Twenty-six years later, academic researchers whose work the OECD has cited and built upon — Thomas Tørsløv, Ludvig Wier, and Gabriel Zucman — estimate that roughly 36 to 40 percent of all multinational corporate profits are still being booked in low-tax jurisdictions. In absolute dollars, the figure has grown. The proportion has not meaningfully declined.
This gap is not what usually gets explained. The explanation on offer — in policy documents, in op-eds, in politicians’ speeches — is that tax havens are the work of rogue jurisdictions exploiting international law, that reform is proceeding but is technically complex, and that with enough institutional coordination the problem will eventually yield. This explanation is flattering to the institutions offering it and incorrect in one important respect: it describes the system as a failure. What the evidence actually describes is a system that is working.
The word worth examining here is not “legal” or “routine.” It is “legitimate.” The international tax haven system operates entirely within laws written by the states that most loudly condemn it. It serves the coherent development interests of some of the world’s smaller economies. It functions as the ordinary infrastructure of international business — not as a criminal underground but as the plumbing behind every major multinational’s annual report. If the institutions exist, the technical knowledge exists, the will is publicly declared to exist, and the thing is not illegal — what, precisely, is missing?
What a Tax Haven Actually Is
The popular image of a tax haven is a Caribbean island with a brass-plate office and a numbered account. This image is not wrong exactly — it describes part of the system — but it misdirects attention away from where the system’s center of gravity actually lies.
In 2017, Javier Garcia-Bernardo, Jan Fichtner, Frank Takes, and Eelke Heemskerk published a study in Scientific Reports that mapped global corporate ownership structures to identify which jurisdictions function as conduits and which function as sinks. The distinction matters. Sink offshore financial centers — the Cayman Islands, Jersey, Bermuda, the British Virgin Islands — are endpoints: capital arrives there and stays, warehoused in holding structures with minimal ongoing activity. Conduit offshore financial centers are pass-throughs: Ireland, the Netherlands, Luxembourg, Singapore. Capital moves through them, routed toward sinks, and they derive their role from a combination of low-to-zero withholding taxes, extensive bilateral tax treaty networks, and legal structures purpose-built for intra-group transactions.
The finding that gets least attention follows from this taxonomy directly: OECD member states are at the center of the system, not at its margins. Ireland, the Netherlands, and Luxembourg are not outliers tolerating harmful competition against the wishes of the international community. They are the international community — its members, its partners, signatories to its frameworks, participants in its reform processes.
The mechanics are not exotic. A multinational corporation incorporates an Irish subsidiary and licenses its intellectual property to it. That Irish subsidiary then charges royalties to operating subsidiaries in Germany, France, and the United States — jurisdictions where the actual customers, employees, and factories are located. The royalty payments move taxable profit from where tax is high to where tax is low. The IP remains nominally in Dublin or in a structure that Dublin facilitates, while the income it generates accumulates at rates far below what the operating country would have charged. Ireland’s headline corporate tax rate of 12.5 percent is, in many structures, not the effective rate that applies.
Treaty shopping adds another layer. The Netherlands has signed bilateral tax treaties with close to a hundred countries capping withholding taxes on dividends, interest, and royalties flowing between them. A company routing a payment through a Dutch holding entity can reduce the withholding tax charged at source from 30 percent to 5 percent or zero. The treaty network is, in this sense, a product that jurisdictions compete to offer — and the Netherlands has built one of the world’s most comprehensive.
The output is measurable and anomalous. Tørsløv, Wier, and Zucman found that foreign affiliates operating in tax havens are an order of magnitude more profitable than domestic firms operating in the same jurisdictions. No genuine economic explanation accounts for this. Companies incorporated in Ireland are not ten times more productive than Irish firms. The profitability premium is an artifact of profit shifting — income relocated, on paper, to where it is taxed least.
By the researchers’ own calculations, Ireland is the world’s largest corporate tax haven by profit-shifting volume. Not the Cayman Islands. Ireland — a eurozone member, an EU member since 1973, a country whose political leaders have spent twenty-six years attending reform summits and signing framework agreements.
