On 1 August 2024 a seven-judge Constitution Bench of the Supreme Court of India decided the SC sub categorisation question that Indian constitutional law had carried unanswered for two decades. In State of Punjab v Davinder Singh, reported as 2024 INSC 562, six judges held that a state may divide the Scheduled Castes named in the Presidential List into sub-groups for reservation, and one judge, Bela M. Trivedi, dissented. The verdict overruled E. V. Chinnaiah v State of Andhra Pradesh, decided on 5 November 2004, which had forbidden the practice entirely. This article defends one claim about that sequence: the twenty years between the two verdicts were a constitutional-architectural deferral, a period in which the architecture of reservation contained a question it could neither answer nor escape, and the 2024 verdict is best understood as the ending of that deferral on stated conditions.

A Question the Architecture Asked in 1950 and Deferred in 2004

The SC sub categorisation question begins with a structural feature of the Constitution that most accounts of reservation pass over quickly. Article 341 empowers the President, after consultation with the Governor of a state, to specify by public notification the castes, races or tribes deemed to be Scheduled Castes for that state. Article 341(2) reserves every later inclusion in or exclusion from that list to Parliament, acting by law. The list is therefore a central act of specification, state by state in content, national in authority. Reservation itself operates through different articles. Article 15(4) permits special provision for the advancement of socially and educationally backward classes and for the Scheduled Castes and Scheduled Tribes. Article 16(4) permits reservation in appointments or posts for any backward class of citizens inadequately represented in the services. Article 335 requires that the claims of Scheduled Castes and Scheduled Tribes be taken into consideration, consistently with the maintenance of efficiency of administration, in appointments to central and state services.

Between the list article and the reservation articles a gap opens, and the whole dispute lives inside that gap. Article 341 names the beneficiaries as a single schedule. Articles 15(4) and 16(4) authorise benefits for classes. If the castes inside a state list differ sharply among themselves in education, occupation and access to state employment, may a state treat the list as a floor rather than a ceiling, and distribute the reserved share unevenly so that the least reached castes receive a protected portion? Or does the act of listing fuse the named castes into one constitutional class, so that any internal division by a state legislature amends the list by another name and usurps Parliament? Every actor in this story, legislator, commissioner, judge and movement leader, was answering one of those two forms of the question.

The three-layered reading of the reservation framework developed for this pillar, set out in the three-layered constitutional reservation framework, places the question precisely. The rights-protective layer of Articles 15, 17 and 46 forbids discrimination and abolishes untouchability. The services-affirmative layer of Articles 16 and 335 distributes public employment. The representational layer of Articles 330, 332 and 334 reserves legislative seats. Sub-categorisation is an operation attempted inside the services-affirmative layer upon material supplied by Article 341, which sits outside the three layers as the schedule-making provision. The architectural puzzle is that the instrument of benefit and the instrument of identification are different instruments, drafted for different purposes, and the Constitution nowhere states in terms whether the second constrains the first.

Two features of the founding settlement sharpened the puzzle. First, the Presidential List inherited the schedule appended to the Government of India (Scheduled Castes) Order of 1936, itself the product of the Lothian Committee’s franchise work and the Poona Pact settlement of September 1932, whose joint-electorate architecture is traced in the 1932 Poona Pact’s architectural inheritance. The list was a political settlement cast in administrative form, not a measurement of comparative deprivation among the castes it named. Second, the founding generation knew the listed castes were unequal among themselves. B. R. Ambedkar, piloting the constitutional scheme as chairman of the Drafting Committee, treated untouchability as a common condition that justified a common schedule, and the Constituent Assembly accepted a single list per state as the price of a workable settlement. Whether that acceptance was a finding of homogeneity or an act of administrative convenience is the historical question underneath the legal one, and the two Constitution Benches of 2004 and 2024 answered it in opposite ways.

The claim defended here is architectural rather than doctrinal. Read as doctrine alone, the sequence is a prohibition followed by a permission, interesting chiefly to lawyers. Read state by state, the sequence fragments into a Madiga story in Andhra Pradesh, a Balmiki and Mazhabi Sikh story in Punjab, an Arunthathiyar story in Tamil Nadu, each with local causes. The architectural reading holds the sequence together: a constitution that lists castes centrally and reserves benefits through the states will generate the sub-categorisation question wherever the listed castes are internally unequal, and it will keep generating that question until a court of final authority either answers it or forbids it permanently. The 2004 bench forbade it and the question returned within two years through commissions, notifications and fresh statutes. The 2024 bench answered it, attached conditions, and converted a recurring constitutional embarrassment into a governed power. That is the sense in which a deferral ended on 1 August 2024.

The Founding Design: A Schedule Built for Identification, Not Distribution

The Presidential List that both verdicts construe was issued as the Constitution (Scheduled Castes) Order of 1950, promulgated by President Rajendra Prasad on 10 August 1950 under Article 341. Its contents were inherited almost entire from the schedule to the Government of India (Scheduled Castes) Order of 1936, which had itself codified the work of the Indian Franchise Committee under Lord Lothian of 1932 and the settlement reached at Poona in September 1932. The inheritance matters to the architectural argument. The 1936 schedule was built to identify the electorates and beneficiaries of a colonial franchise settlement; it answered the question which castes count as depressed classes for defined political purposes in each province. It was not built to answer the question how benefits should be shared among the castes it counted, because the settlement of 1932 had deliberately located the sharing inside joint electorates, where distribution would be settled by politics rather than by schedule.

The Constituent Assembly’s treatment of the schedule between 1947 and 1949 preserved that division of labour. The draft article that became Article 341 was debated as a provision about identification and authority: identification, because the President’s notification would carry the 1936 schedule into the republic with continuity; authority, because the power to vary the list was withdrawn from the provinces and placed with the President in the first instance and with Parliament thereafter. Ambedkar’s defence of the design, as the Drafting Committee’s chairman, turned on distrust of provincial legislatures: a list variable by every provincial majority would become an instrument of provincial electoral arithmetic, and the depressed classes, a minority in every province, would pay for each variation. The centralisation of the list was thus a protective device for the listed castes against state politics. The irony that organises this article’s whole story is visible at the founding: the device designed to protect the listed castes from state legislatures became, in Chinnaiah’s hands, the device that forbade state legislatures to protect the weakest listed castes from the distributive consequences of the list itself.

Three features of the 1950 Order sharpened the later dispute. The first is state specificity: the Order lists castes province by province, and a caste scheduled in one state may be absent from the neighbouring state’s list, so the schedule never described a single all-India population whose internal equality could be presumed. The second is the religion condition in paragraph 3 of the Order, which confined the status, as issued, to persons professing the Hindu religion, extended to Sikhs in 1956 and to Buddhists in 1990; the schedule therefore sorted by a criterion, religious profession, that has no relation to comparative deprivation among the castes sorted. The third is the Order’s silence on gradation. Colonial enumeration had in fact distinguished among the depressed classes repeatedly, in census reports and in provincial lists that ranked castes for educational concessions, and the Madras Presidency had operated graded communal rotations in public employment from the 1920s. The 1950 Order’s refusal to grade was a choice made for the settlement’s workability, and the choice is what the 2024 majority would later cite against the claim that the founders had pronounced the listed castes homogeneous: a list drawn without gradation for reasons of settlement is not a finding that no gradation exists.

The list, moreover, was never the frozen object the prohibition required it to be. Parliament varied it repeatedly under Article 341(2): the Scheduled Castes and Scheduled Tribes Lists (Modification) Order of 1956 recast the schedules for the reorganised states, the religion condition was extended to Sikhs in 1956 and to Buddhists in 1990, and the Scheduled Castes and Scheduled Tribes Orders (Amendment) Act of 1976 removed the area restrictions that had confined many castes’ status to parts of their states. Each variation was an admission that the schedule tracks a changing social reality and must be maintained against it. The 2024 majority’s treatment of the list as an instrument of identification, authoritative and revisable for its own purpose, is faithful to that maintenance history in a way the 2004 treatment, which converted a maintained administrative schedule into an untouchable constituent icon, was not. The point is not that Parliament’s power to vary the list implies a state power to grade it; the dissent is right that it does not, by itself. The point is that the list’s own history refutes the premise of frozen homogeneity on which the prohibition’s rhetoric depended, and a doctrine whose rhetoric is refuted by its subject’s biography will be litigated until a bench says so.

The distributive silence of the founding design was tolerable for as long as the reserved share was small against the listed population and the competition for it was thin. It ceased to be tolerable as higher education expanded, as the services grew, and as the relatively advanced listed castes converted their head start into a durable lead visible in every state’s representation returns. The commissions of the deferral years did not discover a new fact; they documented, state by state, the predictable working of a distribution-blind schedule in a field of unequal castes. The founding design is therefore the first term of the architectural reading: the question was not created by the Madiga movement or by the Punjab preference, it was deposited in the Constitution in 1950 and fell due when the distributive record became impossible to file away.

Before the Prohibition: Two Decades of State Experiment

Sub-categorisation was not invented by the Andhra Pradesh legislature in 2000. States experimented with internal division of the Scheduled Caste quota for a quarter of a century before the Supreme Court spoke in Chinnaiah, and the experiments supply the factual record on which both verdicts were argued. Punjab moved first. Andhra Pradesh moved furthest. Haryana and Tamil Nadu followed with instruments of their own, and the Punjab and Haryana High Court’s treatment of the Punjab provision in 2006 and 2010 built the litigation road that eventually reached the seven-judge bench.

Punjab’s starting point was an executive instruction of 1975. The state government, administering a 25 percent reservation for Scheduled Castes in services and admissions, directed that one half of the reserved vacancies be offered first to candidates of the Balmiki and Mazhabi Sikh castes. The two named groups stood at the bottom of the state’s internal hierarchy of listed castes by every measure the administration used: the Mazhabi Sikhs, descendants of Sikh converts drawn largely from the Chuhra caste, worked overwhelmingly as agricultural labourers and sanitation workers, and the Balmikis occupied a comparable position in the towns. The more advanced listed castes of the state, the Ad-Dharmis and the Ramdasia and Ravidassia castes among them, drew a disproportionate share of reserved posts through better schooling and urban networks. The 1975 instruction was a preference device, not a partition: the Balmiki and Mazhabi Sikh preference operated inside the single quota and lapsed to the general Scheduled Caste pool where candidates were unavailable. Its logic, distribution within a fixed share according to measured deprivation, is the logic of every sub-categorisation instrument since.

