Article 14's three tests don't add up

A litigant who walks into the Supreme Court with an Article 14 challenge today does not know which test the bench will apply. There are three of them. The classification test, traceable through State of West Bengal v Anwar Ali Sarkar and crystallised in Ram Krishna Dalmia v Justice Tendolkar, asks whether a law that draws a line between two groups draws it on an intelligible differentia and whether that differentia bears a rational nexus to the object the law seeks to achieve.¹ The arbitrariness test, launched in E.P. Royappa v State of TN and elevated by Maneka Gandhi v UOI and Ajay Hasia v Khalid Mujib, asks whether the State action is arbitrary in some larger non-comparative sense.³ The proportionality test, imported through Modern Dental College v State of MP and entrenched by K.S. Puttaswamy v UOI and Anuradha Bhasin v UOI, asks four sequential questions about legitimate aim, rational connection, necessity, and balancing. Each test was a response to a specific anxiety. None was designed to displace the others. Half a century after the second test was announced and a decade after the third was formally adopted, the Court has not told us which one controls when they meet on the same set of facts.

The puzzle is not new. Tarunabh Khaitan, writing in the Oxford Handbook of the Indian Constitution, set out the structural difficulty with precision: the classification doctrine and the arbitrariness doctrine answer conceptually distinct questions, the second one was anchored in a misreading of the rule of law, and the Court has shown a long-running unwillingness to pick.¹² The proportionality framework, layered on top in the last decade, has compounded the difficulty rather than resolved it. What follows is a survey of how each test came to be, what each one actually measures, where they collide, and why the Court has not chosen.

The classification test, in its narrow form

The classification test reached the Supreme Court along with the Constitution itself. Its earliest serious application was in Anwar Ali Sarkar, where the Court struck down Section 5(1) of the West Bengal Special Courts Act, 1950, because it allowed the State Government to refer 'any case' to a special court without indicating which cases warranted special treatment. The reasoning is comparative through and through. There were two procedures available — the ordinary criminal procedure and the special-courts procedure — and the law made no intelligible distinction between the cases that should attract one rather than the other. Mukherjea J's opinion noted that even where both procedures might individually satisfy fair-trial requirements, the unguided choice between them itself attracts Article 14.¹

The doctrine was tidied up six years later in Ram Krishna Dalmia v Justice Tendolkar. Das CJ set out the test in two limbs: a permissible classification under Article 14 must be founded on an intelligible differentia which distinguishes persons or things grouped together from those left out, and the differentia must have a rational nexus with the object the law seeks to achieve.² Those two sentences became the dominant Article 14 test for the first quarter-century of the Constitution's life. They are still recited in equality judgments today.

What is striking, on a re-read, is how narrow the test actually is. It asks two questions out of a much larger possible set. It does not ask whether the differentia is normatively acceptable — whether it is the kind of distinction (sex, caste, religion) that ought to require special justification. It does not ask about the importance of the State's objective. It does not ask whether the means chosen are the least restrictive available. It does not ask about real-world impact on the persons or groups affected. It assumes the State's stated objective is the genuine one. And it presumes constitutional validity, with the burden of unconstitutionality on the challenger. Khaitan's review of the inquiry that the test does not make — the questions about disproportionate impact, about presumptively impermissible grounds, about over- and under-inclusiveness, about expressive harm — runs to nearly half his chapter.

That narrowness is what made the test feel obsolete by the early 1970s, and it is what produced the second test.

The arbitrariness test, and the conceptual confusion at its heart

The arbitrariness test came in through a service-law dispute that did not need it. E.P. Royappa v State of Tamil Nadu concerned the transfer of a senior IAS officer in Tamil Nadu out of his cadre post. The Court could have decided the case as a service-law matter or as an administrative-law challenge to mala fide executive action. Bhagwati J chose, instead, to write the now-famous passage about equality and arbitrariness being sworn enemies. The passage is short. The doctrinal consequence is large. It said that any State action that is arbitrary is for that reason a violation of Article 14, with no requirement that the arbitrariness be cashed out as differential treatment between two classes.³

Maneka Gandhi v UOI picked up the formulation four years later, used it to read Article 14 into the structure of Article 21 review, and lent it the authority of a seven-judge bench.⁴ Ajay Hasia v Khalid Mujib closed the loop. Bhagwati J, writing again, said in plain terms that arbitrariness in State action is per se a violation of Article 14 and does not need to be proved through a comparison.⁵ The arbitrariness test had detached itself from comparison altogether.