Approximately two-thirds of Fortune 500 companies are incorporated in Delaware, drawn by a legal environment that offers flexible corporate law, near-total privacy around beneficial ownership, and no corporate tax on income earned outside the state. Delaware is not a secret. It appears on the incorporation papers of some of the world’s most scrutinized companies. It functions as a domestic tax haven in plain sight, run by a US state that participates in every international discussion about the need to eliminate tax havens.
Ireland’s transformation from one of Europe’s poorest economies in the 1980s to one of its wealthiest is inseparable from its tax strategy. That strategy was deliberate, sovereign, and by every applicable legal standard, correct. The point is not moral innocence. It is legal embeddedness, institutional normality, and a development logic that works well for the jurisdiction deploying it, at cost to those that don’t.
The conduit/sink OFC methodology
The 2017 Garcia-Bernardo et al. study in Scientific Reports reconstructed global corporate ownership networks using data from the ORBIS database, which tracks ownership relationships among millions of firms worldwide. By mapping the volume and direction of capital flowing through each jurisdiction, the researchers distinguished between jurisdictions where capital accumulates (sinks) and those it merely passes through (conduits). Cayman Islands, Jersey, and Bermuda absorb capital — it arrives and stays. Ireland, the Netherlands, Luxembourg, and Singapore route capital onward. A payment flowing from a German operating subsidiary to a Cayman Islands holding company might pass through a Dutch or Irish intermediate entity, with each layer reducing withholding tax at the relevant stage. The finding that conduit jurisdictions are almost exclusively OECD members reframes the standard account of tax haven geography.Who Actually Uses Them
That profitability premium belongs to someone. It does not belong to oligarchs routing stolen state assets through shell companies — that is a different, smaller story. It belongs, mostly, to the ordinary corporate infrastructure of international business: the annual reports, the group structures, the transfer pricing policies of the multinationals that appear in every Fortune 500 list and FTSE index.
Intra-group transactions — transfers of goods, services, intellectual property, and capital between subsidiaries of the same parent company — account for the majority of global trade by widely cited measures, a fact the OECD transfer pricing framework exists specifically to address. The rules governing how those transactions are priced — transfer pricing — determine where taxable income lands. The arm’s-length standard, the central principle of international transfer pricing, holds that intra-group transactions should be priced as if they were between unrelated parties at market rates. In practice, for intellectual property — where no genuine market comparison exists — this principle gives enormous latitude for creative valuation.
Google structured its non-US earnings through what became known as the Double Irish with Dutch Sandwich: two Irish-incorporated entities, one managed from Bermuda for tax purposes (making it Irish for US tax law but Bermudian for Irish tax law), and a Dutch holding company between them. Royalties flowed from the first Irish entity through the Dutch entity to the Bermudian-managed second, reducing taxable income at each stage. Reuters reported in early 2019 that Google had shifted approximately $23 billion in profits to Bermuda through this structure in 2017 — in a single year. The rate of tax paid on that income in Bermuda was zero, because Bermuda levies no corporate income tax.
Apple’s achievement was more concentrated. The US Senate Permanent Subcommittee on Investigations reported in May 2013 that Apple had maintained an effective tax rate of approximately 0.06 percent on its non-US profits between 2009 and 2011. This was not a rounding error. It was the output of structures involving Irish subsidiaries that Ireland did not consider tax resident because they were managed from elsewhere, and that the United States did not consider tax resident because they were incorporated in Ireland. They were, for practical purposes, tax resident nowhere. James Hines, writing in the Journal of Economic Perspectives in 2010, documented that tax haven affiliates of US multinationals accounted for more than 20 percent of total US foreign direct investment and nearly a third of the foreign profits of US firms — while representing a small fraction of those firms’ foreign employment and capital assets. The concentration of paper profits bears no relationship to the concentration of actual economic activity.
