I.  The Many Doors Into a Market

Imagine an enterprise resident in one treaty country that wishes to do business in another. It has, before it, a whole menu of ways to go about it. It may open a branch. It may set up a place of management or an office. It may incorporate a subsidiary. It may run a showroom or a sales outlet, keep a warehouse, build a factory or a workshop. It may sink a mine, drill an oil or gas well, work a quarry. It may take up a building site, an installation or an assembly project of sufficient duration. It may even run a representative office that does more than gather information or advertise, one whose functions go beyond the merely preparatory or auxiliary and so amount to a genuine business presence rather than a mere listening post. Each of these is a different way of being present in the foreign market. Each is, in plain commercial terms, a different business model for the same underlying ambition: to earn profit in a country not one's own.

The doors into a market: every form of presence, from branch to agent, pursues the same ambition.

There is one further door, and it is the one that concerns us. Instead of planting any establishment of its own, the enterprise may do business through a person, an agent who acts on its behalf. In this arrangement there need be no office, no branch, no tangible footprint at all. There is only someone, on the ground, who secures the business for the enterprise. And here is the feature that matters: where the agent is acting for the enterprise in this way, it is the enterprise that carries the entrepreneurial risk of the venture. The agent is, in substance, a front, a conduit, an instrument through which the enterprise itself participates in the market. The agent earns his fee; the enterprise keeps the venture's profit, and bears its risk.

So the agent is not something apart from the list of business models. He belongs on it. Trading through a dependent agent is simply one more way of doing business in the other country, sitting alongside the branch and the subsidiary and the showroom, distinguished only by the fact that the enterprise's presence is achieved through a person rather than through bricks and mortar.

II.  The Choice of Door Should Not Decide the Tax

Once we see these as alternative models, a principle almost announces itself. The tax outcome ought not to turn on which door the enterprise walks through. If two enterprises participate in a market to precisely the same economic extent, one through a branch and the other through an agent, it would be incoherent for the source country to tax the first and exempt the second merely because of the form each chose. Neutrality across forms of presence is the bedrock policy of this whole area of law. The source state's taxing right should attach to the substance of market participation, not to the costume the enterprise happens to wear.

The negotiated threshold: where trading with a country ripens into doing business in it.

The permanent establishment is the device by which the treaties give effect to that principle. It is a negotiated threshold, the agreed line that separates mere dealing-with a country from doing-business-in it. Below the threshold, the source state stays its hand and leaves the profit to the residence state. On crossing the threshold, the source state acquires the right to tax, though only so much of the enterprise's profit as is attributable to the establishment. That is the settled architecture of Article 7(1): no permanent establishment, no source taxation of business profits; a permanent establishment, and the source state may tax what is attributable to it.

The agency permanent establishment is the treaties' way of bringing the agent-model across that same threshold. Because trading through a dependent agent is economically a way of doing business in the country, the treaties deem the enterprise to have an establishment there by virtue of the agent's activities. The neutrality principle demands nothing less: if the branch crosses the threshold, so must the economically equivalent agency arrangement. Otherwise the threshold would reward the enterprise that chose the cleverer form, which is the very mischief neutrality exists to prevent.

It is worth being precise about what the dependent agency permanent establishment is, because much of the confusion that follows flows from a hazy grasp of the concept, and precision begins with origins. The permanent establishment was conceived, in the treaty practice that took shape in the early twentieth century and matured through the League of Nations models, as a measure of the degree of presence at which a foreign enterprise's involvement in a country becomes substantial enough to justify taxing its business profits there, in much the way residence justifies taxation of a person. It is, in essence, the dividing line at which mere trading with a country ripens into doing business in it, and the source state's claim ripens with it. Against that backdrop the agency rule follows naturally. In its classical formulation, a person acting in the source state for or on behalf of a foreign enterprise, who is not an agent of independent status, is deemed to give rise to a permanent establishment of that enterprise if he has, and habitually exercises, an authority to conclude contracts in the name of the enterprise. The independent agent, the broker or general commission agent acting in the ordinary course of his own business for many principals, is excluded; the rule reaches only the dependent agent, the person who is, in substance, an arm of the enterprise rather than a trader on his own account. The premise is straightforward: where such a person habitually binds the enterprise in the market, the enterprise has attained that degree of presence, and the deeming fiction places it across the threshold as surely as if it had opened a branch.

