Every argument about reservation in India eventually runs into a missing number. The Other Backward Classes hold 27 percent of central government posts and central educational seats under the settlement that followed the Mandal Commission, yet no census since 1931 has counted how large a share of the population the backward classes form, and the 27 percent itself was fixed in 1980 by arithmetic performed on a census taken in 1931. The caste census debate is the argument about whether the Indian state should repair that missing number, and the claim defended in this article is that the debate operates through three substantive positions on demographic and political evidence: a demographic-justice position that wants reservation rationalised against counted population shares, an architectural-stability position that defends the layered compromises through which reservation has been built since 1932, and a federal-implementation position that holds caste counting to be legitimate work for state governments rather than for the decennial census of the Union. Each position answers a different question, each rests on a different body of evidence, and the state surveys published in Bihar in October 2023, in Karnataka through the Hegde Commission report of April 2025, and in Telangana after the survey of November 2024 test all three at once.

The standard commentary on the caste census debate sorts the participants by party. One side of that commentary describes a demand raised by the Congress and its allies in the INDIA coalition after 2021, pressed by Rahul Gandhi in campaign speech after campaign speech, and answered by a Bharatiya Janata Party that first resisted the demand in affidavits and parliamentary replies and then, on 30 April 2025, accepted caste enumeration inside the decennial census through a decision of the Cabinet Committee on Political Affairs. That telling is accurate as chronology and empty as analysis. A second standard telling sorts the material by dataset: the Socio Economic and Caste Census of 2011, whose caste returns were collected and never published; the Bihar survey of 2023, which published its tables within ten months of fieldwork; the Karnataka enumeration whose compiled report waited a decade for a cabinet table. That telling is accurate as bibliography and equally empty, because a dataset acquires force only inside an argument about what counting is for. This article follows neither sorting. It reconstructs the three substantive positions in the strongest form their own advocates and scholars have given them, tests each against the documentary record from the colonial census volumes of 1871 to 1931 through the unpublished SECC returns to the state survey reports of 2023 to 2025, and closes the pillar on caste, reservation and Mandal politics by showing how the debate gathers every earlier argument in the pillar into a single question: what does the republic owe to a number it has declined to collect for ninety years?

The Colonial Count: Caste in the Census from 1871 to 1931

Caste entered Indian governance as a counted thing. The first attempt at an all-India count, the census of 1871 to 1872, was non-synchronous, assembled province by province, and already organised around caste as a primary classification of the population. The decennial series that began with the synchronous census of 1881, conducted under Census Commissioner W. C. Plowden, treated caste as the skeleton of Indian society and printed caste tables for every province. Bernard Coady and later historians of the census, among them Padmanabh Samarendra in his studies of census categories, documented how the act of enumeration forced fluid, local, occupation-linked identities into fixed printed columns; a name spoken to an enumerator became a row in a provincial table, and the row hardened into an administrative fact that later petitioners cited back to the state. Communities responded to the printed table as a prize. The census decades produced floods of memorials from caste associations demanding higher varna placement, a literature that Bernard Cohn analysed in his 1987 essay on the census and social classification, and the demand itself proved that Indians understood enumeration as a ranking exercise long before independence.

The ranking became explicit in 1901. Census Commissioner H. H. Risley, whose ethnographic theories treated caste as a racial hierarchy, directed that castes be classified by social precedence as recognised by native public opinion, an instruction that converted the census schedule into an official ladder of honour. Risley’s 1901 classification provoked organised protest, memorials and counter-memorials, and it fixed the template that every later caste count has had to answer: a state list of castes is never only a list, because its order and its category boundaries distribute recognition. The census of 1911 under E. A. Gait retreated from Risley’s explicit precedence ordering, and the census of 1921 under J. T. Marten wrestled with returns in which religious conversion, sanskritising name changes and enumerator guesswork produced figures that the commissioners themselves distrusted in print. Each decennial volume contained a chapter of caveats, and each caveat is evidence in the present debate, because the counter-case against caste enumeration has quoted the colonial commissioners’ own doubts since 1951.

When was caste last counted across all of India?

The census of 1931, conducted under Census Commissioner J. H. Hutton, was the last enumeration to record and publish the caste of every person counted in British India. Hutton’s published tables supplied the demographic base from which every later estimate of backward-class population was extrapolated. No later all-India exercise has repeated that count.

The Mandal Commission’s 52 percent figure of 1980, the most consequential population estimate in the republic’s distributive politics, descends directly from Hutton’s tables, adjusted forward across five decades. Every argument about the age of India’s caste data begins at this date.

Hutton’s volume deserves attention because both camps in the present debate claim him. The demographic-justice camp points to the completeness of the 1931 operation: a colonial state with paper schedules and no computing machinery counted caste for some 350 million people and printed the results. The architectural-stability camp points to Hutton’s own misgivings, recorded in his report and in his 1946 book Caste in India, about the reliability of returns in which the same community appeared under dozens of names and in which status claims inflated some counts and deflated others. Both readings are faithful to the record. The 1931 census proved that caste could be counted at continental scale, and its commissioner warned in the same volumes that what had been counted was partly an artefact of the counting. A serious treatment of the caste census debate has to hold those two truths together, because the state surveys of 2023 to 2025 reproduced both of them: Bihar’s enumerators coded the raw returns into just over two hundred categories, and Karnataka’s compiled tables carried 1,351 categories, a difference that is itself a finding about how enumeration manufactures the object it measures.

The count of 1941 closed the colonial sequence in farce and warning. Census Commissioner M. W. M. Yeatts conducted the 1941 enumeration under wartime economy orders; caste was recorded on the schedules, and the tabulation of caste was then abandoned to save money and paper as the Second World War consumed the administration. The returns were never processed into published tables. The episode matters beyond antiquarian interest. It established, six years before independence, the pattern that independent India would repeat with the SECC in 2011: the Indian state proved capable of collecting caste data and then declining, for reasons presented as administrative, to convert the collection into public knowledge. The 1941 schedules mouldered unpublished. When the Constituent Assembly’s generation took charge of the census, the memory of 1941 and the caveats of 1931 sat on the table beside the ambition to build a casteless citizenship, and the decision of 1951 chose between them.

The Decision of 1951: Enumeration Stopped, and the Reasoning Left on Record

The census of 1951, the first of independent India, was conducted under Census Commissioner and Registrar General R. A. Gopalaswami under the Census Act of 1948. Its schedules asked for caste only where the Constitution required it: members of the Scheduled Castes and Scheduled Tribes were enumerated as such, because Articles 330 and 332 reserved legislative seats for them and Article 341 and 342 empowered the President to specify the lists. For the rest of the population the caste question was dropped. Gopalaswami’s general report recorded the policy and its rationale in terms that the architectural-stability position still quotes: the government of the new republic held that a general caste count would perpetuate the divisions the Constitution sought to dissolve, that the returns of 1931 had been corrupted by organised status-claiming, and that a state committed to equality of citizenship should not sort its citizens by birth in its master statistical instrument. Prime Minister Jawaharlal Nehru’s government treated the omission as a founding choice, of a piece with the decision to build the republic on individual citizenship rather than corporate community, and every decennial census from 1961 to 2011 repeated the omission without a parliamentary vote ever being taken to reverse it.

The omission coexisted, from the start, with an elaborate caste machinery elsewhere in the state. The Constitution’s own text required caste lists for the Scheduled Castes and Scheduled Tribes. Article 340 authorised commissions on the backward classes, and the first such commission, chaired by Kaka Kalelkar and appointed on 29 January 1953, reported in 1955 with lists built substantially on the census of 1931 and on state records of uneven quality. The Kalelkar Commission’s experience, examined in the account of the first Backward Classes Commission, demonstrated the cost of the 1951 decision within four years of its taking: a state that refused to count caste in its census found itself identifying backwardness from a twenty-year-old colonial table supplemented by testimony. The southern and western states filled the gap with their own commissions and their own lists, so that by the 1960s India possessed a dense archipelago of state-level caste knowledge floating on an empty national sea. That archipelago is the factual foundation of the federal-implementation position: caste knowledge in India has always been produced by states, because the Union chose not to produce it.

The deferral hardened into doctrine through repetition. The census of 1961 under Asok Mitra refined economic and linguistic tabulation and left caste aside. The census of 1971 and 1981 did the same. When the Mandal Commission was appointed on 1 January 1979, its members confronted the deferral as an operational problem, and the method they built in response, reconstructed in the reading of the Mandal Commission’s methodology, was an instrument for estimating a population they were forbidden by circumstance to count. Their 52 percent figure for the Other Backward Classes was produced by subtracting the counted and estimated shares of forward castes, Scheduled Castes, Scheduled Tribes and non-Hindu communities from the population recorded in 1931, adjusted forward by half a century of growth. The figure carried the authority of a commission and the fragility of a subtraction, and both qualities passed into the politics of 1990, when Prime Minister V. P. Singh’s government implemented the 27 percent recommendation on 7 August 1990 and the streets filled with protesters who could ask, with documentary justice, in what year the beneficiaries had last been counted. The answer, 1931, was fifty-nine years old.

What did the census of 1941 do with its caste returns?

The census of 1941 recorded caste on its household schedules and then, under wartime economy orders issued to Census Commissioner M. W. M. Yeatts, abandoned the tabulation of those returns. The caste data were never processed into published tables, and the schedules passed unpublished into the archives.

The episode left a standing precedent in the Indian statistical record: collection without publication. The Socio Economic and Caste Census of 2011 repeated the pattern seventy years later, at far greater expense and with the same official silence at the end of it.