Pension funds — the retirement savings vehicles of ordinary workers in the United States, the United Kingdom, and the Netherlands — routinely use Cayman Islands special purpose vehicles for investment in private equity and infrastructure. The Cayman structure provides pass-through tax treatment that avoids double taxation on returns. Sovereign wealth funds do the same. This is legal, widespread, and essentially invisible in public debate about offshore finance, which tends to focus on the corporation or the billionaire and ignore the pension beneficiary at the other end of the same chain.
The Double Irish was closed to new entrants from 2015 and fully phased out by 2020. This is cited, routinely, as evidence that reform works. What is cited less often is that successor arrangements — structures using Ireland’s participation exemption, its knowledge development box, and its treaty network — have in large measure replaced it. The mechanism changed; the function did not.
The Double Irish with Dutch Sandwich
The structure operated through three entities. Company A was incorporated in Ireland but managed from Bermuda; under Irish law at the time, tax residence followed management, making Company A Bermuda-resident for Irish purposes. Under US tax law, however, it was Irish-incorporated and therefore foreign. Company A held the intellectual property. Company B was a Dutch holding company. Company C was the second Irish entity, genuinely Irish-resident, which held the licence to exploit the IP in markets outside the United States. Royalties collected from operating subsidiaries worldwide flowed to Company C, which paid royalties to Company B (reducing Irish taxable income), which paid royalties to Company A (now in Bermuda, taxed at zero). The Dutch entity reduced the withholding tax that would otherwise apply to the royalty flow. By the time income reached Bermuda, it had been stripped of taxable character at every stage. Ireland closed this structure to new entrants in 2015 under pressure from the European Commission and the OECD. The knowledge development box, introduced simultaneously, offered preferential rates on IP income for compliant structures — a narrower version of the same logic.The Reform Theatre — BEPS, Pillar Two, and What They Don’t Do
The Base Erosion and Profit Shifting project, launched by the OECD and G20 in 2013 and culminating in a fifteen-point action plan in October 2015, represents a genuinely ambitious piece of international institutional work. By 2024, the Inclusive Framework had grown to more than 145 member jurisdictions. By the OECD’s own accounting, more than 130 harmful preferential regimes have been amended or abolished since the process began. The Country-by-Country Reporting requirements introduced under BEPS Action 13 have generated more tax authority data on multinational profit allocation than at any previous point in history. These achievements are real.
What BEPS did not do is close the gap. The action items are not legally binding. They are not self-executing. They require domestic implementation by national legislatures that are subject to the lobbying of the multinationals most affected by reform. More fundamentally, BEPS does not address the underlying incentive structure. Countries remain entirely free to set their statutory corporate tax rates at whatever level they choose. The framework addresses the most egregious abuses — structures that were nominally compliant but openly artificial — without touching the core mechanism of rate competition.
The arm’s-length transfer pricing standard, which BEPS preserved and in some respects reinforced, rests on a legal fiction: that a multinational’s payment from its Irish subsidiary to its Bermudian parent for use of intellectual property is comparable to a market transaction between unrelated parties. For most IP, no comparable market transaction exists. The fiction is necessary because the alternative — formulary apportionment, splitting multinational profits according to where sales, employment, and assets actually are — would require a fundamental restructuring of international tax law that major economies have consistently declined to pursue.
Pillar Two, the second of the two-pillar solution agreed in 2021 and effective January 1, 2024 for qualifying jurisdictions, introduced a global minimum corporate tax rate of 15 percent applying to multinationals with revenues above €750 million. The ambition was to eliminate the rate competition at the bottom of the market — if every jurisdiction imposes a top-up tax bringing the effective rate to 15 percent, there is no remaining incentive to route profits to zero-rate jurisdictions.
The substance-based income exclusion substantially qualifies this ambition. Multinationals are permitted to deduct from income subject to the top-up tax an amount equal to 5 percent of the value of their tangible assets plus 5 percent of their payroll costs in each jurisdiction. For a capital-intensive manufacturer with real physical assets and employees in a low-tax jurisdiction, this exclusion materially reduces the top-up tax liability. For a technology or pharmaceutical company whose primary asset is intellectual property, held in a jurisdiction with minimal physical presence, the exclusion does comparatively little — but those are precisely the companies doing most of the profit shifting. The effective floor for many large multinationals sits well below 15 percent.