That classical test, however, was keyed to a single formal act, the conclusion of contracts in the name of the enterprise, and astute structuring learned to avoid it. The commissionnaire was the textbook escape: a person who negotiated and effectively settled the sale, drawing the customer in and fixing the terms, but who signed the contract in his own name rather than the enterprise's, and so fell outside a rule that asked whose name was on the paper. Article 12 of the Multilateral Instrument was the response. It widened the dependent-agent definition in two ways. First, it extended the trigger beyond the person who concludes contracts in the enterprise's name to the person who habitually plays the principal role leading to the conclusion of contracts that are routinely concluded without material modification by the enterprise, language aimed squarely at the commissionnaire who does the real work of the deal whatever the signature page may say. Secondly, it tightened the independent-agent exception, so that a person acting exclusively or almost exclusively for one or more closely related enterprises can no longer shelter behind a claim of independence. The effect was to drag the commissionnaire, and arrangements like it, back within the agency net. We will see in due course that this was a real repair, but a narrow one, and that it left the deeper question untouched.

III.  How Are the Establishment's Profits to Be Computed? The Real Controversy

So far so uncontroversial. The difficulty, and it is the difficulty around which this entire subject turns, begins the moment one asks a simple question: once the enterprise is deemed to have an agency permanent establishment, how much profit is to be attributed to it? For a reader meeting this problem for the first time, it helps to see why the question is hard. The establishment is a fiction; it has no separate accounts, no balance sheet, no profit of its own waiting to be read off a ledger. Its profit must be constructed. And there are two rival ways of constructing it, two schools that have contended for decades and that lie at the root of the whole debate.

The first is the two-taxpayer (or non-zero-sum) approach. On this view, the dependent agent and the dependent agency permanent establishment are two distinct units of taxation. The agent is taxed, under domestic law, on the remuneration he earns for his services. Separately, the foreign enterprise is taxed, under Article 7, on the profit attributable to the establishment, computed by treating that establishment as a hypothetical independent enterprise and allowing, as a deduction, the remuneration paid to the agent. The agent's fee is thus a cost in arriving at the establishment's profit, not a substitute for it.

The second is the single-taxpayer (or zero-sum) approach. On this view, if the agent has been paid an arm's length reward for the functions he performs, the assets he uses and the risks he assumes, then the establishment and the agent collapse into one: tax the agent correctly, and nothing further remains to attribute to the enterprise. The agent's arm's length remuneration is treated as exhausting the source state's claim.

Two schools of thought: the gap between them is the difference between taxing a sliver and taxing the substance.

The gap between the two is not a technicality; it is the difference between taxing a sliver and taxing the substance. An illustration, of the kind set out by the Tribunal in SET Satellite (Singapore) Pte Ltd in 2007 [(2007) 106 ITD 75 (Mum)], makes it vivid. Suppose a foreign enterprise sources goods abroad and sells them into the source country for three million, through a local agent it pays a commission of nine hundred thousand. The agent, after meeting his own costs, is left with a slender net margin, say a thousand, on which he pays his tax. On the single-taxpayer view, that thousand is the end of the source country's claim. But the enterprise's own profit on the venture, after deducting the commission, the cost of the goods, and a fair charge for the handling functions it performs abroad, may run to half a million. On the two-taxpayer view, it is that half a million, the profit the enterprise earns by deploying its goods and its risk into the market through the agent, that is attributed to the establishment and taxed in the source country. One approach reaches a thousand; the other reaches half a million; and the whole of international tax practice on this question is a debate about which is right.

