Separation of powers is not what India has

The textbooks teach a sentence that is wrong. They teach that the Indian Constitution embodies a doctrine of separation of powers, in some Indian version of what Montesquieu and the American framers were doing, with judicial review as its sharpest expression. The doctrine, the textbooks say, has produced the basic structure, the rule in Kesavananda Bharati, the strike-down of the National Judicial Appointments Commission, the periodic judicial scoldings of the executive for ordinance abuse. Separation of powers, on this account, is the central organising principle of the Indian constitutional order. It is what the Court invokes when it draws lines between itself and the other branches. It is what protects the citizen from concentrations of power. I want to argue, against this account, that the Indian Constitution does not separate the branches in any sense the doctrine would recognise. It fuses two of them in Article 75(3). It permits the third, since 1993, to appoint itself. The vocabulary of separation that judicial-review doctrine still reaches for is a borrowed costume — one that fits the American constitutional body and does not fit ours, and that has, by being kept in the wardrobe, distorted the work the Indian Court actually does.

The settled view, in its strongest form

Let me state the case I disagree with before I dismantle it. The settled view does not rest on naivety. It rests on three propositions, and the strongest version of the case takes each of them as far as it can go.

The first proposition is textual. Article 50 of the Constitution directs the State to take steps to separate the judiciary from the executive in the public services of the State. Articles 53 and 154 vest executive power formally in the President and the Governors. Article 121 forbids parliamentary discussion of the conduct of judges in the discharge of their duties except on a motion for removal. Article 122 and Article 212 keep parliamentary and State-legislative proceedings beyond judicial scrutiny on grounds of irregularity of procedure. These provisions, the settled view says, are textual proof that the framers wanted three branches doing three different things, with limits between them written into the document itself.

The second proposition is doctrinal. In Kesavananda Bharati v State of Kerala in 1973, separation of powers was named as part of the basic structure that the constituent power of Parliament could not amend away. In Indira Nehru Gandhi v Raj Narain in 1975, the Court struck down the 39th Amendment partly on the ground that the amendment violated separation of powers by carving out the Prime Minister's election from judicial review.¹⁶ In SR Bommai v Union of India in 1994, separation of powers was again named as an element of basic structure. In I R Coelho v State of Tamil Nadu in 2007, separation was used to justify the conclusion that Ninth Schedule laws are still amenable to fundamental-rights review. The doctrinal accretion is real. The Court has used the phrase, repeatedly, in cases of considerable consequence.

The third proposition is institutional. The Indian judiciary has, on the settled view, behaved as a separate branch. It struck down the NJAC. It has imposed proportionality review on executive action. It has policed the boundaries of delegated legislation in cases like Hamdard Dawakhana and Gwalior Rayon Silk. It has, in Madras Bar Association and successors, ringfenced the upper judiciary's appellate jurisdiction against tribunal encroachment. Whatever the textual reading might be, the institutional reading is plain: the Indian Supreme Court conducts itself as if it occupies one of three constitutionally separated stations, and the other branches, with varying degrees of resentment, treat it as such.

Take these three propositions together and the settled view writes itself. Separation of powers is real, it is doctrinal, it is institutional, and the Constitution embodies it in an Indian register that combines Westminster fusion at the executive-legislative interface with Montesquieu-style isolation of the judiciary at the other end. This is the position I want to argue is wrong — not in its parts, which are mostly accurate as descriptions of what happens, but in the conclusion it asks us to draw about the structural grammar of the document.

What the framers actually wrote

The clearest sustained judicial statement of the structural position is buried in Re Delhi Laws Act 1912 in 1951, a seven-judge bench answering whether delegated legislation was constitutional. Fazl Ali J's plurality opinion contains paragraph 285, in which the bench notes — flatly, without polemic — that 'the Indian Constitution does not expressly vest the different sets of powers in the different organs of the State'.² The same Court, sixty years later, would assert the opposite — that there is 'distinct and rigid separation of powers under the Indian Constitution' — without troubling to identify the provision the rigidity is supposed to reside in.³ Both statements are official; only the earlier one is correct. The only express vesting in the original Constitution is of executive power, in the President under Article 53 and the Governors under Article 154. Legislative power is conferred on Parliament and the State Legislatures in Articles 245 and 246, but the Constitution does not say that this power must be exercised by them alone, and the body of delegated-legislation jurisprudence built on top of Re Delhi Laws takes that omission as a permission. Judicial power is nowhere expressly vested in the courts; it is implied by the existence of the Supreme Court, the High Courts, and their jurisdictions in Articles 32, 131-141, 226, and 227.