The United States compounds the problem. The One Big Beautiful Bill Act, signed in July 2025, renamed the Global Intangible Low-Taxed Income regime the Net CFC Tested Income regime and set its effective rate at 12.6 percent — below the 15 percent Pillar Two floor. The US has not aligned its regime with the international standard. The result is a persistent gap: US multinationals do not face the same effective minimum as their European competitors. This is not a technical accident awaiting correction. It reflects the material interests of US multinationals that benefit from the current architecture — interests that are represented, with considerable effectiveness, in the US Congress.
The Pillar Two substance-based income exclusion
The carve-out is calculated per jurisdiction. A multinational with €100 million of tangible assets and €40 million of payroll in a 9 percent corporate tax jurisdiction can deduct €5 million (5% × €100M) plus €2 million (5% × €40M) from the income base before calculating any top-up tax owed to bring the effective rate to 15 percent. The deduction applies regardless of whether the assets and payroll represent genuine economic substance or were located there specifically to exploit the exclusion. For IP-intensive companies — technology, pharmaceuticals, consumer brands — whose primary profit-generating asset is intangible and whose physical footprint in low-tax jurisdictions is minimal, the carve-out offers limited relief. These are, by the measure of profit-shifting research, the companies most responsible for the gap between statutory and effective tax rates globally. The exclusion is most generous to precisely the companies that most need it least.The Political Economy of Permanent Failure
The scale of what has not changed is worth stating plainly. Tørsløv, Wier, and Zucman estimate that corporate tax revenue lost to profit-shifting has grown from less than 0.1 percent of corporate tax revenues in the 1970s to approximately 10 percent today. The global average statutory corporate tax rate fell from 46.66 percent in 1980 to 26.04 percent in 2025, on a GDP-weighted basis, according to Tax Foundation data. That decline is not a coordination failure — it is the outcome of competition. Every country that lowered its rate did so rationally, from its own perspective, to attract or retain investment. The aggregate result is a race no individual actor had reason to stop.
The Apple/Ireland case illuminates the political economy with unusual clarity. In 2016, the European Commission ruled that Ireland had granted Apple illegal state aid through its tax arrangements — specifically, that the Irish Revenue Commissioners had issued opinions on Apple’s tax structures that amounted to selective treatment not available to other companies. The Commission ordered Ireland to recover approximately €13 billion in unpaid taxes.
Ireland appealed alongside Apple. The General Court overturned the Commission’s ruling in 2020. The Court of Justice of the European Union reinstated it on September 10, 2024. The funds — approximately €14.25 billion, including accumulated interest — were fully paid to the Irish Exchequer by May 2025.
Consider what happened here. The Irish government spent the better part of a decade fighting in court against receiving €13 billion it was legally owed. It deployed considerable public resources and political capital to avoid collecting money from a company that had paid almost none on profits earned using Irish arrangements. This is not corruption. It is not irrational. It is the rational expression of a sovereign-development model that produced genuine results: Ireland’s GDP per capita has grown from among the lowest in Western Europe in the 1980s to among the highest. The country’s economic model depends on attracting foreign direct investment through a combination of EU access, educated English-speaking workforce, and a tax regime that makes Ireland a competitive location for multinational headquarters and holding structures. Collecting €13 billion from Apple, and signaling to every other multinational with an Irish structure that the regime is subject to retroactive revision, was a direct threat to the foundation of that model. The Irish government’s calculation was not difficult: lose €13 billion in a one-time payment, or risk losing the foreign investment base that generates tens of billions annually. They chose to fight.