Two things should be noticed about this move. The first is that the bridge between equality and arbitrariness in Royappa rests on the proposition that the two are 'sworn enemies' because one belongs to the rule of law and the other to royal whim. The bridge is a historical one. But the historical claim, as Khaitan shows, conflates two different principles of the rule of law. Dicey's first principle — that no person can be made to suffer except for a breach of law established before the ordinary courts — addresses arbitrariness in the sense of unrestrained executive whim. Dicey's second principle — that every person, regardless of rank, is subject to the same law — addresses equality in the sense of universal subjection. They are different ideas. A statute that exempts officials from ordinary law violates the second; it is not for that reason arbitrary. The argument that equality entails non-arbitrariness collapses one principle into the other and treats both as flowing from a single source. They do not.

The second observation is sharper. What the arbitrariness test actually polices, in practice, is unreasonableness, not arbitrariness. The two are not the same. A decision is arbitrary when there is no relevant ground to prefer one option over another — a coin-flip choice of the age of consent at sixteen rather than fifteen or seventeen is arbitrary in this strict sense, and yet it is plainly reasonable. A decision is unreasonable when it is grounded in illegitimate or insufficient reasons. The pregnancy-termination rule struck down in Air India v Nargesh Meerza was not arbitrary in the strict sense; the airline had relevant reasons to want to avoid maternity-leave costs. It was unreasonable because those reasons were constitutionally unacceptable. The Court called it arbitrary. The label was inaccurate. What the Court was doing — and has been doing ever since under the heading of arbitrariness — is reasonableness review.

The doctrinal upshot is that Article 14, through the arbitrariness gate, has become a freestanding ground of unreasonableness review of State action. This is what Ajay Hasia set up. It is what successive benches have used to strike down rules, regulations, executive orders, and (occasionally) statutes that they could not knock out under any other constitutional head. As a freestanding ground, it does what Wednesbury unreasonableness does in English administrative law, with the difference that it has the authority of a fundamental right behind it. The constitutionalisation of an administrative-law standard has its own quiet costs. A challenge to unreasonable administrative action that could once have been argued before the High Court on the writ side as a Wednesbury question now has to be framed as a fundamental-rights claim, with all the procedural and remedial consequences that follow. The doctrinal architecture has bent to fit the move.

The Court has, to its credit, been more willing to apply the arbitrariness test to executive action than to primary legislation. The leading authority is State of Andhra Pradesh v McDowell & Co., where Jeevan Reddy J said in 1996 that no enactment can be struck down by simply calling it arbitrary, unreasonable, or unwarranted; some other constitutional infirmity has to be found before invalidating an Act.¹¹ McDowell was reaffirmed in dicta by a five-judge bench in the Re Natural Resources Allocation reference in 2012 and again in Subramanian Swamy v CBI in 2014.

The trouble is that the same Court, in three-judge benches running in parallel, has gone the other way. Malpe Vishwanath Acharya v State of Maharashtra in 1998 held that a statutory provision had become arbitrary by the passage of time. Mardia Chemicals v Union of India in 2004 struck down a provision of the SARFAESI Act on arbitrariness grounds. Neither involved a classification. Neither relied on any other constitutional infirmity. Both did exactly what McDowell said could not be done.