The courts absorbed the deferral into doctrine in 1992. The nine-judge bench in Indra Sawhney v. Union of India, decided on 16 November 1992, upheld the 27 percent reservation for Other Backward Classes in central employment, imposed the 50 percent ceiling on total reservation and required the exclusion of the creamy layer, and it did so while accepting identification of backward classes substantially through caste, working from lists descended from 1931 and from state compilations. The judgment’s majority treated the absence of a fresh caste count as a regrettable gap rather than a constitutional defect, and governments of every composition took the hint: the lists could be administered, litigated and trimmed indefinitely without the underlying population ever being re-measured. Between 1951 and 2010 the demand for a caste count surfaced in Parliament at intervals and was each time deferred by the government of the day, usually on the double ground that the census was a statistical exercise unsuited to a politically charged question and that enumeration would harden the identities it recorded. Both grounds were arguments, not findings, and the scholars who kept the demographic-justice position alive through those decades, among them sociologist Satish Deshpande in essays collected around his 2013 writings on caste and census, answered them on their merits: a state that distributes a quarter of its jobs by caste while refusing to count caste has not transcended caste, it has merely exempted its own allocations from evidence.

Counted and Withheld: The SECC of 2011 and the Politics of an Unpublished Table

The deferral was nearly broken in 2010. In the parliamentary debates of April and May 2010, members across parties, including senior figures of the ruling Congress and the opposition, demanded that the census of 2011 carry a caste question, and Prime Minister Manmohan Singh assured the Lok Sabha that the Cabinet would consider the matter. A Group of Ministers deliberated through the summer of 2010. The Cabinet’s decision, taken in September 2010, split the difference in a way that produced the strangest artefact in Indian statistics: caste would be enumerated, but not by the census and not by the Registrar General. A separate exercise, the Socio Economic and Caste Census, would be run through the Ministry of Rural Development and the state governments, using handheld electronic devices, and its caste returns would be processed outside the census machinery whose statutory protections and tabulation discipline had handled every count since 1881. The compromise satisfied the demand’s letter and denied its substance, because the demand, as its parliamentary advocates stated it, was for caste data with the authority of the census behind it.

The SECC was conducted through 2011 and 2012 across rural and urban India. Its socio-economic returns were published in July 2015 by Finance Minister Arun Jaitley, and they remain a working source on deprivation, housing and occupation. Its caste returns were never published at all. The reason given by the government of the day, in statements by ministers and officials between 2015 and 2018, was that the raw returns contained an unusable profusion of names: respondents had returned caste, sub-caste, surname, gotra and clan descriptions in a volume reported at some 46 lakh distinct entries, and no tabulation could be certified until the entries were coded into a finite list of communities, a task the government assigned first to an expert process under NITI Aayog vice-chairman Arvind Panagariya and then allowed to lapse. The explanation was technically true and politically incomplete. Coding 46 lakh raw strings into a master list is exactly the work census offices perform on occupation and mother-tongue returns in every decennial round; Bihar’s administration performed the equivalent work on its 2023 returns in months. What the SECC episode demonstrated was not that caste returns cannot be tabulated but that a government which does not wish to publish them can always find the tabulation unfinished. The Registrar General’s absence from the exercise, a design choice of September 2010, had removed the one institution professionally obliged to finish.

The afterlife of the SECC sharpened every position in the debate. In September 2021 the Union government filed an affidavit in the Supreme Court, in proceedings arising from Maharashtra’s demand for SECC data to support reservation for Marathas in local bodies, stating that the SECC caste data were flawed, riddled with error and unusable for policy, and that caste enumeration in the forthcoming census was not proposed. The affidavit converted a lapse into a doctrine: the state argued on oath not merely that it had not published a count but that the count it held could not be trusted and that no fresh count should be taken. The demographic-justice camp received the affidavit as documentary proof of its central charge, that the absence of caste data was a policy preference dressed as a statistical finding. The architectural-stability camp read the same document as a late official confession of the enumeration-error problem that Hutton had flagged in 1931. Yogendra Yadav, whose essays in Economic and Political Weekly and in the press through 2021 to 2024 form the most sustained scholarly brief for enumeration, answered the affidavit by separating the two questions it fused: the defects of one badly designed exercise conducted outside the census machinery say nothing about the capacity of the census itself, an institution that tabulates hundreds of mother tongues and thousands of occupations without pleading impossibility. Christophe Jaffrelot, in his analyses of the Bihar survey published from October 2023, made the complementary point from the other end: a state government with a fraction of the Union’s statistical resources had designed, fielded and published a caste count in twenty-one months, which relocated the argument from capacity to will.

The state surveys did not emerge from a vacuum, and their legal footing matters for the federal-implementation position. Census is entry 69 of the Union List in the Seventh Schedule, and Article 246 gives Parliament exclusive competence over it; the Census Act of 1948 governs the decennial operation. A state government cannot conduct a census. It can, however, conduct a survey under its own executive and statistical powers, gathering information for the administration of subjects on the State and Concurrent Lists, including the identification of backward classes for state services and state educational institutions under Articles 15(4) and 16(4). When Bihar’s survey was challenged, the Patna High Court upheld it on 1 August 2023 as a survey within state competence, and the Supreme Court declined to stay its publication, a sequence that converted the federal-implementation position from a debating argument into a litigated and surviving practice. Karnataka’s enumeration had a different pedigree: commissioned in 2015 by the government of Siddaramaiah and executed through the Karnataka State Commission for Backward Classes under H. Kantharaj, its returns were compiled and reviewed over the following decade, and the report associated with Justice K. Jayaprakash Hegde, who chaired the Commission in its later phase, was placed before the state cabinet in April 2025. Telangana’s government under Chief Minister A. Revanth Reddy launched its comprehensive household survey in November 2024 and placed its summary findings before the state assembly in February 2025. Three states, three instruments, three sets of tables: the federal laboratory produced, within twenty months, more published caste data than the Union had released in ninety years.

The Three Positions Mapped: Arguments, Advocates and Tests

The debate’s three substantive positions are set out in Table 1, the three-positions debate map, which names each position’s core claim, its strongest evidence, its characteristic advocates and the test that the state surveys and the census decisions of 2025 apply to it. The map is the article’s findable artifact, and the sections that follow argue each row at the strength its own holders give it.

Table 1: The three-positions debate map of the Indian caste census debate.

Position	Core claim	Strongest evidence	Named advocates and holders	Test applied by events to 2025
Demographic justice	Reservation shares should be revised against counted population shares; a state that allocates by caste must count caste	Mandal’s 52 percent rests on the 1931 census; Bihar’s survey counted backward classes at 63.14 percent against a 27 percent central quantum	Yogendra Yadav’s essays of 2021 to 2024; Satish Deshpande’s writings on caste and census; the Congress and INDIA coalition demand from 2021; the Bihar government of Nitish Kumar	Bihar raised its quota law in November 2023 and the Patna High Court struck the increase down in June 2024, so data alone did not move the ceiling
Architectural stability	The reservation system is a layered settlement of compromises; reopening its quantities against fresh counts invites conflict the settlement was built to contain	The Poona Pact of 1932, the constitutional framework of 1950 and the Indra Sawhney ceiling of 1992 each traded precision for durability	André Béteille’s published arguments against caste enumeration; the Union government’s Supreme Court affidavit of September 2021; defenders of the 50 percent ceiling across parties	The Cabinet decision of 30 April 2025 accepted enumeration while promising continuity of the existing architecture, absorbing the position’s caution into the count itself
Federal implementation	Caste categories are state-specific, so counting belongs with the states that administer state lists; national enumeration is neither required nor well designed for the task	State lists of backward classes differ from the central list and from each other; Bihar, Karnataka and Telangana all fielded surveys under state competence upheld at Patna in August 2023	Federalism scholars writing on the state surveys; the governments of Bihar, Karnataka and Telangana as practitioners; commentators who opposed a Union count while defending state counts	The Union’s April 2025 decision nationalised the count anyway, leaving state surveys as the methodological rehearsal rather than the final instrument

Read as a map, the table already disciplines the argument. The three positions are not three volumes of one opinion. Demographic justice is a claim about what policy owes to evidence. Architectural stability is a claim about what institutions owe to their own settlements. Federal implementation is a claim about which level of the Union’s federal design, the design traced in the asymmetric-federal design, should hold the counting instrument. A participant can hold the first and third together, as the Congress governments of Karnataka and Telangana did by counting in their states while demanding a national count, or the second and third together, as commentators did who defended Bihar’s competence while doubting that any count should revise the quotas. The debate becomes unintelligible only when the positions are collapsed into party labels, which is the failure of the standard commentary this article set aside at the start.

The Demographic-Justice Position: Count What the State Distributes

The demographic-justice position begins from a charge of asymmetry. The Indian state counts caste where counting protects the state: it maintains lists of Scheduled Castes and Scheduled Tribes, counts their populations in every census, and administers their reserved seats and posts against those counts. It declines to count caste where counting might enlarge the claims of the uncounted: the Other Backward Classes, whose share of the population was last measured in 1931 and whose share of central employment was fixed in 1990 at a figure derived from that measurement. Satish Deshpande stated the charge in its canonical form: caste among the privileged operates as an invisible asset, uncounted and unexamined, while caste among the disadvantaged is hyper-visible in lists and litigation; a census that refuses the caste question does not abolish this asymmetry, it ratifies it. Yogendra Yadav’s essays added the democratic limb of the argument. Representation in a democracy, Yadav argued across his post-2021 writings, claims legitimacy from correspondence between the represented and the representatives, and a republic that cannot say how many of its citizens belong to the classes in whose name it reserves cannot test that correspondence; the demand for a caste census is therefore a demand for the minimum statistical condition of democratic self-knowledge, not a sectarian manoeuvre.

The position’s strongest single exhibit is the age of the number at the system’s centre. The Mandal Commission’s estimate that the Other Backward Classes formed 52 percent of the population was, as the comparator-framework reading of the post-Mandal settlement records, the product of the only method available to a commission working in 1980: extrapolation from the census of 1931, the terminal colonial count. The 27 percent quantum was then set not by proportion but by subtraction from the Indra Sawhney ceiling in waiting: 52 percent backwardness, compressed under a judicially enforced maximum of 50 percent total reservation, minus the 22.5 percent already reserved for Scheduled Castes and Scheduled Tribes, left 27 percent as the largest award the architecture could hold. Every term of that arithmetic except the ceiling is a 1931 artefact. The demographic-justice position asks a plain question of it: if the counted share of backward classes in Bihar in 2023 is 63.14 percent, and if the Mandal method applied nationally produced 52 percent from a ninety-year-old base, what exactly does the 27 percent represent, and for how long may a democracy decline to check?

