The judgment in Indra Sawhney v. Union of India is cited in almost every argument about reservation in India, usually as a single sentence: total reservation cannot exceed fifty per cent. That sentence is a fragment of what the nine judges decided on 16 November 1992, and the fragment is the part of the judgment whose authority has since been most heavily contested. The bench answered a set of distinct questions about the equality code of the Constitution, and the answers have had very different fates. One was reversed by constitutional amendment within three years. One was dismantled in two stages by later benches. One survived a direct attack in 2021 and was then held, in 2022, not to bind the very amendment it was invoked against. One, the exclusion of the creamy layer, has never been displaced and has quietly become the most consequential sentence in the case. An account of Indra Sawhney that stops at the ceiling therefore misdescribes the law twice over: it treats the most contested holding as the most secure, and it misses the holdings that govern litigation in the services every working day. This article takes the judgment apart holding by holding, traces each holding to its status after the changes that followed, and states precisely what survives, what was amended away, and what remains open.

The article sits inside a cluster. The legislative history of the Mandal Commission, the decade between its report and its implementation, and the politics of that decade belong to the article on the Mandal Commission and the law of OBC reservation, which owns that story. The wider field, spanning the protective statutes, the commissions and the affirmative provisions together, is mapped in the social justice legislation guide. What belongs here is the doctrine: the questions the nine judges framed, the answers a majority gave, the reasoning that carried those answers, and the sequence of amendments and judgments that has reworked them since. The reader who finishes should be able to state, for any proposition commonly attributed to Indra Sawhney, whether it was held at all, whether it still holds, and which instrument changed it if it does not.

The Text That Did Not Answer Its Own Questions

The interpretive problem begins with the structure of Articles 14, 15 and 16. Article 14 guarantees equality before the law. Article 16(1) guarantees equality of opportunity in public employment, and Article 16(2) forbids discrimination in that field on grounds including caste. Article 16(4) then provides that nothing in Article 16 prevents the State from making provision for the reservation of appointments or posts in favour of any backward class of citizens which, in the opinion of the State, is not adequately represented in the services. Article 15(4) performs a parallel operation for education and social advancement. Article 335 requires that the claims of the Scheduled Castes and the Scheduled Tribes be taken into consideration consistently with the maintenance of efficiency of administration, and Article 46, among the Directive Principles, directs the State to promote with special care the educational and economic interests of the weaker sections.

Read plainly, that text leaves every operative question unanswered. It does not say whether clause (4) is an exception to clause (1), to be read narrowly against the guarantee, or a restatement of what the guarantee already permits, to be read generously with it. It does not fix any maximum share. It does not say whether reservation stops at the door of the service or follows an officer through every promotion. It does not define a backward class, does not say whether caste can identify one, and does not say whether poverty alone can. It does not say what happens to a reserved vacancy that no eligible candidate fills. Each of those silences became a question the courts had to answer, and the answers were built, case by case, over three decades before nine judges attempted a settlement. The settlement is the subject of this article, and the first thing to grasp about it is that the bench was not writing on a clean page. It was choosing among lines of authority that pointed in different directions, and on some questions it overruled decisions that had stood for thirty years.

One phrase of the text has carried more weight than its modest grammar suggests: the class must be one which, “in the opinion of the State”, is not adequately represented in the services. The subjective form was a deliberate drafting choice, locating the operative judgment in the government rather than in the courts, and the first question the case law had to settle was what a court may do with an opinion the Constitution commits to another authority. The answer the doctrine converged on, and which Indra Sawhney confirmed and Nagaraj sharpened, is a division of labour. The opinion is the State’s to form; the court does not substitute its own assessment of adequacy. But the opinion must be formed on material that can bear it, the identification it rests on must satisfy the classification test, and the measure adopted under it must observe the doctrine’s limits. Subjective in form, the clause thus operates objectively in litigation, because a State that cannot produce its material loses. Almost every defeat a reservation measure has suffered in this field, from Balaji’s defective identification to the Maratha statute’s insufficient record, is a defeat at this joint: not a holding that the State lacked the power, but a holding that the State had not done the work the power’s exercise requires.

What exactly was challenged in Indra Sawhney?

Two office memoranda of the Union government were challenged. The first, of 13 August 1990, reserved twenty seven per cent of vacancies in central services and public sector posts, filled by direct recruitment, for the socially and educationally backward classes. The second, of 25 September 1991, modified the first in two respects that the judgment later struck at.

It kept the twenty seven per cent share, added a preference within it for the poorer sections of those classes, and added a further ten per cent for economically weaker sections not covered by any existing scheme. A five judge bench stayed the first memorandum, and the challenge was then placed before nine judges.

Behind the memoranda stood the Commission’s method, which the judgment recites and which the identification holding then judged. The Mandal Commission had assessed backwardness through eleven indicators grouped under three heads: social indicators, carrying the greatest weight, which included the community’s own perception of its standing, marriage practices and the position of women; educational indicators, including literacy and dropout rates measured against the State average; and economic indicators, including asset holding and housing. A caste scoring above the halfway mark on the weighted scale was classified as backward, and on that method the Commission identified some three thousand seven hundred castes and communities, estimated at about fifty two per cent of the population, as socially and educationally backward, recommending for them a reservation of twenty seven per cent, a figure chosen precisely because the existing reservation for the Scheduled Castes and the Scheduled Tribes, together with the new share, would approach but not cross the fifty per cent line that Balaji had indicated. The figure of twenty seven per cent is therefore not an assessment of the backward classes’ share of the population; it is the residue the ceiling left available, and the majority’s decision to test the identification by the Commission’s indicators rather than to re perform the census of backwardness is the reason the judgment reads as a work of review rather than of social inquiry. The history of the Commission, of the decade in which its report waited, and of the governments that announced and deferred its implementation belongs to the legislative history article linked above; what the interpretation article must carry is only this, that the Court was reviewing a method, and it is the method’s judicial validation, more than any list, that the sixth holding decided.

The two memoranda matter because the judgment’s holdings answer them piece by piece. The twenty seven per cent figure was the Mandal Commission’s recommendation for the backward classes, and the Court upheld it. The preference for poorer sections inside the backward classes was not given effect, because the majority substituted its own filter, the exclusion of the creamy layer, for the government’s. The ten per cent for economically weaker sections outside the backward classes was struck down, because the majority held that a class for the purposes of Article 16(4) cannot be constituted by economic condition alone. A reader who keeps the three components separate can follow the whole judgment, because almost every holding is the Court accepting, modifying or rejecting one of them. The arguments against the memoranda were the arguments of the equality guarantee itself: that caste is a forbidden ground and cannot be made the basis of a benefit, that a benefit defined by birth contradicts the promise of equal opportunity, and that administrative efficiency, which the Constitution names in Article 335, would be the casualty. The arguments for the memoranda were the arguments of substantive equality: that a guarantee of equal opportunity addressed to a society of entrenched hereditary disadvantage must permit the State to identify the disadvantaged and reserve a share for them, and that the text of clause (4) says so in terms.

Where the Clauses Came From

The text the nine judges read had itself been written in argument with the courts. Article 16(4) stood in the Constitution from its commencement; it was the framers’ own provision, moved through the Constituent Assembly with remarkably little controversy, because service reservation for backward classes was an established administrative practice in several princely states and presidencies long before independence. Article 15(4) has a different origin. In State of Madras v. Champakam Dorairajan, decided in 1951, the Supreme Court struck down the Madras communal order apportioning seats in medical and engineering colleges, because Article 15 as enacted contained no enabling clause comparable to Article 16(4) and the Directive Principles could not override a fundamental right. Parliament answered with the Constitution (First Amendment) Act, 1951, inserting Article 15(4) and thereby writing into the equality code the distinction the whole field still observes: educational and social advancement provision rests on a clause added after judicial defeat, while service reservation rests on a clause the framers supplied themselves. The sequence is the field’s founding precedent for the amendment cycle this article traces repeatedly, in which a judicial statement of the doctrine is met by a legislative judgment that the doctrine should be otherwise, drafted to the width of the disagreement.

Two further pieces of the architecture matter for everything that follows. The first is the lists. Articles 341 and 342 empower the President to specify, for each State, the castes and tribes deemed to be Scheduled Castes and Scheduled Tribes, and reserve subsequent inclusion and exclusion to Parliament by law. Backward classes have no equivalent constitutional list; their identification is an executive and commission driven exercise under Articles 15(4) and 16(4), informed by Article 340, which empowers the President to appoint a commission to investigate the conditions of backward classes. The Mandal Commission was the second commission appointed under that article, and its report, submitted in 1980, is the factual foundation on which the 1990 memorandum and therefore the 1992 judgment rest. The second piece is the asymmetry the list structure creates. For the scheduled classes, membership is fixed by a constitutional instrument and backwardness is presupposed by the listing itself; for the backward classes, membership must be demonstrated class by class and is open to judicial examination. That asymmetry, invisible in most summaries of the case, is what later allowed the Jarnail Singh bench to strike down a data requirement for the scheduled classes as contrary to Indra Sawhney while leaving the evidential discipline for backward class identification entirely intact.

The Road to the Nine Judge Bench

The ceiling has a genealogy, and it does not begin in 1992. In M. R. Balaji v. State of Mysore, decided in 1963 and reported at AIR 1963 SC 649, the Court struck down a Mysore order that reserved sixty eight per cent of seats in medical and engineering colleges. The Court held that backwardness under Article 15(4) must be both social and educational, that caste is a relevant factor but cannot be the sole test, and, speaking generally and in a broad way, that a special provision should be less than fifty per cent, because a provision that consumes most of the seats ceases to be a special provision and becomes the rule. That sentence was an observation about reasonableness, framed with deliberate looseness, and for thirty years lawyers argued about whether it was a rule at all.

In T. Devadasan v. Union of India, reported at (1964) 4 SCR 680, the observation hardened in application. The Union’s carry forward rule allowed reserved vacancies that went unfilled in one year to be carried into the next, and in the year under examination the arithmetic produced a startling result: of forty five vacancies filled, twenty nine had gone to candidates of the Scheduled Castes and the Scheduled Tribes. The majority treated Balaji as having laid down that reservation could not exceed fifty per cent of the vacancies of a year, and struck the carry forward rule down. The case fixed two ideas that Indra Sawhney later had to handle: that the ceiling, if there is one, applies to the vacancies of a single year rather than to the composition of a service over time, and that a rule about unfilled vacancies can defeat a ceiling without ever raising the headline percentage.

The promotion question travelled the opposite way. In General Manager, Southern Railway v. Rangachari, reported at AIR 1962 SC 36, the Court upheld reservation in promotion, reading clause (4) as extending to every stage of a service career. That view held for three decades, was applied in later cases, and shaped the expectations of a generation of officers, which is why the nine judges’ contrary answer in 1992 required a transitional direction and why Parliament’s reversal of it required a constitutional amendment rather than an ordinary statute. In State of Kerala v. N. M. Thomas, reported at (1976) 2 SCC 310, a bench of seven judges held by a majority that Article 16(4) is not an exception to Article 16(1) at all, but an emphatic statement of the classification that clause (1) itself permits; equality, on that reading, is not offended by reservation but expressed through it, provided the classification is rational and the share is not excessive. The facts of Thomas are a useful corrective to the habit of reading it as pure theory. The case concerned a Kerala service rule that exempted employees of the Scheduled Castes and the Scheduled Tribes, for a specified period, from the departmental tests that promotion required, and the question was whether that exemption, a concession inside a career rather than a reservation of posts, fell within the State’s power. In upholding it, the majority reasoned from the structure of the equality code itself, that Articles 14, 15 and 16 form a single scheme in which classification for the benefit of the disadvantaged is an instrument of the guarantee and not a derogation from it. The reasoning mattered beyond its facts because it dissolved the premise on which strict construction of clause (4) rested: if reservation is the guarantee working, the burden of justification rearranges itself, and the questions become the classification questions of rationality, proportion and limit. Indra Sawhney adopted that rearrangement and then supplied, as Thomas had not, the limits.