Andhra Pradesh arrived at the same logic through agitation rather than administration. Through the early 1990s the Madiga caste, one of the two largest Scheduled Castes of the Telugu country alongside the Mala, pressed the charge that the reserved share in the undivided state was being captured by the Malas, who had converted to Christianity and Buddhism in larger numbers, entered mission and government schooling earlier, and accumulated a decisive advantage in professional admissions and gazetted posts. The Madiga Reservation Porata Samithi, founded in 1994 under the leadership of Manda Krishna Madiga, converted that charge into a mass movement demanding categorisation of the state’s 15 percent Scheduled Caste quota. The demand reached the legislature because the movement made it electorally expensive to refuse: Madiga voters were numerous, organised and geographically concentrated, and the counter-mobilisation of Mala organisations, defending the unified quota, made the dispute a defining fault line of Andhra politics in the decade.

The state government’s answer was a commission before it was a statute. In 1996 the government of Andhra Pradesh appointed a one-man commission under Justice P. Ramachandra Raju, a retired judge of the Andhra Pradesh High Court, to identify the groups among the state’s Scheduled Castes that had failed to secure the benefit of reservation. The Raju Commission examined representation data across services and professional colleges and reported that the benefits of the 15 percent quota were distributed with pronounced unevenness: some listed castes held shares of reserved posts far above their share of the Scheduled Caste population, and others were close to absent from the higher services. The commission recommended dividing the 59 castes then on the Andhra Pradesh list into four groups graded by backwardness, with the quota shared among them in fixed proportions. The state first acted by ordinance in 1997 and then, after litigation and a reference to the National Commission for Scheduled Castes, enacted the Andhra Pradesh Scheduled Castes (Rationalisation of Reservations) Act, numbered Act 20 of 2000.

What did the Andhra Pradesh Act of 2000 actually do?

The Act divided the 59 Scheduled Castes of Andhra Pradesh into four groups and fixed each group’s share of the 15 percent quota. Group A, the Relli group, received 1 percent. Group B, with the Madiga castes, received 7 percent. Group C, with the Mala castes, received 6 percent. Group D, the Adi Andhra group, received 1 percent.

The grouping was the statute’s intellectual core, and its grading was the point of maximum political sensitivity. The Raju Commission’s data had placed the Madiga group as more deprived than the Mala group across most indicators, and the 7 to 6 split between Groups B and C wrote that finding into law. Mala organisations, which had argued that grading one large listed caste below another would harden an internal hierarchy the Constitution was meant to dissolve, challenged the Act, and Madiga organisations defended it, so the litigation that became Chinnaiah was from the beginning a dispute between two readings of equality carried by two movements of the listed castes themselves. The Andhra Pradesh High Court upheld the Act by a majority of four to one, and the appeals against that judgment, carried by the parties recorded in the cause title of E. V. Chinnaiah, reached the Supreme Court’s Constitution Bench.

Two other state instruments complete the pre-2004 record and matter for what followed. Haryana, in a series of executive decisions across the same period, experimented with grouping its listed castes for the purpose of quota distribution, an experiment the state returned to with legislation after 2020. Tamil Nadu acted later but on the same pattern: the Arunthathiyar castes of the western districts, Chakkiliyar and allied groups working predominantly in leather work and sanitation, organised through the 2000s behind the argument that the larger listed castes of the state, the Paraiyar and Pallar among them, absorbed the reserved share. A state commission under Justice M. S. Janarthanam examined the claim, and the Tamil Nadu Arunthathiyars Reservation of Seats in Educational Institutions and of Appointments or Posts in the Services under the State Act of 2009 carved 3 percent out of the state’s 18 percent Scheduled Caste quota for the Arunthathiyar group. Tamil Nadu’s instrument survived its first judicial tests where Andhra’s had not yet been finally judged, an early sign that the constitutional question was being answered differently depending on the forum and the framing.

The Madiga Claim and the Mala Counter-Claim at Their Strongest

Sub-categorisation divides the castes it claims to serve, and an evenhanded account must state both sides of the division in the form their holders give them. The Madiga position, carried by the Madiga Reservation Porata Samithi from 1994 and later by the Madiga Dandora movement and its allied organisations, is a claim about the internal distribution of a fixed benefit. Its evidentiary core is the representation record. Across the undivided Andhra Pradesh of the 1990s, studies placed before the Raju Commission and cited by Sukhadeo Thorat in his essays on Scheduled Caste reservation policy showed the Mala group holding a share of reserved professional seats and gazetted posts several times its share of the state Scheduled Caste population, and the Madiga group, comparable in population, holding a fraction of its proportional share. The Madiga argument from that record is not that the Mala acted wrongly but that a single undifferentiated quota in a field of unequal starting positions functions as a ratchet: the relatively advanced convert schooling into posts, posts into networks, and networks into a hereditary advantage inside the reserved category itself. On this account the unified quota does not merely fail the most deprived; it actively compounds their deprivation, and the constitutional promise of reservation is honoured in form and defeated in distribution.

The Madiga claim also carried a historical argument, stated in movement literature and in Anand Teltumbde’s Republic of Caste of 2018, which discusses intra-Scheduled Caste differentiation as a structural feature of the reservation system rather than an Andhra accident. The two great listed castes of the Telugu country entered the twentieth century from different positions: the Malas of the coastal districts encountered mission education, military recruitment and the Adi Andhra movement’s institutions earlier and at greater scale, and the Madigas of the dry Telangana and Rayalaseema districts remained bound for longer to hereditary leather work and village servitude. A reservation system blind to that history, the argument runs, distributes its benefits in proportion to the very advantages history conferred, so that the schedule’s common condition, untouchability, is made to subsidise the schedule’s internal leaders. Teltumbde’s treatment is unsparing toward all sides, toward the state for postponing the question and toward movement leaderships for converting distributive grievance into caste consolidation, and the article’s architectural reading borrows his central insistence: the unevenness is structural, not episodic, and only a structural instrument can answer it.

The Mala counter-position, carried by Mala Mahanadu and allied organisations and argued in the Chinnaiah litigation, answers at three levels. At the level of constitutional principle, it holds that the Presidential List is a single class created by a single constituent act, and that a state which grades the listed castes against each other performs the President’s and Parliament’s function without their authority. At the level of political consequence, it holds that grading fractures the collective political weight of the Scheduled Castes, converting a unified constituency of the oppressed into ranked fragments competing for fixed percentage points, a competition the relatively advanced groups of the general category can only welcome. At the level of remedy, it holds that uneven capture is real but is to be answered by expanding educational provision and enforcing the existing quota in full, not by ranking castes whose relative deprivation shifts across districts and decades faster than any commission’s table can track.

Punjab supplies the Madiga-Mala structure in a second key, and the parallel confirms that the dispute’s engine is distributive rather than regional. The Ravidassia castes of the Doaba, leather workers by hereditary occupation like the Madigas, organised from the early twentieth century through the Ad-Dharm movement, which asserted a separate religious identity for the listed castes of the region, and through the Dera Sachkhand Ballan and its network of temples at home and in the diaspora. Educational and migratory success followed organisation: the Doaba’s listed castes built schools, entered the services and emigrated at rates the Mazhabi Sikh agricultural labourers of Malwa and the Balmiki sanitation workers of the towns did not approach. The Punjab preference of 1975 was the state’s answer to exactly the gap the Madiga movement would name in Andhra twenty years later, and the defence of the unified quota in Punjab was mounted, as in Andhra, by the relatively advanced: Ad-Dharmi and Ravidassia organisations argued that a first preference consuming half the quota on 1975 measurements had become a hereditary entitlement detached from the deprivation that justified it. The names differ, the structure is identical, and the identity is the architectural reading’s point in miniature: wherever the schedule joins castes of unequal starting position to a single undifferentiated share, the share’s internal distribution becomes the politics, and the politics organises the advanced and the deprived into opposing movements carrying opposing theories of equality.

Anupama Rao, in The Caste Question of 2009, supplies the frame in which the dispute is intelligible as something other than a quarrel over posts. Rao’s account of Dalit politics treats the category of the Scheduled Caste as a modern political artefact, assembled by colonial enumeration and nationalist negotiation, whose unity was always a political achievement rather than a social fact. On that account the Madiga and Mala movements are not betrayals of an original solidarity; they are the normal politics of a category whose members were never similarly placed, conducted with the instruments the category itself supplies. The adjudication offered later in this article accepts Rao’s premise and Thorat’s distributive data, and treats the Mala warning about fragmentation as the strongest statement of the costs of the remedy. The question the courts faced was narrower than the politics: not whether the unevenness existed, which the Raju record made hard to deny, but which organ of the Constitution held the power to answer it.

Chinnaiah: The Prohibition of 5 November 2004

The Constitution Bench that decided E. V. Chinnaiah v State of Andhra Pradesh delivered judgment on 5 November 2004. The bench comprised Justices N. Santosh Hegde, S. N. Variava, B. P. Singh, H. K. Sema and S. B. Sinha, with Justice Hegde writing the principal judgment, and the decision is reported at (2005) 1 SCC 394. The court held the Andhra Pradesh Scheduled Castes (Rationalisation of Reservations) Act of 2000 unconstitutional and struck it down, in reasoning that closed the sub-categorisation question for the whole country and for every state, not for Andhra Pradesh alone.

The holding rested on a construction of Article 341. The President’s notification, the bench reasoned, creates the Scheduled Castes of a state as one class, and the castes named in the notification are deemed to be a single homogeneous group for constitutional purposes. Because Article 341(2) assigns every variation of the list to Parliament, a state legislature that divides the list into sub-groups and distributes benefits among them on differential terms varies the list in substance: it treats the President’s specification as raw material for a state re-specification. The bench’s recurring verb was tinkering; the states, it held, cannot tinker with the Presidential List. From that premise the remaining steps followed. Article 16(4) empowers the state to reserve for a backward class inadequately represented, but the Scheduled Castes as listed form one class for that purpose, and a state cannot manufacture sub-classes inside a class the Constitution has fused. The bench drew support from the nine-judge decision in Indra Sawhney v Union of India of 16 November 1992 only to distinguish it: the division of the Other Backward Classes into backward and more backward, accepted in that case, was possible because the OBC category is a state-constructed classification under constitutional supervision, whereas the Scheduled Caste category is a presidential specification under Article 341, and the two instruments are not interchangeable.

Why did the 2004 bench treat the Presidential List as untouchable?

The bench treated the list as untouchable because it read Article 341 as a complete code for identifying Scheduled Castes. On that reading, specification by the President and variation only by Parliament exhaust the field, so a state grading listed castes against each other performs a variation by another name and exceeds its legislative competence under Articles 15 and 16.