Bihar’s survey gave the position its second exhibit, the demonstration effect. A state government decided in June 2022 to count, fielded the first phase in January 2023, completed enumeration by August 2023 and published the headline tables on 2 October 2023: Extremely Backward Classes at 36.01 percent, Other Backward Classes at 27.13 percent, the two together at 63.14 percent, Scheduled Castes at 19.65 percent, Scheduled Tribes at 1.68 percent and the unreserved population at 15.52 percent of a counted population of some 13.07 crore. The sky did not fall. The enumeration error that the counter-case predicted appeared in manageable form, in coding disputes and category challenges that the state processed administratively. Christophe Jaffrelot’s commentary on the survey drew the inference the demographic-justice camp wanted drawn: the Bihar operation proved that the obstacle to a national count since 1951 had never been technique. The survey’s methodological documentation, preserved in the state’s published reports and archives, shows an ordinary large survey competently run: household schedules, caste coding frames prepared in advance, enumerator training, phased verification. Nothing in it required resources beyond those the Registrar General deploys every ten years.

The position’s policy limb follows from its evidence. If backward classes form roughly three-fifths of the population in the state that has counted them, and if the Mandal extrapolation points the same way nationally, then the 27 percent quantum and the 50 percent ceiling together compress the majority’s claim into a minority’s share, and rationalisation means revision: of the quantum upward, of sub-quotas within it toward the most backward, whose case the Rohini Commission on sub-categorisation of the central OBC list, appointed in October 2017 under Justice G. Rohini and reporting in July 2023, documented from within the system’s own records. The demographic-justice position does not promise that revision would be consensual. Its scholars concede the conflict and price it in: Yadav’s essays argue that a conflict conducted over published numbers is democratically superior to a peace maintained over withheld ones. That concession is the position at its strongest, and it is also where the architectural-stability position aims its reply.

The Architectural-Stability Position: A Settlement Is Not a Spreadsheet

The architectural-stability position begins from a different premise about what the reservation system is. In this reading, the system was never a demographic instrument awaiting accurate inputs. It is a layered political settlement, built by negotiation at moments of crisis, and its quantities are treaties rather than measurements. The founding layer was laid at Poona on 24 September 1932, when B. R. Ambedkar surrendered separate electorates for the Depressed Classes under the pressure of M. K. Gandhi’s fast and accepted joint electorates with reserved seats in their place; the 1932 Poona Pact’s architectural inheritance runs through every later layer, because the Pact’s quantities, 148 reserved seats against the Communal Award’s 71, were fixed by bargaining strength and moral emergency, not by any count of the Depressed Classes, who had in fact just been counted in 1931 and whose counted share was not the basis of the bargain. The second layer was laid in the Constituent Assembly, whose reservation provisions in the layered provisions of the constitutional framework balanced Articles 15 and 16 against each other and time-limited legislative reservation to ten years in Article 334, a sunset that has been extended by amendment every decade since, each extension a renegotiation rather than a recalculation. The third layer was laid by the Supreme Court on 16 November 1992, when Indra Sawhney fixed the 50 percent ceiling and the creamy-layer exclusion as the terms on which the Mandal award could stand. At no layer did a fresh population count set the quantities. The stability position concludes that demanding the system behave as if its quantities were measurements mistakes its nature, and that forcing measurement upon it will not rationalise the settlement but dissolve it, since every group armed with a counted share larger than its reserved share acquires a documented grievance, and every group with a counted share smaller acquires a documented exposure.

The position’s scholarly anchor is André Béteille, the sociologist whose essays against caste enumeration, published across the census debates of 2000, 2010 and 2011, argued that a caste count in the census would reify the categories it recorded. Béteille’s argument was anthropological before it was political. Caste as lived is local, contextual and shifting; caste as enumerated is national, fixed and ranked; the census does not photograph the first, it manufactures the second, and citizens then organise to claim the manufactured identities because the state distributes goods through them. The 1901 experience under Risley, when the offer of an official precedence order drew competitive memorialising from every province, is Béteille’s standing historical proof, and the 46 lakh raw caste strings of the SECC are his standing statistical proof: hand a billion people a blank caste line and the returns will describe the politics of self-description, not the structure of society. On this account the enumeration error that the demographic-justice camp treats as a coding inconvenience is the phenomenon itself. The count would not merely risk inaccuracy; it would produce authoritative falsehood, printed in census volumes and cited in court, and the authority is precisely what makes it dangerous.

The position’s institutional limb is the affidavit of September 2021. When the Union government told the Supreme Court that the SECC caste data were defective and unusable and that no caste enumeration was proposed for the forthcoming census, it spoke, whatever its motives, in the stability position’s vocabulary: the social fabric argument, which the Cabinet Committee on Political Affairs would repeat almost verbatim in April 2025 even while reversing the decision, holding that enumeration handled as a political instrument creates doubts in society. Defenders of the 50 percent ceiling across the party spectrum add the prudential limb. The ceiling has survived since 1992 because it caps the total claim that counted disadvantage can make on open competition; a census that returned backward-class shares above 60 percent in state after state would make the ceiling the single most contested rule in Indian public law, and the litigation and agitation required to defend or break it would consume the administrative calm in which reservation actually delivers its benefits to the individuals who clear its examinations each year. The stability position therefore prices the count not at its field cost, which is small, but at its settlement cost, which it holds to be the peace of the system itself.

The position has a genuine weakness, and its honest holders concede it. A settlement whose quantities cannot be checked against the population decays, over generations, from treaty into fiction, and the defence of the 27 percent as a compromise of 1990 loses force as the compromise recedes from living memory and the population it governs is succeeded by grandchildren. The stability argument explains why the quantities should not be revised by simple proportion; it does not explain why they should never be re-measured at all, and the distinction between measuring and revising is the ground on which the Cabinet’s April 2025 decision eventually stood: enumeration inside the census, quantities untouched until a future government chooses otherwise. Whether that distinction can survive the publication of the numbers is the sharpest open question the position faces, and the federal-implementation position proposes to postpone it indefinitely by keeping the numbers provincial.

The Federal-Implementation Position: Count Where the Lists Live

The federal-implementation position starts from the observation that caste, as an administrative category in India, is not one system but dozens. The central list of Other Backward Classes, maintained for Union services and Union educational institutions, differs from every state list; state lists differ from each other, because the communities that compose backwardness are regional formations with regional histories. A community classified as backward in Bihar may be unlisted across the border in Uttar Pradesh; the same name may denote a prosperous trading group in one state and a landless labouring group in another; synonyms, sub-caste segmentation and migration multiply the mismatches. The identification machinery of the republic has always recognised this geography. The Scheduled Castes and Scheduled Tribes lists are themselves state-specific under Articles 341 and 342, and the Supreme Court’s judgment of 1 August 2024 in State of Punjab v. Davinder Singh, which permitted sub-classification within the Scheduled Castes, extended the logic: the operative units of disadvantage are sub-state and state-level populations, as the post-2024 architectural-permission framework records. A national caste schedule, the federal position argues, must either flatten this geography into categories coarse enough to be meaningless or reproduce it at a scale no single coding frame can govern, which is the lesson the SECC’s 46 lakh raw returns taught at national scale and the state surveys avoided by working within one list’s territory at a time.

From that premise the position builds its positive case. The states are not waiting for permission to know their populations; they are constitutionally equipped to count for their own purposes, and three of them did. Bihar’s survey was designed against Bihar’s own backward-classes architecture, in which the Extremely Backward Classes, a category the state constructed from the Karpoori Thakur formula of 1978, form a distinct administrative population; a national census schedule would have had no column for the distinction that matters most to Bihar’s reservation politics. Karnataka’s enumeration grew from the state’s own backward-classes commission tradition running back to the Havanur Commission of 1975, and its compiled tables answer Karnataka’s questions about its own lists. Telangana’s survey of November 2024 was framed by its government as an instrument for designing the state’s welfare and representation measures, and its summary, placed before the assembly in February 2025, reported Backward Classes at a little over 46 percent of the population counted on the state’s own definitions. Each survey, on the federal reading, is methodologically superior to a national count for the decisions its government actually faces, because the categories counted are the categories administered. The comparative frame of the four-dimensional comparative framework for state political settlements supplies the wider point: India’s states are distinct political settlements with distinct caste formations, and a statistical instrument that treats them as sampling strata of a single national caste order misdescribes the federation it serves.

Does a caste count by itself change any reservation quota?

A count changes no quota by its own force. Reservation quantities in India are set by executive order and legislation within the constitutional limits the Supreme Court fixed in Indra Sawhney on 16 November 1992. A census or survey table enters that process only as evidence a government chooses to act on or a court chooses to weigh.

Bihar’s sequence proves the separation. The survey published in October 2023 counted backward classes at 63.14 percent; the state legislature raised total reservation to 65 percent in November 2023; the Patna High Court struck the increase down on 20 June 2024 for want of the exceptional justification the ceiling requires. The number was real, and the quota still had to survive law. Counting and allocating are distinct acts of state, and the debate’s confusion of the two is the source of half its heat.