Thomas supplied the theory. What it did not supply was a stable set of limits, and a five judge bench in K. C. Vasanth Kumar v. State of Karnataka in 1985 produced separate opinions pointing in several directions, including the suggestion that reservation should be reviewed periodically and should not harden into permanence. By the end of the 1980s the field had a theory, a contested ceiling, an ageing carry forward rule, a settled-seeming promotion rule, and no single authoritative statement reconciling them. That is the state of the law the nine judges inherited.

The Judgment: Bench, Opinions and Shape

The bench of nine consisted of M. H. Kania, Chief Justice, and M. N. Venkatachaliah, S. Ratnavel Pandian, T. K. Thommen, A. M. Ahmadi, Kuldip Singh, P. B. Sawant, R. M. Sahai and B. P. Jeevan Reddy, JJ. The judgment is reported at 1992 Supp (3) SCC 217. By a majority of six to three, the Court upheld the twenty seven per cent reservation for the backward classes, subject to the conditions the majority attached. B. P. Jeevan Reddy J. wrote the principal majority opinion for himself and for Kania CJ and Venkatachaliah and Ahmadi JJ. P. B. Sawant and S. Ratnavel Pandian JJ. wrote separately and concurred in the result. T. K. Thommen, Kuldip Singh and R. M. Sahai JJ. dissented. The internal differences inside the majority matter, and one is worth flagging at the outset: Pandian J., while agreeing that the memoranda should substantially stand, declined to accept that any maximum ceiling on reservation could be fixed at all, so the ceiling is a majority holding within the majority, not a unanimous position of the six.

The Court framed a series of questions and answered them in sequence, and the discipline of this article is to keep those answers separate, because later law treated them separately. Parliament could reverse the promotion answer without touching the ceiling. A later bench could discard the data requirement that Nagaraj built on the promotion amendments without touching the creamy layer. A still later bench could overrule the assumption that the Scheduled Castes form a single indivisible class without touching any of the eight answers directly. The erosion of Indra Sawhney is not a general fading of its authority. It is a series of specific operations performed on specific holdings, and each operation has its own instrument, its own date and its own reasoning. Stating the eight holdings cleanly is therefore the necessary first step.

The argument before the bench covered the whole field, and its shape explains the judgment’s structure. The petitioners, led by the advocate Indra Sawhney whose name the case carries, advanced three principal contentions: that the extension of reservation to the backward classes in the form proposed violated the guarantee of equality of opportunity; that caste is not a reliable indicator of the backwardness the clause requires; and that the efficiency of public institutions, the interest Article 335 names, would be put at risk by reservation on the proposed scale. Around those contentions ranged the arguments of the Union and the intervening States, defending the memoranda as the overdue implementation of a commission’s findings, and the arguments of parties who would have gone further than the memoranda in one direction or the other. The case had also a procedural peculiarity that shaped its reception: a five judge bench had stayed the operation of the 1990 memorandum at an early stage, so the policy the nine judges finally upheld had been announced, suspended and argued over for more than two years, in a period of intense public agitation that the judgment itself does not describe but which forms the background against which its transitional directions were framed.

A multi opinion judgment of nine judges poses a further problem that any honest account must state: the majority is a different assembly on different questions. Six judges upheld the memoranda in substance, but on the promotion question eight of the nine agreed; on the ceiling, the operative proposition commanded the principal opinion and the general concurrence of Sawant J. but not Pandian J.; on the character of clause (4), the facet reading commanded the principal line of authority descending from Thomas. Later courts extract the law of such a case from the conclusions on which the requisite majority converged question by question, and the Constitution Bench in the Maratha case, when it treated the ceiling as the declared law binding under Article 141, was performing exactly that extraction. This is why the holdings are stated in this article as eight separate propositions rather than as a single ratio. The judgment has no single ratio. It has a lattice of propositions with different majorities and, as events proved, different life expectancies.

The Eight Holdings of Indra Sawhney, One by One

The first holding is the foundation: reservation for backward classes in public employment is constitutionally permissible, and the proportion notified, twenty seven per cent for the socially and educationally backward classes in central services, was upheld. The majority rejected the argument that the identification of the beneficiary classes was a political act immune from review, and equally rejected the argument that any caste-linked benefit was forbidden. The identification of a backward class is a justiciable exercise: the Court will examine whether the class identified is genuinely backward and whether the material relied on can support the conclusion. On the material before it, the Mandal identification, tested against the indicators the Commission had used, survived. This holding has never been disturbed, and everything else in the judgment is a set of conditions on the power it recognises.

The second holding created the creamy layer. Within any backward class there are sections that have advanced socially, educationally and economically to the point where their continued inclusion defeats the purpose of the classification: a person who has risen beyond the backwardness that founds the class is, for this purpose, no longer of it. The majority directed that such advanced sections, described in the judgment as the creamy layer, be excluded from the benefits of reservation, and left the criteria of exclusion to be worked out by an expert committee and the government. The criteria the Union issued in 1993 rested principally on the status and posts held by a candidate’s parents, with income and property tests alongside, which is why the exclusion has always operated as a status filter more than as a simple means test. The holding is doctrinally the most original in the case, because it converts a class-level classification into an individual-level inquiry inside the class, and it is the holding whose reach later benches kept extending.

The design of the 1993 criteria shows what the doctrine takes advancement to mean. The core exclusions attach to office and occupation rather than to wealth alone: the children of persons holding constitutional posts, of officers of the higher civil services, of senior ranks in the armed forces, and of established professionals and large property holders are excluded by the position of the family, on the reasoning that institutional power and social capital, once acquired, compound across a generation in ways a salary figure does not capture. Around that core sits an income and wealth test for the remaining families, a threshold the Union has revised upward in stages as prices and salaries moved. The structure explains a persistent confusion in public argument, where the creamy layer is discussed as if it were an income ceiling and nothing more. It is not. A candidate may be excluded at an income below the threshold because of a parent’s office, and a candidate from a family with a high but transient income may survive the filter if no status exclusion bites. The doctrine’s target is not the prosperous individual but the family whose advancement is structural, and its administrative instrument, the non creamy layer certificate, has generated a large procedural law of its own about whose income counts, over what period, and before which authority.

The third holding is the ceiling. The majority held that the total reservation, of posts or of seats, should not ordinarily exceed fifty per cent, and that a breach could be justified only in extraordinary situations arising out of exceptional circumstances, with reasons recorded. The illustrations the judgment gave were of populations in far flung and remote areas whose conditions kept them out of the mainstream of national life; the exception was designed for geography and isolation, not for political arithmetic. Two features of the holding are routinely lost in summary. The first is that the ceiling was stated as an incident of the equality guarantee rather than as a rule of arithmetic: a reservation beyond half, the reasoning runs, converts the exception into the rule and denies the open competition that Article 16(1) promises to everyone else. The second is that the majority itself declined to treat its figure as a discovery in the text. The figure descends from Balaji’s broad-way observation, and its justification in 1992 was stare decisis and workability as much as deduction.

Why does the fifty per cent ceiling apply year by year?

Because reservation is administered against vacancies, and the equality injury the ceiling guards against is suffered by the candidates of a single recruitment year. The majority therefore held that in any given year the total of reserved appointments, including vacancies carried forward from earlier years, must remain within the ceiling.

The carry forward of unfilled reserved vacancies is valid in itself; what the ceiling forbids is their accumulation defeating the limit in the year of filling. A service examined over decades might show a balanced composition while a single year’s recruitment, swollen by accumulated backlog, offered open candidates almost nothing; it is the year, not the service, that the candidate experiences.

That is the fourth holding: the carry forward rule survives, but subject to the ceiling in the year of filling. Devadasan’s outcome is thereby reversed in its reasoning while its anxiety is preserved. The rule that unfilled reserved vacancies may be carried forward was not struck down; the proposition that the ceiling does not apply to the accumulated result was rejected. The practical consequence is the accounting problem that dominates service litigation: whether a vacancy is a fresh vacancy of the year or a backlog vacancy carried into it, against which point of the roster it is to be counted, and whether the year’s total has stayed within the limit. Parliament later revisited this holding too, and its answer, in 2000, was to remove backlog vacancies from the ceiling’s arithmetic altogether.

One scope question the ceiling’s summary form conceals is what the ceiling counts. It counts vertical reservation: the shares allotted to the scheduled classes, the scheduled tribes and the backward classes as compartments of the whole. It does not count, by addition, the horizontal reservations that cut across those compartments, for women, for persons with disabilities, for ex servicemen and the like, because a horizontal reservation is satisfied within each vertical compartment rather than added on top of it: a woman appointed against a horizontal point occupies a point of the compartment she belongs to, and the vertical total is unchanged. The distinction was recognised in the working of the doctrine from the start and was restated when a later bench, in litigation over reservation for persons with disabilities in promotion, held that the Indra Sawhney bar on promotion reservation arose in the field of Article 16(4) and does not disable a horizontal provision resting on a different constitutional footing. The practical importance is not small. Public argument habitually adds every percentage in sight and announces that the ceiling is breached, in services where the arithmetic, correctly kept, shows nothing of the kind; and litigation has repeatedly turned on whether a given preference is a vertical reservation within the ceiling’s field or a horizontal provision outside the addition. The question is one of characterisation, and it is answered by asking within which compartments the benefit operates, not by asking how large it is.

A worked illustration, with invented round numbers used only to show the mechanism, makes the difference between the two regimes concrete. Suppose a service recruits against one hundred vacancies in a year, with an aggregate reservation of forty nine per cent across the scheduled classes and the backward classes, and suppose twelve reserved vacancies from earlier years stand unfilled for want of eligible candidates. Under the 1992 rule, the appointing authority could fill at most fifty of the year’s posts from the reserved categories in total; if it cleared the twelve backlog vacancies, only thirty eight of the year’s own reserved points could be filled, and the balance of the year’s reserved entitlement itself rolled forward, so that a persistent shortfall of eligible candidates converted the ceiling into a ratchet that held the reserved share of every year’s intake below its nominal percentage. Under Article 16(4B), the twelve backlog vacancies form a separate class: they may all be filled, and the year’s own reservation is computed against the year’s vacancies without reference to them, so the reserved share of the year’s total intake may lawfully pass fifty per cent in a backlog clearing year. Neither regime changes the nominal percentage of reservation, which is why the argument between them was conducted as an argument about accounting. But the difference in outcomes, compounded across the services of the Union and the States, is the difference between a ceiling that bites on every year’s intake and a ceiling that bites only on the vacancies of the year in which the entitlement arose, and it is the kind of difference that service litigation, which is arithmetic before it is doctrine, is fought over.