The prohibition’s reach exceeded the Andhra statute in three ways, and each excess became a pressure point during the deferral. First, the bench’s homogeneity premise was stated as constitutional fact: the listed castes are a homogeneous class. Yet the list itself, on its face, enumerates castes of vastly different occupation, region and ritual standing, and the state’s own Raju data, accepted for the purpose of argument, showed distribution inside the quota to be radically unequal. A premise that the empirical record contradicted could hold the system only by authority, not by persuasion. Second, the bench treated the Punjab-style preference device and the Andhra-style graded partition as the same act, although a lapsing first preference inside a common pool and a fixed percentage partition differ in their effect on every other listed caste’s prospects. Third, the prohibition left the acknowledged unevenness without any constitutional addressee: Parliament could amend the list but possessed no instrument for distributing benefits within it, and the states possessed the distribution instruments but were forbidden to aim them. The deferral of the next twenty years is the record of that unaddressed grievance accumulating.

Justice Sinha’s concurring reasoning added a warning that reads, across the interval, as the deferral’s epitaph: a classification frozen at the moment of notification cannot track social reality, and a doctrine that forbids the state to respond to measured internal inequality within the reserved category would eventually have to answer for the inequality itself. The majority closed that door in 2004. The point for the architectural reading is that Chinnaiah did not resolve the tension between Article 341 and Articles 15 and 16; it suppressed one side of it. Suppressed questions in a working constitution do not disappear. They migrate into commissions, ordinances, electoral promises and fresh litigation, and the migration is the subject of the next two sections.

Punjab’s Parallel Road: Preference, Statute, Repeal by Judgment

Punjab’s sub-categorisation history runs alongside Andhra’s and ultimately supplied the vehicle that carried the question back to the Supreme Court. The state’s internal Scheduled Caste politics differ from the Telugu country in composition and vocabulary. Punjab holds the highest Scheduled Caste share of any major state, roughly a third of its population by the Census of 2011, and its listed castes divide along lines of religion and occupation that the Andhra grading never faced: Mazhabi Sikhs and Balmikis at one pole of deprivation, Ad-Dharmi and Ravidassia castes, organised around the Ravidassia religious movement and comparatively advanced in schooling and overseas migration, at the other. The Punjab settlement’s own longer history, including the Ad-Dharm movement of the 1920s and the state’s distinctive caste-party alignments, is treated in the Punjab state-political context, and it explains why the 1975 preference for Mazhabi Sikhs and Balmikis commanded support across Punjab’s party spectrum for decades.

The 1975 executive instruction, described in the previous section, operated without serious legal challenge for a generation. Chinnaiah changed its legal foundations overnight: if states could not sub-divide the Presidential List, a first-preference instruction favouring two named castes inside the quota was indefensible, and in 2006 the Punjab and Haryana High Court struck the Punjab preference down on Chinnaiah’s authority. The state legislature answered with a statute. The Punjab Scheduled Castes and Backward Classes (Reservation in Services) Act of 2006, in Section 4(5), provided that fifty percent of the vacancies of the quota reserved for Scheduled Castes in direct recruitment shall be offered to Balmikis and Mazhabi Sikhs, by way of first preference, from among the Scheduled Caste candidates, subject to their availability. The drafting chose the preference model deliberately: no separate sub-quota was carved out, the common pool remained, and the preference lapsed where the preferred candidates were not available. The state defended the provision as the mildest instrument that could answer the distributive record, and its defenders argued that a lapsing preference could not be a tinkering with the list even on Chinnaiah’s own terms.

The Punjab and Haryana High Court disagreed. In 2010, in the litigation that bears Davinder Singh’s name as respondent, a division bench struck Section 4(5) down, holding that Chinnaiah governed every form of intra-list differentiation, preference included. The state appealed. What followed shows how a deferred architectural question travels. In August 2014 a three-judge bench of the Supreme Court, hearing the Punjab appeal, recorded doubt about Chinnaiah’s premises and referred the matter to a Constitution Bench, framing the reference around the relations among Articles 16, 338 and 341 and the reasoning of Indra Sawhney. On 27 August 2020 a five-judge Constitution Bench headed by Justice Arun Mishra, with Justices Indira Banerjee, Vineet Saran, M. R. Shah and Aniruddha Bose, held in the reference that Chinnaiah had not been correctly decided: in the five judges’ view, the 2004 bench had misread the Constitution’s equality code, the Scheduled Castes were not a homogeneous class for the purpose of distributing reservation benefits, and a state answering measured internal inequality with sub-classification acted within the object of Articles 15 and 16 rather than against Article 341. Because a five-judge bench cannot overrule a coordinate bench, the 2020 decision referred Chinnaiah to a still larger bench. The deferral had entered its final stage: the prohibition stood in law, stood condemned by a bench of equal strength, and awaited a bench of greater strength to bury or restore it.

What separates sub-classification from tinkering with the Presidential List?

Sub-classification distributes the benefits of reservation among castes that all remain on the list. Tinkering, the act Chinnaiah forbade, would add a caste to the list or remove one from it. The first operation leaves the President’s specification intact and works on Articles 15 and 16. The second changes who counts as a Scheduled Caste at all.

Punjab’s road also displays the political economy that kept the question alive during the deferral. Every Punjab government of the period, Congress and Akali Dal alike, defended the Balmiki and Mazhabi preference in court and in the assembly, because the named castes were organised, electorally weighty in the Doaba and Malwa regions, and able to punish abandonment. The counter-pressure came from within the listed castes: Ad-Dharmi and Ravidassia organisations, whose relative advancement was the preference’s justification, argued that first preference to half the quota on the strength of deprivation measured decades earlier had outlived the data and become an entitlement of its own. That argument, that distributive instruments freeze the hierarchy they measure, is the Mala argument in Punjab dress, and its presence inside the beneficiary state itself is evidence that the deferral was not a north-south curiosity but a general property of the architecture wherever the list contained unequal castes. The state-level settlement comparisons in the federal-architecture context frame the larger pattern: India’s federal design repeatedly pushes unresolved distributive questions down to the states, and the states repeatedly hand them to commissions.

The Comparative Mirror: Sub-Classification Inside the OBC Architecture

The strongest structural evidence for the architectural reading is comparative. The Constitution’s other great beneficiary category, the Other Backward Classes, has operated with internal differentiation for decades without doctrinal crisis, and the contrast isolates the variable that produced the Scheduled Caste deferral: not differentiation itself, but the Article 341 form of the Scheduled Caste list.

Bihar’s government of Karpoori Thakur divided the state’s backward classes in 1978, reserving a distinct share for the Extremely Backward Classes within the backward-class quota, and the Mungeri Lal Commission’s two-annexure design, separating the most backward from the backward, became the template for OBC sub-categorisation across the Hindi belt. Indra Sawhney in 1992 expressly approved the principle: a beneficiary class may be divided into backward and more backward where the internal differences warrant it, and the creamy layer exclusion, built for the OBCs in the same judgment, presupposes that a beneficiary category is internally stratified and that the stratification is constitutionally cognisable. At the Union level the Justice G. Rohini Commission, appointed in October 2017 under Article 340 to examine sub-categorisation within the central OBC list, reported in 2023 on the distribution of the 27 percent central quota among the listed OBC communities, documenting a concentration of benefits in a minority of the listed castes closely parallel to the Raju Commission’s Andhra findings of 1996. No court struck the Rohini exercise down at the threshold, no doctrine held the OBC list a homogeneous class, and the Union government received the report as a policy document to be acted on politically.

Set the two records side by side and the architectural variable stands alone. The distributive facts are the same in kind: internal inequality among listed beneficiary castes, documented by commissions, pressed by movements of the least reached. The doctrinal treatment is opposite: thinkable, indeed routine, inside the OBC category; forbidden for twenty years inside the Scheduled Caste category. The only feature that differs is the form of the list. The OBC lists are state and central executive constructions under Articles 15(4), 16(4) and 340, revisable by the authority that made them. The Scheduled Caste list is a presidential specification under Article 341, variable only by Parliament. Chinnaiah’s prohibition is therefore exactly as wide as the Article 341 form and no wider, which is what the architectural reading predicts: the deferral was generated by a drafting form, the schedule’s centralisation, interacting with a distributive reality the form could not express. The 2024 verdict can be read, on this account, as the court redrawing the boundary between the two: Article 341 keeps its monopoly over identity, and the distributive articles recover, for the Scheduled Castes, the internal-differentiation power they had always carried for every other beneficiary class.

The comparison also disciplines the fears. OBC sub-categorisation has been litigated, revised and politically contested for four decades, and the category has not dissolved: the Bihar two-tier design survived changes of government, the Rohini documentation entered public debate without fracturing the OBC constituency it measured, and the states that grade their backward classes continue to field unified backward-class politics at election time alongside the grading. The Mala and Ad-Dharmi warning, that grading dissolves collective weight, is a real mechanism and the comparative record shows it operating as friction rather than as dissolution, a cost to be weighed in each scheme’s design rather than a prohibition to be imposed on all schemes in advance. That weighing, case by case on data, is the permission’s method, and the OBC record is the permission’s longest-running pilot study.

The Deferral Years: Commissions, Workarounds and Frozen Data

Between November 2004 and August 2024 the sub-categorisation question was constitutionally closed and politically open. The interval is best read as a sequence of attempts to answer the distributive grievance without performing the forbidden act. States appointed commissions to measure internal inequality they were not permitted to remedy. Legislatures tested instruments drafted to look like something other than sub-division. Movements kept the demand in electoral circulation. Each attempt deposited material, data, draft statutes, reasoned reports, that the 2024 bench would eventually use, so the deferral was not empty time: it was the period in which the permission’s conditions were assembled.

Punjab appointed the period’s emblematic commission. In 2006, as the High Court struck the 1975 instruction down, the state government constituted a commission under Justice Usha Mehra, a retired judge of the Delhi High Court, to examine the representation of the state’s Scheduled Castes in services and to advise on equitable distribution within the quota. The commission’s inquiry rehearsed the Punjab record: Mazhabi Sikh and Balmiki representation in the higher services stood far below the groups’ share of the state’s Scheduled Caste population, and the commission’s recommendations supported a protected share for the most deprived listed castes, in substance the policy Chinnaiah forbade the state to enact. The commission device had become a ritual of the deferral: a state would document inequality with judicial thoroughness, file the documentation, and wait for the doctrine to move. Karnataka ran a parallel exercise through the commission the brief for this pillar records under Justice Khwaja Mohammad Faruq in 2005, and Karnataka returned to the question with the Justice A. J. Sadashiva Commission, whose 2012 report recommended internal reservation among the state’s 101 listed castes on the basis of a caste-wise survey; successive Karnataka governments received the report and deferred the decision, the purest instance of the period’s pattern of measurement without remedy.

