Bench hunting in Article 32

Two writ petitions raise the same constitutional question. They are filed in the same week, by petitioners on the same side of the issue, against the same respondent. One lands before a bench that has historically read the right narrowly; the other before a bench that has historically read it broadly. Both benches pass orders; the orders conflict. There is no doctrine in the Indian Supreme Court's jurisprudence that tells us which order prevails. There is no listing rule that prevented the two petitions from going to different benches in the first place. There is no obligation on the Court — none I have been able to find — to club the petitions, or to refer the conflict to a larger bench, or to do anything at all about the inconsistency. The two orders stand, and the litigants whose case is on appeal from one of them will have to argue, on remand, that the order before them is the binding one. The Court's silence is the doctrine.

This is bench hunting in Article 32, and my argument is simple. It is not an abuse of an otherwise orderly system. It is the predictable use of a system the Court has refused to discipline. Every defence of the present arrangement turns on the inviolability of the Chief Justice's administrative discretion as master of the roster³; every cost of the present arrangement falls on the litigants who do not have the resources to game the listing list. The silence is a feature for the powerful and a bug for the rest. I want to make the case for treating the silence as a constitutional problem, not a managerial one.

How bench hunting actually works

The technique is not subtle. A skilled Supreme Court counsel — and 'skilled', in this context, means experienced in the Court's listing patterns — knows roughly which benches are likely to be sympathetic to a given line of argument. A Special Leave Petition or an Article 32 writ is filed; mentioning is sought; if listing falls before an unsympathetic bench, the matter is withdrawn before any judicial application of mind takes place; a fresh petition is filed, with marginal redrafting, often through a different advocate-on-record. The system permits this. Article 32 imposes no procedural bar on multiple filings of materially similar petitions by different petitioners. The Court's Rules do not require disclosure of cognate proceedings already pending. The Registry's listing software, by all accounts, picks the bench by random draw within the categories defined by the Chief Justice's roster.

On its face, randomness sounds like a defence against bench hunting. It is the opposite. Randomness, combined with the freedom to refile, means that a petitioner with the resources to file in volume can simply wait. The third or fifth or eighth petition will, on the law of averages, land before a friendlier bench. The petitioner without those resources, who files once and accepts what comes, is the one who loses. The architecture treats the two as equivalent litigants. The outcomes treat them as anything but.

The other half of the architecture is the practice of sitting in division benches of two. Raeesa Vakil's chapter on jurisdiction in the Oxford Handbook of the Indian Constitution describes the consequence with precision²: the 'inconsistencies that arise in decisions by disparate benches of two judges have led to conflicts in constitutional interpretation', and the problem is 'closely related to the practice of the Court of sitting in division benches of two'. The arithmetic is grim. The Court sits, on most days, in twelve or thirteen division benches. Each bench is a coordinate authority. Each bench can pronounce on Article 32 questions. None binds the others, and the standard mechanism for resolving conflicts among coordinate benches — reference to a larger bench — is itself discretionary. The result is the structural condition for bench hunting: many doors, no door-keeper, and no obligation to close any of them.

The 2017 medical-college-admissions episode⁴, which led directly to the January 2018 press conference by the four senior-most puisne judges⁵, was not an aberration. It was the moment the system showed its hand. A petition raising questions about the Court's own probity in admissions of a particular medical college was listed before a bench whose composition raised questions of propriety; another bench's order was passed, then circumvented; a Constitution Bench was constituted at short notice to declare that the Chief Justice's listing power was unchallengeable as master of the roster³. What is striking, on rereading Asok Pande, is how thin the legal reasoning is. The bench did not say that bench-listing decisions should not be reviewable; it said they could not be reviewable, because the alternative would produce administrative chaos. The argument is consequentialist, not doctrinal. And a consequentialist argument from chaos can be made for almost any unchecked power.

The legal status of the listing list

This is, in my reading, the most important thing about Asok Pande: it left the listing list with a constitutional status that no other administrative document has. The Chief Justice's roster is not a statute, not a constitutional provision, not even a published rule under Article 145. It is a circulated document, revised at intervals, by which the Chief Justice tells the Registry which categories of cases go to which benches. It is, structurally, a routing slip. The Court's own holding is that it cannot be challenged. The framers of the Constitution made the Supreme Court's appellate jurisdiction subject to Article 134A certificates, to bench-strength rules under Article 145, to a constitutional review procedure under Article 137 — they imposed every conceivable structural check on the Court's powers in the substance. The listing of cases, the one mechanism by which all of those substantive checks are operationalised in any given case, they did not address. The Court has filled the silence by holding that the silence is binding.

