Identifying Illegal Subsidies

Identifying Illegal Subsidies

69 Am. U. L. Rev. 479 (2019).

* Edwin S. Cohen Distinguished Professor of Law and Taxation, University of Virginia School of Law. This Article was written in part while the author was professor in residence at the International Bureau of Fiscal Documentation, Amsterdam. The author thanks Reuven Avi-Yonah, John Brooks, Quinn Curtis, John Duffy, Lilian Faulhaber, Josh Fischman, Jens Frankenreiter, Brian Galle, Sriram Govind, Werner Haslehner, Andrew Hayashi, Georg Kofler, Sarah Lawsky, David Lenter, Mitchell Kane, Rebecca Kysar, Michael Livermore, Helena Malikova, Tarcísio Magalhães, Omri Marian, Tom Nachbar, Lee Osofsky, Fadi Shaheen, Dan Shaviro, Paul Stephan, Bret Wells, and workshop participants at Boston College, BYU, Fordham, the IBFD, Luxembourg University, the Max Planck Institute for Tax Law and Public Finance, Northwestern University, Georgetown University, University of Houston, University of Pennsylvania, University of Virginia, and Vienna University of Economics and Business Administration. I owe special thanks to Michael Knoll and Wolfgang Schön for reading multiple drafts. Librarians at the UVA Law Library, especially Xinh Luu, provided excellent research support, as did my research assistants, Justin Aimonetti, Emily Ostertag, and Griffin Peeples. This Article benefitted from discussions with tax practitioners and Commission officials; it represents only my own views. © 2019, Ruth Mason, all rights reserved.

This Article argues that current methods for identifying illegal tax subsidies trigger the well-known conceptual difficulties of tax-expenditure analysis. To avoid these problems—particularly the irresolvable conflict over the correct baseline for measuring tax expenditures and tax subsidies—this Article advocates the “internal consistency test” as a superior method for identifying illegal subsidies. Developed by the U.S. Supreme Court to evaluate the compatibility of state taxes with the dormant Commerce Clause, the internal consistency test easily can be adapted to the subsidy context.

Introduction

In 2016, the European Commission dropped a bombshell—it ordered the U.S. technology giant Apple to repay Ireland subsidies of $14.5 billion, plus interest. The Commission’s decision enforced the prohibition on state aid found in the Treaty on the Functioning of the European Union (TFEU). The state-aid rules prevent European Union (EU) Member States from distorting market competition by granting subsidies that are protectionist or that operate as import or export subsidies.1Commission Notice on the Notion of State Aid, art. 5, 2016 O.J. (C 262) 1, 27–40 [hereinafter 2016 Notice] (describing state aid); Consolidated Version of the Treaty on the Functioning of the European Union of Dec. 13, 2007, art. 107(1), 2012 O.J. (C 326) 47, 91 [hereinafter TFEU] (prohibiting state aid). The Commission found that Ireland illegally subsidized Apple by allowing it to pay too little Irish tax.2Commission Decision 2017/1283 of Aug. 30, 2016, On State Aid Implemented by Ireland to Apple, 2017 O.J. (L 187) 1, 109 (EU) [hereinafter Apple].

