Tax treaties were created for a simple purpose: the same income should not be taxed twice—in the source country and in the recipient country. But a network of thousands of bilateral treaties with different rates has itself become a resource. Through a conveniently located intermediary company, one could reach a benefit to which the ultimate recipient had no right. This is treaty shopping—and it is precisely against this that the OECD built a separate line of defense in the BEPS project.
Where Treaty Shopping Came From
The idea of a treaty network goes back to the League of Nations model conventions of the 1920s and was consolidated in the OECD Model Tax Convention of 1963. Withholding tax rates on dividends, interest, and royalties differ across treaties: zero in some cases, five or fifteen percent in others. As soon as there are many rates, arbitrage appears—a payment can be routed through a jurisdiction with the broadest and most favorable network of agreements.
The OECD identified the problem long before BEPS: the 1986 report on conduit companies already described shell companies whose sole function was to pass income through themselves and reduce withholding tax at source. No systemic response followed at that time. It appeared in 2015 in Action 6 of the BEPS plan—"Preventing the Granting of Treaty Benefits in Inappropriate Circumstances." This measure became one of the four minimum standards, mandatory for all participants in the Inclusive Framework.
Concept
Treaty shopping is the use of the tax treaty network to circumvent their intended purpose. An investor from country A invests in country C through an intermediary company in country B, because B and C have a favorable treaty: reduced withholding tax on dividends, interest, or royalties. The company in B itself often exists only on paper.
The classic scheme is an intermediate holding in a jurisdiction with a broad treaty network and low withholding tax at source. For decades, dividend and interest flows passed through such conduits, losing minimal tax along the way.
What MLI Did
The response was the MLI: the convention was adopted on 24 November 2016 and the first signing ceremony took place on 7 June 2017; it entered into force on 1 July 2018. Its idea is elegantly engineered: with one document, countries introduced common anti-avoidance provisions into thousands of bilateral treaties at once, without renegotiating them one by one. The OECD's current list includes 107 signatory jurisdictions, and it modifies covered agreements automatically.
The mechanics of the MLI are structured as an overlay on treaties. The text of each agreement remains unchanged, and the convention applies on top of it: a country lists the treaties it wants to cover (Covered Tax Agreements) and records its positions and reservations on each article. A provision for a specific pair of states arises where their positions coincide—this is called matching. The OECD's current summary says the MLI covers around 1,950 bilateral treaties, with the first effects taking place from 1 January 2019.
Minimum Standard: PPT and LOB
BEPS Action 6 made protection against treaty shopping a minimum standard. A treaty must now contain a statement in the preamble that it is not intended to create opportunities for non-taxation, and one of the mechanisms: principal purpose test (PPT) or limitation on benefits (LOB). LOB works as an objective test—the benefit is available only to qualified persons based on ownership criteria, public status, or active business. PPT evaluates the purpose of the transaction. Most MLI countries chose PPT; US treaties traditionally rely on LOB.
PPT is the principal purpose test, a general anti-abuse rule. A treaty benefit is not granted if obtaining it was one of the principal purposes of the transaction or structure. The test has two elements. Subjective: the tax authority, weighing all the facts, reasonably concludes that the tax benefit was one of the main purposes. Objective: the taxpayer has the right to show that the benefit in these circumstances is consistent with the object and purpose of the relevant treaty provisions. Most countries chose PPT as the shortest path to close the minimum standard.
LOB—limitation on benefits—works through formal criteria. Access to the benefit is granted to qualified persons: companies with sufficient connection to the country of residence, public companies, structures with real active trade or business, sometimes through derivative benefits. The MLI offers a simplified version (simplified LOB), while detailed LOB is traditionally embedded in US treaties. The minimum standard allows three combinations: PPT alone, PPT together with simplified LOB, or detailed LOB plus a separate mechanism against conduit structures.
Beneficial Owner and the Danish Cases
In parallel, there is a long-standing treaty requirement: the benefit belongs to the beneficial owner of the income; a transit recipient has no right to it. In 2019, the EU Court in the so-called "Danish cases" denied conduit companies benefits on dividends and interest, directly calling artificial chains an abuse. This established the principle: a structure without the right to dispose of income does not receive treaty protection.