The anatomy of the gap, on the SET Satellite (2007) illustration.

It was the two-taxpayer approach that the Tribunal adopted in SET Satellite in 2007 (supra), holding that an arm's length payment to the agent does not extinguish the enterprise's liability on the profit attributable to the establishment. (The competing approaches, with the authorities and scholarship on each side, are comprehensively marshalled in the later decision in Asia Today Limited, to which I return.) The choice between the two schools has consequences that reach well beyond arithmetic, and it is to those consequences, and to where each school leads, that I now turn.

IV.  The Threshold That Triggers Nothing

Consider where the single-taxpayer approach leads when it is elevated into a rule that the agent's arm's length fee always exhausts the source state's claim, which is the prevailing position in Indian law. The agent is a separate taxpayer, taxable on his remuneration whether or not any permanent establishment exists. So if crossing the threshold adds nothing to what was taxable anyway, the threshold has triggered precisely nothing. We have built an elaborate doctrine, deemed an establishment into existence, located it, and litigated it, only to announce that its existence is a matter of complete tax indifference. A threshold on crossing which nothing changes is not a threshold. It is an ornament.

A threshold on crossing which nothing changes is not a threshold; it is an ornament.

This is the absurdity at the centre of the matter, and it is worth stating without euphemism. On the single-taxpayer rule, there is, admittedly, a permanent establishment; and there is, admittedly, nothing attributed to it. The two propositions sit side by side, and their coexistence is incoherent. The whole point of recognising the agency permanent establishment was to bring the foreign enterprise's market profit within the source state's reach. To then say that the enterprise owes nothing is to grant the threshold with one hand and abolish its consequence with the other. It denies the very base that the threshold was constructed to create. The two-taxpayer approach, whatever its difficulties of computation, at least avoids this absurdity: it gives the establishment a content, the entrepreneurial surplus, answering to the reason the establishment was deemed to exist at all.

Why, then, does the single-taxpayer approach attract adherents at all? Its great virtue, and in candour its only one, is simplicity. It spares the administration the genuinely hard task of constructing the establishment's profit, and Baker, its most distinguished proponent, has frankly acknowledged that the merit he claims for it is precisely that it avoids an attribution exercise he regards as speculative. Simplicity is a real value in a tax system. But it cannot be bought at the price of incoherence. A method whose simplicity consists in taxing nothing where the law says something is taxable has purchased its ease by abandoning its purpose.

V.  The Dependent Agent Is Not the Establishment

The confusion at the root of the single-taxpayer rule is the equation of two distinct things: the dependent agent, and the dependent agency permanent establishment. They are not the same. It is because there is a dependent agent that the foreign enterprise is deemed to have an establishment. The agent is the cause; the establishment is the consequence; and the two are taxed in different hands on different income.

The agent is the cause; the establishment is the consequence. They are taxed in different hands on different income.

Under Article 7, the taxable person is the foreign enterprise. What is taxed is the enterprise's profit attributable to the establishment. The remuneration paid to the agent is, from the enterprise's standpoint, an expense, not income, and it is allowed as a deduction in computing the establishment's profit. The agent, separately and under domestic law, is taxed on that remuneration as his income. To say that taxing the agent on his fee discharges the enterprise's liability is therefore to allow the tax paid by one person, on his own income, to extinguish the liability of a different person on different income. That is not a doctrine of attribution. It is a confusion of taxpayers, and it is the same confusion, viewed from another angle, that produces the hollow threshold.