This Article uses controversy over Apple3Id. and other recent EU state-aid cases to explore a defect common to many anti-subsidy regimes that limit states’ ability to use subsidies to interfere with private competition, including World Trade Organization (WTO) rules, soft-law agreements, and even the U.S. Constitution. Each regime applies not only to cash subsidies and regulation but also to taxation.4See, e.g., Dan T. Coenen & Walter Hellerstein, Suspect Linkage: The Interplay of State Taxing and Spending Measures in the Application of Constitutional Antidiscrimination Rules, 95 Mich. L. Rev. 2167, 2169 (1997) (discussing dormant Commerce Clause); Lilian V. Faulhaber, The Trouble with Tax Competition: From Practice to Theory, 71 Tax L. Rev.311, 330 n.99 (2018) (discussing international soft law); Ruth Mason, Common Markets, Common Tax Problems, 8 Fla. Tax Rev. 599, 624–28 (2007) (comparing treatment of discriminatory taxes and subsidies under U.S. and EU law); John O. McGinnis & Mark L. Movsesian, The World Trade Constitution, 114 Harv. L. Rev. 511, 546 n.203 (2000) (discussing WTO). Anti-subsidy regimes typically rely on tax-expenditure analysis to identify subsidies delivered through the tax law or tax administration.5Stanley Surrey introduced the tax-expenditure concept to the United States. Stanley S. Surrey, Pathways to Tax Reform: The Concept of Tax Expenditures (1973); Stanley S. Surrey & Paul R. McDaniel, Tax Expenditures (1985). Under this approach, a state confers a tax subsidy when it deviates from its own generally applicable domestic law or procedure to reduce taxes for particular enterprises, such as exporters or multinationals. Special tax reductions could take the form of reduced tax rates, tax deductions, tax credits, or the like. This tax-expenditure approach to identifying subsidies works well when both the domestic law baseline and the “special” or deviating provisions are readily identifiable.

But this Article argues that the approach becomes intractable when the subsidy reviewer and the accused state disagree over how to define the baseline from which tax expenditures (and therefore illegal subsidies) can be measured. This baseline problem is familiar to the tax-expenditure debate, and despite the enormous importance of the tax-expenditure concept to tax policy analysis, fifty years of study has brought little progress in finding a neutral tax baseline against which tax expenditures can be judged.6See, e.g., Staff of J. Comm. on Taxation, 110th Cong., A Reconsideration of Tax Expenditure Analysis (Comm. Print 2008) [hereinafter 2008 JCT Report] (reviewing “seemingly endless debates about what should and should not be included in the ‘normal’ tax base” and reviewing various academic attempts to resolve the issue).

As just one example, tax-expenditure analysis fails in cases involving “structural” tax rules.7See infra Section II.A. Structural rules are the background rules of the tax system, including the taxable unit, the accounting period, progressive tax rates, and the like. Under a tax-expenditure approach that relies exclusively on domestic law to formulate the baseline, the subsidy adjudicator would incorporate all structural rules into the baseline with the result that no structural rules would be regarded as tax expenditures, and, therefore, no structural rules would be regarded as conferring illegal subsidies. But an approach that regards all structural rules as permissible, no matter their actual effects on cross-border commerce, is underinclusive.8Structural provisions can convey tax benefits. See, e.g., 2008 JCT Report, supra note 6, at 10 (giving deferral of tax on foreign income and the debt-equity distinction as examples of structural rules that “materially affect economic decisions”). For more examples, see infra Section IV. At the same time, however, it is unclear how a subsidy adjudicator ought to handle structural rules under tax-expenditure analysis if such rules will not be automatically incorporated into the reference base.

This Article argues that the Commission’s need to evaluate tax provisions that were not easily cognizable under traditional tax-expenditure analysis—including structural rules—led it to adopt a new approach to identifying illegal subsidies in recent cases. Instead of evaluating Member State tax rules against a baseline consisting of the challenged state’s own generally applicable tax law, the Commission began to evaluate Member State tax rules against external norms.9See, e.g., Apple, supra note 2, ¶ 441. In some cases, the Commission used an internationally accepted norm; in others, the Commission judged Member State taxes against its own view of good tax policy.

Failure to adequately explain its departure from the reference-law approach left the Commission vulnerable to criticism that it exceeded its institutional authority. Member States officials—especially those from the small states targeted for state-aid review—argued that judging national tax rules by reference to external benchmarks threatened tax diversity in the European Union and invaded the reserved tax powers of the Member States.10See, e.g., Advisory Bd., Fed. Ministry of Fin., Tax Benefits and EU State Aid Control: The Problem of and Approaches to Resolving the Conflict of Jurisdiction with Fiscal Autonomy 5, 10 (2017) (Ger.) (report, prepared by academics for the German finance ministry, complaining of incursion by the Commission into reserved state tax powers); Kelyn Bacon, European Union Law of State Aid 5 (2017) (“State aid is being used partly as a tool to incentivise tax harmonisation.”). Under long-standing interpretations of the TFEU, states are free to have tax laws that differ from each other, as long as they avoid tax rules that function equivalently to tariffs and import/export subsidies.11See infra Section I.A. The TFEU, therefore, does not completely eliminate tax competition, nor does it prevent so-called tax mismatches, cross-border tax advantages (and disadvantages) that arise from differences between states’ rules.12See infra Section I.A.