The fifth holding is the one with the shortest life. The majority held that reservation under Article 16(4) is confined to initial appointment and does not extend to promotion. The reasoning was that the handicap of backwardness is addressed at entry, that to reproduce the preference at every stage of a career would compound it, and that efficiency under Article 335 weighs more heavily as posts ascend. On this question the bench was nearly united: eight of the nine judges held reservation in promotion impermissible, and Indra Sawhney overruled Rangachari to the extent it had held the contrary. The majority added a transitional direction of real human importance: reservations in promotion that already existed were to continue for five years from the judgment and then cease, a direction that protected settled expectations while announcing the rule’s end. That five year window is the space in which Parliament acted.

The sixth holding governs identification. A backward class cannot be identified solely by an economic criterion; a reservation constructed on poverty alone, outside the apparatus of social and educational backwardness, was not a classification Article 16(4) authorised as the text then stood, and on that ground the ten per cent quota for economically weaker sections in the 1991 memorandum was struck down. Caste, by contrast, is a permissible starting point: a caste can be a class, and where a caste is socially and educationally backward, identification may begin from it, but caste cannot be the only indicator, and other material, occupation, education, economic condition and representation, must be applied to confirm or displace the starting point. The majority also accepted that a backward class may be divided into backward and more backward for the distribution of benefits, a proposition that returns, in a different setting, thirty two years later. The holding is a compound: it licenses caste as evidence while forbidding both caste as a complete answer and poverty as a complete answer.

Identification on this model is a continuing exercise, not a single act, and the majority institutionalised it. The judgment directed that a permanent expert body be established to examine the lists of backward classes, to entertain complaints of over inclusion and under inclusion, and to keep the identification under review as conditions changed. Parliament enacted the National Commission for Backward Classes Act, 1993 to give that direction statutory form, and the body the Act created spent two and a half decades advising on the composition of the Central List before the Constitution (One Hundred and Second Amendment) Act, 2018 repealed the statute and re founded the Commission as a constitutional body under Article 338B. The direction matters to the doctrine because it locates the lists’ legitimacy in a standing expert process rather than in the political judgment of the government of the day, and because the litigation history of the lists, communities seeking entry, communities resisting the entry of others, communities seeking removal of castes they say have advanced, runs through the advisory record the Commission built. When the Maratha litigation turned on whether the Gaikwad Commission’s material could sustain the Maharashtra statute, the Court was applying, to a State commission, the same discipline the 1992 judgment had designed for the Union’s: identification is expert work, its conclusions are entitled to deference, and deference is exhausted where the material does not carry the conclusion.

The seventh holding is the theoretical keystone: Article 16(4) is not an exception to Article 16(1) but a facet of it, an emphatic statement of the classification that the equality guarantee itself permits. The consequence is not rhetorical. If clause (4) were an exception, it would be construed strictly, the burden would rest on the State throughout, and every reservation would begin as a suspect derogation from a right. As a facet, reservation is one of the instruments by which the equality code pursues its own object, and the judicial inquiry is the classification inquiry: is the class genuinely within the clause, is the share proportionate to the purpose, and are the doctrine’s internal limits, the ceiling and the creamy layer, observed. The majority here consolidated N. M. Thomas against the older Balaji framing, and this holding has been treated as settled in every later case, including those that dismantled its neighbours.

The eighth holding concerns the boundary of the whole apparatus: there should be no reservation in certain posts and services. The majority instanced the defence services, establishments of research and development, and super specialities in medicine and allied fields, spheres in which the efficiency consideration in Article 335 operates at its strongest and in which the Court considered that the preference could not properly be pressed. The holding was stated at the level of principle rather than as an exhaustive list, and its afterlife has been modest compared with the other seven; it functions in later law chiefly as the acknowledgment, inside the doctrine itself, that the affirmative instrument has zones where it does not run. A later bench, in a case about reservation for persons with disabilities, took care to confine the Indra Sawhney embargo to the Article 16(4) field it arose in, which is a useful reminder that even the judgment’s quieter holdings are read by subject and not as general slogans.

Set out together, the eight holdings form a structure: a power, an internal filter, an external limit, an accounting rule, a boundary at promotion, a method of identification, a theory, and a zone of exclusion. The artifact below states each holding against its later fate, and the sections that follow narrate the operations the table compresses.

The judgment then had to be converted into administration, and the conversion year is part of its meaning. The stay on the 1990 memorandum stood discharged by the decision, but the twenty seven per cent could not operate until the creamy layer, which the majority had made a condition, had criteria; the Union constituted an expert committee to frame them and issued the implementing office memorandum in 1993, adopting the status based exclusions and the income test and applying the reservation to direct recruitment in the central services and public sector undertakings. The interval mattered. Between November 1992 and the 1993 memorandum, the law on paper and the law in the recruitment notices diverged, selections proceeded under the interim arrangements of the litigation, and the first generation of disputes under the new regime concerned not the doctrine but the certificates: whether a candidate’s family fell within an exclusion, and before which authority that question was to be proved. That pattern, in which the celebrated holding is quickly displaced in practice by the unglamorous instrument that implements it, recurs throughout the field, and it is why this article treats office memoranda, rosters and certificates as part of the doctrine rather than as its clerical aftermath.

Holding of the 1992 judgment	What changed it	Status after the changes
Reservation for backward classes is permissible; the twenty seven per cent share upheld	Not reversed; identification remains justiciable and evidence based	Stands
The creamy layer within backward classes must be excluded	Extended to the Scheduled Castes and Scheduled Tribes in the promotion field by Nagaraj and confirmed in Jarnail Singh	Stands and has expanded
Total reservation should not ordinarily exceed fifty per cent, save in extraordinary situations	Declined to be revisited in the Maratha case in 2021; held in Janhit Abhiyan in 2022 not to bind reservation under the new economically weaker sections clauses	Stands for reservation under Articles 15(4), 15(5) and 16(4); its reach beyond those clauses is contested
The carry forward rule is subject to the ceiling in the year of filling	The Eighty-first Amendment, 2000, inserted Article 16(4B), treating carried forward vacancies as a separate class outside the year’s ceiling	Reversed by amendment
Reservation does not extend to promotion	The Seventy-seventh Amendment, 1995, inserted Article 16(4A) for the Scheduled Castes and the Scheduled Tribes; conditions attached by Nagaraj and modified by Jarnail Singh	Reversed by amendment; the field is governed by the amendment and its conditions
A backward class cannot be identified solely by an economic criterion	The One Hundred and Third Amendment, 2019, created a separate economic criterion route under Articles 15(6) and 16(6), upheld in Janhit Abhiyan	Stands within Article 16(4); bypassed by a new clause rather than overruled
Article 16(4) is a facet of Article 16(1), not an exception to it	Not reversed; restated in the later cases	Stands
There should be no reservation in certain posts and services	Not reversed; later benches have confined its field rather than extended it	Stands, narrowly read

Posts, Vacancies and the Roster: The Sabharwal Correction

One refinement arrived quickly, and it governs the daily operation of everything the nine judges decided. Indra Sawhney had framed the ceiling against the vacancies of a year, because the case before it concerned a year’s recruitment. In R. K. Sabharwal v. State of Punjab, reported at (1995) 2 SCC 745, a Constitution Bench relocated the inquiry: reservation, it held, operates with reference to the posts in a cadre and is worked through a running roster of points assigned to the categories in proportion, and the roster continues to operate until the representation of each reserved category in the cadre reaches its prescribed share of the cadre’s strength. The two holdings coexist and must be held together. The ceiling of Indra Sawhney disciplines the intake of a year; the roster of Sabharwal disciplines the composition of the cadre over time; and the characteristic disputes of service litigation arise at the joint between them, where a candidate’s claim depends on whether a particular vacancy is counted against the year’s intake, against a roster point, or against the cadre’s unachieved proportion.

Sabharwal decided a second question whose importance grew with the promotion amendments. The bench held that once the prescribed percentage of reserved posts in a cadre stands filled, the roster mechanism yields to the rule that a reserved category candidate thereafter competes for the vacancies that arise in the ordinary course, the object being representation in the cadre rather than a perpetual preference in every selection. Read with the later cases on seniority, the decision completes the administrative picture the constitutional text never supplies: a percentage in a clause becomes, in operation, a roster in a register, a cadre strength return in a department, and a seniority list that officers litigate line by line. The distance between the constitutional argument, conducted in the vocabulary of equality, and the operative dispute, conducted in the vocabulary of roster points and cadre percentages, is the distance a practitioner in this field crosses daily, and it is the reason the holdings of 1992 are applied far more often as arithmetic than as philosophy.

A companion rule, stated in the 1992 judgment and elaborated immediately after it, governs who counts against the quota at all. A candidate of a reserved category who is selected in open competition on merit, without availing any relaxation or concession attached to the category, is appointed against an open point and not against the reserved share: the reservation is a floor of representation, not a ceiling on the category’s success, and a reserved candidate’s merit selection neither consumes a reserved vacancy nor reduces the number the roster must still deliver to the category. The rule sounds obvious and generates disproportionate litigation, because its application turns on the meaning of availing a concession: a candidate who used an age relaxation, a fee concession or a lowered qualifying standard is treated as having taken the benefit of the category and is counted against it, while a candidate who crossed every open threshold unaided is not, and selection records do not always disclose which occurred. In admissions the same logic produces the doctrine of the meritorious reserved candidate, whose open seat is protected even where taking a reserved seat in a preferred institution would advantage him, a refinement the courts built case by case on the 1992 foundation. Together with the roster and the catch up rule, the own merit principle completes the trio of counting rules through which the constitutional percentages are translated into named individuals on a list, and a surprising share of the case law that cites Indra Sawhney cites it, in substance, for one of these.

The Dissents, and the Majority Inside the Majority

The three dissenting judges, Thommen, Kuldip Singh and Sahai JJ., disagreed with the majority at the root. On their reading, Article 16(4) is a self contained code for reservation rather than an incident of clause (1), and caste cannot do the work the majority allowed it: Sahai J. held that caste cannot be equated with class and that identification built on caste offends Article 16(2); Kuldip Singh J. located backwardness in poverty and economic disadvantage and would have made inadequate representation, rather than caste, the organising test; Thommen J. accepted that caste could figure as one guide among factors such as poverty and illiteracy, but held that the purpose of the clause is relief of disadvantage produced by historical discrimination, not the permanent distribution of posts among communities. The dissents are worth reading against the majority because they isolate the choice the case actually presented. The majority treated caste as evidence of a condition; the dissents treated its use as the entrenchment of the very category the Constitution aimed to dissolve, and predicted that a caste keyed system, once installed, would be administered by caste arithmetic indefinitely. Much of the political argument of the following decades is those two predictions competing.

The more instructive internal difference sits inside the majority. Pandian J. concurred in the outcome while rejecting the ceiling outright, holding that no maximum could be derived from the text and that the line of cases fixing fifty per cent was unsustainable. That position matters because it shows the ceiling was not the price of the result: the memoranda could have been upheld without it, and four judges at most of the six in the majority carried it. When later benches describe the ceiling as the law declared by a nine judge bench, the statement is accurate as a matter of precedent and incomplete as a matter of intellectual history. The ceiling was a majority proposition of a divided Court, resting on an observation of 1963 that its own author called a broad way of speaking, and the judgment’s authority on that point has been maintained by later benches’ refusal to reopen it rather than by any fresh demonstration of its necessity. How much weight that distinction should carry is the live question taken up below.