The United States contains within its own territory the world’s most heavily used domestic tax haven and operates a tax system creating powerful incentives for offshore profit booking. The United Kingdom hosts Jersey, Guernsey, and the Cayman Islands as Crown Dependencies and Overseas Territories. The City of London’s financial services sector depends substantially on capital flows routed through those jurisdictions. France, Germany, and the Netherlands preach reform in OECD working groups while their own multinationals use Dutch and Luxembourg structures extensively. The Netherlands is not embarrassed by its treaty network. Luxembourg’s economy would not exist in its current form without the investment fund structures it offers.
These are not hypocritical positions, exactly. They are the expression of states that have two simultaneous interests: appearing to support reform, and not actually achieving it at a pace that would damage their own actors. The OECD process serves both interests admirably. It generates reports, frameworks, action plans, and inclusive framework membership lists. It creates a narrative of progress. It achieves real but limited results. And it leaves the structural incentives — rate competition, arm’s-length pricing, treaty shopping, intellectual property licensing — essentially intact.
The Apple/Ireland EU state aid case The European Commission's original 2016 decision found that Ireland had issued tax opinions in 1991 and 2007 allowing Apple to allocate the vast majority of its non-US profits to a stateless head office — an entity with no employees and no physical presence — rather than to its Irish branches. The result was that profits from Apple's European, Middle Eastern, African, and Indian operations were effectively untaxed by any jurisdiction. The Commission concluded this constituted selective state aid. Ireland appealed, arguing the Commission had misapplied EU state aid law and misunderstood Irish tax law. The General Court agreed in 2020. The CJEU reversed that judgment on September 10, 2024, restoring the original Commission decision. Payment of approximately €14.25 billion, held in an escrow account during litigation, was transferred to the Irish Exchequer by May 2025. Ireland had spent nine years and substantial legal resources fighting to avoid receiving this sum — a transaction whose perversity is entirely explicable once the underlying sovereign-development logic is understood.
Why It Cannot Change
Genuine reform of the international tax system would require a specific set of conditions. The world’s largest economies — the United States, the United Kingdom, the major EU member states — would need to act simultaneously against the material interests of their most powerful corporations and wealthiest citizens. They would need to accept a period of competitive disadvantage while the transition played out, with no assurance that other jurisdictions would follow. They would need to override the lobbying capacity of entities that can relocate capital, intellectual property, and nominal headquarters across borders in ways that ordinary citizens and small businesses cannot. And they would need to do this with no first-mover advantage — because the country that moves first simply exports its multinationals to the country that hasn’t.
This is not a coordination problem that better institutions could solve. It is a structural feature of a system where the parties controlling the rules are the parties benefiting from them. Ronen Palan, Richard Murphy, and Christian Chavagneux documented in their 2010 Cornell University Press study Tax Havens: How Globalization Really Works that the offshore system was not an accidental byproduct of globalization but an integral part of its design — constructed deliberately, over decades, by precisely the states and financial institutions that nominally oppose it.
Tax avoidance — holding companies, treaty shopping, IP licensing, transfer pricing — is legal. Tax evasion — concealing income, falsifying records, maintaining undisclosed accounts — is not. The distinction is real and should be maintained. The Panama Papers and Pandora Papers documented genuine criminal activity alongside legal avoidance, and the criminal activity deserves criminal prosecution. But the distinction cannot do the ideological work it is usually asked to do: it cannot explain why avoidance at the scale documented by Tørsløv, Wier, and Zucman is somehow separate from the system it depends on. The structures are legal because the states that write tax law wrote them to be legal. The avoidance/evasion line is drawn where it is drawn not by neutral principle but by the same political economy that produced the system.
Corporations and high-net-worth individuals who benefit most from offshore arrangements have the greatest capacity to lobby against reform. They fund research supporting tax competition. They can, if rules change, relocate IP, headquarters, and ultimately capital in ways that impose real costs on the jurisdictions attempting reform. This asymmetry — between those who benefit and those who bear the cost, between those with exit options and those without — is the mechanism that reproduces the system across political cycles and reform initiatives, regardless of which party or ideology holds office.