The federal position’s advocates include scholars of Indian federalism who had opposed a Union caste count on design grounds while defending the states’ competence to count, and its practitioners include the three state governments themselves, two of them Congress governments that simultaneously demanded a national count, a combination the position’s purists regard as incoherent and its pragmatists regard as the federation working as intended: states generate the evidence their lists require, and the Union is pressed, by demonstration, to repair the national gap. The position also absorbed, after 2023, a strand of the stability argument in federal dress. If counted shares must eventually press against the 50 percent ceiling, better that the pressure arrive state by state, in litigation localised to one list and one legislature at a time, as it did at Patna in June 2024, than in a single national confrontation over a single national table. The Cabinet decision of 30 April 2025 overruled the position’s preference while vindicating its demonstration: the Union’s own press statement justified taking enumeration into the census by pointing at the state surveys and criticising their transparency and intent, an official admission that the states had made the running and that the question was no longer whether caste would be counted but by whom.

Why do enumerators record far more castes than official lists contain?

Because a caste return is a self-description before it is a classification. Respondents answer with sub-caste names, surnames, gotras, clan titles, occupational labels and aspirational variants of all of these, and neighbouring households describe the same community in different words.

The SECC of 2011 accumulated some 46 lakh distinct raw entries in this way, and Bihar’s coding teams spent months collapsing the state’s returns into just over two hundred categories. The gap between raw returns and publishable categories is not fraud or failure; it is the standard distance between speech and a schedule, which census offices close for language and occupation in every round. What requires explanation is not the gap but the occasions on which governments have cited the gap as a reason to publish nothing at all.

The state surveys also expose the federal position’s own unresolved difficulty, and candour requires naming it. Three state counts conducted on three definitions produce three incommensurable tables. Bihar’s 63.14 percent backward share, Telangana’s figure of a little over 46 percent and Karnataka’s compiled distributions cannot be added into a national picture, because the categories counted are not the same categories. A litigant in the Supreme Court arguing a national question, a finance commission weighing backwardness across states, or a Parliament debating the central list still possesses, after all three surveys, no national number. The federal position answers that national questions about caste are mostly fictitious, since the benefits and lists at stake are overwhelmingly state-administered; the demographic-justice position answers that the central list, the 27 percent and the ceiling are national facts whatever one thinks of them, and that a federation which counts everything except the thing its largest distributive programme turns on has made a choice, not a discovery. The adjudication between those answers is where this article’s claim is tested, and the post-2024 trajectory supplies the test.

The State Laboratories: What Bihar, Karnataka and Telangana Actually Did

Bihar’s caste-based survey is the debate’s central document, and it rewards the methodological attention that the brief for this article, following the survey’s own documentation, insists upon. The survey landed in the state whose politics had been organised since 1990 by the post-1990 Bihar three-party architecture, itself a product of the Mandal decade, and its tables were read there as a census of the coalition that governed the state. The state cabinet resolved on 2 June 2022 to conduct a caste-based survey through the General Administration Department, after the Union government had made clear in Parliament and in its September 2021 affidavit that no caste enumeration would accompany the deferred decennial census. Fieldwork ran in two phases, from 7 to 21 January 2023 for household listing and from April to August 2023 for the detailed enumeration, covering some 13.07 crore persons. The published tables of 2 October 2023 distributed that population as follows: Extremely Backward Classes 36.01 percent, Other Backward Classes 27.13 percent, Scheduled Castes 19.65 percent, Scheduled Tribes 1.68 percent and unreserved categories 15.52 percent, with the largest single community, the Yadavs, at 14.26 percent. The fuller socio-economic tables followed in the report laid before the state legislature on 7 November 2023. Two features of the operation carry analytical weight beyond the percentages. The first is speed: a state statistical apparatus, neither large nor lavishly funded, moved from cabinet resolution to published tables in sixteen months. The second is survival: the survey withstood challenge in the Patna High Court, which upheld it on 1 August 2023, and the Supreme Court declined to halt publication, so the tables entered public life carrying judicial clearance that the SECC, designed to avoid exactly such scrutiny, never obtained.

The survey’s methodological documentation, preserved in the state’s published reports and in the working papers of the exercise, is the under-cited primary source of this debate, because it records the decisions that determine what a caste count is: the construction of the coding frame from the state’s own lists before fieldwork began, the instruction that enumerators record the respondent’s stated caste verbatim and code centrally rather than classify at the doorstep, and the phased verification by which coding disputes were resolved. Scholars and research teams who work with such schedules after the fact depend on that documentation surviving intact and consultable, a custody problem of the kind examined in the account of how journalists use VaultBook to protect sources, where the chain of custody over field material decides whether later researchers can trust what they read. Bihar’s published sequence also exposed the distance between counting and allocating that the stability position predicts. The state legislature in November 2023 raised reservation in state services and education to 65 percent, plus the 10 percent for Economically Weaker Sections, citing the survey; the Patna High Court struck the increase down on 20 June 2024, holding that the state had not made out the exceptional circumstances Indra Sawhney requires for breaching 50 percent and that population share alone does not constitute such a circumstance. The demographic-justice camp thus holds a published count and a judicial rebuff in the same hand, and its scholars have had to argue, as Yadav did in commentary through 2024, that the count’s value survives the quota’s defeat because the argument has permanently shifted from whether the backward classes are a majority to what follows from their being one.

Karnataka’s exercise tests the framework differently, because its count long predates the post-2021 debate and its report arrived inside it. The government of Siddaramaiah commissioned a social and educational survey through the State Backward Classes Commission in 2015, when H. Kantharaj chaired the Commission, and fieldwork enumerated the state’s households that year. The compiled report then waited through changes of government and commission leadership for a decade, a delay that itself became evidence: opponents of publication argued the data had decayed past usefulness and that its caste tables, recording 1,351 categories in the compiled schedules, would inflame list politics; supporters answered that the delay demonstrated the political cost of counting, since a report commissioned by one government had become the property of its successors’ caution. Justice K. Jayaprakash Hegde, who chaired the Commission in its concluding phase, submitted the report to the state government in February 2024, and the cabinet took it up in April 2025 amid public argument over its findings and over demands from dominant communities, including Lingayat and Vokkaliga organisations, that its figures understated their numbers. Karnataka’s lesson for the debate is about the half-life of caste data. A count is an event with a date; publish it a decade late and every community can plausibly claim the portrait no longer resembles it, and the stability position gains an argument the demographic-justice position handed it by waiting.

Telangana’s survey, launched on 6 November 2024 by the government of A. Revanth Reddy as a comprehensive social, economic, educational, employment and caste survey, was designed with Bihar’s precedent open on the table. Its fieldwork compressed the enumeration into weeks rather than months, its stated purpose was the design of welfare delivery and representation measures, and its summary findings, placed before the state assembly in February 2025, reported Backward Classes at a little over 46 percent of the population counted, with Scheduled Castes at a little over 17 percent and Scheduled Tribes at a little over 10 percent. The Telangana figures, lower for the backward classes than Bihar’s, illustrate the federal position’s core fact and the national debate’s core problem in a single contrast: Telangana counted on Telangana’s lists and definitions, and its 46 percent and Bihar’s 63.14 percent are both defensible and mutually unusable. What the three surveys together established, beyond their tables, is a sequence of demonstrated capacities and demonstrated limits. Capacity: states can count caste, code it, publish it and defend it in court. Limits: state counts do not aggregate, decay in the drawer, and do not by themselves move a single quota past the ceiling of 1992. Each limit points back toward the national instrument the states were substituting for, and it was into that gap that the Union government stepped in April 2025.

The Census That Was Not Held: The Postponement and the Price of the Gap

The caste argument unfolded, between 2021 and 2025, inside a larger statistical emergency that raised the stakes of every position: for the first time since the decennial series began in 1881, India simply did not hold its census. The count due in 2021 was postponed as the COVID-19 pandemic consumed the administrative machinery, a postponement every observer initially accepted as temporary. The house-listing phase had been scheduled for April to September 2020 and the population enumeration for February 2021; both were deferred by executive decision, and as the pandemic receded the schedule was not restored. Year followed year without a notified census date, and the gap lengthened into the longest in the series’ history, longer than the wartime dislocations that had merely truncated the tabulation of 1941. A state that had counted its people through famine, partition and emergency had stopped counting in peacetime, and every distributive argument in the country began, whether it acknowledged the fact or not, to run on a population portrait from 2011 that receded annually further from the society it described.

The gap’s costs were not confined to caste. The Constitution’s own machinery keys representation to the census: the Eighty-fourth Amendment of 2001 froze the allocation of Lok Sabha and assembly seats among the states until the first census taken after 2026, so the postponed count became the trigger on which the entire delimitation question waits, with southern states publicly anxious that a fresh count will convert their success in population stabilisation into lost parliamentary weight. The Constitution (One Hundred and Sixth Amendment) Act of September 2023, the Nari Shakti Vandan Act reserving one third of seats in the Lok Sabha and the state assemblies for women, was written to take effect only after a census and a subsequent delimitation, which tied the largest expansion of women’s legislative representation since independence to the same missing count. Finance Commission devolution, urban governance boundaries, welfare rolls built on SECC data from 2011, and the simple administrative knowledge of how many people live in India’s cities all drew on a statistical base that aged a year every year without renewal. The caste question travelled inside this convoy. When the Cabinet Committee on Political Affairs decided on 30 April 2025 to include caste enumeration in the forthcoming census, it was deciding about an exercise whose postponement had already become a constitutional bottleneck, and the caste column’s inclusion rode on the republic’s need to count at all.

The postponement also rearranged the debate’s burdens of proof in a way the participants were slow to notice. Through the decades of deferral, the stability position could present the absence of a caste count as the natural state of a functioning system: censuses happened, caste was not in them, governance proceeded. After 2021 that presentation collapsed, because governance was proceeding without any census, and the system’s defenders were obliged to argue that a state may indefinitely govern 1.4 billion people on a portrait from 2011 while treating a caste column as the one addition too dangerous to attempt. The demographic-justice camp exploited the inversion skilfully, folding the caste demand into the broader restorationist demand that the republic resume counting its people. The federal position gained least from the gap and lost most: state surveys multiplied in the vacuum, Bihar’s and Telangana’s among them, but each new state table underlined that the federation’s master instrument stood idle, and that the states were improvising substitutes for a Union function rather than exercising a virtue of federal design. When the Union finally moved, it moved on the whole census, caste included, and the sequence vindicated the article’s organising claim once more: the positions are distinguished by what they think counting is for, and a crisis of counting itself forced each of them into the open at the same moment.