This is a strikingly different architecture from Article I, Article II, and Article III of the United States Constitution. Each American article opens with the same construction — 'The legislative powers herein granted shall be vested in a Congress', 'The executive Power shall be vested in a President', 'The judicial Power of the United States, shall be vested in one supreme Court'. The Indian Constitution opens no such article in those terms. Its framers, all of whom were familiar with the American text, chose not to mirror it. The textual asymmetry is not accidental. It is the most basic structural choice the framers made about how the document would work.

The choice was a Westminster choice. Article 75(3) makes the Union Council of Ministers — that is, the entirety of the executive at the political level — collectively responsible to the House of the People.⁵ Article 164(2) does the same at the State level for the Council of Ministers and the Legislative Assembly. The ministers must be members of one or the other House of Parliament within six months of appointment under Article 75(5). The Prime Minister is constitutionally required to be the head of a Council of Ministers under Article 74(1), the proviso to which now binds the President to act on that Council's advice. In Ram Jawaya Kapur v State of Punjab, Mukherjea CJ for a five-judge bench said it plainly: 'The Cabinet enjoying, as it does, a majority in the legislature concentrates in itself the virtual control of both legislative and executive functions.'⁷ That formulation has never been judicially repudiated. It is the operating description of the system.

Bagehot, writing about the British system in 1873, called the Cabinet a 'hyphen which joins, a buckle which fastens, the legislative part of the State to the executive part'.⁶ Justice Ruma Pal, in the chapter that anchors much of the constitutional-law discussion of this question, treats the description as applying with equal force to India.¹ H M Seervai was even blunter. The Constitution, he wrote, is not based on the separation of powers.⁴ Seervai's view was, until somewhere in the 1980s, the doctrinal orthodoxy. What happened between Seervai and the textbooks that now teach the opposite is the subject of the rest of this essay.

The executive sits inside the legislature

The first reason the separation-of-powers vocabulary does not fit India is that, in the system the framers built, the political executive is a sub-set of the legislature. The Prime Minister is a Member of Parliament. So is every Cabinet Minister. They draft most legislation, they introduce it, they whip their party to vote for it, and they implement it after it passes. The Council of Ministers in Parliament, with a working majority, controls the legislative calendar, the committee structure, the speech list, and the timing of every vote that matters. To call this a separation between executive and legislature is to use the word in a sense that the American framers, drafting against the experience of George III's ministers in a hostile Parliament, would not have recognised.

The point sharpens when one looks at the residual executive power under Article 73 of the Constitution. Article 73 extends the Union's executive power to all matters on which Parliament has power to make laws. In Ram Jawaya Kapur, the Court held that this does not require Parliament to have actually legislated; executive action can proceed wherever Parliament could legislate.⁷ The Indra Sawhney bench confirmed that executive orders in such a field carry the same legal force as parliamentary legislation.¹² Across an enormous range of subject-matters — public employment, education, health policy, much of welfare administration — the Union and the States govern by executive order, on the strength of legislative competence that has never been exercised. A federal executive that can legislate by order in fields the legislature has not even entered is not a separated executive. It is a legislature in another dress.

This is not an oversight. It is constitutional design. Article 77 requires the President to make Rules of Business for the Government of India; those Rules allocate decision-making between ministers and their departmental subordinates and have binding statutory force on executive action. Article 123 permits ordinances. Article 213 permits gubernatorial ordinances. Article 309 permits the executive to make service rules where Parliament has not legislated. Article 320(3) requires consultation with the Union Public Service Commission for service matters but does not bind the executive to its advice. Each provision authorises the executive to do things that, in a system that took separation seriously, would have to be done by the legislature. The Constitution does not call them legislative; it calls them executive. The distinction is doing no real work.

The ordinance power is the sharpest illustration. Article 123 says the President may promulgate ordinances having the same force and effect as Acts of Parliament when Parliament is not in session and the President is satisfied that circumstances exist which render it necessary to take immediate action. The Supreme Court has repeatedly held that the President's satisfaction under Article 123 is not justiciable, though the Krishna Kumar Singh bench in 2017 finally walked back some of the worst implications.¹⁴ Between 1960 and 1980 the Union Government promulgated about two hundred ordinances. The Bihar Government, between 1967 and 1981, kept some ordinances 'alive' for three to fourteen years by repromulgation, which DC Wadhwa v State of Bihar in 1987 finally addressed — by saying that such repromulgation was unconstitutional, but leaving open a window for situations where 'there is too much legislative business' or 'time at the disposal of the Legislature is short'.¹³ Every successive Union and State government has driven through that window. The ordinance power, conceived as an exception, has been worked as a parallel legislative track. The Court, having noticed the practice, has not closed it.