VI.  Vogel, Baker, and the Verdict of the OECD

The point was put with unusual clarity by the late Professor Klaus Vogel, the foremost treaty scholar of his generation, and his words deserve to be set out as he wrote them. Commenting in his Tax Treaty Monitor column (Bulletin for International Taxation, November 2007, at page 475) on the very reasoning under discussion, Vogel met head-on the objection that an agency establishment which does not exist in reality can hardly be the source of profit. Such sceptics, he wrote, should consider that the parent enterprise "as a rule will aim to realize receipts from the contracts concluded by the dependent agent which, in addition to compensating the agent's fee, include a surplus profit, for otherwise the parent would lack any commercial reason for employing the agent." That surplus, he continued, is not earned in the residence country; it is a profit the parent obtains through employing the agent in the country where the profits arise, and inter-nation equity requires that it be taxed in that state.

If a treaty is to achieve that, Vogel reasoned, its drafters must notionally attribute the surplus to a contact in the source state, and this does not require attributing it to any real-world person or object: "In the world of law, a legal concept, a figure of thought, will do." The agency permanent establishment, he said, "is such a figure of thought which makes it technically possible to connect the surplus profit to the agent's state," with the consequence that it is "not only possible, but it is the rule that a profit exceeding the agent's compensation will be submitted to the agent's state." The fiction, in other words, is not a sleight of hand; it is a recognised legal technique for fastening real income to the jurisdiction in which it genuinely arises.

The contrary case was made, ably and in good faith, by Philip Baker, who favoured the single-taxpayer approach. Its appeal to simplicity I have already noted; its substantive concern, which deserves to be stated at its strongest, is that to attribute anything beyond the agent's reward is to credit the establishment with functions the agent never performed and risks it never assumed, inviting an arbitrary allocation of profit untethered from reality.

The verdict of global consensus: the OECD's 2008 and 2010 Reports rejected the single-taxpayer theory.

It is a serious objection, and the attribution exercise is indeed where the two-taxpayer approach must do its most disciplined work. But it did not carry the day, and the institutional verdict is unambiguous. The single-taxpayer theory was not adopted by the OECD. It was examined and rejected in the 2008 Report on the Attribution of Profits to Permanent Establishments, and that rejection was carried forward, unchanged, into the 2010 Report, which crystallised the Authorised OECD Approach on the explicit footing that the dependent agent enterprise and the permanent establishment are analytically distinct, and that profit may be attributed to the establishment over and above the agent's reward. The reason is the one Vogel gave: the surplus is real even though the establishment is notional, and to refuse to tax real income because the vehicle of taxation is a fiction is to let the analytical device defeat the income it was designed to reach.

Two clarifications keep the argument honest. First, a rigorous application of the Authorised OECD Approach may, on particular facts, yield a small or even nil residual where the agent genuinely performs all the functions and bears all the risks. But that is a factual result of asking the right question case by case, not a rule that forecloses the question for all cases in advance. Second, the post-BEPS direction of travel confirms rather than disturbs this. Action 7 set out deliberately to lower the agency threshold and bring more arrangements within it, expressly to shift taxing rights toward source states. It would be incongruous to widen the gateway to source taxation while holding that passing through it changes nothing.

There is a further point in the post-BEPS material that repays attention, for it is sometimes pressed in aid of the single-taxpayer approach and in truth tells against it. The 2018 Additional Guidance on the Attribution of Profits to Permanent Establishments, conscious of the administrative burden of permanent establishments yielding little or no separately computed profit, allows that the tax attributable to the establishment may, as a matter of convenience, be collected from the local agent rather than assessed separately on the enterprise. Those who favour the neutrality doctrine read this as a concession that the agent and the establishment are, after all, one. But it is no such concession; if anything, it quietly affirms the opposite. One can speak of collecting the establishment's tax through or from the agent only because the establishment's tax is something distinct from the agent's own, something that must be collected through the agent precisely because it is not the agent's to begin with. Were the two genuinely one, there would be nothing additional to collect, and no need for a mechanism to collect it. The accommodation thus rests on the very duality it is invoked to deny: the distinctness of the agent and the establishment is embedded in the architecture of the relief itself. What is offered as the single-taxpayer approach's support turns out, on inspection, to presuppose the two-taxpayer premise.