The National Count Decided: April 2025 and the Testing of the Framework

The decennial census due in 2021 had been postponed, first on the ground of the COVID-19 pandemic and then without a notified schedule, leaving India in the longest gap between censuses since the series began in 1881. Through the postponement the caste question travelled with the census question: every statement about when the count would happen carried the subsidiary question of what it would count. The Union government’s position, stated in Parliament in 2021 and defended in the September 2021 affidavit, was that caste beyond the Scheduled Castes and Scheduled Tribes would not be enumerated, on grounds continuous with 1951: administrative complexity, the integrity of the census as a statistical instrument, and the social consequences of official caste sorting. The Congress and its allies made the opposite commitment a central campaign plank through the general election of 2024, with Rahul Gandhi’s formulation that the population’s counted share should govern its share of resources repeated across the campaign, and the coalition’s failure to win that election appeared, to the stability camp, to have settled the matter for another cycle.

On 30 April 2025 the Cabinet Committee on Political Affairs, chaired by Prime Minister Narendra Modi, decided that caste enumeration would form part of the forthcoming census. The Press Information Bureau’s statement of that date grounded the decision in Article 246 and entry 69 of the Union List, criticised state surveys as variable in transparency and intent, and argued that enumeration inside the census would serve the social fabric better than enumeration as a separate political exercise. The reversal is the single most instructive event in the debate’s recent history, because each of the three positions can explain it and each explanation exposes the others. The demographic-justice position reads April 2025 as its vindication: a demand argued from evidence for fourteen years, armed with Bihar’s demonstration that counting was feasible, became irresistible to a government that had sworn an affidavit against it. The architectural-stability position reads the same event as its absorption strategy succeeding: the count was conceded precisely by being placed inside the census, the state’s most rule-bound and least campaign-driven instrument, with no accompanying commitment to revise any quantity, so that enumeration and stability, supposedly opposites, were combined by design. The federal-implementation position reads it as a defeat that confirms its analysis: the Union nationalised the count while denouncing the state exercises that had forced its hand, and the methodological questions the states had answered in their own territories, whose lists, whose categories, whose coding frame, return at national scale unanswered, three decades after the SECC showed what an unanswered coding frame produces.

The framework’s robustness lies in surviving this triple reading. A political-strategic account of April 2025, of the kind the standard commentary prefers, explains the timing by electoral arithmetic and the Bihar assembly election due later that year; it cannot explain why the instrument chosen was the census rather than another survey, why the official justification borrowed the stability position’s social-fabric language verbatim, or why state governments that had already counted greeted the decision as a beginning rather than a redundancy. The three-positions reading explains each: the census was chosen because the architectural argument required the count to be housed in the institution least likely to destabilise the settlement; the justification was borrowed because the stability position supplied the only vocabulary in which a reversal could be presented as continuity; the states greeted it as a beginning because the federal position had always held that state counts were rehearsals whose tables could not aggregate into the national evidence the central list requires. The complication this article was bound to confront is that developments after 2025 may supersede the framework itself, as coding frames are published, as the count proceeds and as litigation over its tables begins. The answer given here is that supersession would itself be legible only through the framework: whatever the eventual census returns, the argument over them will be an argument between counted shares and settled quantities, conducted between the Union that counted and the states whose lists the count must somehow describe. The positions are not predictions. They are the grammar in which the next decade of the debate will be spoken, as the positions of 1932 became the grammar of every later layer, from the Mandal moment’s decisive reframe to the methodologically distinct reservation built for the Economically Weaker Sections.

Closing the Pillar: What Ninety Years of Not Counting Decided

This article closes the pillar on caste, reservation and Mandal politics, and its synthesis is the pillar’s own arc read backwards from the census question. The pillar opened with the Mandal moment of 1990, when a government implemented a commission’s decade-old report and the republic argued about a number, 27 percent, whose demographic root reached back to 1931. It moved to the Poona Pact of 1932, where the architecture of reserved representation was bargained before any count could discipline it, and to the constitutional framework of 1950, which layered rights, services and representation into a structure durable enough to absorb every later shock. It examined the Kalelkar Commission of 1953 to 1955, the first casualty of governing caste without counting it, and the Mandal Commission’s methodology of 1980, the most ingenious estimation machine ever built to avoid a count. It followed the implementation protests of 1990 and the party system they remade, the sub-categorisation question that the Supreme Court’s Davinder Singh judgment of August 2024 carried into the Scheduled Castes, and the EWS amendment of 2019, which built a reservation on income and thereby proved, by exception, that every other reservation in the system still runs on caste categories nobody has counted since the Raj. The caste census debate is where all of those arguments converge, because each of them silently presupposes an answer to the question the debate asks aloud: whether the republic will measure the population in whose name its largest distributive system operates.

Three concrete questions, all live after the Cabinet decision of 30 April 2025, show the framework earning its keep as an instrument rather than a summary. The first is the coding frame: whether the Registrar General will publish, in advance of enumeration, the master list of caste categories into which returns will be coded, and whether claims and objections to that list will be invited as they are for the Scheduled Castes and Scheduled Tribes orders. The demographic-justice position demands publication and objection, on the mother-tongue precedent, as the conditions of a legitimate count. The stability position dreads the objection window as a national mobilisation conducted through memoranda, on the 1901 precedent. The federal position asks whose frame it will be, since a Union frame must somehow digest state lists that divide similar social material into 215 categories in one state and 1,351 in another. The second question is the schedule: the census postponed from 2021 has acquired, through the Eighty-fourth Amendment’s freeze and the women’s reservation Act of September 2023, a train of constitutional consequences waiting on its completion, and the caste column’s fate is bound to a timetable set by delimitation politics as much as by statistics. The third question is harmonisation: whether the state surveys of 2023 to 2025 will be treated as rehearsals whose categories the national frame absorbs, or as rivals whose tables will be cited against the national tables when the two disagree, as disagree they will. Each question is technical in form and, in substance, a re-argument of the three positions at the level of the schedule, the calendar and the list. A reader who can see the positions inside the questions possesses the debate; a reader who sees only parties possesses its noise. The distinction is the pillar’s closing gift to the pillars that follow, on religion, elections, courts and movements, where the same discipline, positions before parties and evidence before labels, will be required again at every step of the argument.

The adjudication this article defends can be stated with the evidence assembled. The political-strategic framing fails because it cannot explain the choice of instruments, in 2010, in 2023 or in 2025. The demographic-evidence framing fails because it cannot explain why the same data invigorate one position and arm another, or why Bihar’s tables produced a statute and a striking-down in eight months. The three-substantive-positions reading succeeds because it treats the debate as what its participants, at their strongest, have always said it is: a disagreement about what the republic owes to counted evidence, argued between a position that would rationalise the settlement against the count, a position that would protect the settlement from the count, and a position that would keep the count where the lists live. The post-2024 trajectory has tested each position and broken none of them. It has instead done something more useful. It has moved the argument from whether caste will be counted, a question closed in principle at Patna in August 2023 and in the Cabinet room in April 2025, to the questions the counting itself cannot answer: whose categories, which coding frame, what the counted shares will be permitted to change. Those are the questions with which the pillar closes, and they are, recognisably, the questions of 1931 and 1951 returned in modern dress, proof that in the politics of caste and counting, India does not finish its arguments. It enumerates them, defers them, and meets them again in the next decade’s tables. The tables, when they come, will be read in the states whose surveys rehearsed them, in the courts whose ceilings await them, and in the commissions whose lists they will be asked to revise, and each reading will be conducted in the grammar of the three positions, because that grammar is the debate’s durable product.

The Demand Rebuilt: From the Parliament of 2010 to the Campaign of 2024

The demand for a caste count has a parliamentary biography that the party-political commentary flattens, and recovering it matters because the demand’s carriers changed across three decades in ways that map the three positions. In the 1990s the demand belonged to the parties of the Mandal coalition and to the parliamentary left, raised in debate whenever the census approached and shelved by governments of the Congress and of the United Front alike. The implementation decade had taught every governing party the same lesson that the 1990 Mandal implementation protests burned into the record: numbers about caste, once official, acquire owners, and owners present bills. The census of 2001 passed without a caste question after a debate in which the government of Atal Bihari Vajpayee weighed the proposal and declined it, and the census of 2011 approached with the demand again in the hands of backbenchers rather than cabinets.

The debates of 2010 altered the demand’s ownership. When members of the Lok Sabha pressed the matter in April and May of that year, the pressure crossed the floor: Lalu Prasad Yadav and Mulayam Singh Yadav carried it from the Mandal parties, Sharad Yadav carried it from the Janata Dal (United), and members of the Congress and the Bharatiya Janata Party spoke for it in numbers large enough that Prime Minister Manmohan Singh’s assurance of Cabinet consideration was a recognition of arithmetic rather than a concession of principle. The Group of Ministers that deliberated included the finance, home, law and social justice portfolios, and its work product, the September 2010 compromise that diverted caste into the SECC, bears rereading as a document of the stability position operating inside a government publicly committed to the count: enumeration was conceded in form, denied the census’s machinery, and thereby denied the timetable, the statute and the finishing obligation that make a count a count. The scholars who later dissected the episode, among them political scientist Sudha Pai in her writings on the post-Mandal parties, treated the SECC as the demand’s burial by administration, and the fourteen years that followed bore the reading out.