The dissents have had a stranger afterlife than most dissents are granted. Their central objection, that identification by caste entrenches what it should dissolve and that economic condition is the honest marker of disadvantage, lost in 1992 and was then substantially enacted in 2019 by the constitutional route the majority’s logic left open: when Parliament created an economic criterion reservation, it did not contradict the sixth holding on its own ground but built beside it, in new clauses the dissenters’ philosophy could inhabit. And the dissenters’ second theme, that the unit of the clause is the genuinely disadvantaged person rather than the community as a bloc, returned from the opposite direction in 2024, when the Davinder Singh majority’s insistence that listed castes are not homogeneous, and the plurality’s demand that advanced families within the lists be filtered out, enacted within the doctrine the individualising pressure the dissents had applied to it from outside. A dissent’s function in a case of this size is to hold the losing analysis in a citable form until the law is ready for it; on the evidence of three decades, the Thommen, Kuldip Singh and Sahai opinions performed that function twice.

Parliament Answers: The Four Amendments

The judgment’s five year transitional window for promotion reservation was an invitation, and Parliament accepted it before the window closed. The Constitution (Seventy-seventh Amendment) Act, 1995 inserted Article 16(4A) with effect from 17 June 1995, empowering the State to make provision for reservation in matters of promotion, with the same inadequacy condition, in favour of the Scheduled Castes and the Scheduled Tribes. The holding it reversed was thereby reversed in terms: the bar on promotion reservation had rested on a reading of Article 16(4) itself, so nothing short of amending the Article could displace it, and each such amendment remains answerable to the limits described in the basic structure doctrine explained, which is why the four amendments of this period were all, in time, carried into court.

Why could the promotion bar be removed only by constitutional amendment?

Because the bar was not a rule of policy or of a statute but the Supreme Court’s construction of the constitutional text itself. A legislature cannot overrule a court’s reading of the Constitution by ordinary law; the reading stands until a larger bench departs from it or the constitutional text is changed.

The holding that Article 16(4) stops at initial appointment could therefore be displaced only by writing a new clause into Article 16, which is what the Seventy-seventh Amendment did, confined, deliberately, to the Scheduled Castes and the Scheduled Tribes.

The clause the amendment inserted is worth reading closely, because its drafting is the disagreement reduced to text. Clause (4A) provides that nothing in Article 16 shall prevent the State from making provision for reservation in matters of promotion to any class of posts in the services in favour of the Scheduled Castes and the Scheduled Tribes which, in the opinion of the State, are not adequately represented. Every element answers a sentence of the judgment: the words “in matters of promotion” answer the entry only holding; the confinement to the two scheduled classes answers the judgment’s own observation that the case for the scheduled classes stood on a different footing of historical disability; and the retention of the inadequacy condition, borrowed from clause (4), keeps the power conditional in form, which is what later allowed the courts to convert the condition into an evidentiary discipline without striking the clause down. An amendment drafted against a judgment is a dialogue in a strict sense, and this one is the clearest example in the Constitution of a legislature accepting a court’s premises while rejecting its conclusion.

The remaining three amendments worked through the judgment’s other pressure points as litigation exposed them. The Constitution (Eighty-first Amendment) Act, 2000 inserted Article 16(4B), under which a reserved vacancy carried forward is considered as a separate class of vacancy, to be filled in a succeeding year, and is not counted with the vacancies of that year when the fifty per cent ceiling is determined; the fourth holding’s accounting rule was thereby reversed for the backlog. The Constitution (Eighty-second Amendment) Act, 2000 added a proviso to Article 335 permitting relaxation in qualifying marks, or lowering of the standards of evaluation, in favour of the Scheduled Castes and the Scheduled Tribes in promotion. The Constitution (Eighty-fifth Amendment) Act, 2001 rewrote clause (4A) so that reservation in promotion operates with consequential seniority, and gave that rewriting retrospective effect from 17 June 1995, the date clause (4A) itself had taken effect. Consequential seniority was the point on which the case law of the 1990s had gone the other way: in Union of India v. Virpal Singh Chauhan and the Ajit Singh Januja decisions, the Court had applied a catch up rule under which a reserved category officer promoted earlier through the quota lost seniority when a general category officer senior to him was promoted later. The amendment displaced that rule for the two scheduled classes, so that the seniority gained on a reserved promotion is retained as a consequence of the promotion itself.

The catch up rule repays a concrete statement, because its abolition is otherwise unintelligible. Take two officers of the same entry cohort: the first belongs to a reserved category, the second does not, and the second stands senior in the feeder grade. A reserved vacancy in the promotion grade arises first, and the first officer is promoted into it while the second waits for a vacancy in the open stream. Some years later the second officer is promoted. Under the catch up rule, the second officer’s original seniority reasserts itself at that moment: he is placed above the first officer in the promotion grade, as though the interval in which the first officer held the higher post had not disturbed the order between them, and when the next promotion falls to be made, it is the second officer who is considered first. The rule thus gave the reserved officer the post but not its durable consequence, and over a career the difference compounds, because every subsequent selection runs off the seniority list. Consequential seniority reverses the compounding: the seniority the first officer acquired by being promoted when he was promoted stays with him, and the second officer, on his later promotion, takes his place below. Which of these is the just arrangement is precisely the kind of question on which the Constitution, after 2001, has made the choice itself for the scheduled classes, with retrospective effect, and the amplitude of feeling the question generates among serving officers is explained by the arithmetic: it is not about one promotion but about the order of every promotion that follows.

Read as a sequence, the four amendments show Parliament treating the judgment not as an adversary to be defied but as a text to be edited. Each amendment is drafted to the width of a specific holding: promotion, backlog accounting, qualifying standards, seniority. None of them touched the ceiling, the creamy layer or the identification holding, which is itself a datum about the political constitution of the period: the holdings Parliament reversed were those that distributed identifiable benefits to organised constituencies of serving officers and aspirants, while the holdings that would have required a frontal assault on the judgment’s authority were left alone. The amendments were then challenged together, and the challenge produced the second great restatement of the field.

Nagaraj and Jarnail Singh: Conditions Built, Then Partially Dismantled

In M. Nagaraj v. Union of India, reported at (2006) 8 SCC 212, a Constitution Bench of five judges in October 2006 upheld all four amendments and dismissed the basic structure challenge to them. The amendments survived because clause (4A) and clause (4B) are enabling provisions: they empower the State without compelling it, and a power exercised within limits does not damage the equality code merely by existing. But the bench attached to the exercise of the power the conditions that have governed the field since. A State that chooses to reserve in promotion must have quantifiable data showing the backwardness of the class, its inadequacy of representation in the services, and the bearing of the measure on administrative efficiency under Article 335. The fifty per cent ceiling was reaffirmed as a limiting condition. And the creamy layer principle, which Indra Sawhney had fashioned for the backward classes, was carried into the Scheduled Castes and the Scheduled Tribes in this field: Nagaraj treated its extension as following from the equality principle itself.

Why is reservation not a fundamental right of the candidate?

Because the provisions that authorise reservation are enabling provisions addressed to the State, not entitlements conferred on the citizen. As Nagaraj stated the position, no citizen holds a fundamental right to claim reservation, and no State is under a constitutional compulsion to provide it; the Constitution empowers, and then disciplines the exercise of the power.

What the Constitution does is to permit the State to make the provision where the conditions for it exist. The practical consequence is large. A challenge to a reservation policy is ordinarily a challenge to excess or to a missing condition, not the enforcement of an entitlement, and a decision to withdraw or restructure a scheme is tested as policy under the classification doctrine rather than as the deprivation of a right.

The form of Nagaraj’s inquiry matters as much as its conditions. The bench tested the amendments by asking whether they damaged the identity of the equality code, applying to service reservation amendments the discipline that the basic structure doctrine, restated the previous year by the nine judge bench in I. R. Coelho, applies to constitutional change generally: an amendment may widen or adjust a provision, and the question is whether what it leaves standing is recognisably the same constitutional commitment. The conditions Nagaraj attached, the ceiling, the creamy layer, the compelling reasons of backwardness, inadequacy and efficiency, were presented as the features that preserved the identity while the width changed. That framing explains both the decision’s strength and its vulnerability. Its strength is that the amendments were saved by a test any future amendment in the field must also pass. Its vulnerability is that the conditions were derived not from the text of clause (4A), which speaks only of the State’s opinion of inadequate representation, but from the bench’s account of what the equality code requires, and one of them, as the next bench showed, had been derived against the authority of Indra Sawhney itself.

Nagaraj’s data requirement on backwardness did not survive a decade. The complaint against it was precise: for the Scheduled Castes and the Scheduled Tribes, backwardness is not a fact to be freshly proved in every service and every decade, because the lists under Articles 341 and 342 rest on that very determination, and Indra Sawhney itself had proceeded on it. In Jarnail Singh v. Lachhmi Narain Gupta, reported at (2018) 10 SCC 396 and decided on 26 September 2018, a Constitution Bench of five judges, in a judgment delivered by Nariman J., declined to refer Nagaraj to a larger bench and held that its requirement of quantifiable data on the backwardness of the Scheduled Castes and the Scheduled Tribes was contrary to Indra Sawhney and need not be satisfied. The rest of Nagaraj was left standing: the State must still place quantifiable data on the inadequacy of representation of the class in the service, and must still satisfy Article 335. On the creamy layer, Jarnail Singh held squarely that the principle applies to the Scheduled Castes and the Scheduled Tribes in the context of promotion, and answered the objection that exclusion tinkers with the Presidential lists by pointing out that the filter operates on individuals within the list and alters no entry in it.

The litigation that carried the question to the 2018 bench shows the data discipline operating as a practical veto. In the years after Nagaraj, State promotion schemes and departmental instructions were challenged across the country on the single ground that the State had not first collected the quantifiable data the judgment demanded, and several were quashed by High Courts on that ground alone, without any finding that representation was adequate or that the scheme was otherwise unjust. The Jarnail Singh proceedings themselves grew out of such a quashing: a State’s promotion instructions, issued in reliance on clause (4A), had been struck down for failure to satisfy the Nagaraj conditions, and the appeals required the Constitution Bench to decide whether the conditions, as a package, stated the law. The bench’s severance of the package is therefore best understood as a calibration performed under pressure of consequences. The backwardness limb had frozen schemes for a class the Constitution itself treats as backward; the inadequacy limb and the efficiency condition were retained because they perform work the clause’s own text contemplates. A doctrine that cannot be complied with is not a discipline but a prohibition wearing discipline’s clothes, and the 2018 bench, while professing and practising fidelity to both of its predecessors, in substance refused to let the 2006 conditions become the prohibition that Indra Sawhney had imposed and Parliament had removed.

The two decisions repay reading as a pair, because they show the doctrine correcting itself by distinguishing the propositions a judgment actually decided from the propositions later attributed to it. Nagaraj converted the amendment’s conditions of opinion, the State’s satisfaction about inadequate representation, into a judicially enforced evidentiary discipline, and added a backwardness inquiry the text had not demanded for the scheduled classes. Jarnail Singh pruned the addition that contradicted the larger bench, retained the discipline that the text supported, and converted the creamy layer’s extension into settled law for the promotion field. A government servant’s promotion roster, a State’s cadre review, and a tribunal’s inquiry into a promotion dispute all run on this pair of cases far more than on the nine judge judgment that set the sequence in motion. For anyone advising on a promotion dispute, the general position described here is no substitute for advice on the individual case, and the cadre specific data on which these conditions turn must be examined in the service’s own record before any step is taken.