Add to all this the volume of delegated legislation that the executive produces under enabling statutes — the rules under the Income Tax Act, the regulations under the SEBI Act, the notifications under the Customs Act, the rules under the IT Act, the price control orders under the Essential Commodities Act, the notifications under the Foreign Trade Act. Justice Pal quotes the Supreme Court's own admission, in Cynamide India, that with the proliferation of delegated legislation 'there is a tendency for the line between legislation and administration to vanish into an illusion'.⁸ The doctrine of excessive delegation that the Court has theoretically maintained since Re Delhi Laws Act 1912 has, by Justice Pal's reckoning and by mine, almost never been deployed to strike down an executive order on that ground.¹¹ The only outer limit the Court has recognised is the rule against 'abdication and self-effacement' by the concerned legislative body, and even that limit has been read in such a way that statutes which delegate broad powers to fix taxes, fix prices, extend the coverage of legislation to new geographical areas, and decide when statutes come into force have all been upheld. A separation of legislative and executive power that permits the executive to legislate in the legislature's name, on the legislature's say-so, on subjects the legislature did not specify, is not a separation. It is a sub-contract.

The judiciary appoints itself

The second reason the separation-of-powers vocabulary does not fit is that, since 1993, the Indian higher judiciary has been the only major court in any democratic system that appoints itself. The Second Judges Case in 1993 invented the collegium of the Chief Justice of India and the senior-most judges as the body whose recommendation on judicial appointment is binding on the President.⁹ The Third Judges Case in 1998 fixed the collegium at five for Supreme Court appointments and three for the Chief Justice's consultations on High Court appointments. The text of Articles 124 and 217 still says the President appoints judges after consultation with the Chief Justice and such other judges as he may think necessary. The text now means the opposite of what it says: the executive consults the judges, the judges decide, the executive ratifies.

In 2014 Parliament passed the Ninety-ninth Amendment to the Constitution and the National Judicial Appointments Commission Act, replacing the collegium with a six-member NJAC consisting of the Chief Justice, the two senior-most Supreme Court judges, the Union Minister of Law, and two eminent persons selected by a panel consisting of the Prime Minister, the Chief Justice, and the Leader of the Opposition. The amendment passed both Houses with overwhelming majorities. Sixteen State Legislatures ratified it. A five-judge bench of the Supreme Court struck it down 4-1 on 16 October 2015 in Supreme Court Advocates-on-Record Association v Union of India.¹⁰ The reasoning of the majority was that judicial independence, as a feature of the basic structure, requires that judges of the higher judiciary be appointed by a body in which the judiciary holds the controlling voice. The NJAC, the majority said, did not give the judiciary that controlling voice. Chelameswar J, in dissent, said something different. He said the collegium was opaque, captured by personal preference and back-scratching, untransparent in its working, and unaccountable in its outcomes — and that nothing in the basic-structure doctrine required the constitutional position to be the way the Second and Third Judges Cases had made it.

The position now is that the higher judiciary in India is the only branch of the State whose recruitment is decided by itself. Civil services are recruited through the UPSC. Members of Parliament are elected. Ministers are drawn from those elected representatives. Judges of the higher judiciary are chosen by judges of the higher judiciary, ratified by the Council of Ministers on pain of stalling the process, and appointed by the President on the ratified recommendation. The American Supreme Court is appointed by the President with the advice and consent of the Senate; the UK Supreme Court is selected by a five-member Commission of which only the Court's own chair and one nominated judge sit; the Canadian Supreme Court is appointed by the Governor in Council on the Prime Minister's recommendation after a parliamentary committee process. India's collegium is the global outlier.

A judiciary that selects its own members is not, in any standard separation-of-powers sense, separated from itself. The collegium is the judiciary's executive arm, the judiciary's appointing authority, the judiciary's grievance committee, and the judiciary's promotions board. Its decisions are not justiciable in any meaningful sense, because the body that would adjudicate them is composed of the same judges. The 'master of the roster' rule — under which the Chief Justice alone assigns benches — adds an internal executive function to the same office, and the Court's polyvocal structure means that the same body can take inconsistent positions on its own institutional role in cases decided weeks apart.¹⁵ The Court does not separate adjudication from administration; it concentrates both in the senior-most judges who, in their administrative capacity, decide who hears what case and, in their adjudicative capacity, decide what the law is. The position is a long way from anything Montesquieu would have called separation.