The position today, on the most charitable reading, is that the arbitrariness test is available against executive action and delegated legislation, that it is in principle unavailable against primary legislation, and that the latter restriction has been honoured in the breach often enough that it cannot be relied on. This is not a doctrinal position. It is a doctrinal stand-off — and one that Janhit Abhiyan v UOI revived without resolving, when both majority and dissent slid between classification language and arbitrariness language at different points of the same opinion.¹⁰

The third test: proportionality, and where it came from

For most of the period in which the classification and arbitrariness tests were being argued out, the Court occasionally invoked 'strict scrutiny' or 'proportionality' as labels, but treated them, as Khaitan notes, as rhetorical cloaks for a continued deferential standard of review. That changed in 2016. Modern Dental College & Research Centre v State of MP was a challenge to the regulation of admissions in private medical colleges in Madhya Pradesh. The Court, in a judgment by Sikri J, formally adopted the four-pronged proportionality test as a matter of Indian constitutional law: a measure that restricts a right must (i) be designated for a proper purpose, (ii) bear a rational connection to that purpose, (iii) be necessary in the sense that no less restrictive alternative would achieve the same result, and (iv) on balance produce benefits commensurate with the harm.⁷

The four-pronged structure was adapted from the German constitutional tradition by way of the Israeli Supreme Court and the South African court. Modern Dental College was framed primarily as a freedom-of-occupation case under Article 19(1)(g), but the Court applied the test to the equality challenge under Article 14 as well. A year later, K.S. Puttaswamy v UOI entrenched proportionality as the operative test for fundamental-rights infringements generally; the Aadhaar majority in 2018 ran the four steps as the spine of its analysis, including under Article 14.⁸ By Anuradha Bhasin v UOI, the Court was applying proportionality almost reflexively, including to Article 14 challenges, while declining to say what relationship the test bore to the older two.⁹

What proportionality offers, theoretically, is a structured inquiry that incorporates many of the questions the classification test does not ask: about the importance of the State's purpose, about whether the means chosen really do advance that purpose, about whether less restrictive means were available, and about whether the harm to the right is justified by the gain to the public interest. In other words, it offers some of the depth of analysis that Khaitan's chapter says the classification test ought to incorporate. If the Court were committed to the proportionality framework, the classification test could in principle be folded into its first two steps and the arbitrariness test into the third or fourth. None of that consolidation has happened. The three tests run alongside each other, with no rule of priority.

It is worth pausing on what proportionality, as imported, actually does to Indian doctrine. In the German tradition from which it descends, proportionality presupposes a settled hierarchy among the steps: a measure that fails the necessity step is unconstitutional regardless of its showing on balancing. In the Israeli tradition, the steps are sequential but the balancing step does most of the heavy lifting. The Indian Court has adopted the four-step structure without specifying how the steps relate to one another in cases of partial failure. A measure that is rationally connected to a legitimate aim but fails the necessity step has, on the German reading, lost the case. On a soft reading of the Indian formulation, that conclusion is not yet automatic. The transplant has been incomplete on the technical side, and that incompleteness is part of why proportionality has not crowded out the older two tests.

Where the three meet, and where they don't

An equality challenge to a statute that classifies persons today can in principle be analysed through three different lenses. The classification test asks whether the differentia is intelligible and the nexus rational. The arbitrariness test asks whether the law, classification or no, can be defended on reasonable grounds. The proportionality test asks the four sequential questions. Each one can produce a different result on the same statute.

Take the EWS reservation upheld in Janhit Abhiyan v UOI.¹⁰ The 103rd Amendment, which introduced the ten per cent reservation for economically weaker sections, drew a line between the economically weaker among the unreserved and the economically weaker among SC/ST/OBC categories. On the classification test, the line is intelligible and bears some nexus to the stated objective of providing relief to those who do not benefit from existing reservations. On the arbitrariness test, the question is harder: is it reasonable to exclude the SC/ST/OBC poor from a benefit aimed at the poor? On the proportionality test, the question is harder still: is the means chosen necessary, and does it produce benefits commensurate with the harm to the equality of the excluded? The majority and the dissent in Janhit Abhiyan reached opposite conclusions; the gap between them is not a gap of disagreement on facts but a gap on which test was being applied. The judgments shifted between classification language and arbitrariness language as it suited each opinion. The 50 per cent ceiling that Indra Sawhney v UOI framed in 1992 as part of the constitutional balancing exercise within Article 14 was treated by some opinions as foundational and by others as a modifiable judicial policy.⁶

The same drift can be traced in Anuradha Bhasin, where the four-pronged proportionality structure was applied but the question of whether the older two tests had been displaced was not addressed. It is traceable in Puttaswamy's Aadhaar majority, which recited proportionality without saying whether the classification test was now subsumed within it. It is traceable in service-law decisions where the Court continues to invoke Royappa and Ajay Hasia as if proportionality had not been adopted. The pattern is not one of test, but of selection of test.