The Haryana strand shows the deferral operating inside a single quota dispute across the whole period. Haryana’s listed castes divide between groups associated with sanitation and leather work, organised in later years under the banner of the Deprived Scheduled Castes, and the comparatively advanced Chamar group, whose share of the state’s reserved posts the deprived groups’ organisations documented as disproportionate across the 2000s and 2010s. The state legislated in 2020 to reserve one half of the Scheduled Caste quota in direct recruitment for the Deprived Scheduled Castes group, and the statute was carried into the Supreme Court with the Punjab and Tamil Nadu instruments, so that Haryana’s local arithmetic became part of the record on which the seven judges ruled. The sequence matters beyond Haryana: a state that had taken no part in the original Chinnaiah litigation found its statute suspended in the same doctrinal freeze, which is the behaviour of an architectural constraint rather than of a case-specific holding. The freeze bound states that had never been parties, on a question their own commissions had answered in the opposite direction, and every year it bound them it converted more of the political system to the view that the doctrine, and not the distributive record, was the obstacle.

Other states found workarounds that tested Chinnaiah’s edges. Tamil Nadu’s Arunthathiyar Act of 2009, described earlier in this article, distributed 3 percent of the 18 percent quota to a named group of castes and survived challenge in the Madras High Court, defended on the ground that it apportioned the benefits of reservation without altering the list, a distinction Chinnaiah’s blanket language did not obviously allow. Bihar chose a different instrument altogether: acting on the Bihar State Mahadalit Commission of 2007, the government of Nitish Kumar designated a Mahadalit category covering, in its first form, all but one of the state’s 22 listed castes, and directed development missions, housing and scholarships toward the category without partitioning the service quota, a welfare-side answer to the same measured inequality that stayed clear of the forbidden act by never touching Articles 15(4) and 16(4) at all. Haryana legislated in 2020 for the Deprived Scheduled Castes, grouping its listed castes and reserving a share of the Scheduled Caste quota for the group, and the Haryana and Tamil Nadu instruments travelled with the Punjab appeal into the seven-judge reference, so that by February 2024 the court was not judging two statutes but a national pattern of state improvisation under a prohibition.

The scholarly record of the deferral years matters to the adjudication because it kept the two empirical propositions of the dispute in public view. Thorat’s essays, drawing on National Sample Survey and service-representation data, documented that the internal distribution problem was general rather than regional: in state after state a small number of listed castes accounted for a large share of reserved higher posts, and the least reached castes were those bound to sanitation work, leather work and agricultural labour. Teltumbde, in Republic of Caste and in his essays for the Economic and Political Weekly, pressed the harsher corollary: a reservation regime that cannot differentiate among its beneficiaries will be administered as if they were identical, and the pretence of identity is itself a distributive decision in favour of the advanced. Against both, the defenders of the Chinnaiah settlement, including counsel for the Mala and Ad-Dharmi organisations in successive litigations, maintained the institutional argument: differentiation by percentage is a one-way ratchet, each graded settlement creates constituencies for its own permanence, and the only protection the least numerous castes possess is the rule that the list cannot be opened from below. The 2024 bench heard every one of these positions across three days of argument in February 2024, argued by the Attorney General, the Solicitor General and senior counsel for the states and the intervening caste organisations.

The Hearing: Three Days of Argument in February 2024

The seven-judge bench heard the reference and the connected appeals across 6, 7 and 8 February 2024, and the argument’s structure maps the three readings this article adjudicates. The State of Punjab, defending Section 4(5) of its 2006 Act, argued the specification point in its cleanest form: Article 341 exhausts itself in identifying the Scheduled Castes, the identification binds every authority, and nothing in the identification speaks to distribution, which Articles 15(4) and 16(4) govern; a state that distributes by measured need among the identified castes disobeys no provision. The Union of India, through the Solicitor General, supported the states’ power in principle, treating sub-classification as an instrument of the substantive equality the reservation articles exist to serve, and joined the states in asking that Chinnaiah be overruled. The Attorney General assisted the court on the constitutional structure, and the bench’s questions across the three days returned repeatedly to two points: whether the homogeneity attributed to the list was a proposition of fact open to evidence, and what discipline would govern a state that invoked the power.

The opposition’s argument, carried by counsel for the respondents and for the intervening organisations of listed castes that defended the unified quota, stated the Mala position at the level of constitutional law. The list, they argued, is a constituent act with constituent consequences: the listed castes accepted a common schedule in 1950 on the footing of a common condition, and the benefits attached to the schedule attach to the class the schedule creates. To grade the class from below, by state legislation, is to reopen the constituent settlement with ordinary legislative instruments, and to do so in the arena, state politics, from which the founders had deliberately withdrawn the list. The respondents also pressed the empirical objection in its strongest form: the data on which grading would rest are state-assembled, dated and district-variable, and a court that licenses grading on such data licenses a permanent litigation of caste demography in which every percentage point is a judgment-proof political fact. The dissent of Justice Trivedi adopts the core of each of these submissions, which is why the dissent reads as the prohibition’s most durable defence.

Two exchanges in the hearing foreshadowed the judgment’s shape. The first concerned N. M. Thomas. The 1975 case, in which a seven-judge bench upheld Kerala’s promotional reservations for Scheduled Caste and Scheduled Tribe employees, had described the listed castes as a class for the purposes of Article 16, and Chinnaiah had built its homogeneity premise on that description. Counsel for the states took the bench through the Thomas text to show that class language there served a comparison between the listed castes and the general category, and said nothing about relations among the listed castes themselves; the Chief Justice’s eventual judgment adopts exactly that reading. The second concerned limits. Pressed on whether a state could reserve the entire Scheduled Caste quota for its most deprived caste, counsel for the states conceded a line at total exclusion, and the concession entered the concurrences as doctrine: Justice Gavai’s judgment states it as a rule that no sub-group may be granted the whole of the reserved share. The hearing is therefore part of the evidence for the architectural reading. The permission was not a mood; it was drafted in argument, condition by condition, out of the failure modes of the prohibition and the fears of its defenders.

Davinder Singh: The Permission of 1 August 2024 and Its Conditions

The seven-judge Constitution Bench in State of Punjab v Davinder Singh comprised Chief Justice D. Y. Chandrachud and Justices B. R. Gavai, Vikram Nath, Bela M. Trivedi, Pankaj Mithal, Manoj Misra and Satish Chandra Sharma. The bench heard the reference, together with the Punjab, Haryana and Tamil Nadu appeals, over three days in February 2024 and delivered judgment on 1 August 2024. Six judges held sub-classification of the Scheduled Castes constitutionally permissible and overruled Chinnaiah. Justice Trivedi dissented. The opinions divide into the Chief Justice’s judgment, written for himself and Justice Misra, separate concurrences by Justices Gavai, Vikram Nath, Pankaj Mithal and Satish Chandra Sharma, and the dissent. The judgment is reported at 2024 INSC 562 and (2024) 4 SCC 869.

The majority’s core reasoning dismantles the homogeneity premise on which the prohibition stood. The Chief Justice’s judgment holds that the Presidential List carries no constitutional fiction of homogeneity: Article 341 identifies the castes entitled to the Constitution’s protective and affirmative provisions, and the castes so identified have never been, as a matter of historical or sociological fact, a single undifferentiated class. The lists themselves record castes separated by occupation, region and ritual rank, and the judgment traces the error in Chinnaiah to a reading of State of Kerala v N. M. Thomas, decided on 19 September 1975, in which the description of the Scheduled Castes as a class distinct from the general category hardened, across citation, into a description of the class as internally homogeneous. A class for one constitutional purpose, the majority holds, is not thereby homogeneous for every purpose: Article 14 permits, and the substantive equality underlying Articles 15 and 16 can require, classification within a class where the members of the class are dissimilarly situated for the purpose in hand. Sub-classification that includes no caste and excludes no caste does not touch Article 341(2): the list stands entire, and the state works on the distribution of benefits under Articles 15(4) and 16(4), provisions whose text, the judgment notes, speaks of classes and of inadequate representation and contains no bar on looking within the listed castes.

The permission is conditional, and the conditions are the verdict’s operative legacy. The Chief Justice’s judgment lays down that a state sub-classifying the Scheduled Castes must ground the exercise in quantifiable data: empirical evidence of the inter se backwardness of the sub-groups and of the inadequacy of representation of the less advanced groups in state services. Sub-classification must rest on a rational principle connected to that purpose, and it is open to judicial review on the data. Justice Gavai’s concurrence adds the distributive guardrails: the state must justify preferential treatment on empirical material, and it cannot set apart the whole of the reserved share for one sub-group, excluding the remainder of the list entirely; sub-classification is legitimate as a device for reaching the least reached, not as a device for transferring the quota between organised constituencies. The concurrences converge on the proposition that the measure of the exercise is inadequate representation, the standard Article 16(4) itself supplies, applied group by group rather than to the schedule as an undifferentiated whole.

What conditions did the 2024 verdict attach to sub-categorisation?

The majority attached three conditions. The state must produce empirical, quantifiable data showing that a named sub-group is more backward and inadequately represented in services. The differentiation must rest on that data and remain open to judicial review. And the exercise cannot hand the entire Scheduled Caste quota to a single sub-group, shutting out the rest of the Presidential List.

The majority also mapped the instruments a state may choose, and the map matters for the schemes that followed. The opinions distinguish two models of sub-classification. In the preference model, the state directs that vacancies within the Scheduled Caste quota be offered first to the more backward sub-groups, the Punjab design of 1975 and 2006, and the quota remains a single pool whose order of filling is regulated. In the exclusive model, the state sets apart defined shares of the quota for defined sub-groups, the Andhra design of 2000 and the Telangana and Andhra designs of 2025, and each sub-group’s share is its own unless it goes unfilled. Both models were held available, subject to the data condition and to the bar on total exclusion, and the choice between them was treated as a legislative judgment about administrative workability rather than a constitutional preference of the court. The distinction explains the otherwise puzzling variety of the implementation record: states were not improvising at random in 2024 and 2025, they were selecting from a menu the verdict had written, and the litigation each scheme attracted has been litigation about the selection’s evidentiary basis rather than about the menu itself.