What 'separation' has been doing in the doctrine

The honest reader will ask at this point what work the separation-of-powers vocabulary is doing in Indian doctrine if the structure is, in fact, what I have described. The answer is that it is doing two things, and both of them are things the Court could do more cleanly without the borrowed vocabulary.

The first is institutional-competence work. When the Court tells the executive that it cannot determine the guilt of a particular individual by Order — as the Court held in State Bank's Staff Union (Madras Circle) v Union of India — what it is really saying is that adjudication of an inter-partes dispute is something the executive is institutionally ill-equipped to do. The Court has, in fact, said as much. Reaching for separation of powers as the doctrinal label adds nothing analytically. Institutional competence — the proposition that disputes between identified parties require an adjudicator independent of either party — does the work directly. The same is true of the body of cases on tribunals. The R Gandhi line of cases on the National Company Law Tribunal and the National Tax Tribunal are framed in separation-of-powers language, but the actual reasoning is about institutional competence to hear appeals against High Court orders and about whether the executive can compose a body that supplants High Court jurisdiction. The work is real; the label is misleading.

The second is institutional-protection work. When the Court strikes down a constitutional amendment that purports to immunise the Prime Minister's election from judicial review, as the Court did in Indira Nehru Gandhi v Raj Narain, what it is really saying is that the rule of law requires final adjudicative authority to be located somewhere outside the political branches.¹⁶ The American constitutional vocabulary calls this separation of powers because the American Constitution genuinely separates the three branches at the threshold. The Indian Constitution does not so separate them at the threshold. What the Indian Court is doing — and doing rightly, in my view — is something more specific: it is preserving the institutional position of an independent adjudicator within a system that was otherwise built on legislative-executive fusion. The vocabulary of separation makes the work harder to defend than it should be, because it invites the legitimate retort that the Indian Constitution does not separate. The vocabulary of institutional protection does not invite that retort.

One more cost: the borrowed vocabulary makes it harder for the Court to acknowledge the parts of the system that have outgrown any conceivable separation principle. The collegium is the clearest example. If the Court were operating on a vocabulary of institutional competence and independence, it could honestly say that judicial independence is a value, that the collegium is one possible way of securing that value, that other systems secure the same value through different institutional arrangements, and that the design of the appointment system is, within the limits of independence, a legislative choice. The NJAC judgment did not say this. It said that the collegium was constitutionally required, because separation of powers is part of the basic structure and the NJAC violated it. That conclusion is sustainable only if you believe — as I do not — that the framers wrote a doctrine of separation of powers into the Constitution. If they did not, the conclusion has no anchor.

The counter-argument, taken honestly

The strongest reply to all this is one I have to take seriously, because it is the reply Justice Pal herself ends her chapter with, and she is the most authoritative recent commentator on this question. The reply runs: yes, the Indian Constitution does not write Montesquieu-style separation into its text; yes, the architecture is fusion at the executive-legislative interface and self-selection in the judiciary; but the post-1993 Supreme Court's invocation of separation of powers is not a description of the original text. It is a doctrinal construct, built up over decades of constitutional interpretation, that serves the same functional purpose that separation of powers serves in the American system — namely, the prevention of dangerous concentrations of power. The reply is that the framers may not have written separation in, but the basic-structure jurisprudence has read it in, and the reading has been salutary.

I take this seriously. The basic-structure jurisprudence has been salutary. The Court that struck down the 39th Amendment was right to do so. The Court that struck down parts of the 42nd Amendment in Minerva Mills was right to do so. The Court that brought the President's Rule provision under judicial review in SR Bommai was right to do so. The Court that constrained the Ninth Schedule in IR Coelho was right to do so. None of this is in dispute. The question is whether 'separation of powers' is the right vocabulary in which to describe what the Court has done in these cases, or whether the vocabulary obscures the actual reasoning. My claim is that it obscures.

Take Minerva Mills specifically. The Court struck down clauses (4) and (5) of Article 368 because they purported to make constitutional amendments beyond judicial review and to give the Directive Principles supremacy over the Fundamental Rights for a class of laws. The Court used basic-structure reasoning. Was the operative reason 'separation of powers'? It was not. The operative reason was that the Court would not accept an amendment that ousted judicial review of the validity of constitutional amendments themselves, because that ouster would have left no institutional check on the amending power. That is institutional-competence reasoning. The Court was the institution best placed to test constitutional amendments against the rest of the Constitution; the amendment that removed that test left no other institution to do the testing. The vocabulary of separation adds nothing here. The vocabulary of institutional check does the analytical work.