The demand’s second rebuilding came from the states and from the data famine of the pandemic years. The census due in 2021 was postponed on pandemic grounds and then left unscheduled, an interruption without precedent in the series since 1881, and the postponement itself became an argument: a republic that could not say when it would next count its people could hardly be surprised that its states began counting on their own account. Bihar’s cabinet resolution of 2 June 2022 was the hinge. Nitish Kumar, whose political formation ran through the Mandal decades and whose government rested on a coalition with the Rashtriya Janata Dal, framed the survey in the demographic-justice vocabulary from the start: the state would count because policy without counted shares was guesswork wearing the dress of principle. The Congress, repositioning after 2021 toward an explicit social-justice platform, adopted the demand at national level; Rahul Gandhi’s campaign formulation through 2023 and the general election of 2024, that a caste census would reveal how the country’s population shares map onto its shares of institutional power, carried the argument into the integrated three-tensions framework of the period’s politics, where the census demand functioned as the evidentiary limb of a larger claim about representation. The Bharatiya Janata Party’s responses through those years were themselves a study in the stability position’s public form: ministers defended the 2021 affidavit, questioned the motives and methods of state surveys, and warned against dividing society by enumeration, while the party’s own state units in Bihar had voted for the survey resolution and its national leadership left itself the room that April 2025 eventually used.

Two features of the rebuilt demand distinguish it from its 1990s ancestor and explain its eventual success. The first is its evidentiary modesty: the demand of 2010 and 2023 asked for a count, not for a quota, and its advocates learned to say so, separating enumeration from revision in a way that let the stability camp’s moderate wing retreat from opposing knowledge to opposing consequences. The second is its federal proof: by the time the national argument peaked, Bihar had counted, Karnataka’s report existed in a government drawer, and Telangana was preparing to count, so the demand no longer rested on the prediction that counting was possible. It rested on exhibits. A government resisting the demand after October 2023 had to argue not that caste could not be counted but that it should not be counted by the one institution built for counting, and that argument, as the Cabinet’s reversal of April 2025 conceded, had become the weakest of the three positions’ public forms.

The Science of the Count: Coding Frames, Error and the Manufacture of Categories

Underneath the political argument runs a technical one, and the caste census debate cannot be adjudicated without it, because each substantive position smuggles in a theory of what enumeration does to the thing enumerated. A census is not a camera. It is a schedule: a finite list of questions, a finite set of codes for the answers, an army of enumerators trained to map speech onto codes, and a tabulation that publishes the codes as if they were the population. For most census variables the gap between speech and code is an inconvenience. For caste it is the whole problem, because caste names in speech number in the tens of thousands, shift across districts, carry rank claims inside their very spellings, and answer to administrative lists that were themselves built by commissions taking testimony. The colonial commissioners knew this and printed their doubts: J. T. Marten’s 1921 report worried over returns in which religious movements instructed adherents to record themselves under new names, and Hutton’s 1931 volumes catalogued the inflation and deflation of counts by organised community campaigns. The error problem is therefore not a discovery of the SECC’s critics. It is a documented property of caste enumeration since 1901, and the genuine question the technical debate poses is comparative: whether the errors of counting exceed the errors of not counting, which are also documented, in a 52 percent national estimate extrapolated from 1931 and in state lists assembled from testimony of uneven vintage.

The coding frame is where the technical and the political fuse. Before a single enumerator knocks, someone must decide the list of categories into which returns will be coded, and that decision determines the published shape of society. Bihar’s frame, built from the state’s own lists, produced 215 categories and a portrait in which five great blocs compose the population. Karnataka’s frame, accreted through decades of commission work since the Havanur Commission reported in 1975, produced 1,351 categories and a portrait of fine-grained segmentation. Neither portrait is the territory; each is the shadow of a frame. The demographic-justice position treats the frame as an engineering problem: build it from the lists, publish it in advance, invite claims and objections as the Scheduled Castes lists are revised, and let the count proceed. The stability position treats the frame as the danger itself: the moment a national frame is drafted, every community in India acquires a once-in-a-generation incentive to organise for its placement, its spelling, its merger or its separation, and the drafting process becomes a national caste mobilisation conducted through memoranda. The 1901 precedent, in which Risley’s precedence order drew exactly that mobilisation, is the stability camp’s permanent exhibit, and the demographic-justice camp’s answer, that the republic survived 1901 and conducts list revisions routinely, is its permanent rejoinder. The federal position dissolves the dilemma by scale: a state frame mobilises a state’s communities, a bounded and governable politics, while a national frame mobilises all of them at once against a single drafting table in Delhi.

The SECC’s failure, examined technically, was a failure of frame rather than of fieldwork. The exercise of 2011 collected free-text caste returns without first settling the coding frame into which they would be resolved, and it collected them through rural development machinery rather than through the census organisation whose entire craft is frame management. Census offices handle equivalent problems in every round: India returns hundreds of mother tongues, which the census codes into a published language table through a rationalisation process that linguists criticise and the state performs anyway; it returns thousands of occupations and hundreds of religious self-descriptions, including sects and movements that officials must fold into published categories. The mother-tongue precedent is the demographic-justice camp’s strongest technical argument, because it shows the Indian census already manufacturing a contested social classification, language, through exactly the coding discipline the SECC skipped, and publishing the result without social collapse. The stability camp counters that language lacks what caste possesses, a distributive machinery of quotas waiting to convert each published share into a claim, and that the analogy therefore fails at the point of consequence. Both arguments are serious, and the adjudication between them turns on a judgment about institutions rather than about statistics: whether a coding frame published in advance, under the Census Act’s authority and the Registrar General’s discipline, disciplines the claims it will generate, or merely franchises them.

A word is owed to the enumerators, the debate’s invisible workforce, because the feasibility argument is finally about them. A decennial census of India fields some three million enumerators and supervisors, overwhelmingly schoolteachers and local officials seconded to census duty, each responsible for a block of households, each trained on a schedule whose questions must be askable at a doorstep in minutes. The caste question’s difficulty at the doorstep is real and should not be minimised by either camp: the respondent may state a name the enumerator has never heard, may give a varna label instead of a caste, may name a sect or a gotra, may refuse, or may answer strategically in the hearing of neighbours. Bihar’s design answered the doorstep problem by separating statement from classification, instructing enumerators to record the answer verbatim and reserving all coding to trained teams working later against the frame. The design is neither exotic nor new; it is how the census already treats open-ended returns. Its adoption in a state survey, and its success in producing tables that survived judicial scrutiny at Patna in August 2023, removed the last purely operational objection to a national count and left the argument where this article has located it throughout: among positions about purpose, not among claims about capacity.

One further technical matter deserves record, because the state surveys brought it into the open: the decay of counts. A census is a portrait with a date, and caste portraits decay as communities rename, merge, migrate and reclassify. Karnataka’s report, compiled from a 2015 enumeration and considered by the cabinet in April 2025, was attacked by community organisations precisely on decay grounds, that a decade-old portrait misstates present shares. The attack has force, and it cuts in a direction its authors may not intend: decay is an argument for counting regularly, inside a decennial series whose repetitions correct each other, rather than for counting once by survey and litigating the portrait for a generation. The 1931 count’s ninety-year afterlife is what a single uncorrected portrait becomes. A caste column restored to the decennial census would be wrong in the ways all census columns are wrong, and then, ten years later, less wrong in public, which is the most that any statistical system promises and the least that a distributive system of this scale should accept.

The Count Demanded from the Other Side: Maharashtra, the Marathas and the Litigation of Absence

The caste census debate is usually narrated as a demand of the backward classes against a reluctant state, and the Maharashtra sequence complicates that narration usefully, because there the demand for caste data came from a dominant community seeking entry into the backward lists, and the missing data defeated it in court. The Maratha community of Maharashtra, politically dominant in the state for generations and heavily represented in its legislatures and cooperative institutions, pressed from the 1980s onward for classification as backward. The state government legislated Maratha reservation in 2014 on the strength of the Narayan Rane committee’s findings, and the Bombay High Court stayed the measure. A fresh statute of November 2018, the Socially and Educationally Backward Classes Act, granted 16 percent reservation on the basis of the report of the Maharashtra State Backward Classes Commission under Justice M. G. Gaikwad, and the Bombay High Court upheld a trimmed version in June 2019. On 5 May 2021 a five-judge bench of the Supreme Court struck the Act down in Jaishri Laxmanrao Patil v. Union of India, holding that the Marathas were not shown to be a socially and educationally backward class by the standard required, that no exceptional circumstance justified breaching the 50 percent ceiling, and, in the judgment’s most consequential limb, that after the Constitution (One Hundred and Second Amendment) Act of 2018 the states had lost the power to identify backward classes for their own lists, a reading Parliament reversed within months through the Constitution (One Hundred and Fifth Amendment) Act of August 2021.

The data question ran through the Maharashtra litigation like a fault line. The Gaikwad Commission had worked from surveys, sample studies and the state’s records, in the absence of any caste count since 1931, and the Supreme Court’s majority treated the evidentiary basis of backwardness as exactly the point on which the state’s case was weakest. Maharashtra then reached for the SECC: in the proceedings of 2021 over reservation in local-body elections, the state sought the SECC caste data to demonstrate the backward population’s share, and the Union government’s affidavit of September 2021, refusing the data as defective and unusable, was filed in that very case. The sequence deserves its prominence in this article because it dissolves the debate’s standard casting. The affidavit that the demographic-justice camp cites as proof of bad faith was filed against a demand from a state government seeking data to enlarge a dominant community’s quota, and the data’s absence injured, in that courtroom, a claim that many caste-census advocates would themselves have opposed. Absence of evidence is not a partisan weapon; it is a general condition, and it defeats weak claims and strong ones without discrimination. The Maratha agitation that followed, led by Manoj Jarange Patil from August 2023, with its demand for Kunbi certificates that would route Maratha families into the existing OBC category through historical records from the Nizam’s era, was an attempt to solve by archive what could not be solved by count: if the state would not enumerate caste, families would document it one certificate at a time from the record rooms of the former Hyderabad State. A politics reduced to scavenging century-old revenue records for proof of identity is a portrait of the data famine from an angle the parliamentary debate rarely supplies.