The Doctrine Travels: Admissions and Local Bodies

Indra Sawhney was a services case, decided under Article 16(4), and one measure of its authority is how completely its apparatus migrated to the Constitution’s other affirmative provisions. The migration to admissions came through Article 15(5), inserted by the Constitution (Ninety-third Amendment) Act, 2005 to empower special provision for the backward classes and the scheduled classes in admission to educational institutions, including private unaided institutions other than minority institutions. When Parliament then enacted the Central Educational Institutions (Reservation in Admission) Act, 2006, reserving twenty seven per cent of seats in central institutions for the backward classes, the challenge reached a Constitution Bench in Ashoka Kumar Thakur v. Union of India, reported at (2008) 6 SCC 1. The bench upheld the amendment and the Act, and did so by importing the 1992 discipline wholesale: identification of the beneficiary classes by the Mandal derived lists, the fifty per cent ceiling as the governing limit, and the exclusion of the creamy layer as a condition of the scheme’s validity, with observations favouring periodic review of the lists so that identification would not ossify. A judgment about government jobs had become, within sixteen years, the operating manual for university admissions, because the Court treated the ceiling and the filter as incidents of the equality code rather than as rules of one clause.

One textual subtlety, decided by the 1992 bench and easily missed, governs the boundary between the services doctrine and the admissions doctrine. Article 15(4) speaks of socially and educationally backward classes, while Article 16(4) speaks of backward classes without the qualifying adverbs, and the majority held that the expressions are not interchangeable: the class for services is identified by backwardness assessed with reference to the employment context, including inadequacy of representation, which is a condition written into clause (4) itself and absent from Article 15(4). The practical consequence is that a community’s presence on a list for one purpose does not automatically settle its position for the other, and challenges to educational reservation and to service reservation run on related but distinct records. Later benches, in carrying the apparatus across, have generally elided the distinction in favour of a common discipline of ceiling, filter and evidence, but the textual point survives in the case law and explains why the lists, though overlapping, are maintained separately for admissions and for appointments.

The migration to local bodies confirms the same pattern from the other direction. Articles 243D and 243T, inserted by the Seventy-third and Seventy-fourth Amendments, reserve seats in panchayats and municipalities for the scheduled classes in proportion to population and permit State law to reserve for backward classes. In K. Krishna Murthy v. Union of India, reported at (2010) 7 SCC 202, a Constitution Bench upheld the scheme and attached to backward class reservation in local bodies the two disciplines this article has traced: the State must rest the reservation on quantifiable data concerning inadequacy of representation in the local bodies themselves, and the aggregate of vertical reservation must observe the fifty per cent ceiling. The judgment is instructive about what the ceiling is. The bench applied it to elected office, a field far from the service context in which Balaji and Indra Sawhney framed it, and applied it as a ceiling on the aggregate rather than on any single category, which is why the local body cases are the cleanest statement in the later case law that the rule governs totals within its field. Each migration also marks the field’s boundary. The doctrine travels where the affirmative provisions travel; it does not govern benefits that are special provisions without being reservations, and it does not govern the distinct economic criterion clauses of 2019, as the next section shows.

The Ceiling Put to the Test: The Maratha Case and Janhit Abhiyan

Between the nine judge judgment and the two Constitution Benches of 2021 and 2022, the ceiling did its work largely out of the Supreme Court’s sight, in the High Courts, and the pattern of that work explains the confidence with which the Maratha bench later spoke. State legislatures, acting under sustained community pressure for inclusion or enhancement, repeatedly enacted reservation statutes whose totals ran well beyond fifty per cent, and the High Courts, applying Indra Sawhney as settled law, struck down or stayed such enactments where the State could not demonstrate the extraordinary circumstances the doctrine requires, treating the community’s size, its political organisation and its electoral weight as the ordinary facts of democratic politics rather than as exceptions to a constitutional rule. The litigation rhythm became standard: an agitation, a statute, a challenge, a stay, and a commission appointed to build the record that should have preceded the statute. By the time the Maharashtra law reached New Delhi, the question the Constitution Bench had to answer was therefore not really whether one State’s share could pass fifty per cent; the High Court line had answered that for a generation. The live question was the one the States’ counsel pressed with full force, whether the rule itself, being judge made and textually unanchored, should be sent back to a bench larger than the one that made it. The history that follows is the history of that question and of its partial successor.

The ceiling’s later history is two judgments pointing in opposite directions. The Maharashtra statute had been built to survive the doctrine: the State first constituted a commission under its backward classes commission machinery, chaired by M. G. Gaikwad, which reported that the Maratha community was socially and educationally backward and inadequately represented, and the Socially and Educationally Backward Classes Act, 2018 then reserved sixteen per cent for the community. The Bombay High Court sustained the Act in a trimmed form, reducing the operative shares, and the appeals reached the Constitution Bench with the commission’s record as the centre of the case. In Jaishri Laxmanrao Patil v. Chief Minister of Maharashtra, decided on 5 May 2021, a Constitution Bench of five judges examined the statute in the form in which it stood after the High Court’s judgment, reserving twelve per cent of seats in education and thirteen per cent of posts in public employment for the Maratha community, additions that carried the State’s total reservation well beyond fifty per cent. The bench’s treatment of the ceiling was unanimous and emphatic. It declined, unanimously, to refer Indra Sawhney to a larger bench or to revisit it; it held that Nagaraj contains no ratio permitting the ceiling to be crossed on the strength of quantifiable data; and it held that no extraordinary circumstances within the meaning of the 1992 judgment had been made out, since the fact that backward classes form a large share of a State’s population is the ordinary condition of Indian society and not an exception to it. The statute was struck down as violating Articles 14 and 16. In the same judgment the bench held, by a majority of three to two, that the Constitution (One Hundred and Second Amendment) Act, 2018 had transferred the power to identify socially and educationally backward classes to the President alone, a reading Parliament reversed within the year by the Constitution (One Hundred and Fifth Amendment) Act, 2021, which wrote the States’ power to maintain their own lists into Article 342A in express words. The identification strand of that episode belongs to the field’s map; its place here is as evidence of the same cycle this article traces, in which a judicial reading is answered by an amendment drafted to the width of the disagreement.

The Maratha reasoning on the ceiling deserves its strongest statement, because it is the strongest defence of the holding in the later case law. The bench treated the fifty per cent figure as having acquired, through three decades of application, the character of a constitutional recognition: to abandon it, Bhushan J. wrote, would be to move toward a society organised on caste rather than on equality, and the demand for certainty in a rule that structures every recruitment in the country is itself a constitutional value. Stare decisis did the work. The question was not whether nine judges would fix fifty per cent afresh on an empty page, but whether a figure applied in lakhs of selections, embedded in the roster systems of every government in the country, should be unsettled on the argument that it was contestable at birth.

The arithmetic of the Union’s own services shows how completely the ceiling organised the system built under it. Reservation at the Centre settled at fifteen per cent for the Scheduled Castes and seven and a half per cent for the Scheduled Tribes, shares fixed by reference to population, and the Mandal twenty seven per cent for the backward classes was calibrated, as the Commission itself explained, to leave the aggregate at forty nine and a half per cent, inside the line. The system therefore has, by design, almost no headroom: any new vertical claim of any size collides with the ceiling immediately, which is why the politics of reservation at the Centre since 1992 has been a politics of the margin, of sub division within fixed shares and of the search for a route outside the counted field. The economically weaker sections reservation took the second route, and its ten per cent stands outside the forty nine and a half, so that the Union’s aggregate, stated as a single figure, is fifty nine and a half per cent, a number that is unconstitutional as a description of one system and accurate as a description of two systems sharing a service. That doubleness is the Janhit Abhiyan holding expressed as arithmetic, and it is the cleanest available demonstration that the ceiling, whatever else it remains, is no longer a ceiling on reservation as such.

Tamil Nadu marks the boundary of that defence. The State reserves sixty nine per cent in the aggregate under a statute of 1993 which the Constitution (Seventy-sixth Amendment) Act, 1994 placed in the Ninth Schedule. In I. R. Coelho v. State of Tamil Nadu in 2007, a bench of nine judges held that enactments placed in the Ninth Schedule after 24 April 1973 are open to challenge where they damage the basic structure, including the equality principle, so the Tamil Nadu share operates under a shelter that is real but qualified. The episode shows the ceiling’s actual mode of operation across the federation: it is national doctrine, the shares are State choices, and the places where a State choice exceeds the doctrine are managed through the amendment and basic structure machinery rather than through the ordinary law of the services.

Then, on 7 November 2022, the ground shifted. In Janhit Abhiyan v. Union of India, a Constitution Bench of five judges upheld the Constitution (One Hundred and Third Amendment) Act, 2019 by a majority of three to two. The amendment had inserted Articles 15(6) and 16(6), enabling reservation of up to ten per cent for economically weaker sections other than the classes covered by the older clauses; the challenge, and the series’ own examination of the amendment in the article on the 103rd Amendment and EWS reservation, turned on the exclusion built into that definition and on the ceiling, since the new ten per cent operates in addition to existing reservation and takes the aggregate beyond fifty per cent wherever the existing share is at the ceiling. Maheshwari, Trivedi and Pardiwala JJ., in separate concurring opinions, upheld the amendment, and on the ceiling the majority position was that the fifty per cent rule is not inflexible and is not a command of the basic structure: it is a rule developed for reservation under Articles 15(4), 15(5) and 16(4), and reservation founded on economic criteria under the new clauses is a distinct exercise that the rule does not govern. Bhat J., dissenting for himself and Lalit CJ, would have struck the amendment down on the ground of the exclusion, and treated the breach of the ceiling as part of the same vice: to permit the ceiling to be passed by the device of a new clause, in his reasoning, is to convert a right to equality into a right to reservation distributed among compartments.

The three majority opinions reached that result by distinguishable routes, and the routes matter for what the case is authority for. Maheshwari J. treated reservation as an instrument of affirmative action directed at forms of disadvantage beyond the caste based backwardness of the older clauses, and held that the ceiling, not being a feature of the basic structure, could not be deployed to strike down a constitutional amendment that pursued equality by a new route. Trivedi J. located the ceiling as a rule internal to reservation under Articles 15(4), 15(5) and 16(4) and held the new clauses to stand outside that field altogether, so that no breach of the rule arose. Pardiwala J. emphasised the narrowness of basic structure review of amendments and the enabling character of the new provisions, which compel no State to reserve and fix a maximum rather than a mandate. The convergence is on outcome; the divergence is on whether the ceiling was inapplicable, non binding on amendments, or simply not a basic structure norm, and a future bench asked to apply Janhit will have to choose among those strands. The dissent, by contrast, is a single opinion, and its coherence is its strength: for Bhat J., the exclusion of the scheduled and backward classes from a benefit defined by poverty, in a country where poverty and caste disadvantage substantially coincide, made the classification itself an act of exclusion, and the ceiling’s breach was the same defect expressed numerically. A reader who wants the strongest statement that the ceiling retains constitutional force after 2022 will find it in the dissent, which is a measure of how far the centre of the doctrine has moved.

The two judgments are not formally inconsistent, and stating why they are not is the precise way to state what the ceiling is after the two judgments. The Maratha bench applied the ceiling to a State law made under the old clauses and found no exception; the Janhit majority confined the ceiling to those clauses and held a constitutional amendment creating a new head of reservation to be outside its field. A State legislature that crosses fifty per cent under Article 16(4) without extraordinary circumstances still legislates unlawfully, as Maharashtra learned. What the ceiling no longer does, after 2022, is operate as a general axiom of the equality code that every form of reservation, from whatever source, must satisfy. Its status has moved from a principle of the Constitution to a rule of a defined field, and much of the argument since has been about whether that movement is a clarification or a demotion.