Perceptions that the Commission invaded the reserved powers of the states provoked sharp criticism in recent cases. For example, in the Amazon13Commission Decision 2017/6740 of Oct. 4, 2017 on State Aid Implemented by Luxembourg to Amazon, 2018 O.J. (L 153) 1 [hereinafter Amazon]. state-aid case, Luxembourg accused the Commission of engaging in “covert fiscal harmonisation . . . thereby infringing the exclusive competence of the Member States in . . . taxation.”14See, e.g., Case T-816/17, Luxembourg v. Comm’n, 2018 O.J. (C 72) 38, ¶ 14. Similarly, the Irish government appealedApple to defend against “the encroachment of EU state aid rules into the sovereign Member State competence of taxation.”15Irish Department of Finance, Explanatory Memorandum for the Members of the Oireachtas Dáil Debate of Government Motion on the Apple State Aid Case, 13 (Sept. 7 2016). Officials from the United Kingdom have been particularly sharp in their criticism. Campbell Bannerman, a U.K. member of European Parliament, called Apple an “EU power grab” and a “tax trespass,” while his fellow member Steven Woolfe called it “an attack on the tax sovereignty of EU nation states by the back door.”16J.P. Finet, Apple Ruling Is EU Power Grab, U.K. and Irish Lawmakers Say, 83 Tax Notes Int’l 1019, 1019–20 (2016) (quoting pre-Brexit-vote EU parliamentary debates). Some commentators even connected Brexit directly to the Commission’s Apple decision.17Max Bearak, How the E.U.’s Ruling on Apple Explains Why Brexit Happened, Wash. Post (Aug. 30, 2016), https://www.washingtonpost.com/news/worldviews/wp/2016/08/30/how-the-e-u-s-ruling-on-apple-explains-why-brexit-happened. Because many of the recent cases involved U.S. multinationals, U.S. lawmakers also weighed in to complain about the Commission’s arrogation of power.18The Senate Finance Committee complained that the Commission was “using State aid . . . to achieve a ‘reform agenda.’” Letter from U.S. S. Comm. on Fin. to Jacob Lew, U.S. Sec’y of the Treasury, U.S. Dep’t of the Treasury 1 (May 23, 2016), https://www.finance.senate.gov/imo/media/doc/Hatch,%20Wyden,%20Portman,%20Schumer%20Continue%20Push%20for%20Fairness%20in%20EU%20State%20Aid%20Investigations.pdf [https://perma.cc/E46Z-QRGW]; Alexander Lewis, Apple’s Income Should Be Taxed in U.S., Not Ireland, Lew Says, 83 Tax Notes Int’l 862, 862 (2016) (quoting Department of the Treasury Secretary Jack Lew as asserting that the Commission was “overrid[ing] national tax law authority”); U.S. Dep’t of the Treasury, The European Commission’s Recent State Aid Investigations of Transfer Pricing Rulings, at 1 (2016) (arguing that the Commission applied EU law too broadly in the state-aid cases).

