hierarchy.

Ambedkar acknowledged that this was one possible interpretation, but distinguished these groups from the untouchable communities for whom he was seeking special political representation (Ambedkar, 1928).

When questioned about criminal tribes and aboriginal and hill tribes, he explained that there was insufficient social interaction between some of these groups and the main body of Hindus to determine whether they should be regarded as untouchable. He therefore did not propose to include them automatically within the category. 

Evolution of the term ‘Depressed Classes’

The term "depressed classes" originated as a borrowed English label used by Christian missionaries and colonial officials in the late 19th century to describe India's poorest and most marginalized groups. Between 1901 and 1917, the phrase functioned as a loose "catch-all" category that encompassed untouchables, aboriginal tribes, wandering communities, and even some classes of Muslims (Chandrachud, 2023).

This broad classification was adopted by prominent figures such as Gokhale, Annie Besant, and eventually Mahatma Gandhi, who initially viewed the condition of these groups as a social rather than a political issue. However, the 1918 Montagu–Chelmsford reforms fundamentally shifted this perspective by promising legislative representation, which necessitated a more precise administrative definition to determine exactly who should be counted (Chandrachud, 2023).

The 1921 Census served as the definitive pivot in this evolution, officially shifting the criterion for the depressed classes from general poverty to "ritual impurity". Although the census acknowledged a lack of prior definitional clarity, it effectively established untouchability as the standard for enumeration (Hutton, 1933).

As mentioned previously in this essay, Marten estimated the total population of these classes at between 55 and 60 million, with 52.7 million actually recorded. While the 1921 exercise was successful in fixing a specific ritual criterion for the Depressed Classes, it didn’t produce a stable count. Wildly varying estimates from contemporary groups like the Hartog and Franchise Committees demonstrated that a reliable statistical consensus remained elusive (Hutton, 1933).

Impact of 1921 census on 2nd Roundtable Conference

The significance of the 1921 census outlived its immediate administrative purpose. A decade later, during the deliberations of the Second RTC in 1931, Sir Geoffrey Corbett, the Secretary to the British Indian delegation of the Second RTC, noted in his memorandum on the Punjab, ‘the figures of the 1921 census have been taken, because the district communal figures of the 1931 census are not yet available’ (Secretary of State for India, 1932). Although the 1931 Census had been conducted, only the provincial totals had been published by the time the conference met.

The proceedings of the Minorities Committee during the Conference reveal how census statistics had become central for voicing competing claims over political representation. Muslims justified their demand for a majority in the Punjab legislature on the grounds that they constituted a numerical majority in the province, while Sikhs and Hindus invoked the same demographic data to urge for safeguards against permanent Muslim dominance. 

Here we see that census data became the basis for framing political rights, and legislative representation. Each new census possessed the potential to alter the constitutional balance by changing the officially recognised proportions of India's socio-religious communities.

British Prime Minister, Ramsay Macdonald announced the Communal Award (1932) which granted separate electorates for Hindus, Muslims and ‘untouchables’ for the new federal legislatures. However, M.K. Gandhi began demanding substituting separate electorates with joint electorates consisting of reserved seats.

Gandhi’s political influence within the Indian nationalist movement placed considerable pressure on B.R. Ambedkar, the leader of the Depressed Classes, who accepted his demands, as a compromise. This resulted in the Poona Pact (1932) and later the Government of India Act (1935) followed it.

This example still serves as an instance to show how census data had acquired political significance and became an instrument by different interest groups beyond the state to make claims for their rights and representation.

(Authors: Nishitha Mandava, is an independent researcher and research consultant for Center for Legislative Education and Research, FLAME University, Pune;

Gaurav Kalyani works as Research Associate for the India State Stories Project at the Center for Legislative Education and Research, FLAME University, Pune;

Shivakumar Jolad works as Associate Professor (Public Policy), and is the Chair of Center for Legislative Education and Research and Director India State Stories, FLAME University, Pune)

(Author Contributions: Nishitha did research and primary writing, Gaurav contributed to writing and edited the article, Shivakumar contributed to conceptualization, research, data visualization)

Bibliography

Ambedkar, B. R. (1980). Evidence before the Indian Statutory Commission on 23rd October 1928. In V. Moon (Ed.), Dr. Babasaheb Ambedkar: Writings and speeches (Vol. 2, pp. 459-485). Education Department, Government of Maharashtra.

Banerjee-Dube, I. (2014). A History of Modern India: (1st edn). Cambridge University Press.

Cassan, G. (2010). British law and caste identity manipulation in colonial India: The Punjab Alienation of Land Act [Working paper].

Chandrachud, A. (2023). These Seats Are Reserved - Caste, Quotas and the Constitution of India. Penguin Random House India.

Hutton, J. H. (1933). Census of India, 1931, Vol.1-India, Part I-Report, Appendix I - Exterior Castes. Manager of Publications, Delhi.

Marten, J. T. (1924). Census of India, 1921: Volume I, India — Part I, Report. Calcutta: Superintendent Government Printing, India.

Roy, P. (2015). Bhadralok/Bhadramahila. In G. Dharampal-Frick, M. Kirloskar-Steinbach, R. Dwyer, & J. Phalkey (Eds.), Key Concepts in Modern Indian Studies (pp. 21–23). NYU Press. http://www.jstor.org/stable/j.ctt15zc7zj.13

Secretary of State for India. (1932). Indian Round Table Conference (Second Session), 7th September 1931–1st December 1931: Proceedings (Cmd. 3997). His Majesty's Stationery Office.