VII.  Indian AOA Reservation

A particular argument has gained currency among many eminent domain experts, and it must be confronted directly because it has become the principal prop of the neutrality doctrine. It runs as a syllogism. The two-taxpayer reasoning, the argument goes, rests on the Authorised OECD Approach; India has formally rejected that Approach; therefore its foundation falls away, and what remains, by default, is the single-taxpayer approach under which the agent's arm's length remuneration exhausts the source state's claim. Stated baldly, the contention is that because India rejected the AOA, India is committed to the very thing the AOA was designed to displace.

India against itself: the rejection of the AOA points toward more attribution to the source state, not less.

The argument is ingenious, and, in my understanding, fallacious. The fallacy lies in treating India's rejection of the AOA as though it pointed downward, toward less attribution, when in truth it points upward, toward more. India did not discard the AOA in order to tax the foreign enterprise on a thinner base. It discarded the AOA because, in India's view, the Approach attributes too little to the source state. The reason is a matter of record. India's stated position, set out in its Position on the OECD Model and elaborated in the Central Board of Direct Taxes' Report of the Committee on Profit Attribution to Permanent Establishments (2019), is that it does not accept the Authorised OECD Approach of attributing profits on the basis of functions, assets and risks alone, because that supply-side method ignores the demand side, the market, the sales, the customer base, that India regards as a genuine contributor to the profit. India therefore favours a mixed approach that apportions to the source country a share of the enterprise's profits by reference to both supply and demand, and its treaties deliberately retain the pre-2010 and UN-Model formulations, which preserve the apportionment and direct-accounting methods that the 2010 AOA abandoned.

Once that is understood, the syllogism collapses. A rejection lodged on the ground that the source state should receive more cannot be enlisted as authority for giving it less, still less for giving it nothing beyond the agent's fee. The cause-and-effect the argument asserts simply does not exist. India's quarrel with the AOA is that function-asset-risk attribution is too modest a measure of the source state's due; to convert that quarrel into a charter for the single-taxpayer approach, the most modest measure imaginable, is to make India's reservation prove the opposite of what it says. There is no causal link between rejecting the AOA and embracing the single-taxpayer approach. There is only a verbal coincidence in the word "rejection," which is permitted to smuggle in a direction it does not carry.

And there is a deeper flaw still, which closes the trap. The single-taxpayer approach was never a part of the AOA that a rejection of the AOA could preserve. It was the very theory the AOA condemned, examined and rejected by the OECD as the author of the Approach. The practitioner who invokes India's rejection of the AOA to resurrect the single-taxpayer approach is therefore in an impossible posture: he discards the framework, yet seeks to keep, as a default lurking beneath it, the one component that framework expressly threw out. The single-taxpayer approach is doubly orphaned, disowned by the OECD that built the AOA, and unsupported by the India that rejected it. It is the residue of a misread precedent, and nothing more.

It is worth pausing on how isolated the neutrality doctrine thus stands. So far as I have been able to ascertain, no jurisdiction in the world holds, as a rule of law, that the taxable profit of an agency permanent establishment is confined to the remuneration of the agent. The OECD rejected that proposition as fundamentally incorrect. India rejected the AOA for conceding too little, not too much. The international debate, vigorous as it is, is a debate about how much to attribute to the establishment above the agent's reward, and how to compute it, never about whether the agent's reward is the ceiling. India's neutrality doctrine is the solitary instance of that ceiling being raised to the dignity of binding law, and it is isolated not only from the practice of other nations but from the declared policy of India's own tax administration.

VIII.  The Judicial Analysis- The Unanswered Questions

It is against that background that the Indian judicial position must be assessed, and here I must write with care and I confine myself to conceptual architecture.