Academic critics argue that the Commission is using state aid to mandate “single taxation,” an aspirational tax policy goal that ensures taxation of one hundred percent of a multinational’s global income.19See Reuven S. Avi-Yonah, Advanced Introduction to International Tax Law 59–65 (2015) (giving history of the single-tax principle); Daniël S. Smit, Recovery of Fiscal State Aid: Always Somewhere?, Kluwer Int’l Tax Blog (Oct. 3, 2016), http://kluwertaxblog.com/2016/10/03/recovery-of-fiscal-state-aid-always-somewhere [https://perma.cc/3GN5-Y9DT] (characterizing the Commission’s Apple decision as holding Ireland “responsible to ensure single taxation”). Typically, critics of the recent cases make both the harmonization and the single-tax claim; these claims are closely related. See supra note 10 and accompanying text. Not all commentators who regard the Commission as enforcing single taxation oppose this interpretation. See, e.g., Pierpaolo Rossi-Maccanico, Fiscal State Aids, Tax Base Erosion and Profit Shifting, 24 EC Tax Rev. 63 (2015). Controversial comments by Commissioner of Competition Vestager on the recent U.S. cases lent credence to fears that the Commission interpreted state aid to mandate single taxation. Margrethe Vestager, Comm’r, European Comm’n, Speech to EU Parliament TAXE 2 Committee 2 (Apr. 4, 2016), http://www.europarl.europa.eu/cmsdata/99649/Vestager_Speaking%20Points%20TAXE%204%20April%20NEW.pdf [https://perma.cc/2DQC-5U9M] [hereinafter Commissioner Vestager Speech to TAXE] (lamenting the “loophole” that opens when a state’s allocation rules result in untaxed profits). Single taxation may seem like a worthy—even uncontroversial—goal, and countries increasingly coordinate their laws to promote single taxation.20See infra Section V.B. But the idea that all of a multinational’s income must be subject to taxation is fundamentally inconsistent with the notion that each state’s definition of income may differ from that of its neighbors. Thus, single taxation and state tax autonomy are incompatible. In the absence of international harmonization, tax mismatches will lead inevitably to tax gaps (and overlaps). In contrast with these criticisms, this Article suggests that, rather than a blatant power grab or an attempt to impose single taxation, the Commission’s use of external benchmarks in recent cases represents a response to the deficiencies of tax-expenditure analysis as a methodology for identifying tax subsidies.

Regardless of how it ended up there, the Commission now finds itself in a double bind. Benchmarking tax subsidies exclusively by reference to domestic law is underinclusive; for example, it regards structural rules as incapable of conveying state aid, regardless of their actual effects on cross-border commerce. But benchmarking by norms replaces policy preferences enacted by elected representatives with the policy preferences of the unelected Commission. Furthermore, because it mistakes mismatches for state aid, benchmarking by norms is overinclusive.

This Article offers an escape from the double bind of tax-expenditure analysis. The European Commission and other subsidy adjudicators could use the U.S. Supreme Court’s internal consistency test to review tax laws. The Supreme Court developed the internal consistency test to analyze dormant Commerce Clause challenges to state tax rules, including structural provisions.21See Ruth Mason, Made in America for European Tax: The Internal Consistency Test, 49 B.C. L. Rev. 1277 (2008) (arguing that the test could be used in fundamental freedoms cases). Under the test, the Supreme Court assumes all states apply the challenged state’s law.22Id. at 1283 (citing Okla. Tax Comm’n v. Jefferson Lines, Inc., 514 U.S. 175, 185 (1995)). If cross-border tax disadvantages persist despite hypothetical harmonization, they unconstitutionally discriminate against cross-border commerce.

The internal consistency test easily can be adapted for subsidy analysis by looking for cross-border tax advantages rather than disadvantages. The test offers several benefits compared to tax-expenditure analysis. First, the assumption embedded in the test—that all states apply the challenged state’s rule—has the effect of hypothetically harmonizing Member State tax laws. As a result, if the cross-border tax advantage disappears under the harmony assumption, the Commission can safely conclude that it arose from a tax mismatch, not from discriminatory subsidization by a single state. By preventing the Commission from mistaking tax mismatches for state aid, internal consistency could help the Commission avoid false positives.