I do not believe the framers intended this. Vakil's chapter records that the Constituent Assembly believed 'the Supreme Court would evolve principles on the manner of the use of its appellate jurisdiction', and that the assembly was confident that 'a process of internal restraint would ensure that the High Court would grant certificates sparingly, just as the Supreme Court would grant leave cautiously'. The expectation was that the Court would discipline itself. The Court has not. As HM Seervai put it, in a sentence Vakil quotes and that I borrow here⁷: 'to coin a phrase is not to indicate a standard'. The Court has coined the phrase 'master of the roster'. It has not indicated a standard for the master's discretion.

The point is sharpened by Article 32 itself. Article 32 is the only fundamental right that is itself a remedy. Ambedkar's well-known description of it as 'the very soul of the Constitution and the very heart of it'¹, set out by Gopal Subramanium at the head of his chapter on writs, was not rhetoric. The reason the framers gave the Supreme Court direct writ jurisdiction was that they did not want fundamental-rights enforcement to be the gift of a discretionary administrative gatekeeper. They wanted petitioners to be able to walk into the apex court when their rights were violated. They did not, it is fair to say, want the listing list to be the gatekeeper that Article 32 had been designed to bypass.

The Court itself has, at times, acknowledged the problem of unequal access. In Kanubhai Brahmbhatt v State of Gujarat¹⁰, the Court directed that petitioners should ordinarily exhaust the High Court remedy under Article 226 before invoking Article 32. The reasoning was, in part, equality: a Supreme Court that hears every fundamental-rights matter directly favours litigants with the resources to file in Delhi. The direction has been honoured selectively. In practice, well-resourced petitioners walk in under Article 32; less-resourced petitioners are routed back to the High Courts; the standard for which route a petitioner takes is the same standard that governs bench-listing — undisclosed and unreviewable. The two practices are connected. Both are discretionary procedural gates that operate to the benefit of those who know how to use them. Both have a constitutional patina and an administrative reality.

Why every reform has stopped short

It is not as if the problem has been invisible. The Court itself, in Mathai v George⁶, referred to a larger bench in 2010 the question of broad guidelines for the exercise of discretion under Article 136. The reference is still pending. The Law Commission has, on several occasions, recommended structural reforms — a Constitution Bench permanently at Delhi, cassation courts in four regions, more judges. Successive senior advocates and retired judges have observed, in print, that the practice of two-judge benches deciding constitutional questions is a problem. Justice Ruma Pal, quoted by Vakil, warned that 'this element of emotional subjectivity in the assessment of what constitutes an injustice would necessarily result in greater uncertainty in the outcome of a proceeding before the Supreme Court'. None of this has been translated into binding doctrine. The reforms that the Law Commission has proposed have not been taken up; the Court's own self-reference has been allowed to lapse; the press-conference of January 2018 produced no procedural change. The reform energy has consistently fallen short of the institution being asked to reform itself.

I want to be precise about what 'stopped short' means. There has been reform at the margins. The Court's Rules have been amended, registry procedures have been computerised, the listing has moved online. What there has not been is a doctrine that says the Chief Justice's listing power is constrained by anything. The 'master of the roster' framing in Asok Pande rules out judicial review of listing decisions; the absence of any binding listing-policy text means that there is also no internal constraint. The gap between the two is the space in which bench hunting operates. Closing the gap requires the Court to do one of three things: publish a binding listing policy; submit listing decisions to review; or refer the question to a constitution bench under Article 145(3) and let it lay down guidelines. None of these has happened. The Court has, in effect, declared that the question is not justiciable, and then declared that it is also not for it to answer outside of justiciability. Both halves of the answer cannot be right.

It is worth pausing on a related procedural innovation, because it tells us how the Court treats unwritten rules elsewhere. In Rupa Ashok Hurra v Ashok Hurra⁸, the Court created the curative-petition jurisdiction — a 'second review', as Vakil calls it, that contravenes the express terms of Article 137 and Order XL of the Supreme Court Rules. The Court justified it as an inherent power to do justice. It has since been reluctant to admit curative petitions in practice, but the doctrinal point stands: the Court can, when it wishes, invent procedural tools without seeking legislative or constitutional sanction. If the Court can write rules to expand its own jurisdiction, it can write rules to discipline its own listing. The asymmetry is striking. Tools for the Court's reach are invented at will; tools for the Court's restraint are deferred indefinitely.

The steel-manned defence of the present arrangement

The strongest case for the present system is institutional and I want to set it out at length, because an argumentative essay that ignores its target's strongest claim is just shouting.