The reply will then say: well, what difference does the vocabulary make? If the Court arrives at the right outcome, why fight about the label? It makes this difference. The vocabulary of separation lets the Court frame its decisions as if they flow from a constitutional architecture that pre-exists the decisions themselves. The vocabulary of institutional check, or institutional competence, would force the Court to defend the institutional position on its own merits — to say why this institution and not that one should be the final adjudicator, why this kind of question is one the political branches should resolve, why this kind of decision requires an independent body. The NJAC judgment is the clearest illustration of the cost. Had the majority engaged with the question on institutional-competence grounds, it would have had to defend the collegium as the best institutional arrangement for securing judicial independence. It would have had to engage with Chelameswar J's evidence about how the collegium actually works. It would have had to consider whether a body designed differently — say, a commission with a non-judge majority but with judicial veto over individual appointments — could secure the same value. Instead, it said that the collegium follows from separation of powers, that the NJAC violated separation, and that the matter was therefore foreclosed by basic-structure doctrine. That reasoning will not survive serious scrutiny.

What follows from naming it correctly

If I am right that the Indian Constitution does not embody separation of powers but a Westminster-style fusion with a self-selecting judiciary, three things follow for how judicial-review doctrine should be done.

The first is that the Court should stop reaching for the vocabulary of separation when it means something else. When the Court tells the legislature that it cannot retroactively reverse an individual judicial decision inter partes, the doctrinal ground is not separation but the rule of law and the bar against legislative adjudication of particular disputes. When the Court tells the executive that it cannot suspend the operation of a statute the legislature has enacted, the ground is not separation but the supremacy of the statutory text and the limits on the executive power conferred under Article 73. When the Court tells the legislature that it cannot oust judicial review of constitutional amendments, the ground is not separation but the structural requirement of an independent test of validity. Each of these grounds is available without the borrowed vocabulary. Using the vocabulary makes the reasoning harder to defend; abandoning it makes the reasoning clearer.

The second is that the doctrines built on top of the borrowed vocabulary should be re-examined. The doctrine of excessive delegation, as Justice Pal documents, has had almost no operative content; the Court has not in living memory struck down delegated legislation on that ground alone.¹¹ The doctrine of judicial review of ordinances has, despite the protective language of Krishna Kumar Singh, allowed Article 123 and Article 213 to be used as a parallel legislative track without effective judicial constraint.¹⁴ The doctrine of separation against tribunals has produced the R Gandhi line, which is real, but the doctrinal anchor of those cases is institutional competence, not separation. If the Court wanted to do the work of regulating the executive-legislative overlap with the seriousness the system requires, it would need a different vocabulary — one that named the overlap honestly and tested specific instances of it against specific institutional values.

The third is that the collegium needs to be defended, if it is defensible, on its merits and not on the borrowed vocabulary. The Court that struck down the NJAC owes the Indian public a doctrinal account of why the collegium produces better judicial appointments than the alternative arrangements democratic systems use, why its opacity is compatible with the rule-of-law values it is supposed to protect, and why the institutional defects Chelameswar J catalogued do not require constitutional correction. The Court has not given that account. It has, instead, said that the matter is settled by basic structure and the doctrine of separation of powers. That answer is not adequate. Until the Court gives a better one — or until the political branches force the question again — the position of the higher judiciary as a self-selecting body will continue to sit awkwardly in a Constitution that, on every other dimension of institutional power, requires accountability to the electorate or to its representatives.

The doctrine of separation of powers, as it is taught and as the Court reaches for it, describes a constitutional architecture the framers did not build. The framers built something different. They built a Westminster executive-legislative fusion, supplemented by an extensive system of executive law-making, on top of a judiciary whose independence they secured through tenure and salary protections and whose appointment they left to political-judicial consultation. The post-1993 collegium and the post-1980 explosion of delegated legislation have moved the system further from any conceivable doctrine of separation, not closer to it. The Court has continued to use the vocabulary anyway, because the vocabulary licenses the conclusions it wants to reach. The license is a counterfeit. The reasoning the Court actually relies on — institutional competence, the rule of law, the independence of adjudication, the structural requirement of a non-political test of validity — is sound, and is what the Court should be saying instead. Until it does, the Indian constitutional vocabulary will continue to teach a doctrine the document does not contain, and the work the doctrine is meant to do will continue to be done by reasoning the doctrine does not name.

Get the week's biggest decisions, distilled.

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