Homogeneity Falls: Chinnaiah, the Reference, and Davinder Singh

The 1992 judgment’s identification holding assumed, without deciding, that the classes it discussed were units. For the Scheduled Castes, that assumption hardened into doctrine in E. V. Chinnaiah v. State of Andhra Pradesh, decided in 2004 by a bench of five judges and reported at (2005) 1 SCC 394. An Andhra Pradesh statute had apportioned the Scheduled Castes quota among groups within the State’s list, on the ground that some listed castes were capturing a disproportionate share of the benefit. The Court struck the statute down: the Presidential list under Article 341, it held, constitutes the Scheduled Castes as a single homogeneous class, the list can be altered only by Parliament, and a State that distributes the quota unequally among listed castes is in substance amending the list, which it has no power to do. For two decades that holding closed the question of sub classification within the scheduled lists, and State attempts to prefer the most backward among the listed castes were litigated in its shadow, including a Punjab statute reserving half the Scheduled Castes quota for specified castes, whose defence eventually produced the reference that undid Chinnaiah.

The vehicle that carried the question to seven judges was a Punjab statute. The Punjab Scheduled Castes and Backward Classes (Reservation in Services) Act, 2006 directed that, of the vacancies reserved for the Scheduled Castes, half be offered in the first instance to two specified groups within the list, the Balmikis and the Mazhabi Sikhs, on the State’s assessment that those communities were the most disadvantaged within the category and were being out competed for the quota by better placed listed castes. The Punjab and Haryana High Court struck the provision down on the authority of Chinnaiah, holding the State incompetent to prefer some listed castes over others, and the State’s appeal, Civil Appeal No. 2317 of 2011, remained pending while the authority it challenged came to be doubted. A bench of three judges, hearing the appeal, recorded its disagreement with Chinnaiah’s homogeneous class premise and referred the question for an authoritative answer; a Constitution Bench of five in turn considered that the five judge decision in Chinnaiah could be revisited only by a larger bench, and the matter was placed before seven.

In State of Punjab v. Davinder Singh, decided on 1 August 2024 and reported at 2024 INSC 562, a bench of seven judges held by a majority of six to one, Bela M. Trivedi J. dissenting, that Chinnaiah was wrongly decided and is overruled. The majority’s route is worth tracing rather than merely noting. Article 341 empowers the President to specify the castes deemed to be Scheduled Castes in relation to a State; its function is identification, and it creates no fiction that the castes so identified are, for every legal purpose, a single homogeneous class. The list itself discloses heterogeneity: the castes on it differ among themselves in the degree and the texture of the disadvantage they carry, and a Constitution that pursues substantive equality cannot forbid the State from seeing differences that are there. Classification within the list is therefore tested like any other classification under Article 14 and under the affirmative provisions: it must rest on a rational principle that indicates relative backwardness, and it must be supported by empirical data showing the inadequate representation of the group preferred. Two limits were fixed. A State may not allocate the entire reserved quota to one sub group, thereby excluding the rest of the list; and the quota, however it is distributed inside it, remains a Scheduled Castes quota throughout, because sub classification distributes a benefit and does not rewrite the list, which Parliament alone can amend.

Sub classification itself was not new; only its application to the scheduled lists was. Within the backward classes, the 1992 judgment had expressly permitted division into backward and more backward for the distribution of benefits, and State practice had long run ahead of the permission: Bihar’s settlement of 1978, associated with the government of Karpoori Thakur, divided the backward classes of that State into two annexures, the extremely backward and the backward, with separate shares of the State’s reservation for each, an arrangement that has structured Bihar’s politics and its service composition ever since. The Davinder Singh majority could therefore present its conclusion as the extension of a familiar technique to the one set of lists from which Chinnaiah had excluded it, and the dissent could present the same conclusion as the removal of the one protection the scheduled lists had retained against the technique. The contrast between the two settings is the analytical heart of the disagreement. Sub division among backward classes operates on lists that governments themselves make and remake; sub division among the scheduled classes operates on a list that the Constitution withdrew from government hands. Whether that difference is one of kind or only of institutional history is the question the case actually decided, and the majority’s answer, that identification and distribution are distinct operations, is a choice between those characterisations rather than a deduction from the text.

Trivedi J.’s dissent states the contrary position in its strongest form, and it deserves statement rather than summary dismissal. On the dissenting view, the list under Article 341 has a constitutional finality that the majority’s reasoning dissolves: specification of the Scheduled Castes was entrusted to the President, and after the initial notification to Parliament alone, precisely in order to withdraw the composition of the list from State level politics, and a State that may distribute the list’s benefits unequally among its entries holds, in substance, a power over the list’s practical content. The uniformity of treatment of the listed castes across the country, on this view, is not an incidental feature of the scheme but its point: the framers had before them the option of State wise variation and chose a single national discipline for the scheduled lists, reserving change to the one legislature in which every State is represented. The majority’s answer, that identification and distribution are distinct operations, and the dissent’s answer, that the distribution is the substance of which identification is the form, are both serious constitutional arguments, and the difference between them is a difference about where the Constitution locates a power, which later litigation in the States that act on the judgment will have to work out in application.

How did the Rangachari ruling differ from the position taken in 1992?

Rangachari, decided by a Constitution Bench in 1962, upheld reservation in promotion, reading the power in Article 16(4) as running through the whole of a service career and treating promotion as one of the appointments or posts the clause covers. Indra Sawhney overruled that reading and confined the clause to entry.

The majority of the nine judges held that clause (4) speaks to entry into the service, that the equality injury the clause remedies is exclusion from the service altogether, and that to extend the preference to every promotion would multiply the remedy beyond the injury while pressing hardest on the efficiency interest that Article 335 protects at the higher levels of a service.

What the Judgment Did Not Decide

The discipline of a judicial interpretation article includes the negative space, because in this field the propositions most confidently asserted are often propositions no bench has adopted. Indra Sawhney did not decide the validity of reservation in admissions: the memoranda concerned services, the Article 15(4) discussion is reasoning by parallel, and the admissions apparatus was settled only in 2008 on the service doctrine’s terms. It did not decide anything about reservation in the private sector, which no clause then or since has authorised as reservation, as distinct from the special provisions and incentives that governments may create. It did not fix creamy layer criteria, which it deliberately remitted to expert and executive determination. It did not hold that the lists of backward classes, once settled, require no revision; its direction for a permanent expert body presupposes revision, and several of its opinions treat periodic review as part of the scheme’s legitimacy. It did not consider sub classification within the Scheduled Castes, a question that reached the courts through a State statute a decade later and was answered in 2004 on the text of Article 341 rather than on anything the nine judges had said. And it did not decide that fifty per cent is a command of the basic structure, a characterisation no majority in the case adopted and which the 2022 majority expressly declined to endorse. Two further items of negative space are worth recording for the examination candidate, in whose papers the overclaim is a standing trap. The judgment did not pronounce on the validity of any State list of backward classes, the challenge before it being confined to the Union’s memoranda, and it did not address the position of minority educational institutions in relation to special provision, a subject the constitutional text later handled by exempting such institutions from the admissions clause altogether. Keeping the negative space in view is not pedantry. Each of those undecided propositions has been asserted in argument as though decided, and the first task of the lawyer or the examiner confronting such an assertion is to ask for the paragraph in which nine judges are supposed to have decided it.

The Creamy Layer: Three Questions That Are Not One Question

No phrase in this field generates more confident error than the creamy layer, because the same words answer three different legal questions with three different degrees of settledness, and accounts that do not separate them are wrong in a way that misleads practitioners. The first question concerns the backward classes. There, the exclusion of the creamy layer is the original 1992 holding, restated in every later case, and it is settled law: an identification of a backward class that does not filter out its advanced sections is vulnerable, and the State’s criteria for the filter are themselves justiciable. The second question concerns the Scheduled Castes and the Scheduled Tribes in promotion. There, the exclusion is also settled law, but by a different route: Nagaraj extended the principle in 2006, and Jarnail Singh in 2018 confirmed the extension as a holding, grounding it in Articles 14 and 16(1) and answering the Presidential list objection. A promotion policy for the scheduled classes that ignores the filter is, after 2018, exposed on that ground.

The third question is whether the creamy layer principle applies to the Scheduled Castes and the Scheduled Tribes generally, in admission and in initial appointment and not only in promotion. That question is open, and the events of 2024 did not close it. In Davinder Singh, Gavai J. held in his separate opinion that the State must evolve a policy for identifying the creamy layer even among the Scheduled Castes and the Scheduled Tribes and excluding it from the benefit of affirmative action, reasoning that to place the child of a family that has risen through reservation into high office in the same bracket as the child of a family in manual work in a village defeats the equality the provisions exist to serve. Three colleagues agreed with him: Vikram Nath J. held that the principle applicable to the backward classes should extend to the Scheduled Castes, with criteria that may differ; Pankaj Mithal J. agreed and added that the exercise should be periodic and that reservation should in his view be confined to the first generation that has risen by it; Satish Chandra Sharma J. held that identifying the creamy layer among the scheduled classes ought to become a constitutional imperative for the State. Four of the seven judges therefore expressed the view. But the view was not the Court’s decision. The question before the bench was sub classification; the operative holdings concern the permissibility, conditions and limits of sub classification, and nothing in the disposal of the appeals turned on the creamy layer. The lead opinion of Chandrachud CJ, written for himself and Manoj Misra J., decided the case on the classification ground and did not make any creamy layer exclusion a condition of what it upheld. The observations are accordingly the considered views of a plurality of the bench, carrying the persuasive weight that such views carry, and capable of being developed by a later bench into law, as the Jarnail Singh bench developed Nagaraj. They are not a holding that any State must apply a creamy layer filter to the scheduled lists, and any account that reports the 2024 judgment as having applied the creamy layer to the Scheduled Castes misstates the record in the one respect the case will most often be cited for.

The Ceiling Debate, Stated at Full Strength on Both Sides

Because this article’s charter requires contested positions to be given in their own strongest form, the ceiling argument should be set out as its two camps make it, with the named holders of each.

The case for treating the ceiling as binding doctrine runs as follows. Equality of opportunity is a right held by every citizen, and a reservation system without a ceiling has no internal stopping point: each extension is justified by the previous one, and the open share can shrink toward nothing without any single step appearing unreasonable. The figure of one half is the only line that has ever been stated by a court of final authority, it was fixed by the largest bench to consider the question, and a Constitution Bench in 2021, invited expressly to unsettle it, refused unanimously. The counsel who argued for the reference in the Maratha case, Mukul Rohatgi and Kapil Sibal among them, were answered on stare decisis grounds that were not technical: a rule that has structured every public recruitment for three decades is a fact of the constitutional landscape, and litigants, governments and candidates have ordered their affairs on it. On this view, the ceiling is the content the Court has given to the word “special” in a special provision: a provision that takes more than half has ceased to be a provision for a class within a general system of equal opportunity and has become a distribution of the whole by category.