One further element of the principal judgment fixes the permission inside the constitutional text with care. The court reads Article 335, the efficiency clause, as a direction to consider the claims of the Scheduled Castes and Scheduled Tribes consistently with administrative efficiency, and holds that the clause informs the exercise of Articles 16(1) and 16(4) without cutting down the sub-classification power: the claims to be considered are the claims of the listed castes, and where the record shows one sub-group’s claims systematically unconsidered in practice, a distributive instrument that restores consideration serves the clause’s purpose. The reading is characteristic of the judgment’s method. Each provision that had been deployed for twenty years as a reason the question could not be asked, Article 341, Article 335, the equality code of Article 14, is taken in turn and shown to address a different subject from the one the prohibition assigned it. Chinnaiah is overruled provision by provision, not merely displaced by a larger bench’s preference, and the thoroughness is a measure of how completely the 2004 reasoning had occupied the field.

A further strand of the majority reasoning deserves separate notice because it reaches back to the pillar’s founding materials. The judgment reads the reservation provisions through the substantive-equality line that runs from N. M. Thomas through Indra Sawhney, in which Justice Jeevan Reddy’s majority opinion of 1992 accepted the division of a beneficiary class into backward and more backward where the internal differences were real. Chinnaiah had confined that principle to the Other Backward Classes on the strength of the Article 341 specification; the 2024 majority holds the confinement was the error, since the specification governs identity, not distribution. In pillar terms, the verdict locates sub-categorisation inside the services-affirmative layer of the three-layered constitutional reservation framework, subject to that layer’s own tests, and denies that the schedule-making provision silently overrides the layer from outside. The reservation-political setting in which the older case law formed, including the Mandal settlement and the post-1989 reservation-political context, is part of the judgment’s own historical account of why distribution within beneficiary classes had been thinkable for the OBCs for three decades while the question was treated as unthinkable for the Scheduled Castes.

The Deferral Mapped: The Article’s Artifact

The trajectory this article defends can be set out as a single map, and it is reproduced here as Table 1, the 2004 to 2024 architectural-deferral trajectory map, so that the sequence of prohibition, deferral and permission can be cited as one object. Read the three rows as constitutional states of a single question: the state of prohibition created by Chinnaiah, the state of deferral in which states improvised around the prohibition and a coordinate bench condemned it, and the state of permission created by Davinder Singh, whose conditions specify the evidence a state must carry before it differentiates. The map is the article’s findable artifact, and the sections before and after it supply the record each row compresses.

Phase	Instrument and date	Architectural content	Governing limit
Prohibition	E. V. Chinnaiah v State of Andhra Pradesh, 5 November 2004, five-judge bench	The Presidential List is one homogeneous class; state sub-division of the list is unconstitutional tinkering under Article 341	States retain reservation implementation; no intra-list differentiation in any form
Deferral	Usha Mehra Commission, Punjab, 2006; Tamil Nadu Arunthathiyar Act, 2009; Punjab Act Section 4(5) struck down, 2010; five-judge reference opinion, 27 August 2020	States measure internal inequality and test workarounds; a coordinate bench holds Chinnaiah wrongly decided and refers it upward	Prohibition stands in law; remedy waits on a larger bench
Permission	State of Punjab v Davinder Singh, 1 August 2024, seven-judge bench, 6 to 1	States may sub-classify the listed castes for reservation under Articles 15(4) and 16(4); Article 341 fixes identity, not distribution	Empirical data on inter se backwardness and inadequate representation required; no sub-group may take the whole quota; exercise open to judicial review

Table 1 repays one caution. Its rows are states of doctrine, and doctrine is not the whole of the events: the deferral row contains movements, elections and commission reports that no table of cases captures, and the permission row opens implementation contests that the table cannot foresee. The map’s purpose is narrower. It fixes the dates and instruments on which the article’s claim depends, so that a reader who disputes the claim can dispute it at its joints: the characterisation of Chinnaiah as deferral-creating rather than question-answering, the reading of the commission years as accumulation rather than drift, and the treatment of the 2024 conditions as the verdict’s centre rather than its decoration.

The Evidentiary Burden: What a State Must Measure

The data condition is the verdict’s hinge, and it deserves analysis as an instrument rather than as a slogan. The majority requires quantifiable data on two quantities: the inter se backwardness of the sub-groups into which the state proposes to divide its list, and the adequacy of each sub-group’s representation in the services of the state. Each quantity conceals a methodological decision that the verdict leaves to the states and their commissions, and each decision is a future litigation surface.

Backwardness, in the commission tradition the verdict draws on, is a composite: literacy and years of schooling, presence in higher and professional education, occupational distribution with particular weight on hereditary stigmatised work, landholding and housing, and access to the institutions through which the other goods are reached. The Raju Commission of 1996 built its Andhra grading from such a composite, and the Punjab exercises from 1975 onward rested on the occupational and educational position of the Mazhabi Sikh and Balmiki castes. The composite method has a known weakness, named by critics of every commission that has used it: the indicators are correlated, so a caste deprived on one is deprived on most, and the ranking the composite yields can be sensitive to the weights the commission assigns, which are judgment calls wearing the dress of measurement. The verdict’s answer is not to prescribe weights but to make the exercise reviewable: a court may test whether the data exist, whether they are empirical rather than asserted, and whether the grouping bears a rational relation to them. That is a real but limited discipline, and Justice Trivedi’s dissent identifies the limit exactly: review for rationality will not catch a grouping engineered at the margins for electoral effect, because the margins are where the data are softest.

Representation, the second quantity, is harder than it looks, and the difficulty is instructive. A sub-group’s share of reserved posts can be computed only against its share of the state Scheduled Caste population, and the population denominator is the problem: the Census of India publishes counts for individual Scheduled Castes, and the census of 2011 supplies the last complete denominator, already a decade old when the verdict was delivered and older still against the schemes built on it. The Socio-Economic and Caste Census of 2011 collected caste returns outside the main census, and the Union government’s decision not to publish those returns left the states to commission surveys of their own, as Telangana and Karnataka did after 2024, or to reason from service records and admission returns whose coverage is partial. Thorat’s essays on Scheduled Caste reservation policy have long insisted on the underlying point: representation must be measured grade by grade, since a sub-group may approach proportionality in Class IV posts and remain nearly absent from Class I, and a quota distributed on aggregate figures can certify as adequate a representation that is, at the levels where authority sits, a vacancy. The commissions of the permission era have generally accepted the grade-wise method, and the acceptance is one of the deferral’s quiet gains: the evidentiary practice of 2025 is markedly more exact than the practice Chinnaiah judged in 2004, because twenty years of forbidden measurement taught the states to measure as if a hostile court were reading, which, in the event, one was.

A final feature of the burden deserves emphasis because it distinguishes the Indian instrument from cruder quota politics. The data condition ties the state’s hands in both directions. A state may not grade without data, and a state confronted with data of the Raju kind cannot plead Chinnaiah’s ghost as a reason to file the report and do nothing; the political cost of inaction against a published commission record is a cost several state governments of the deferral years paid at elections. The permission thus operates as a ratchet toward measurement-based distribution, and the measurement, once published, acquires a public life beyond the scheme it justifies: caste organisations cite commission tables in every later dispute, and the tables become the common factual ground on which the distributive argument proceeds. Whatever the fate of particular schemes, that change in the argument’s factual basis is the verdict’s most concrete legacy.

The Creamy Layer Strand: Four Judges and a Political Refusal

The strand of Davinder Singh that travelled furthest in public argument was not the holding but an observation attached to it. Justice Gavai’s concurrence held that the state should evolve a policy for identifying a creamy layer within the Scheduled Castes and Scheduled Tribes and exclude it from the benefits of reservation, on the ground that a sub-group’s advanced families, once lifted by one generation of reserved posts into the higher services and professions, stand in a different position from the sub-group’s remaining households, and that treating the two as one beneficiary repeats inside the sub-category the error sub-classification was designed to correct. Three further judges, Justices Vikram Nath, Pankaj Mithal and Satish Chandra Sharma, expressed agreement with the application of a creamy layer principle to the Scheduled Castes and Scheduled Tribes in their concurrences, Justice Mithal adding the proposal that reservation be confined, as a matter of policy reconsideration, to the first generation of a family’s beneficiaries and subjected to periodic review. Four of the seven judges therefore placed a creamy layer exclusion on the constitutional table, a position with no counterpart in Chinnaiah and no basis in the pleadings’ central question.

The observation’s legal status is precise and often misstated: the creamy layer passages are concurring opinion, not the operative holding of the court. The holding of Davinder Singh is that sub-classification is permissible on empirical justification. No state was directed to exclude a creamy layer, no criteria were laid down, and the Chief Justice’s judgment, which carries the plurality reasoning, treats the creamy layer considerations as part of the policy field a state may survey rather than as a condition of valid sub-classification. The distinction matters for the architectural reading. The permission the verdict created is a permission to differentiate among sub-groups on data. A creamy layer rule would differentiate within sub-groups, household by household, along an economic axis, and would import into the Scheduled Caste scheme the instrument built for the Other Backward Classes in Indra Sawhney, whose application to the Scheduled Castes the founding settlement had rejected on the ground that untouchability is a status, not an income band.

The political system’s answer arrived within days and was unambiguous. On 9 August 2024 the Union Cabinet, as reported from the briefing of the Minister for Information and Broadcasting, Ashwini Vaishnaw, recorded that there is no provision for a creamy layer in Scheduled Caste and Scheduled Tribe reservation in the Constitution as framed by B. R. Ambedkar, and that the government would proceed on the constitutional position. Representatives of the major parties, including parties that supported sub-classification itself, separated the holding from the observation in the same terms. The episode is instructive for the article’s claim: the deferral ended on the question the architecture had posed, differentiation among listed castes, and the verdict’s attempt to open a second question, differentiation within them, was refused by the political branches in the name of the same founding settlement. Whether that refusal holds as states draft their sub-classification schemes is an implementation question, and it is recorded here as disputed ground rather than as settled doctrine.

Did the 2024 verdict itself impose a creamy layer on the Scheduled Castes?

The verdict imposed no creamy layer. Four of the seven judges, led by Justice Gavai’s concurrence, urged that a state evolving sub-classification should identify and exclude a creamy layer within the Scheduled Castes and Scheduled Tribes. That urging is concurring opinion, not the court’s holding, and the Union Cabinet publicly declined it on 9 August 2024.