This is the inconsistency the title of this essay points at. The three tests do not add up because the Court has not added them up. There is no reported decision of a constitution bench that says: where these tests apply to the same fact, this is the order of priority, this one subsumes those two, this is the residual ground left for those. There is also no reported decision that says they cohabit and are interchangeable. There is silence, and selection on the facts.

An empirical pattern: the test follows the result

What does the case-law selection of test look like when read in bulk? A rough pattern emerges, though I want to put it tentatively. When the Court is inclined to uphold State action, it reaches for the classification test, which is deferential by design, formal in execution, and asks the State for very little. When the Court is inclined to strike down State action without locating the infirmity in another constitutional provision, it reaches for the arbitrariness test, which has no internal limits and can absorb almost any concern. When the Court is inclined to engage seriously with both sides — a stronger version of judicial review than either of the older tests offers — it reaches for proportionality.

This pattern, if it holds, is not an accident of inconsistency. It is the consequence of having three tests at different levels of deference and intensity available simultaneously, with no principle for choosing among them. The bench can reach for the test that fits the result it intends to reach. That is the freedom Khaitan's analysis traces back to the elevation of an administrative-law standard into a constitutional test in Royappa — what got smuggled in along with non-arbitrariness was a doctrinal flexibility that no formal constitutional text supplied.

Two specific observations make the pattern more concrete. First, the arbitrariness test almost never appears as the basis for upholding a statute. It is overwhelmingly an invalidating test — when invoked, it produces strike-downs. Second, when the Court wants to uphold State action that runs into difficulty on the classification test, it has shown a willingness to redescribe the test as a 'reasonable nexus' inquiry and to defer heavily to the State's stated objective. The two-limb test of Ram Krishna Dalmia can be made to do almost any work depending on how anxiously its second limb is applied. The choice of how anxiously to apply it is, in practice, the choice of result.

The third observation, and the one most likely to draw objection, is about reservations doctrine. The Article 14 analysis in Indra Sawhney v UOI, the Mandal nine-judge bench, was structured as a balancing exercise within the equality clause itself: formal equality on one side, social-justice provisions on the other, with the Court holding that total reservations should not cross 50 per cent in any given year. That balancing is in spirit closer to proportionality than to either of the older two tests, but proportionality had not yet been formally adopted in 2016, and the Court did not name what it was doing. Janhit Abhiyan in 2022 had to revisit the same balance with the proportionality framework now available, but did not actually use it. The 50 per cent ceiling was treated by some opinions as a fixed point of the Article 14 analysis and by others as a floating one. The disagreement is at one level a disagreement about whether the ceiling is itself basic structure. At another level, it is a disagreement about which Article 14 test is doing the work.

The institutional defence of overlapping tests

The argument I have been making — that three uncoordinated tests produce inconsistent doctrine — is not the only available reading. There is a serious institutional defence of the present arrangement, and it deserves a hearing.

The defence runs roughly like this. Indian equality jurisprudence covers a vast range of State action: tax classifications, service rules, regulatory schemes, criminal procedures, reservations, statutory presumptions, allocation of natural resources, regulation of education, regulation of speech. No single test, calibrated to a single intensity of review, will fit all of these contexts. A test that is deferential enough for tax law will be too thin for a challenge to a fundamental restriction on speech. A test demanding enough for the speech challenge will be unworkable in tax law, where every classification is to some degree under-inclusive or over-inclusive. The three tests, on this defence, function as a graded toolkit. The Court picks the test that fits the gravity of the State action and the intrusiveness of the regulation.