The defence has three pillars. The first is volume. The Supreme Court of India hears more matters per working day than any apex court in a comparable jurisdiction. Nick Robinson's data, cited by Vakil, shows that 84.6 per cent of the docket is special leave appeals⁹; the Court's daily mentioning list runs to a thousand items in some weeks. Any rule that requires the Court to club cognate matters, or to refer conflicts to larger benches, or to publish a binding listing policy, would, on this view, simply jam the machinery. The Chief Justice's discretion in listing is what keeps the Court moving. A rule against bench-listing discretion would be a rule that the Court must grind to a halt.

The second pillar is the institutional logic of the master-of-the-roster doctrine. Some person must decide which case goes to which bench. The Constitution names the Chief Justice as the head of the Court. Locating the listing power in the Chief Justice is not arbitrary; it is the institutional consequence of the constitutional design. If the listing power were subject to judicial review, the reviewing court would itself need a list, and a master to manage it, and the regress is endless. The Chief Justice's discretion is, on this view, an institutional necessity that the Constitution has implicitly accepted by making the Chief Justice the head of the Court.

The third pillar is the empirical claim that bench hunting, while real, is overstated. Most petitions reach the bench they are randomly assigned to. Most withdrawals before listing happen for reasons other than bench-listing strategy. The cases in which bench hunting actually changes outcomes are, on this view, a small number of high-profile matters that draw disproportionate attention. To redesign the listing system around those cases is to redesign the system around a tail risk. The right response to bench hunting, on this account, is the response any institution makes to outlier conduct: targeted oversight in the cases where it matters, not structural reform of the routine.

This is the steel-manned defence. I think it has real weight, particularly the first pillar; the volume problem is real, and any reform proposal has to take it seriously. But the defence fails for three reasons that I want to take in turn.

Where the defence fails

The volume argument proves too little. It is true that the Court hears more matters than any comparable apex court. It does not follow that the answer to volume is unchecked listing discretion; it follows that the answer is some combination of jurisdictional reform (the cassation-court proposal, fewer special leaves on routine appeals), more judges, and more transparent listing rules. The volume argument is, in fact, an argument for structural reform of the kind the Court has resisted, not an argument against the structural reform of listing. To run the volume argument the way the defenders of the present system run it — as a reason to leave the Chief Justice with unchecked discretion — is to use a problem of input as a justification for opacity of process. Those are different problems with different solutions.

The institutional-necessity argument proves something narrower than it claims. It is true that some person must decide which case goes to which bench. It does not follow that the decision must be unreviewable, unwritten, and uncontestable. Most apex courts, including those in jurisdictions where listing is centralised in a single judicial officer, publish their listing principles. The German Federal Constitutional Court has published rules on the allocation of cases between its two senates. The US Supreme Court's procedure for assigning cases to opinion-authorship is governed by published norms. The South African Constitutional Court allocates cases by predetermined rotation, with departures requiring published reasons. None of these courts has dissolved into administrative chaos. The infinite-regress argument advanced in Asok Pande — that any review of the listing power would itself need a list and so on — is a debating point, not a constitutional argument. Every administrative power in a constitutional order is reviewable on grounds of mala fides and arbitrariness; there is no reason listing should be the unique exception, and the Constitution does not say it is.

The empirical claim is the weakest of the three. The bench-hunting cases that get attention are the politically salient ones, but they are the visible tip of a larger pattern. The pattern is visible to anyone who watches the Court's listing pages over time: the same kinds of corporate writs end up before the same kinds of benches; the same kinds of fundamental-rights petitions end up before the same kinds of benches; the matching is not random. I do not have, and could not have, public data that establishes this rigorously — the Court does not publish bench-assignment data in a researchable form, which is itself part of the problem. But the absence of data is not evidence that the practice is rare; it is evidence that the practice is unobserved. And the cases where bench hunting demonstrably changes outcomes — the 2G review, the medical-college-admissions matter, several environmental writs on coal and forest clearances — are not a 'tail risk' in any sense the public-law literature would recognise. They are the cases that define what the Court's Article 32 jurisdiction means.

There is a fourth defence sometimes offered, weaker than the three above but worth naming. It is that the Court has, in practice, been a good guardian of rights, and that whatever the procedural opacity, the substantive record is what matters. I think this defence collapses on inspection. An institution that has produced good outcomes under an indefensible procedure has not made the procedure defensible; it has, in spite of the procedure, done some of what was asked of it. That is a defence of the institution's character, not of the institution's rules. And the question my essay raises is about the rules. A rule that permits bench hunting is a bad rule even if the Court that operates the rule is, at most times, a decent court. The two assessments are independent. The first is about procedure; the second is about personnel. Conflating them is how procedural problems get protected.

What ought to change

I want to propose three concrete things, not as a programme of reform (the Court is unlikely to adopt any of them on its own) but as a way of locating what the doctrine should be.