The Tribunal in SET Satellite (Singapore) Pte Ltd  [TS-5214-ITAT-2007(Mumbai)-O] adopted the two-taxpayer approach: an arm's length payment to the agent does not extinguish the foreign enterprise's liability on the profit attributable to the establishment. The Bombay High Court reversed in 2008, in the judgment reported as Set Satellite Singapore Pte Ltd v DDIT  [TS-5893-HC-2008(Bombay)-O] and the resulting position is that arm's length remuneration to the agent renders the establishment wholly tax-neutral. With the greatest respect, and in full acceptance of the binding force of that judgment, the conceptual difficulty is the one this article has traced, and it has two strands.

Does it conflate the agent with the enterprise?

As shown above, Article 7 taxes the enterprise, not the agent, and the agent's fee is the enterprise's expense, not its income. Treating the agent's tax as the enterprise's discharge collapses two taxpayers into one and, with them, the cause into the consequence. The deeming fiction of the agency establishment exists precisely to keep them apart.

Does it read Morgan Stanley beyond what it decided?

Misapplying Morgan Stanley: a holding from the service-PE setting transplanted into the agency-PE setting.

Hon’ble Supreme Court’s judgment in Morgan Stanley [ 284 ITR 260 (SC)] concerned a service permanent establishment. The Supreme Court held that arm's length payment for the services exhausted the attribution, provided the transfer pricing analysis captured all the functions and risks, and it expressly preserved the contrary case where the analysis does not. The distinction matters because the entrepreneurial-risk point, muted in the service-PE setting, becomes critical in the agency setting. In a service PE, the captive unit genuinely performs the functions and may well assume the risks for which it is paid, so that an arm's length service fee can, intelligibly, capture most of what is attributable. An agency establishment is different in kind, for it carries an entrepreneurial risk that an agent, by his very nature, can never assume, a distinction I develop below. To transplant Morgan Stanley whole from the one setting into the other is to overlook the very feature on which attribution depends.

Whatever the conceptual misgivings, the doctrine is now applied as settled law, and the benches have followed it loyally once it was laid down. The detailed exposition is the Tribunal's decision in Asia Today Limited  [TS-620-ITAT-2021(Mum)], which marshals both competing approaches and the scholarship on each side, follows the Bombay High Court as binding, and yet, tellingly, preserves the conceptual reservation on the face of the order, recording, through the observations in Delmas France judgment [TS-5029-ITAT-2012(Mumbai)-O], that the tax-neutrality theory "may not indeed be wholly unqualified, at least on a conceptual note." A doctrine whose own loyal application feels obliged to carry a reservation of this kind is not one that has been laid to rest; it is one awaiting the higher reckoning that is now at hand. The line has continued without interruption to the present. As recently as May 2026, in FedEx Express International B.V. case, the existence of the dependent agency permanent establishment being conceded, deleted an attribution of some seventeen crore rupees on the footing that, the transactions with the Indian agent-entities having been found to be at arm's length, no further profit could be attributed to the establishment, resting on Morgan Stanley and E-Funds. The case is a clean illustration of the doctrine in routine operation, and of the very conflation under discussion: the reasons given, that the agent was adequately remunerated and had in fact returned more than the amount attributed, are answers about the agent's reward, not about the enterprise's surplus.

The cumulative effect is a doctrine that, so far as I am aware, no comparable jurisdiction embraces: the permanent establishment as a threshold whose crossing produces nothing. And it leaves unanswered the plainest question of all. If the establishment yields only the agent's remuneration, and the agent is taxable on that remuneration regardless of any establishment, then what work is the establishment doing? Why does the law define it, locate it, and litigate it, if its existence is a matter of complete tax indifference? A threshold that triggers nothing is not a threshold at all.

IX.  The Unfinished Task

Two pieces of work therefore remain. The first is to restore, in the attribution of profit to an agency establishment, the substantive enquiry that both the Authorised OECD Approach and India's own preferred mixed approach share, namely that the source state taxes what the enterprise genuinely earns through its presence, in place of a rule of automatic neutrality that taxes nothing. The two camps differ on the measure, function-asset-risk on the one hand, a supply-and-demand apportionment on the other, but they agree on the premise the neutrality doctrine denies: that there is something there to attribute. The surplus may be large or small on the facts; what cannot be right is to refuse to look for it.