The decline in statutory corporate tax rates from 46.66 percent to 26.04 percent over forty-five years is not a policy failure. It is the observed output of states rationally competing for a mobile tax base. The Pillar Two minimum is an attempt to arrest this dynamic at 15 percent rather than wherever competition alone would take it — but at 15 percent, with the substance-based exclusion reducing the effective floor further, and the United States having locked in 12.6 percent through the OBBBA, the race is paused rather than ended.
Closing
The OECD published its criteria for harmful tax competition in 1998. In 2000, it named 82 targets. In 2013, it launched BEPS. In 2021, it agreed a global minimum tax. In 2024, that minimum came into force. The institutional architecture is the most sophisticated in the history of international tax coordination, and it has been built by the same states whose multinationals account for the largest share of profit shifted to low-tax jurisdictions.
Genuine reform would require the United States, the United Kingdom, the European Union’s largest economies, and their most significant trading partners to act in concert against the stated preferences of the most powerful actors in each of their economies, with no first-mover advantage and significant near-term cost to whoever moves first. It would require abolishing the arm’s-length transfer pricing standard and replacing it with formulary apportionment — a measure that would fundamentally redistribute taxable income and that every major economy has declined to pursue. It would require Ireland to abandon the development strategy that turned it from a country of net emigration into a net destination. It would require the United Kingdom to fundamentally alter the relationship between the City and the Crown Dependencies. It would require the United States Congress to override the preferences of the multinationals that fund it.
This is not impossible. It is impossible in the political economy that actually exists — the one that created the system in the first place.
The system is not failing to function. Twenty-six years after the OECD listed 82 targets for elimination, after fifteen action points and a global minimum and 145 Inclusive Framework members, roughly 40 percent of global multinational profits still flow through low-tax jurisdictions because the states that condemn the system built it to carry that load, and they have never had a genuine reason to change the design.
Gen AI Disclaimer
Some contents of this page were generated and/or edited with the help of a Generative AI.
Media
Key Sources and References
Garcia-Bernardo, J., Fichtner, J., Takes, F.W., and Heemskerk, E.M. “Uncovering Offshore Financial Centers: Conduits and Sinks in the Global Corporate Ownership Network.” Scientific Reports 7, no. 1 (2017): 6246.
Hines, James R. “Treasure Islands.” Journal of Economic Perspectives 24, no. 4 (2010): 103–126.
Irish Times. “Apple’s €14bn Escrow Account Wound Down.” July 15, 2025.
OECD. Harmful Tax Competition: An Emerging Global Issue. Paris: OECD Publishing, 1998.
OECD. Towards Global Tax Co-operation: Progress in Identifying and Eliminating Harmful Tax Practices. Paris: OECD Publishing, 2000.
OECD. Action Plan on Base Erosion and Profit Shifting. Paris: OECD Publishing, 2013.
OECD. Base Erosion and Profit Shifting: 2015 Final Reports. Paris: OECD Publishing, 2015.
One Big Beautiful Bill Act, Pub. L. No. 119-21, 139 Stat. (2025). Signed July 4, 2025.
Palan, Ronen, Richard Murphy, and Christian Chavagneux. Tax Havens: How Globalization Really Works. Ithaca: Cornell University Press, 2010.
Reuters. “Google Shifted $23 Billion to Tax Haven Bermuda in 2017, Filing Shows.” January 3, 2019.
Tax Foundation. Corporate Tax Rates around the World, 2025. Washington, DC: Tax Foundation, 2025.
Tørsløv, Thomas R., Ludvig S. Wier, and Gabriel Zucman. “The Missing Profits of Nations.” NBER Working Paper No. 24701, National Bureau of Economic Research, 2018. Revised and published in Review of Economic Studies 90, no. 3 (2023): 1499–1534.
US Senate Permanent Subcommittee on Investigations. Offshore Profit Shifting and the U.S. Tax Code — Part 2 (Apple Inc.). Hearing, May 21, 2013. Washington, DC: US Government Printing Office, 2013.
Lena Martin
Doing economics. Occasionally mathematics. Avoiding algebraic topology on purpose.
English 
Español 
Français 
Deutsch 
日本語 
Polski