First, the Chief Justice's listing policy ought to be a published, binding document. Categories of matters that go to which benches; the principles by which a matter is to be clubbed with a cognate matter; the procedure for the Chief Justice to reassign a matter mid-way through. None of these would prevent the Chief Justice from exercising judgment in particular cases. All of them would force the judgment to be exercised against a published baseline. The 'master of the roster' framing in Asok Pande would then mean what it should have meant — a presumption in favour of the listing decision, rebuttable by showing that the decision violates the published policy. That is a workable doctrine, and it is one most administrative-law systems have already arrived at.

Second, there should be a duty of disclosure on petitioners filing Article 32 writs. If a cognate matter has been filed before, or if a petition on materially similar facts has been withdrawn before listing, the second petitioner should be required to say so. The standard is not invented; it is the standard already applied in the Court's commercial jurisdiction for the disclosure of related proceedings. Extending it to Article 32 would not eliminate bench hunting, but it would force the practice into the open, where the Court could see it and act on it. The Court's existing power to dismiss for forum-shopping abuse would then have something to bite on.

Third — and this is the one that the Court has the smallest appetite for — the practice of two-judge benches deciding novel Article 32 questions should be ended by self-imposed rule. Vakil's chapter records that the framers expected questions of constitutional interpretation to be heard by 'a bench of a minimum of five judges' under what became Article 145(3); the practice that has emerged, of two-judge benches deciding such questions and then declining to refer them to larger benches, is a departure from the constitutional design. A self-imposed rule that any matter raising a substantial question on Article 32 must be heard by at least a three-judge bench would do most of the work. It would reduce the number of doors a litigant can choose between. It would force the Court to acknowledge that an Article 32 question is a constitutional question, not an SLP-level question. And it would, by raising the bench strength, raise the cost of bench hunting to the point where the technique becomes uneconomic.

The democratic stakes

I want to close with what I think is the underlying stake, because the bench-hunting debate is often treated as a technical or housekeeping question, and it is not.

Article 32 was designed as the direct route by which a citizen whose rights have been violated reaches the apex court. The choice of design — direct writ, no exhaustion of remedies required, no certification needed — was a constitutional choice about who controls access to rights-enforcement. The framers made the choice in favour of the citizen and against any administrative gatekeeper. The present listing practice has reinstated, by indirect means, the gatekeeper the framers refused to install. The citizen who can afford to file three petitions and withdraw two reaches the bench she wants. The citizen who can afford one petition reaches the bench she draws. The first citizen has, by economic means, the Article 32 the framers designed; the second citizen has, by the same means, something less. The Court's silence on listing is what made this possible.

It is sometimes said, in defence of the unwritten regime, that the Court's discretion has been exercised in good faith and that the writing of rules would only invite litigation about the rules. I find this defence unpersuasive on its own terms. The discretion has been exercised in good faith in many cases; it has also, on the evidence of the January 2018 press conference by Justices Chelameswar, Gogoi, Lokur and Joseph, been exercised in some cases in ways that the senior-most judges of the Court thought worth raising in public. The press conference was an extraordinary event. Four judges of the Supreme Court of India do not stand on a lawn and brief the press about case-assignment unless they have concluded that the internal channels are inadequate. The Court that received that briefing did not respond by writing rules; it responded by reaffirming the master-of-the-roster doctrine in Asok Pande. The response told us how the Court intends to handle internal dissent on listing — by closing the question rather than answering it.

The point is not that the listing problem is unfixable. It is that the institution that would have to fix it has chosen, repeatedly, not to. The argument for writing the rules has not become weaker over the last twenty-five years; if anything, it has become stronger, as the Court's docket has expanded, as the consequences of bench-assignment in commercial and political matters have grown larger, and as the gap between the Court's substantive jurisprudence on Article 32 and its procedural practice on Article 32 has widened. If the rules are not written, the Court will, sooner or later, be forced to answer the question that the press-conference judges were asking: who actually runs the Supreme Court? An institution that cannot answer that question publicly is an institution that has already conceded something the Constitution did not authorise it to concede.

What I have argued is not that bench hunting is the worst thing in Indian constitutional practice. It is not. There are larger pathologies — the executive's capture of the appointments process, the legislature's evasion of substantive debate, the routine criminalisation of political dissent. Bench hunting is a smaller problem than any of those. But it is a problem the Court itself can fix, with rules the Court itself can write, without any cooperation from Parliament or the executive. The problem is not jurisdictional. The problem is institutional will. The Court has, for twenty-five years, declined to write the rules that would constrain its own listing discretion. The decision not to write the rules is itself a decision. It is the one I think the Court should be willing to revisit.

Get the week's biggest decisions, distilled.

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