The Maharashtra sequence also supplied the debate’s most quoted judicial sentence on evidence. The Constitution bench in Jaishri Laxmanrao Patil treated the Gaikwad Commission’s empirical record, assembled from samples and secondary sources because no count existed, as insufficient to carry the exceptional claim the state was making, and it reaffirmed Indra Sawhney’s ceiling in the same breath. Read beside the Patna High Court’s judgment of 20 June 2024, the two decisions frame the counted republic from opposite ends: at Patna a count existed and the quota failed for want of legal exceptionality; in the Maratha case the quota failed for want, in part, of a count. Between them they describe the corridor within which any future caste census will operate, wide enough for evidence to matter and narrow enough that evidence alone decides nothing.

The Baseline Under the Microscope: What the 1931 Count Actually Contains

Because so much of the debate leans on 1931, the baseline itself merits the same sceptical reading the article has applied to every other document. Hutton’s census was a formidable operation, and its caste tables were also, by its commissioner’s own account, an negotiated artefact. The schedules recorded the caste each respondent stated; the tabulation offices then had to fold the stated names into the provincial caste lists, resolving synonyms, sectarian renamings and status-inflated variants by administrative decision. Hutton printed the difficulties rather than hiding them, including the campaigns by which communities instructed their members to return themselves under sanskritised names or under a varna label rather than a caste name, campaigns that swelled some categories and emptied others for reasons having nothing to do with births and deaths. The 1931 tables also rest on the boundaries and definitions of 1931: they count the population of British India and the princely states under the categories those administrations maintained, and the Mandal Commission’s use of them required a chain of adjustments, for Partition, for the reorganisation of states, for the conversion of communities across religions, each adjustment a judgment wearing the dress of arithmetic. None of this makes the baseline useless. It makes it a historical document rather than a measurement of the present, and the demographic-justice position is at its most persuasive when it says exactly that: the objection to governing from 1931 is not that Hutton failed but that ninety years of unrepeatable change separate his tables from the population they are asked to describe.

The baseline’s second life, as a symbol, matters as much as its contents. For the stability camp, 1931 functions as the cautionary terminal point: the last count is remembered as the high-water mark of colonial caste governance, and its non-repetition as the republic’s deliberate act of self-denial, renewed by each generation’s census commissioners. For the demographic-justice camp, the same date functions as an accusation: a colonial power counted caste to govern a subject population, and the free republic, which distributes benefits by caste on a scale the Raj never attempted, declined to count at all. Both symbolisms are available in the record, and the choice between them is a choice about what the republic is for, which is why the argument has never been settled by statisticians. It will be settled, if the April 2025 decision is executed, by the unglamorous sequence of a published coding frame, a trained enumerator corps, a tabulation office and a set of printed tables, after which the debate this article has mapped will resume over the tables’ contents with the three positions intact and the evidence, for the first time since 1931, on the table.

What a Count Would Have to Settle: Sub-Categorisation, the Creamy Layer and the Life of Lists

A national count, if it comes, will not arrive at a quiet system. It will arrive at a system whose internal questions have multiplied beyond the capacity of its inherited data, and the caste census debate is best closed by enumerating what the count would be asked to settle. The first question is sub-categorisation. The Rohini Commission’s report of 31 July 2023 documented, from the Union government’s own appointment and admission records, that the benefits of the 27 percent central OBC reservation concentrate heavily: a small set of communities takes a large share of the posts and seats, while hundreds of listed communities record none across years. The Commission’s remedy, a division of the quota among groups of communities ranked by their demonstrated share, requires exactly the population and outcome data that only enumeration supplies, and the Supreme Court’s Davinder Singh judgment of 1 August 2024, permitting sub-classification within the Scheduled Castes on evidence of inter se inequality, extended the same evidentiary logic across the reservation system: sub-classify if you can prove the inequality, and proof, in a court-governed system, means counted data. The era of governing reservation by inherited list and impression is being closed by the judiciary from one side and by the census demand from the other, and the two closures are the same event seen from different rooms.

The second question is the creamy layer, the exclusion, imposed by Indra Sawhney in 1992 and administered since through income thresholds revised at intervals, of the socially and economically advanced within the backward classes. The creamy layer’s thresholds are set by executive decision informed by committee reports, and its operation generates a steady litigation over whose income counts and which posts confer status. Enumeration bears on it obliquely but really: a census that counted caste alongside occupation, education and assets, as the SECC’s design attempted and as Bihar’s socio-economic tables of November 2023 partially performed, would let governments see the internal stratification of each listed community rather than inferring it from salary certificates. The third question is the lists themselves. Central and state lists of backward classes are amended through commission recommendations and political decision, and entries are added far more readily than they are removed, because removal requires a government to tell a community that its counted disadvantage has ended. The Kalelkar Commission’s experience in 1955, when its own chairman’s covering letter disowned the caste criterion his commission had applied and the report was shelved, stands at the head of this history, as the first Backward Classes Commission’s record shows: a list-making exercise without a census beneath it degenerates into testimony management. A decennial count would not depoliticise the lists. It would give their revision an external reference, a population against which claims of growth, shrinkage and graduation could be checked, which is precisely why list politics both demands the count and fears it.

The fourth question is the one the pillar’s whole arc has been approaching: the relationship between counted shares and the 50 percent ceiling. Indra Sawhney’s ceiling was laid down in a data famine, as a rule of prudence for a system whose beneficiary population had been estimated at 52 percent by extrapolation. A census returning counted backward-class shares, in a country where the one modern state count found 63.14 percent, will place the ceiling under a pressure it was designed without the numbers to withstand. The stability position predicts the pressure and fears it; the demographic-justice position predicts it and welcomes the argument it forces into the open; the federal position expects the pressure to be absorbed state by state, in litigation like the Patna judgment of 20 June 2024, where the ceiling held against a counted majority and the republic absorbed the holding without disorder. That last exhibit deserves its weight in the adjudication. Bihar counted, legislated, and was struck down, and the system’s sky remained in place, which suggests that the settlement the stability position defends is hardier than its defenders claim, and that the count the demographic-justice position demands is less explosive than its opponents fear. The debate’s three positions, tested against the record assembled in this article, converge on a narrower disagreement than the one with which the republic began: not whether caste exists as a political fact, not whether it can be counted, but what counted caste will be permitted to change. On that question the pillar closes, and the census, when it comes, will be only the beginning of the argument.

Frequently Asked Questions

Q: What is the caste census debate?

The caste census debate is the argument over whether the Indian state should enumerate the caste of its population, beyond the Scheduled Castes and Scheduled Tribes whom every census since 1951 has counted. Three substantive positions organise the debate. The demographic-justice position holds that a state distributing 27 percent of central posts by caste must count caste, and that reservation shares should be tested against counted population shares. The architectural-stability position holds that reservation quantities are layered political settlements, from the Poona Pact of 1932 to the Indra Sawhney ceiling of 1992, which fresh counts would destabilise. The federal-implementation position holds that counting belongs with the states, whose caste lists differ from each other and from the central list.

Q: What did the Bihar caste survey find?

The Bihar caste-based survey, whose headline tables the state government released on 2 October 2023, counted a population of some 13.07 crore and distributed it as follows: Extremely Backward Classes 36.01 percent, Other Backward Classes 27.13 percent, the two backward groups together 63.14 percent, Scheduled Castes 19.65 percent, Scheduled Tribes 1.68 percent and unreserved categories 15.52 percent. The largest single community recorded was the Yadavs at 14.26 percent, followed among backward groups by Kushwahas at 4.27 percent and Kurmis at 2.87 percent. The fuller socio-economic tables were laid before the state legislature on 7 November 2023, and the state government cited the findings when it raised reservation to 65 percent that month.

Q: What is the SECC 2011?

The SECC 2011 is the Socio Economic and Caste Census, an enumeration conducted through 2011 and 2012 by the Ministry of Rural Development with the state governments, separate from the decennial census of the Registrar General. It was created by the Union Cabinet’s decision of September 2010, which answered parliamentary demands for a caste count in the census of 2011 by commissioning a parallel exercise instead. The SECC collected household data on deprivation, occupation, housing and caste. Its socio-economic findings were published in July 2015. Its caste returns, which accumulated some 46 lakh distinct raw entries, were never tabulated or published, and the Union government told the Supreme Court in September 2021 that the caste data were defective and unusable.

Q: Why was caste enumeration stopped?

Caste enumeration stopped with the census of 1951, the first count of independent India, conducted under Registrar General R. A. Gopalaswami. The government of Prime Minister Jawaharlal Nehru decided that the census would record caste only for the Scheduled Castes and Scheduled Tribes, whose enumeration the Constitution’s reservation provisions required. The reasoning recorded in the census reports and government statements of the period was threefold: a general caste count would perpetuate the divisions the republic sought to dissolve; the returns of 1931 had been corrupted by organised status-claiming by caste associations; and a state founded on equal citizenship should not sort its citizens by birth in its master statistical instrument. Every decennial census from 1961 to 2011 repeated the omission.

Q: What did the Karnataka caste census report?

The Karnataka caste census report is the compiled outcome of the social and educational survey the state commissioned in 2015 through its Backward Classes Commission, then chaired by H. Kantharaj, which enumerated the state’s households that year. Its compiled schedules recorded 1,351 caste categories, and its tables distributed the state’s population among the categories the state’s own reservation architecture administers. The report waited a decade for consideration, passing through changes of government and commission leadership. Justice K. Jayaprakash Hegde, who chaired the Commission in its concluding phase, submitted the report to the state government in February 2024, and the state cabinet took it up in April 2025, amid public dispute over its figures from communities that held themselves undercounted.

Q: What is the Justice Jayaprakash Hegde Commission?

The Justice Jayaprakash Hegde Commission is the Karnataka State Commission for Backward Classes under the chairmanship of Justice K. Jayaprakash Hegde, a former judge, in the phase during which the state’s caste enumeration reached its conclusion. The Commission inherited the survey commissioned in 2015 and executed under an earlier chairman, H. Kantharaj, and carried responsibility for the compiled report through a decade of governmental change and public controversy. Justice Hegde submitted the report to the Karnataka government in February 2024. The state cabinet’s consideration of the report in April 2025, and the objections raised by organisations of the Lingayat and Vokkaliga communities among others, made the Commission’s work the Karnataka chapter of the national caste census debate.