The second task is conceptual and prospective, and it begins with a candid view of how the agency rule has lately been reformed. The deep substance that rule reaches for is the location of entrepreneurial risk: is the enterprise itself bearing the venture's risk and using the local presence as the instrument of its own participation, or has that risk genuinely passed to an independent operator dealing at arm's length? This is precisely the conceptual flaw that a coordinate bench identified in Delmas France. The agency establishment, it observed, inherently assumes an entrepreneurship risk in respect of which the agent can never be compensated, because, while the establishment inherently bears that risk, an agent by his very nature cannot assume it; and to that extent there is a subtle but real line of demarcation between the dependent agent and the dependent agency permanent establishment. That observation goes to the root of the matter. An agent is paid for performing functions, employing assets and bearing the limited risks of an agent; he is not, and cannot be, the bearer of the enterprise's entrepreneurial risk. The entrepreneurial risk, and the entrepreneurial reward that answers to it, remain with the enterprise. A doctrine that treats arm's length payment to the agent as exhausting the source state's claim therefore overlooks the one thing the agency establishment was meant to capture, the enterprise's own risk-bearing presence in the market, which no remuneration of the agent can ever reach. The historical drafting of the rule, keyed to the authority to conclude contracts in the name of the enterprise, was always a narrow proxy for that deeper substance, and it let economically identical arrangements escape on a formal distinction.

Article 12 of the Multilateral Instrument repaired part of this, extending the rule to a person who habitually plays the principal role leading to the routine conclusion of contracts, and tightening the independent-agent exception. That was a genuine advance, but it was a reactive one, and its reactive character is the heart of the difficulty. The BEPS project did not stand back and ask what the agency permanent establishment is fundamentally for, and where that purpose had been eroded. It reacted to the concern most prominent at the moment, the commissionnaire arrangement, thrown into relief by litigation such as the Zimmer case [(2010) 12 ITLR 739 (Conseil d'État, France)], and it fashioned a fix tailored to that concern. But the commissionnaire is not the disease. It is a symptom, one manifestation of a deeper malady, namely a rule that fixes on the form of contract-conclusion while remaining blind to the entrepreneurial risk that is the real substance. Treating the symptom leaves the malady intact. By widening the contract-conclusion test rather than reorienting the rule around entrepreneurial risk itself, the Multilateral Instrument perfected the proxy rather than reaching the principle, and in doing so it ensured that fresh manifestations would keep surfacing. So long as the root is left untouched, the law is condemned to a perpetual game of catch-up, recognising each new arrangement only after it has emerged and done its mischief. The risk-stripped distributor is merely the next symptom waiting in line behind the commissionnaire, but that is only when the next wave of PE reforms translates into a reality. As of now, those things are clearly outside the ambit of inherently limited scope of the coverage of dependent agency permanent establishment in the tax treaties.

Symptom versus disease: the commissionnaire was one symptom; the risk-blind threshold is the malady.

The real solution, therefore, cannot be found by patching symptoms one at a time. It must be sought by returning to first principles, by asking why the permanent establishment clause was conceived at all, what commercial world it was built to address, and how profoundly the models of doing business have changed since that world. The concept was forged for an age of warehouses, branches and travelling agents. It is now asked to govern an age in which an enterprise can saturate a market, and harvest its profit, while arranging its affairs so that nothing it formally does trips the inherited threshold. A rule designed for the first age cannot be made fit for the second by successive ad hoc amendments, each chasing the last clever structure. It can be made fit only by reconstructing the threshold on its own foundational purpose: the taxation, in the place where the economic activity occurs, of the enterprise that bears the risk and reaps the reward of that activity. This is the same disquiet that animates the digital-economy debates and the work beyond them. The dependent agency permanent establishment is the oldest instance of the problem, and it remains, at home, unresolved.