Scholarly reaction divided on the strand in terms that reproduce the article’s central dispute. Thorat’s published responses to the verdict welcomed the holding’s distributive logic and treated the creamy layer suggestion with caution, on the ground that the population of Scheduled Caste households that could be described as advanced is very small against the backlog of unfilled reserved posts, so that exclusion would ration a scarcity rather than correct a capture. Teltumbde read the observation more darkly, as evidence that the court’s social imagination, applied to the Scheduled Castes, defaults to the OBC template. Defenders of the concurrences answered that a sub-classification jurisprudence built on inadequate representation cannot coherently stop at the sub-group boundary: if data may show a caste under-represented, data may equally show a stratum within a caste fully represented, and a doctrine that accepts the first finding and forbids the second protects an internal elite in the vocabulary of solidarity. The adjudication in this article records the strand as the verdict’s most consequential open flank, precisely because the political refusal of August 2024 settled nothing in law: any state scheme that omits a creamy layer may be challenged on the concurrences, and any scheme that includes one will be challenged on the Cabinet’s constitutional claim, and the architecture will have to speak again.

The Dissent at Its Strongest: Justice Trivedi’s Objection

An argued adjudication owes the losing side its strongest form, and Justice Bela M. Trivedi’s dissent, the sole dissent on the bench, states the case against sub-classification more exactly than the prohibition’s defenders had stated it in 2004. The dissent begins where the majority begins, with the text. Article 341, in the dissent’s reading, is not a mere identification device: the President’s notification, issued after consultation and alterable only by Parliament, has the character of a constitutional settlement of the question who the Scheduled Castes are, and the homogeneity the majority dismisses as fiction is in truth a deliberate legal equality among the listed castes, each of which entered the schedule on the same ground of untouchability. A state that ranks the listed castes by measured backwardness substitutes its own scale for the settlement’s premise, and does so, the dissent observes, with instruments, state commissions and state data, that the constitutional design withheld from the states when it centralised the list.

The dissent’s second strand is institutional and predictive. Sub-classification, on the dissent’s account, hands state governments a distributive instrument of great electoral sensitivity, the power to enlarge one caste’s share of a fixed quota at another’s expense before an election, checked only by litigation after the fact on data the state itself assembles. The majority’s conditions, empirical justification and judicial review, are in the dissent’s view protections on paper: courts review reasonableness, not caste demography, and a determined government can manufacture a record. The dissent also reads Indra Sawhney against the majority: the nine judges of 1992, addressing the backward classes, expressly confined their reasoning and excluded the Scheduled Castes from its scope, and a bench of seven in 2024 that extends the OBC logic to the schedule is not applying Indra Sawhney but revising it without the authority of a larger bench. On each strand the dissent concludes that the remedy for intra-list inequality, if remedy there is to be, lies with Parliament under Article 341(2) and with the constitutional commissions under Articles 338 and 338A, not with state legislatures under Article 16(4).

The majority’s answer, which this article finds decisive on the text and incomplete on the institution, runs as follows. On the text: Article 341(2) governs inclusion in and exclusion from the list, and sub-classification does neither; the dissent must therefore locate the prohibition not in 341(2) but in an implied premise of equal treatment among listed castes, and implied premises cannot defeat the express distributive powers of Articles 15(4) and 16(4), whose own trigger, inadequate representation, is meaningless if representation may not be measured group by group. On the institution: the majority concedes the risk of electoral misuse and answers it with the reviewable-data condition, which converts the state’s evidentiary burden from a formality into the litigation battleground itself, as the Punjab and Andhra sequences already show. Where the dissent retains force is in its warning about frozen data. A sub-classification built on a commission’s findings in one decade can harden into entitlement in the next, and the verdict contains no sunset mechanism. That warning is carried into the implementation record examined next, and it is the strongest reason to treat the post-2024 schemes as experiments under review rather than as a settlement achieved.

Implementation Inside the Permission: The State Schemes of 2024 and 2025

The verdict’s conditions made the commission the central institution of the permission era, exactly as the commission had been the central institution of the deferral. The difference is that the post-2024 commissions write with a destination: their tables are drafted to survive judicial review under the empirical-justification condition, and their reports are the evidentiary record on which state statutes will stand or fall. The documentary output is large, technical and mostly published as gazette annexures and assembly papers; researchers assembling the state records side by side, the Punjab material against the Telangana tables against the Karnataka survey, can open the spreadsheet and document annexures through an office file viewer for Excel, DOCX and PPTX files, which is how the comparative record of the implementation period is most practically audited. Four state processes define the period, and the brief for this article directs attention to three of them by name.

Telangana moved first and most completely. In October 2024 the government of Chief Minister A. Revanth Reddy constituted a one-member commission under Justice Shamim Akhtar, a retired judge of the Telangana High Court, to recommend the grouping of the state’s 59 Scheduled Caste sub-castes for reservation. The commission toured the districts, received representations from the sub-caste organisations whose rival claims the verdict had made justiciable, and reported in February 2025. The Telangana legislature enacted the Telangana Scheduled Castes (Rationalisation of Reservations) Act in April 2025, distributing the state’s 15 percent quota among three groups: Group I, the fifteen most deprived sub-castes, received 1 percent; Group II, eighteen sub-castes including the Madiga group, received 9 percent; Group III, twenty-six sub-castes including the Mala group, received 5 percent. The Madiga organisations, whose three-decade movement had begun the national story, received the largest single share and objected that the grouping still understated the Raju-era finding of Madiga deprivation; Mala organisations announced a challenge. The statute is the permission era’s first full exercise: a rationalisation act, consciously titled after the Andhra Act of 2000 that Chinnaiah destroyed, defended by its drafters as the 2000 design rebuilt on the evidentiary foundation the 2024 verdict demanded.

Andhra Pradesh, the state of origin, followed with its own reconstruction, on the terrain left by the post-2014 Andhra-Telangana context. The government of N. Chandrababu Naidu appointed a commission under Rajeev Ranjan Mishra, a retired Indian Administrative Service officer, in late 2024 to re-examine the categorisation of the state’s 59 listed sub-castes, and in April 2025 the state promulgated the categorisation by ordinance and then by legislation, again in three groups in place of the 2000 Act’s four: the most deprived group at 1 percent of the 15 percent quota, a middle group at 6.5 percent and the most advanced group at 7.5 percent. The symmetry with Telangana is instructive and not accidental: the two successor states of the legislature that passed the 2000 Act rebuilt its architecture within eight months of the permission, each on a fresh commission record, each expecting the challenge the other invited. In Punjab the post-2024 process worked in the opposite direction, from statute to repair: the provision struck in 2010 and vindicated in principle in 2024 had to be re-enacted in conformity with the verdict’s conditions, and the state government of Bhagwant Mann committed to a fresh sub-classification of its 25 percent quota for the Mazhabi Sikh and Balmiki groups on updated representation data, with the rival listed castes’ organisations demanding that the data precede the percentages. Haryana notified its division of the state’s 20 percent quota in late 2024, reserving one half for the Deprived Scheduled Castes grouping identified by its state commission against the Other Scheduled Castes, and Karnataka, where the Justice H. N. Nagamohan Das Commission was constituted in November 2024 to survey the state’s listed castes, received its commission’s report in 2025 recommending internal reservation shares, with the governing Congress divided in public between the report’s early adopters and ministers urging a further census of the listed castes first.

The Telangana sequence shows the permission era’s procedural rhythm at its most complete, and it generalises. The state moved by ordinance in February 2025, immediately on the Akhtar Commission’s report, and carried the scheme into an Act in April 2025 when the legislature sat: commission, ordinance, statute, challenge. The challenge came at once, from Mala organisations arguing that the grouping understated Mala deprivation in the northern districts and from smaller sub-castes arguing that Group I’s 1 percent repeated, inside the new architecture, the tokenism the architecture existed to correct; the Telangana High Court was asked to test the scheme against the verdict’s data condition, which is precisely the litigation the majority designed the condition to host. In Andhra Pradesh the Mishra Commission’s grouping drew the mirror-image objection, Madiga organisations arguing that the middle group’s 6.5 percent against the advanced group’s 7.5 percent inverted the Raju Commission’s ordering of the two great castes. The objections cancel neither scheme; they demonstrate the permission working as specified, with distribution contested on published data before courts instructed to read it. A comparison with the prohibition era is the point: between 2004 and 2024 such objections had no forum, because the schemes themselves were unconstitutional at the threshold; after August 2024 the same objections are the ordinary business of reservation law, which is what the ending of a deferral looks like in practice. The wider party-system setting of these state contests, in which every major party holds both beneficiary and objector castes inside its coalition, is the terrain of the three-reading adjudication framework, and the state-party consequences of Mandal-era categorisation politics are traced in the post-Mandal party system study that follows this article in the pillar’s sequence.

Two large states frame the implementation period’s outer boundary, because both hold the question open rather than answered. Uttar Pradesh, with the largest Scheduled Caste population of any state, received a sub-categorisation blueprint before the permission existed: the Social Justice Committee under Hukum Singh, reporting in 2019, recommended dividing the state’s 21 percent quota among graded groups of its listed castes, and the recommendation waited five years for a doctrine that would allow it. After August 2024 the state government committed itself in principle to categorisation on the committee’s lines, and the familiar alignment formed at once, organisations of the state’s most deprived listed castes pressing for the percentages and organisations of its advanced listed castes demanding a fresh caste enumeration before any percentage is fixed. Bihar’s government, whose Mahadalit device had answered the same inequality from the welfare side, announced after the verdict that the state would examine sub-categorisation of the quota itself, setting the Mahadalit Commission’s categories against the service-representation record as the starting evidence. Neither state had enacted a scheme within the period this article surveys, and the pair is recorded deliberately: the permission’s geography is uneven, the largest theatres are the slowest, and the verdict’s consequences will be decided as much in Lucknow and Patna as in the states that legislated first.

The evidence base of the permission era deserves a methodological note, because the verdict’s data condition is only as strong as the data states can actually produce. The last full caste enumeration of the listed castes in a published census is old, the Socio-Economic and Caste Census of 2011 collected caste returns that the Union government declined to publish, and state commissions therefore work from service records, admission records and household surveys of their own design. The Telangana and Karnataka exercises commissioned fresh surveys for exactly this reason. A critic of the permission, and there are named critics in the state debates, can therefore attack any scheme at its foundation: the inter se backwardness of castes that share a district and an occupation can be estimated, and the estimate can be honest, and the estimate will still be contestable in ways a census figure is not. The verdict anticipated the objection by making the data reviewable rather than conclusive; the implementation period is discovering what that standard costs in litigation time.