Q: Will India have a national caste census?

India’s Union government decided on 30 April 2025 that caste will be enumerated in the forthcoming decennial census. The decision was taken by the Cabinet Committee on Political Affairs chaired by Prime Minister Narendra Modi, and the Press Information Bureau’s statement grounded it in Article 246 and entry 69 of the Union List, which make census a Union subject. The statement argued that enumeration inside the census would serve the social fabric better than separate state surveys, whose transparency and intent it questioned. The decision reversed the position the same government had defended in a Supreme Court affidavit of September 2021. The census itself, due in 2021 and postponed, had no completed enumeration behind the decision, so the count was authorised before its schedule and coding frame were public.

Q: How did the colonial censuses between 1871 and 1931 classify caste?

The colonial censuses treated caste as a primary classification of the Indian population from the first non-synchronous count of 1871 to 1872 onward. Each decennial volume printed caste tables province by province, and communities petitioned the census authorities for favourable placement, understanding the printed table as an official ranking. The census of 1901, under Commissioner H. H. Risley, made the ranking explicit by directing that castes be ordered by social precedence as recognised by native public opinion, a scheme that drew organised protest across India. Later commissioners retreated from explicit ordering; E. A. Gait in 1911 and J. T. Marten in 1921 recorded their doubts about returns distorted by status claims and name changes, doubts printed in the census volumes themselves and quoted in the debate ever since.

Q: Who was J. H. Hutton?

J. H. Hutton was the Census Commissioner who conducted the census of 1931, the last enumeration to record and publish the caste of the whole population of British India. An administrator and anthropologist, Hutton supervised an operation that captured caste returns for some 350 million people with paper schedules, and his published tables became the demographic base for every later estimate of backward-class population, including the Mandal Commission’s extrapolation of 1980. Hutton also recorded, in his census report and in his 1946 book Caste in India, serious misgivings about the reliability of caste returns, in which the same community appeared under many names and organised status claims inflated or deflated counts. Both camps in the caste census debate cite him, for the count’s feasibility and for its artefacts.

Q: What is the legal difference between a decennial census and a state caste survey?

The decennial census is a Union operation. Census stands at entry 69 of the Union List in the Seventh Schedule, Article 246 gives Parliament exclusive competence over it, and the Census Act of 1948 governs its conduct under the Registrar General, with statutory duties of response and confidentiality. A state caste survey is a different instrument. A state government cannot conduct a census, but it may gather information under its own executive and statistical powers for administering state subjects, including the identification of backward classes for state services and institutions under Articles 15(4) and 16(4). The Patna High Court upheld Bihar’s survey on that footing on 1 August 2023, and the Supreme Court declined to stay publication of its results.

Q: Why were the caste returns of the SECC never released as official tables?

The caste returns of the Socio Economic and Caste Census were never released because the work of converting them into publishable tables was begun and then allowed to lapse, and no institution owned the obligation to finish it. Respondents had returned caste, sub-caste, surname and clan descriptions amounting to some 46 lakh distinct entries, which required coding into a finite list of communities before any tabulation could be certified. The government assigned the task to an expert process associated with NITI Aayog vice-chairman Arvind Panagariya, and ministers stated between 2015 and 2018 that the data remained under examination. Because the exercise had been designed outside the Registrar General’s census machinery in September 2010, no census office carried the professional duty to complete the tabulation, and in September 2021 the Union government told the Supreme Court the data were defective and unusable.

Q: How many caste categories did the Bihar survey enumerate?

The Bihar survey’s published tables organised the state’s returns under 215 caste categories, the figure carried in the survey’s own documentation of its coding frame. The number is itself a finding about enumeration method. Enumerators recorded each respondent’s stated caste verbatim at the doorstep, and the coding teams then collapsed the raw returns, which contained synonyms, sub-caste names and variant spellings for the same communities, into the finite category set against which the percentages of 2 October 2023 were computed. The contrast with Karnataka’s 1,351 compiled categories shows how much of a caste count’s result is determined before fieldwork begins, in the construction of the frame into which a population’s self-descriptions will be sorted.

Q: How many caste categories did the Karnataka enumeration record?

The Karnataka enumeration’s compiled schedules recorded 1,351 caste categories, a figure that entered the public debate when the report associated with Justice K. Jayaprakash Hegde’s chairmanship of the State Backward Classes Commission was taken up by the state cabinet in April 2025. The survey behind the figure was commissioned in 2015 and executed under H. Kantharaj’s chairmanship of the Commission. The category count reflects Karnataka’s own list architecture, built through the state’s backward-classes commissions since the Havanur Commission of 1975, and the fine segmentation of the state’s communities into sub-castes and synonyms. Set beside Bihar’s 215 categories, the Karnataka figure demonstrates that two competent Indian enumerations can divide similar social material into category systems differing by a factor of six.

Q: What did the Telangana survey launched in November 2024 set out to measure?

The Telangana survey, launched on 6 November 2024 by the government of Chief Minister A. Revanth Reddy, set out to measure the social, economic, educational, employment and caste composition of the state’s households in a single comprehensive exercise, as an evidence base for designing welfare delivery and representation measures within the state’s competence. Its fieldwork compressed enumeration into weeks, and its summary findings, placed before the state assembly in February 2025, reported Backward Classes at a little over 46 percent of the population counted, with Scheduled Castes at a little over 17 percent and Scheduled Tribes at a little over 10 percent. The figures were computed on Telangana’s own lists and definitions, and the state government presented them as an instrument for its own policy rather than as a substitute national count.

Q: What was the Rohini Commission asked to examine?

The Rohini Commission, appointed by the President on 2 October 2017 under Article 340 and chaired by Justice G. Rohini, a former Chief Justice of the Delhi High Court, was asked to examine the sub-categorisation of the Other Backward Classes on the central list. Its question was distributive: within the 27 percent reserved for the backward classes in central services and institutions, a small number of communities had captured a disproportionate share of the posts and seats, while hundreds of listed communities had received none. The Commission consulted state governments and commissions, analysed appointment and admission records, and submitted its report on 31 July 2023. Its findings strengthened the argument, heard throughout the caste census debate, that aggregate quotas conceal internal inequality which only finer enumeration can expose.

Q: How did the Mandal Commission use the census of 1931 to reach its 52 percent estimate?

The Mandal Commission reached its estimate that the Other Backward Classes formed 52 percent of the population by building on the census of 1931, the last count to record caste for the whole population. Working in 1979 and 1980 without any later caste data, the Commission took the population shares recorded in 1931 for the communities its eleven indicators classified as backward, and cross-checked them against the residual method: the total population minus the counted and estimated shares of the Scheduled Castes, Scheduled Tribes, forward castes and non-Hindu communities. The 52 percent figure was therefore an extrapolation across half a century of population growth, carrying the authority of the only comprehensive caste count ever published in India and the fragility of a fifty-year-old base, a combination the caste census debate has argued over since 1990.

Q: Why do state lists of backward classes differ from the central list?

State lists differ from the central list because backwardness in India is regional in composition. The central list governs admission and appointment in Union institutions and is assembled from state-level recommendations, while each state list governs that state’s own services and institutions and reflects the findings of its own backward-classes commissions, from Karnataka’s Havanur Commission of 1975 to Bihar’s commissions behind the Karpoori Thakur formula of 1978. Communities are regional formations: a group classified as backward in one state may be unlisted or differently classified in the neighbouring state, and the same name may cover groups of very different standing across a border. Articles 341 and 342 make even the Scheduled Castes and Scheduled Tribes lists state-specific, so the entire architecture of caste classification in the Constitution is built state by state.

Q: What did the Bihar legislature do to reservation quotas after the survey was published?

After the survey’s headline tables were published on 2 October 2023 and its socio-economic report was laid before the legislature on 7 November 2023, the Bihar legislature passed bills raising reservation in state government services and educational institutions from 50 percent to 65 percent, in addition to the 10 percent reserved for Economically Weaker Sections. The increase redistributed the total toward the groups the survey had counted: the Extremely Backward Classes rose to 25 percent, the Other Backward Classes to 18 percent, the Scheduled Castes to 20 percent and the Scheduled Tribes to 2 percent. The state government defended the increase in court as a proportionate response to counted population shares, making the episode the first full test of what a published caste count does to the Indra Sawhney ceiling.

Q: What did the Patna High Court decide about Bihar’s 65 percent reservation law in June 2024?

On 20 June 2024 the Patna High Court struck down Bihar’s increase of reservation to 65 percent, holding that the state had not made out the exceptional circumstances which the Supreme Court’s judgment in Indra Sawhney requires before the 50 percent ceiling may be breached. The court held that population proportion by itself does not constitute such a circumstance, and that the state had not demonstrated, by the standard the ceiling imposes, why the counted shares in the survey of 2023 compelled a departure from the rule that had governed since 1992. The decision separated the two steps the caste census debate often fuses: the state had successfully counted its castes, and the count had survived challenge, but the quota built on the count still had to satisfy a legal test that counting alone could not pass.

Q: Which office conducts the decennial census of India, and under what statute?

The decennial census of India is conducted by the Office of the Registrar General and Census Commissioner, within the Ministry of Home Affairs, under the Census Act of 1948. The Act empowers the Union government to appoint a Census Commissioner and superintendents of census operations in the states, makes response to the enumerator’s questions a legal duty, and protects the confidentiality of individual returns. The same office maintains the civil registration system and conducts the demographic surveys that run between census rounds. The institutional fact matters to the caste census debate: the Socio Economic and Caste Census of 2011 was deliberately placed outside this office and its statute, in the Ministry of Rural Development, and its unfinished caste tabulation is the standing exhibit for what that placement costs.