The Danish cases are six disputes that the Grand Chamber of the Court of Justice of the EU resolved on February 26, 2019, with two judgments. Four concerned interest under the Interest and Royalties Directive (cases C-115/16, C-118/16, C-119/16, and C-299/16), two concerned dividends under the Parent-Subsidiary Directive (C-116/16 and C-117/16). Danish companies paid interest and dividends to parent structures in Luxembourg, Cyprus, and Sweden, which almost immediately forwarded the income further—to ultimate owners outside the EU.
The Court made two conclusions that changed practice. First, the general principle of prohibition of abuse of EU law applies to directives directly—even when there is no special anti-avoidance provision in national law. Second, the beneficial owner is the one who economically controls the income and decides its fate; a transit link is denied this role. The Court listed the signs of a conduit directly: immediate redirection of payment, absence of own activity, back-to-back financing.
Which Structures Are Under Scrutiny
Classic transit structures came under pressure from PPT and the Danish cases. A holding company that receives dividends and immediately passes them up; a finance company issuing a loan for exactly the amount received back-to-back; a royalty conduit passing licensing payments through a jurisdiction with zero withholding. The signal is the coincidence of incoming and outgoing flows, absence of people and decisions on site, absence of any function other than transferring income.
A working structure is based on real substance. The intermediary company has a board of directors that actually makes decisions, has personnel, an office, its own risk, and independent functions—what the OECD describes through economic substance and CIGA. Jurisdictions like the Netherlands, Luxembourg, and Cyprus remain useful exactly to the extent that the company has this substance. Formal registration without people and decisions no longer retains treaty benefits.
Where This Is Heading
The trend has settled: treaty benefits are based on substance and business purpose, while formal signs of residence take a back seat. PPT has become the default norm in almost all updated treaties, and its subjective formulation leaves tax authorities broad discretion. On top of this, Pillar Two with a global minimum tax of 15% is layered: the gain from zeroing the withholding rate will be picked up by the top-up tax anyway, and the point of aggressive routing falls. Transparency works in parallel—CRS and automatic exchange remove the very secrecy on which conduits relied.
Conclusion for International Structures
For the owner of an international structure, the conclusion is practical. The route of dividends and interest is built around real business logic and substance, and the treaty rate becomes a consequence of this logic. Intermediate jurisdictions like the Netherlands, Luxembourg, and Cyprus remain workable provided that the company has real substance and independent functions.
Q/A
Our Cyprus holding passes dividends straight up — does the treaty rate survive?
No, that is the textbook conduit. In the Danish cases the EU Court listed its markers directly: immediate redirection of the payment, absence of own activity, back-to-back financing. The beneficial owner is the party that economically controls the income and decides its fate; a transit link has no claim to the treaty benefit, however tidy its residence certificate.
Tax was not our main reason — does the PPT still bite?
It does. The test asks whether obtaining the benefit was one of the principal purposes, not the only or the dominant one, so commercial motives sitting alongside a tax motive do not rescue the arrangement. The one way out is the objective limb: showing that granting the benefit in those circumstances is consistent with the object and purpose of the relevant treaty provisions.
We hired a local director and rented an office — is that enough substance?
Not on its own. What counts is not the address but a board that actually makes decisions, personnel, own risk and independent functions — what the OECD describes through economic substance and CIGA. Formal registration without people and decisions no longer holds a treaty benefit.
Was our treaty modified by the MLI at all?
Not necessarily. A provision arises only where both states listed the treaty as a Covered Tax Agreement and their positions on the article matched; if one side did not notify it or entered a reservation, the change never reaches it, so check the pair in the OECD matching database. But an untouched treaty is no refuge: domestic anti-avoidance rules and the EU prohibition of abuse of rights operate independently.
If the structure survives the PPT, is the saving still there?
Often not. Pillar Two's 15% global minimum tax sits on top of the treaty rate: the top-up tax picks up whatever zeroing a withholding rate produced, so aggressive routing buys risk without the return. What still works is a route built around real business logic, with the treaty rate as a consequence of that logic.