The case for treating the ceiling as flexible runs as follows. The figure appears nowhere in the constitutional text; Balaji offered it as a broad way of speaking about reasonableness, and the 1992 majority adopted it without a derivation, which is why Pandian J. could decline it while upholding the memoranda. The judgment itself admits extraordinary situations, which concedes that the line is a rule of administration rather than a principle, since principles do not have exceptions for remoteness. The Janhit Abhiyan majority’s holding, that the ceiling is not inflexible and does not bind a reservation founded on a distinct ground under a new clause, is on this view not a demotion but a recognition of what the rule always was: a working discipline for the backward classes provisions, mistaken by repetition for a theorem of equality. Its proponents add the demographic point the Maratha bench rejected as a matter of law but which remains in the political argument: in States where the classes eligible for reservation form a large majority of the population, a fifty per cent ceiling reserves the smaller half of opportunity for the larger half of the people, and the equality that result serves is formal rather than substantive. Both camps argue from equality. They differ on whether equality is injured by the share taken or by the share withheld, and that difference is a value judgment dressed as a doctrine, which is why the contest has survived every bench that has tried to end it.

One further argument belongs to the flexible camp and should be recorded because it is the critique most often made of the doctrine in the legal literature: that the ceiling is a judicial invention in the strong sense, a number selected by a court to mark the boundary of its own tolerance and then constitutionalised by repetition. On this view, the derivation offered across the cases is circular. Balaji inferred a limit from the premise that a special provision must remain special; Indra Sawhney inferred the same limit from Balaji; the Maratha bench inferred its bindingness from its age. At no point, the critique runs, did any bench derive fifty, rather than forty or sixty, from a property of the equality guarantee, and a figure without a derivation is a preference. The binding camp’s best answer is the one Bhushan J. gave in the Maratha case, and it is a jurisprudential rather than a mathematical answer: the figure, he held, was arrived at on the principle of reasonableness and itself achieves the equality Articles 14, 15 and 16 jointly enact, so that its justification lies not in the number but in the equilibrium the number names, between a guarantee addressed to individuals and a remedy addressed to groups. Whether that answer satisfies depends on whether a reader thinks constitutional doctrine may rest on an equilibrium judgment at all; much of the law of Articles 14 to 16, including the classification test itself, rests on nothing else.

The two camps also divide, less visibly, on a question of fact that neither has been able to close. The binding camp’s equilibrium assumes that open competition above the halfway line in fact delivers a reasonable share of opportunity to candidates of the reserved classes on merit, so that the ceiling costs those classes little; the flexible camp’s demographic argument assumes the contrary, that open competition reproduces the distribution of educational capital and therefore of caste advantage, so that the ceiling is the operative limit on the classes’ advancement. That is an empirical dispute about selection outcomes, and it is in principle answerable from the services’ own records, cadre by cadre and examination by examination, in a way the value dispute is not. The records exist; a full public accounting of them does not, which is one reason the argument has been conducted for thirty years at the level of principle. A court asked to move the ceiling in either direction would be entitled to ask both sides for that accounting before it asked them for their philosophy.

What Remains Open, and What the Doctrine Requires in Practice

The unsettled residue can be stated with some precision. The application of the creamy layer to the Scheduled Castes and the Scheduled Tribes outside promotion awaits a bench for which it is the question; the observations of 2024 make its eventual arrival more likely and do not effect it. The ceiling’s field after Janhit Abhiyan is defined by clause rather than by subject, and whether a future amendment could route a backward classes reservation itself around the ceiling by drafting it as a new clause is a question no bench has answered; the Maratha unanimity suggests the reception such a device would meet, and suggestion is all it is. Sub classification within the scheduled lists is permissible in principle and unbuilt in practice: States must construct the empirical record the majority demanded, of relative backwardness and inadequate representation among listed castes, and the raw material for such a record is the field’s oldest deficiency. The decennial census has not enumerated caste in full since 1931; the caste returns gathered with the Socio Economic and Caste Census of 2011 were not published, the government stating that the data contained errors on a scale that made release unwise; and the identification exercises on which the doctrine insists therefore rest on commission estimates of varying vintage and on State surveys conducted for particular litigations. The first generation of sub classification laws will be litigated on the quality of those surveys, and the benches hearing them will be applying, to evidence of a kind the system has not generated for ninety years, a standard of empirical rigour that the system set for itself in 1992 precisely because the older evidence was thin. The data discipline of Nagaraj and Jarnail Singh, inadequacy of representation shown by quantifiable material, service by service, remains the working condition of promotion reservation and the ground on which most promotion schemes will actually be attacked or defended.

For a State government drafting in this field, the case law amounts to a discipline that can be stated in advance of any particular measure. Build the identification record first, through a commission or an expert process whose material will bear the conclusion, because a measure that arrives in court without its record arrives disarmed. Fix the share against the ceiling of its own clause, and if the measure must pass the ceiling, build the extraordinary circumstances case as a separate and documented exercise, knowing after the Maratha judgment that population share alone will not constitute it. Attach the creamy layer filter to any backward class measure from the outset, with criteria that are status based at the core and published, because the filter’s absence is the cheapest ground of challenge available to a petitioner. In promotion, assemble the cadre wise data on representation before the scheme issues, and record the Article 335 consideration expressly, because the Jarnail Singh conditions are tested on the file and not on assertion. In sub classification, construct the inter se record among the listed castes, relative backwardness and representation shown group by group, and stop short of the whole quota for any one group. None of this guarantees survival; the field’s history is of measures failing on their records despite sound powers. But the converse holds as a rule of the field: no measure in this line of cases has survived on power alone, and several have failed on record alone.

For the practitioner, the doctrine resolves into a small set of working rules that follow from the holdings and their fates rather than from the judgment’s reputation. Test every reservation measure first by its source clause, because the ceiling’s application follows the clause. Test identification by evidence, because from Balaji to Davinder Singh the Court’s steady demand has been for material rather than assertion. Test any promotion scheme against the Jarnail Singh conditions and the creamy layer filter, in that order. Test any sub classification against the two limits of 2024, the rational principle with its empirical support, and the prohibition on allocating the whole quota to one group. Readers who maintain a working file on this doctrine, the kind of file in which the eight holdings, the amending Acts and the later judgments must be kept in correspondence, can open the judgment extracts, spreadsheets and memoranda side by side in an office file viewer for Excel, Word and PowerPoint documents, which is the format in which service records and cadre data in this field are almost always circulated. The aspirant’s route through the same material, organised by the questions the civil services papers actually ask, is set out in the study treatment of social justice laws for the UPSC Mains.

Verdict: A Judgment Surviving in Pieces

The One Test answer this article set itself was that the reader should be able to say what the 1992 judgment held on each of its questions and what each holding is worth after thirty two years of amendment and appeal. The audit is not flattering to the judgment’s reputation. Of its eight holdings, the promotion bar was reversed by amendment inside three years and its field rebuilt by two later benches; the carry forward accounting was reversed by amendment; the identification holding survives inside its own clause and was bypassed outside it by a new clause; the homogeneity assumption on which its identification discussion silently rested was overruled in 2024; and the ceiling, its most famous sentence, was reaffirmed in 2021 as binding law for the clauses it arose under and held in 2022 to be no one’s basic structure and no bar to a reservation drafted elsewhere. What stands untouched is the architecture rather than the limits: reservation as a facet of equality, identification as a justiciable and evidential exercise, and the creamy layer as the doctrine’s internal conscience, extended by later benches further than its authors took it. That is the series thesis in a single case. A judgment’s authority is usually assumed from its citation count; traced holding by holding, the operative law turns out to be a smaller and stranger residue than the headline, and the residue, not the reputation, is what a court will apply.

There is a practical discipline in citing this case, and it follows from everything traced here. A citation of Indra Sawhney is incomplete, and in argument slightly misleading, unless it carries three attachments: the clause under which the proposition was held, because the ceiling’s field is clause bound; the instrument that later touched the holding, because four amendments and at least four Constitution Benches have redistributed the judgment’s parts among different authorities; and the distinction between what the bench held and what some of its members said, because the case’s most repeated proposition in later debate, the creamy layer for the scheduled classes generally, belongs to the second category. Journalists compressing the doctrine into a paragraph, aspirants compressing it into an answer, and advocates compressing it into a ground of challenge all face the same temptation, to let the case stand for the whole field. The field is the case plus its thirty two years of replies. Stated with its replies attached, Indra Sawhney remains what it has been since 1992: the unavoidable starting point of every argument in Indian reservation law, and the sufficient ending point of almost none of them.

Frequently Asked Questions

Q: What did Indra Sawhney decide?

A bench of nine judges, by a majority of six to three, upheld the reservation of twenty seven per cent of central service posts for the socially and educationally backward classes and attached the conditions that define the doctrine. The Court held that the advanced sections of a backward class, the creamy layer, must be excluded; that total reservation should not ordinarily exceed fifty per cent save in extraordinary situations; that reservation is confined to initial appointment and does not extend to promotion; that a backward class cannot be identified by economic criterion alone, though caste may be a starting point; that the carry forward rule operates subject to the ceiling in each year; that Article 16(4) is a facet of Article 16(1) and not an exception to it; and that there should be no reservation in certain posts and services.

Q: Where does the fifty percent ceiling come from?

The figure enters the case law in M. R. Balaji v. State of Mysore in 1963, where the Supreme Court struck down a Mysore order reserving sixty eight per cent of seats and observed, speaking generally and in a broad way, that a special provision should be less than fifty per cent. In T. Devadasan the following year the observation was applied as a rule against a carry forward arrangement that had filled twenty nine of forty five vacancies from the reserved categories. Indra Sawhney adopted the figure in 1992 as the ordinary limit on total reservation, justifying it as an incident of the equality guarantee: a reservation that takes more than half converts the special provision into the rule and empties the promise of equal opportunity made to everyone else.

Q: Did Indra Sawhney allow reservation in promotions?

No. The majority held that Article 16(4) authorises reservation at initial appointment and does not extend to promotion, and on that question eight of the nine judges agreed. The judgment overruled the 1962 decision in Rangachari, which had upheld promotion reservation for three decades. Because settled expectations were involved, the majority directed that promotion reservations already in existence continue for five years from the judgment and then cease. Parliament acted inside that window: the Seventy-seventh Amendment of 1995 inserted Article 16(4A), permitting reservation in promotion for the Scheduled Castes and the Scheduled Tribes, so the bar survives as the general rule for the backward classes while the field it was aimed at is governed by the amendment and the conditions later attached to it.

Q: Which amendments overturned parts of Indra Sawhney?

Four amendments of the 1990s and 2000s reversed specific holdings. The Seventy-seventh Amendment, 1995 inserted Article 16(4A), restoring reservation in promotion for the Scheduled Castes and the Scheduled Tribes against the judgment’s bar. The Eighty-first Amendment, 2000 inserted Article 16(4B), treating carried forward backlog vacancies as a separate class outside the year’s fifty per cent ceiling, against the judgment’s accounting rule. The Eighty-second Amendment, 2000 added a proviso to Article 335 permitting relaxation in qualifying marks and lowering of standards of evaluation in promotion for the two scheduled classes. The Eighty-fifth Amendment, 2001 rewrote clause (4A) so that promotion reservation operates with consequential seniority, retrospectively from 17 June 1995, displacing the catch up rule the courts had applied in the intervening case law.

Q: What did Nagaraj and Jarnail Singh add?

M. Nagaraj in 2006 upheld the four promotion amendments against a basic structure challenge and attached the conditions on which their exercise depends: a State reserving in promotion must have quantifiable data showing the backwardness of the class, its inadequacy of representation, and the measure’s consistency with administrative efficiency under Article 335, and the creamy layer principle and the fifty per cent ceiling apply as limits. Jarnail Singh in 2018 modified that settlement in one respect and confirmed it in another. The Constitution Bench held that the requirement of quantifiable data on the backwardness of the Scheduled Castes and the Scheduled Tribes was contrary to Indra Sawhney and need not be satisfied, while confirming that the creamy layer principle applies to those classes in promotion and that data on inadequacy of representation remains necessary.