Why the Deferral Lasted Twenty Years

A deferral of this length requires explanation, because the grievance was documented early and the doctrine was doubted early. Three mechanisms, acting together, held the prohibition in place long after its premise had ceased to persuade.

The first is the arithmetic of precedent. Chinnaiah was a five-judge decision. A three-judge bench that doubted it, as the Punjab appeal’s bench did in August 2014, could only refer. A five-judge bench that condemned it, as the Arun Mishra bench did in August 2020, could only refer upward again. The seven-judge bench had then to be constituted, listed and heard, in a court whose Constitution Bench capacity is consumed by a queue of references on federalism, money bills and personal law. Six years separated the first judicial doubt from the reference opinion, and four more the reference opinion from the hearing. The deferral’s second decade is substantially a story of bench arithmetic and docket scarcity, and it illustrates a general property of the architecture: a wrong answer delivered by a large bench is the most durable object in Indian constitutional law, because the machinery for correcting it is the scarcest resource in the system.

The second mechanism is political shelter. The prohibition served every incumbent’s convenience. A state government facing rival caste movements could commission an inquiry, publish its sympathy, and plead the Supreme Court’s bar in the same week, retaining the support of the advanced listed castes whose share the status quo protected and of the deprived castes whose demand it could claim to share. Chinnaiah functioned, for twenty years, as a subsidy to governmental indecision, and the subsidy was bipartisan: Congress, Akali, Telugu Desam, YSR Congress, DMK and Janata Dal (United) governments all drew on it in their turn. The movements understood the arrangement, which is why the Madiga organisations’ mature strategy targeted the doctrine rather than any single government, filing and funding the litigation that kept the reference alive through the 2010s. When the permission finally arrived, it arrived against the preference of most sitting governments for the sheltered position, and the speed of the 2024 and 2025 state legislation shows how much prepared demand the shelter had been holding back.

The third mechanism is the data deficit. The prohibition relieved the states of the need to maintain the evidentiary record that a permission would require, and the record duly decayed: the caste-wise census material of 2011 went unpublished at the centre, state service returns were not maintained caste-wise in most administrations, and the commissions of the period worked from fragments. By 2020 the five-judge bench could condemn the doctrine, but any bench permitting sub-classification had to answer Justice Trivedi’s later question in advance: on what data would states act, and who would check it? The four years between the reference opinion and the hearing were used, in part, by the states to rebuild the record, the Haryana and Tamil Nadu statutes and their supporting material travelling with the appeal as exhibits of feasibility. The verdict’s data condition is the deferral’s final product: the doctrine fell when, and arguably because, the evidentiary practice had been rebuilt to a standard a court could govern. Deferral, on this account, was not merely delay. It was the period in which the permission’s operating conditions were, painfully and by improvisation, manufactured.

Three Readings of the Same Twenty Years, Adjudicated

The state-specific reading of the sub-categorisation question treats each state dispute as a local product: Andhra’s from the Madiga movement’s strength, Punjab’s from the Balmiki and Mazhabi Sikh preference’s longevity, Tamil Nadu’s from the Arunthathiyar organisation of the western districts. The reading is accurate as political history and insufficient as explanation. It cannot say why states with unlike caste geographies, unlike party systems and unlike movement histories produced the same instrument in the same decades, nor why the disputes synchronised around the two verdicts. Local causes explain the timing of each dispute; they do not explain the recurrence of the form.

The legal-doctrinal reading treats the sequence as a contest between two constructions of Article 341, homogeneity against specification, conducted in the Supreme Court across twenty years and settled by bench strength. This reading captures the courtroom reality and misses the engine. Doctrine did not generate the question; the distributive record did, in commission after commission, and the doctrine of 2004 held as long as it did because no bench of sufficient strength was assembled to revisit it, not because its premise persuaded the states that lived under it. A purely doctrinal account also cannot explain the verdict’s most distinctive feature, the data condition, which is the court reaching outside doctrine to govern the evidentiary practice of legislatures.

The constitutional-architectural reading defended in this article subsumes both. The architecture, a central schedule joined to state distributive powers, generates the question wherever the schedule’s castes are unequal; the state disputes are the question’s local instances, and the doctrinal contests are the question’s juridical form. The reading also predicts what the other two cannot: that the question survived its prohibition, that its permission arrived with evidentiary conditions rather than a simple reversal, and that its next contests will be fought over data and grouping inside the permission, not over the permission itself. The verdict’s location inside the larger reservation settlement, including its relation to the integrated three-tensions framework, confirms the pattern this pillar has traced from the Poona Pact forward: in Indian reservation politics the durable settlements are architectural, and each generation inherits the previous generation’s architecture as the terrain of its own disputes. The integrated framework for this pillar’s adjudications is set out in the three-reading adjudication framework.

What the Permission Leaves Open

Four questions remain open inside the permission, and an evergreen account should name them without pretending to settle them. The first is the creamy layer question, left by the concurrences and refused by the Cabinet in August 2024, which will return the first time a state scheme is challenged for protecting an advanced stratum within a favoured sub-group. The second is the sunset question, Justice Trivedi’s strongest institutional point: the verdict imposes no periodic re-measurement, so each state’s grouping will age on its commission’s data until a legislature or a court reopens it, and the history of the 1975 Punjab preference, defended on 1970s data into the 2000s, shows how long an unrevised distribution can live. The third is the Scheduled Tribe question: the Davinder Singh bench addressed the Scheduled Castes and, in its reasoning, the Scheduled Tribes in parallel terms, and the appeals before it included tribal sub-classification disputes, but the operative litigation and the state schemes have concerned the Scheduled Castes, and the tribal lists, with their own internal inequalities and their protection under the Fifth and Sixth Schedules, will test the permission’s logic on different ground.

The tribal test will differ from the caste test in ways the data condition will feel. The tribal lists are organised around community and territory rather than around a hereditary occupation hierarchy, the inequalities among listed tribes run along lines of remoteness, language and exposure to mining and displacement, and several states have long operated informal gradations among tribes, the particularly vulnerable tribal groups category at the Union level being a standing administrative recognition that the tribal schedule is internally unequal. A state that moves to grade its tribal list will therefore argue from an existing practice of differentiation that the Scheduled Caste field lacked, and the Madiga-style movement politics of the caste case has its tribal counterpart in the demands of smaller tribes against the dominant tribes of states such as Jharkhand, Chhattisgarh and Odisha. The permission’s logic travels; its politics will have to be rebuilt locally, which is the architectural reading’s expectation rather than its embarrassment.

A fifth open question concerns the referee institutions. Articles 338 and 338A charge the National Commissions for the Scheduled Castes and Scheduled Tribes with monitoring the safeguards and inquiring into deprivation of rights, and the commissions were, across the deferral, the addressees of the references and representations the states generated when they could not legislate. The Andhra government referred its Raju proposals to the National Commission for Scheduled Castes in the 1990s, and commission opinion was cited by both sides in the Chinnaiah litigation. In the permission era the commissions’ role is unsettled: a state scheme’s data will be built by a state commission, reviewed by a court, and debated in a legislature, and the national commissions possess the standing to evaluate the whole field comparatively, state against state, in a way no single forum can. Whether they take that role, publishing the comparative representation record against which individual schemes can be judged, will determine much of the permission’s administrative quality, and the question is recorded here as open because the commissions’ post-2024 practice is still forming. The fourth is Parliament’s silence: Article 341(2) remains available, and a Parliament that wished to govern sub-classification nationally, setting data standards and review intervals, could do so, and the states’ occupation of the field since August 2024 is partly a function of that silence.

The review petitions against Davinder Singh were dismissed by the Supreme Court in October 2024, the bench recording that it found no error apparent in the judgment, so the permission stands as the settled starting point of the next period rather than as an interim position. For this pillar’s longer argument the sequence closes a circle opened at Poona in 1932: the settlement that fused the depressed classes into a single schedule for electoral purposes created, nine decades later, the distributive question the schedule could not answer, and the answer, when it came, preserved the schedule entire and governed the distribution beneath it. What ended on 1 August 2024 was a twenty-year deferral, and what began is a governed argument: about data, about grouping, about the internal order of the schedule, conducted by commissions, legislatures and courts inside a permission whose conditions are themselves the argument’s rules. That is the constitutional-architectural fact of the SC sub categorisation question, and it is why the verdict of 2024, and not the prohibition of 2004, states the durable law of the subject.

Frequently Asked Questions

Q: What is SC sub-categorisation?

SC sub-categorisation is the division of the Scheduled Castes named in a state’s Presidential List into sub-groups, with the reserved share in education and public employment distributed among the sub-groups in fixed proportions or by preference. Its purpose is distributive: where some listed castes have advanced far beyond others, a single undifferentiated quota tends to be captured by the relatively advanced, and sub-categorisation reserves a protected portion for the least reached castes. The Andhra Pradesh Act of 2000 graded 59 castes into four groups. The Supreme Court forbade the practice in 2004 and permitted it, on empirical justification, in 2024.

Q: What was the 2024 Davinder Singh verdict?

State of Punjab v Davinder Singh, decided on 1 August 2024 and reported at 2024 INSC 562, is a seven-judge Constitution Bench judgment that held, by six votes to one, that states may sub-classify the Scheduled Castes within the Presidential List for the purposes of reservation. Chief Justice D. Y. Chandrachud wrote the principal judgment for himself and Justice Manoj Misra, with separate concurrences from Justices B. R. Gavai, Vikram Nath, Pankaj Mithal and Satish Chandra Sharma, and a dissent from Justice Bela M. Trivedi. The verdict overruled the 2004 Chinnaiah prohibition and attached conditions of empirical data and judicial review.

Q: What is the E.V. Chinnaiah verdict?

E. V. Chinnaiah v State of Andhra Pradesh, decided on 5 November 2004 and reported at (2005) 1 SCC 394, is the five-judge Constitution Bench judgment that struck down the Andhra Pradesh Scheduled Castes (Rationalisation of Reservations) Act of 2000. The bench, comprising Justices N. Santosh Hegde, S. N. Variava, B. P. Singh, H. K. Sema and S. B. Sinha, held that the Presidential List under Article 341 constitutes a single homogeneous class, that only Parliament may vary it, and that a state sub-dividing the list for reservation purposes tinkers with the list unconstitutionally. The prohibition governed India for twenty years.

Q: Why was the 2004 ruling overturned?