There is real force to this. Indian fundamental-rights doctrine has always been less rigid than the American or German equivalents about formal tiers of scrutiny. The classification test was designed for tax-law-style review, where deference to legislative line-drawing is appropriate. Proportionality, with its explicit balancing step, is the right test for a privacy infringement or a speech restriction. Treating the three as a continuum rather than a hierarchy lets the Court adjust its standard of review without inventing new categories every time. From the bench's point of view, this is doctrine that does the work doctrine is supposed to do.

The cost of the institutional defence is two-fold, and worth naming. The first cost is to the litigant, who cannot predict the standard of review that will be applied to their case and so cannot frame their pleadings to a known target. The second cost is to the law, which is supposed to develop through reasoned engagement with prior doctrine. When a bench in 2024 quotes Ram Krishna Dalmia for a tax matter, Ajay Hasia for an executive-action matter, and Modern Dental College for a regulation-of-rights matter, without saying which one would govern if all three were applicable, the case-law develops in three parallel streams that never meet. The doctrine is then less a body of reasoned law than a set of available registers.

The institutional defence concedes the unprincipled selection and asks us to value the flexibility it produces. Whether the trade is worth making depends on how much one values predictability against how much one values the Court's discretion to fit the test to the case. There is no neutral ground from which to settle that question.

Where this leaves the doctrine

The classification test of Anwar Ali Sarkar and Ram Krishna Dalmia still controls when the Court wants it to. The arbitrariness test of Royappa, Maneka Gandhi, and Ajay Hasia still controls when the Court wants it to. The proportionality test of Modern Dental College, Puttaswamy, and Anuradha Bhasin controls increasingly often, and is being layered on top of the older two without any rule of priority. The legal profession has internalised this and now litigates in the alternative — pleading classification, arbitrariness, and proportionality together, and waiting to see which one the bench selects.

Two things would clarify the doctrine. The first is a constitution-bench reference that frames the three tests directly and rules on the relationship among them. McDowell tried to do this for two of the three, and was honoured selectively. A nine-judge or eleven-judge bench could settle it for all three. The second is the kind of analytical work that academic commentary like Khaitan's chapter performs — naming the conceptual confusion at the heart of the arbitrariness doctrine, distinguishing arbitrariness from unreasonableness, locating each test in its proper analytical home. That work cannot be done from the bench in the course of deciding a particular dispute. It has to be done before, so that when the dispute arrives, the available tools are clear.

Until either of those happens, the answer to the question of which Article 14 test governs a given case is going to remain what it is now. It is going to depend on the bench, on the State action, on the result the Court is moving toward, and on which of the three labels best legitimates that result. That answer is workable, in the sense that the Court continues to dispose of equality challenges every working day. It is not, in any tighter sense, a doctrine of equality. It is three doctrines that share a number — Article 14 — and not much else.

What I want to add, by way of closing, is a quieter observation. The three tests were not designed in isolation by a planning committee. Each one emerged in response to a felt difficulty with what came before. The classification test was the early Court's attempt to discipline a generously worded equality clause with a borrowed American framework. The arbitrariness test was Bhagwati J's attempt to rescue the equality clause from the formality of that framework. The proportionality test is the present Court's attempt to bring Indian rights doctrine into conversation with the comparative literature of the last forty years. Each move was made for reasons that were defensible at the time. The cumulative result is a doctrinal landscape that no single judge could have planned and that no one is now in a position to tidy up without a constitution-bench reference that the registry has not yet listed.

This is a recurring shape in Indian constitutional law. A doctrine is announced, gets layered with successor doctrines responding to its limits, and then has to be applied in cases where the layers contradict one another. The basic-structure doctrine has the same shape. So does the doctrine on personal liberty. So does the doctrine on the relationship between fundamental rights and directive principles. The Court has, over time, learned to live with these layered architectures by treating them as a set of available registers rather than as inconsistent positions to be resolved. Whether that strategy is sustainable for Article 14, given the weight Article 14 now carries in the public-interest litigation that flows through the Court, is a question that the next decade of constitutional doctrine will have to answer.

Get the week's biggest decisions, distilled.

A weekly digest for lawyers, students and the legally curious. Every Friday morning. Free.