Q: Is the fifty percent ceiling still binding?

Within its field, yes. The ceiling governs reservation under Articles 15(4), 15(5) and 16(4), and a Constitution Bench in the Maratha case in 2021 refused unanimously to revisit it or to refer it to a larger bench, striking down a State law that crossed it without extraordinary circumstances. What changed in 2022 is the ceiling’s reach. In Janhit Abhiyan the majority held that the rule is not inflexible, is not a command of the basic structure, and does not govern the reservation for economically weaker sections created by Articles 15(6) and 16(6). The ceiling therefore binds a State legislating under the backward classes clauses, as Maharashtra learned in 2021, and does not operate as a general limit on every form of reservation the Constitution may authorise.

Q: What did the 2024 sub-classification judgment decide?

In State of Punjab v. Davinder Singh, decided on 1 August 2024, a bench of seven judges held by six to one that sub classification within the Scheduled Castes is constitutionally permissible, overruling E. V. Chinnaiah, which had held the Presidential list to be a single homogeneous class that States could not divide. The majority held that Article 341 identifies the listed castes without declaring them homogeneous, that a State may prefer the more backward groups within the list where the preference rests on a rational principle indicating relative backwardness and on empirical data showing inadequate representation, and that a State may not allocate the entire reserved quota to one sub group. The judgment distributes benefits within the list; it does not alter the list, which only Parliament can amend.

Q: Does the creamy layer principle apply to Scheduled Castes?

In one field, settledly; in general, not by any holding. For the backward classes the exclusion of the creamy layer is the original 1992 rule. For the Scheduled Castes and the Scheduled Tribes in promotion, the principle was extended by Nagaraj in 2006 and confirmed as a holding by Jarnail Singh in 2018, on the basis of the equality guarantee and without altering the Presidential lists. For admission and initial appointment generally, no bench has held the principle applicable. In the 2024 sub classification case, four of the seven judges, led by Gavai J., expressed the view that the State should evolve a policy to identify and exclude the creamy layer among the scheduled classes, but those were observations in separate opinions, not the Court’s decision, and the question remains open.

Q: Who wrote the principal majority judgment in Indra Sawhney?

B. P. Jeevan Reddy J. wrote the principal opinion, for himself and for M. H. Kania, Chief Justice, and M. N. Venkatachaliah and A. M. Ahmadi JJ. P. B. Sawant and S. Ratnavel Pandian JJ. wrote separate opinions concurring in the result, Pandian J. differing from the principal opinion on the ceiling, which he would not have fixed at all. T. K. Thommen, Kuldip Singh and R. M. Sahai JJ. dissented. The bench of nine is reported at 1992 Supp (3) SCC 217 and delivered judgment on 16 November 1992. Because the majority is a different combination on different questions, the case is properly cited holding by holding rather than as a single ratio, and later benches, extracting its law, work from the conclusions on which the requisite majority converged on each question.

Q: What did the dissenting judges in Indra Sawhney argue?

Thommen, Kuldip Singh and Sahai JJ. dissented from the root. They treated Article 16(4) as a self contained provision for reservation rather than an incident of the general equality guarantee, and they rejected caste as the organising instrument of identification: Sahai J. held that caste cannot be equated with class and that caste based identification offends Article 16(2); Kuldip Singh J. located backwardness in poverty and economic disadvantage and would have made inadequate representation the principal test; Thommen J. accepted caste only as one guide among factors such as poverty and illiteracy, insisting that the clause exists to relieve disadvantage produced by historical discrimination. Their shared warning was that a caste keyed system, once installed, would entrench the categories the Constitution aimed to dissolve and would be administered by caste arithmetic indefinitely.

Q: What did E. V. Chinnaiah decide, and why was it overruled?

E. V. Chinnaiah v. State of Andhra Pradesh, decided in 2004 by five judges and reported at (2005) 1 SCC 394, struck down an Andhra Pradesh statute that had apportioned the Scheduled Castes quota among groups within the State’s list. The Court held that the Presidential list under Article 341 constitutes the Scheduled Castes a single homogeneous class, that the list may be altered only by Parliament, and that a State distributing the quota unequally among listed castes in substance amends the list. The bench of seven in Davinder Singh overruled it in 2024 because the premise could not be sustained: the list itself contains castes of differing degrees of disadvantage, Article 341 performs identification and creates no fiction of homogeneity, and treating a heterogeneous group as uniform defeats the substantive equality the affirmative provisions pursue.

Q: Why did Bela Trivedi J. dissent in the 2024 sub-classification case?

Trivedi J. held that sub classification by a State is in substance an alteration of the Presidential list, which Articles 341 and 342 reserve to the President’s notification and thereafter to Parliament alone. On her reasoning, the specification of the Scheduled Castes was deliberately withdrawn from State level politics and given a single national discipline, so that the composition of the list, and with it the equal treatment of every caste on it, would not vary with State governments; a State power to distribute the list’s benefits unequally among its entries gives the State, in practical effect, the power over the list that the text denies it, and deprives the less favoured listed castes of a benefit the Constitution assigned to the class as a whole. The dissent also warned that the majority’s route fragments the uniformity the listing provisions were designed to secure.

Q: What is consequential seniority, and where does it come from?

Consequential seniority is the rule that a reserved category officer promoted earlier under reservation in promotion keeps the seniority acquired by that promotion when a senior general category officer is promoted later. It comes from the Eighty-fifth Amendment, 2001, which rewrote Article 16(4A) with retrospective effect from 17 June 1995. Before the amendment, the courts had applied a catch up rule, in Virpal Singh Chauhan and the Ajit Singh Januja cases, under which the later promoted general officer recovered his original seniority and was placed above the reserved officer in the higher grade. The amendment displaced that rule for the Scheduled Castes and the Scheduled Tribes, so seniority follows the fact and date of promotion. Its validity, as part of clause (4A), was upheld in Nagaraj subject to the conditions that case attached.

Q: How does Article 335 qualify reservation in the services?

Article 335 provides that the claims of the Scheduled Castes and the Scheduled Tribes shall be taken into consideration consistently with the maintenance of efficiency of administration, in appointments to services and posts. It operates as the equality code’s own internal qualification: the claims are mandatory to consider, and the consideration is disciplined by the efficiency interest the same sentence names. Indra Sawhney drew on it for the promotion bar and for the exclusion of certain posts. Nagaraj converted it into an evidentiary condition, requiring a State reserving in promotion to show the measure’s bearing on overall administrative efficiency, a condition Jarnail Singh left standing. The Eighty-second Amendment, 2000 added a proviso permitting relaxation in qualifying marks or lowering of standards of evaluation in promotion for the scheduled classes, a textual acknowledgment that the efficiency qualification and the affirmative provisions must be read together rather than as rivals.

Q: Why did the Maratha reservation fail in the Supreme Court in 2021?

The Constitution Bench in Jaishri Laxmanrao Patil struck down the Maharashtra statute on two connected grounds. First, the fifty per cent ceiling: the additions for the Maratha community carried the State’s total well beyond the ceiling, and the bench held unanimously that no extraordinary circumstances within the meaning of Indra Sawhney had been made out, since the circumstance relied on, that backward classes form most of the State’s population, is the ordinary condition of Indian society and cannot be its own exception; the bench also refused, unanimously, to refer the ceiling to a larger bench. Second, the record: the Gaikwad Commission’s material, examined with the deference commission reports receive, did not in the Court’s view sustain the conclusions drawn from it. The statute accordingly violated Articles 14 and 16 and was held ultra vires.

Q: How can Tamil Nadu lawfully reserve sixty-nine per cent?

Tamil Nadu’s aggregate of sixty nine per cent rests on a State statute of 1993, the Tamil Nadu Backward Classes, Scheduled Castes and Scheduled Tribes (Reservation of Seats in Educational Institutions and of Appointments or Posts in the Services under the State) Act, which Parliament placed in the Ninth Schedule by the Constitution (Seventy-sixth Amendment) Act, 1994. Enactments in the Ninth Schedule were historically shielded from challenge on fundamental rights grounds, which is the mechanism by which the share has operated outside the ceiling’s ordinary discipline. The shield is qualified rather than complete: in I. R. Coelho in 2007 a bench of nine judges held that laws placed in the Schedule after 24 April 1973 are open to challenge where they damage the basic structure, including the equality principle. The Tamil Nadu position is therefore a working exception with a contested perimeter, not a refutation of the ceiling.

Q: Did Indra Sawhney allow backward classes to be divided into backward and more backward?

Yes. Among the majority’s identification holdings was the proposition that a backward class may be divided, for the distribution of the benefits of reservation, into backward and more backward classes, so that the more disadvantaged groups within the beneficiary class are not out competed by its advanced sections. The permission was the doctrinal companion of the creamy layer: both rest on the refusal to treat a beneficiary class as uniform when its internal differences are the very thing the classification exists to address. State practice had anticipated the point, Bihar’s settlement of 1978 having divided its backward classes into two annexures with separate shares. The 2024 judgment on sub classification within the Scheduled Castes proceeds on the same refusal of uniformity, applied to a different list under a different article.

Q: In which posts did Indra Sawhney say reservation should not operate?

The majority held that there should be no reservation in certain posts and services, instancing the defence services, establishments engaged in research and development, and super specialities in medicine and allied fields. The holding was stated as a principle rather than an exhaustive list, and its ground was Article 335: in spheres where the efficiency interest operates at its strongest, the preference could not properly be pressed. It is the least litigated of the eight holdings. Later benches have read it within its field rather than extending it; a bench dealing with reservation for persons with disabilities in promotion took care to hold that the Indra Sawhney embargo, arising under Article 16(4), does not govern a provision resting on a different constitutional footing.

Q: What did the Court direct about a permanent body for the backward classes lists?

The majority directed that a permanent expert body be established to examine the lists of backward classes, to consider complaints of over inclusion and under inclusion, and to keep the identification of backward classes under continuing review as conditions changed. The direction institutionalised the judgment’s view that identification is expert and evidential work rather than a one time political settlement. Parliament gave the direction statutory form in the National Commission for Backward Classes Act, 1993, and the Commission advised on the composition of the Central List for two and a half decades. The Constitution (One Hundred and Second Amendment) Act, 2018 then repealed that statutory footing and re founded the Commission as a constitutional body under Article 338B, where it stands alongside the commissions for the Scheduled Castes and the Scheduled Tribes.

Q: Does the fifty per cent ceiling apply to reservation in local bodies?

Yes. In K. Krishna Murthy v. Union of India, reported at (2010) 7 SCC 202, a Constitution Bench upheld the reservation scheme of Articles 243D and 243T for panchayats and municipalities and attached to it the disciplines of the services doctrine. Reservation for backward classes in local bodies, which State law may provide, must rest on quantifiable data regarding the inadequacy of representation of those classes in the local bodies themselves, and the aggregate of vertical reservation must observe the fifty per cent ceiling. The case is the clearest statement in the later case law that the ceiling governs totals within its field and travels with the affirmative provisions into institutional settings far from the service context in which Balaji and Indra Sawhney framed it, although the political character of local office led the bench to mark differences of application from the services cases.