The seven-judge bench of 2024 held that Chinnaiah rested on a false premise: the Presidential List was never a finding that the listed castes are homogeneous, and the description of the Scheduled Castes as a class in the 1975 N. M. Thomas judgment concerned their position against the general category, not their internal sameness. Once the premise fell, the prohibition fell with it, because Articles 15(4) and 16(4) authorise distribution by class and by inadequate representation and contain no bar on measuring representation within the schedule. A five-judge bench had already held Chinnaiah wrongly decided in August 2020 and referred it upward.

Q: What is the Madiga-Mala distinction?

The Madiga and the Mala are the two largest Scheduled Castes of the Telugu-speaking states, and the distinction between them drove the sub-categorisation movement. The Mala of the coastal districts encountered mission schooling, military recruitment and the Adi Andhra movement’s institutions earlier, and entered professional education and the services in numbers well beyond their population share. The Madiga, concentrated in Telangana and Rayalaseema and bound for longer to hereditary leather work and agricultural servitude, held a small fraction of their proportional share of reserved posts by the 1990s. The Madiga movement from 1994 demanded a graded partition of the quota on that record.

Q: How does sub-categorisation work in Punjab?

Punjab’s instrument is a preference rather than a partition. An executive instruction of 1975 directed that half the Scheduled Caste vacancies be offered first to Balmiki and Mazhabi Sikh candidates, the state’s most deprived listed castes. After the Punjab and Haryana High Court struck the instruction down on Chinnaiah’s authority in 2006, the Punjab Scheduled Castes and Backward Classes (Reservation in Services) Act of 2006 enacted the same first preference in Section 4(5), with the preference lapsing to the general Scheduled Caste pool where preferred candidates are unavailable. The High Court struck the section down in 2010, and the state’s appeal became Davinder Singh.

Q: What is creamy-layer-exclusion in SC reservation?

Creamy-layer exclusion is the proposal that the advanced stratum within the Scheduled Castes, families already lifted by reservation into the higher services and professions, be identified and excluded from further benefit, extending to the Scheduled Castes the device the Supreme Court built for the Other Backward Classes in Indra Sawhney in 1992. In Davinder Singh, four judges in concurrence, led by Justice B. R. Gavai, urged states to evolve such an exclusion. It formed no part of the court’s holding, and on 9 August 2024 the Union Cabinet publicly declined it, holding that the Constitution contains no creamy layer provision for Scheduled Caste reservation.

Q: What is Article 341 of the Indian Constitution?

Article 341 empowers the President, after consultation with the Governor of a state, to specify by public notification the castes, races or tribes that shall be deemed Scheduled Castes for that state, and reserves every later inclusion in or exclusion from the list to Parliament acting by law. The lists are state-specific: a caste may be scheduled in one state and not in another. The article governs identity, who counts as a Scheduled Caste for the Constitution’s protective and affirmative provisions. The Davinder Singh majority held that it does not govern the distribution of reservation benefits among the castes it names.

Q: Who are the Balmikis and Mazhabi Sikhs in the Punjab quota dispute?

The Balmikis are a Scheduled Caste of Punjab concentrated in the towns, historically associated with sanitation work and descended in large part from the Chuhra caste. The Mazhabi Sikhs are Sikh converts drawn from the same Chuhra background, settled predominantly as agricultural labourers in the Malwa countryside. Punjab’s administrations from 1975 onward treated the two groups as the state’s most deprived listed castes, far behind the Ad-Dharmi and Ravidassia castes in schooling, service representation and overseas migration. The state’s 25 percent Scheduled Caste quota therefore carried a first preference for Balmiki and Mazhabi Sikh candidates, the provision at the heart of the Davinder Singh litigation.

Q: What did the Ramachandra Raju Commission recommend in Andhra Pradesh?

The government of Andhra Pradesh appointed Justice P. Ramachandra Raju, a retired judge of the Andhra Pradesh High Court, in 1996 to identify the groups among the state’s Scheduled Castes that had failed to secure the benefit of reservation. The commission examined representation in the services and in professional colleges and found the capture of the 15 percent quota by a minority of listed castes documented and severe. It recommended dividing the 59 castes on the state list into four groups graded by backwardness and fixing each group’s share of the quota, the design the legislature enacted as the Andhra Pradesh Scheduled Castes (Rationalisation of Reservations) Act of 2000.

Q: What is the Madiga Reservation Porata Samithi?

The Madiga Reservation Porata Samithi, the Madiga Struggle Committee for Reservation, is the organisation founded in 1994 under Manda Krishna Madiga to demand categorisation of the Scheduled Caste quota in Andhra Pradesh. The Samithi and its successor formations, including the Madiga Dandora movement, carried the demand through mass mobilisation across the Telugu districts, converting a distributive grievance into an electoral force that no state government could ignore. Its campaign produced the Raju Commission of 1996 and the Act of 2000, and its leaders remained central parties to the categorisation politics of Telangana and Andhra Pradesh after the 2024 verdict reopened the question.

Q: How did Tamil Nadu protect the Arunthathiyar share of the SC quota?

Tamil Nadu acted on the report of a state commission under Justice M. S. Janarthanam, which examined the representation of the Arunthathiyar castes, Chakkiliyar and allied groups of the western districts working predominantly in leather work and sanitation, and found them close to absent from the benefits of the state’s 18 percent Scheduled Caste quota. The Tamil Nadu Arunthathiyars Reservation of Seats in Educational Institutions and of Appointments or Posts in the Services under the State Act of 2009 reserved 3 percent of that quota for the Arunthathiyar group. The Act survived challenge in the Madras High Court and formed part of the national pattern the seven-judge bench reviewed.

Q: What is the Mahadalit category created in Bihar?

The Mahadalit category is a Bihar state designation, created on the recommendation of the Bihar State Mahadalit Commission of 2007 under the government of Nitish Kumar, covering in its first form all but one of the state’s 22 Scheduled Castes, those judged extremely deprived on the commission’s indicators. Unlike Andhra’s graded partition, the Mahadalit designation directed development missions, housing, land and scholarship programmes toward the category without dividing the service quota itself. The device answered the same measured inequality that drove sub-categorisation elsewhere while staying clear of the Chinnaiah prohibition, since it touched welfare delivery rather than the reservation articles.

Q: Why did a five-judge bench refer Chinnaiah for reconsideration in 2020?

Hearing the Punjab appeal against the striking down of Section 4(5) of the Punjab Act of 2006, a five-judge Constitution Bench headed by Justice Arun Mishra held on 27 August 2020 that Chinnaiah had not been correctly decided. The bench reasoned that the Scheduled Castes are not a homogeneous class for the distribution of reservation benefits, that a state answering measured internal inequality acts within the object of Articles 15 and 16, and that the 2004 bench had misread the equality code and the scope of Indra Sawhney. A bench cannot overrule a decision of a coordinate bench, so the 2020 judges referred Chinnaiah to a larger bench, which produced the seven-judge hearing.

Q: What did Justice Bela Trivedi argue in her dissent in Davinder Singh?

Justice Bela M. Trivedi, the sole dissenter, held that Article 341 is a complete settlement of who the Scheduled Castes are, carrying a deliberate legal equality among the listed castes, and that state grading substitutes a state scale for that settlement. She argued that the nine judges in Indra Sawhney had expressly excluded the Scheduled Castes from the sub-classification reasoning applied to the Other Backward Classes, so the majority was revising Indra Sawhney without a larger bench’s authority. She warned that state-assembled data and post-facto judicial review are weak checks on an instrument of high electoral value, and located the remedy with Parliament under Article 341(2).

Q: What did Justice B. R. Gavai say about the creamy layer in 2024?

Justice B. R. Gavai’s concurrence in Davinder Singh held that the state must justify preferential treatment within the Scheduled Castes on empirical data showing inadequate representation, that no state may hand the entire reserved share to one sub-group, and that the state should evolve a policy to identify a creamy layer within the Scheduled Castes and Scheduled Tribes and exclude it from reservation benefits, since a family raised by one generation of reserved posts into the higher services stands differently from the families the quota has not reached. Justices Vikram Nath, Pankaj Mithal and Satish Chandra Sharma expressed agreement with applying a creamy layer principle.

Q: Can a state give the entire SC quota to one sub-group after Davinder Singh?

A state cannot. The Davinder Singh majority permitted sub-classification as a device for distributing the reserved share among the listed castes on evidence of inter se backwardness and inadequate representation, and Justice Gavai’s concurrence, accepted in this respect across the majority opinions, states the limit expressly: the whole of the quota cannot be set apart for a single sub-group, excluding the remainder of the Presidential List. A scheme that transferred the entire share would cease to be classification within a class and would operate as exclusion, the act Article 341(2) reserves to Parliament, and would fail the verdict’s own conditions on judicial review.

Q: How many Scheduled Castes were divided by the Andhra Pradesh Act of 2000?

The Andhra Pradesh Scheduled Castes (Rationalisation of Reservations) Act of 2000 divided the 59 castes then entered on the state’s Presidential List into four groups, on the grading recommended by the Ramachandra Raju Commission. Group A, the Relli group of castes, received 1 percent of the 15 percent state quota. Group B, containing the Madiga and allied castes, received 7 percent. Group C, containing the Mala and allied castes, received 6 percent. Group D, the Adi Andhra group, received 1 percent. The four-to-six-to-seven distribution among the larger groups was the statute’s contested centre, litigated to the Supreme Court as Chinnaiah.

Q: What happened to the review petitions against the Davinder Singh verdict?

Review petitions against State of Punjab v Davinder Singh were filed by parties dissatisfied with the permission, including organisations of listed castes that had defended the Chinnaiah settlement and litigants opposed to the creamy layer observations. In October 2024 the Supreme Court dismissed the petitions, recording that it found no error apparent on the face of the judgment warranting review. The dismissal left the seven-judge holding as the settled law of sub-classification: states may proceed on empirical justification, subject to challenge of individual schemes on their data and design, and the constitutional question itself is closed unless a future bench of greater strength reopens it.

Q: How did Telangana legislate sub-categorisation after the 2024 verdict?

Telangana appointed a one-member commission under Justice Shamim Akhtar, a retired judge of the Telangana High Court, in October 2024 to recommend the grouping of the state’s 59 Scheduled Caste sub-castes. The commission toured the districts, took representations from the sub-caste organisations and reported in February 2025. The legislature then enacted the Telangana Scheduled Castes (Rationalisation of Reservations) Act in April 2025, dividing the 15 percent quota among three groups: fifteen sub-castes in Group I received 1 percent, eighteen in Group II, including the Madiga group, received 9 percent, and twenty-six in Group III, including the Mala group, received 5 percent.