Second, economic analysis has shown that the second step of the internal consistency test, which considers the impact on cross-border commerce of the harmonized rule, reveals whether the challenged rule functions equivalently to a tariff or an import or export subsidy.23Internal consistency correctly identifies tax measures with effects equivalent to import and export subsidies and tariffs. Brief of Michael S. Knoll & Ruth Mason as Amici Curiae in Support of Respondents at 9–24, Comptroller of Treasury v. Wynne, 135 S. Ct. 1787 (2015) (No. 13-485) [hereinafter Knoll & Mason Wynne Brief]; Brief of the Tax Economists as Amici Curiae in Support of Respondents at 23–27, Comptroller of Treasury v. Wynne, 135 S. Ct. 1787 (2015) (No. 13-485) [hereinafter Tax Economists’ Wynne Brief]; see Michael S. Knoll & Ruth Mason, How the Massachusetts Supreme Judicial Court Should Apply Wynne, 78 St. Tax Notes 921 (2015) (explaining how to use the internal consistency test to evaluate allocation rules); Michael S. Knoll & Ruth Mason, The Economic Foundation of the Dormant Commerce Clause, 103 Va. L. Rev. 309 (2017); Ryan Lirette & Alan D. Viard, Putting the Commerce Back in the Dormant Commerce Clause: State Taxes, State Subsidies, and Commerce Neutrality, 24 J. Law & Pol’y 467 (2016). Because this is the precise effect the state aid rules aim to prohibit, internal consistency is a reliable test for state aid.

Third, internal consistency applies the same way to every tax rule—structural or non-structural. By dispensing with the need to identify a baseline—be it the state’s own “normally” applicable law or an external norm—internal consistency completely avoids a major area of dispute between the Commission and the Member States.

Part I briefly explores EU law, emphasizing the goals of state-aid enforcement. Part II presents the Commission’s traditional tax-expenditure approach to identifying state aid, which relies on a domestic-law reference benchmark. Using recent examples, Part III presents the double bind of tax-expenditure analysis: benchmarking by reference law causes the Commission to err, but benchmarking by norms invades the reserved powers of the states. Part III also explains the types of cases that the traditional approach fails to address and argues that these failures drove the Commission to adopt external norms in recent cases. Finally, Part III presents arguments against benchmarking by external norms, including that it is unpredictable and illegitimate. Part IV advocates the internal consistency test as an alternative to tax-expenditure analysis, and it details the advantages of internal consistency over tax-expenditure analysis in identifying illegal subsidies. In addition, Part IV discusses the limits of the internal consistency approach, including that it would leave the European Union more vulnerable to tax competition from mismatches than would evaluating Member State rules against external norms. Tax mismatches—including differences in tax rates, definitions of tax residence, definitions of debt and equity, and rules for allocating income from cross-border transactions—present classic tax arbitrage opportunities. Such mismatches may seem undesirable, but the EU Treaties do not forbid them. Part V considers legislative options for ameliorating tax competition in the European Union.

I.   Balancing Regulatory and Market Competition in State Aid

The prohibition of state aid has appeared—nearly unaltered—in every EU treaty since the Treaty of Rome in 1958, and it applies to all areas, not just tax.24See Claire Micheau, State Aid, Subsidy and Tax Incentives under EU and WTO Law 57 (2014) (detailing the stability of the language and noting that alterations to the text merely changed the word “Community” to “Union”). Article 107 of the Treaty on the Functioning of the European Union (TFEU) prohibits state aid in the European Union:

Save as otherwise provided in the Treaties, any aid granted by a Member State or through State resources in any form whatsoever which distorts or threatens to distort competition by favouring certain undertakings or the production of certain goods shall, in so far as it affects trade between Member States, be incompatible with the internal market.25TFEU, supra note 1.

The TFEU obliges the Commission to monitor state aid and ensure abolition of incompatible aid.26Id. art. 108. This Part briefly describes the prohibition of state aid and its enforcement by the Commission.

A.   State-Aid Goals and Limits

The Member States of the European Union retain significant tax powers under the TFEU. Because the European Union can legislate on taxes only by unanimous agreement of the Member States, taxes largely remain creatures of the Member States, and each state’s system differs from the others.27See id. art. 115 (requiring unanimity); id. art. 4 (“[C]ompetences not conferred upon the Union in the Treaties remain with the Member States.”). Despite the high bar of unanimity, the states have legislatively harmonized value-added taxes across the EU, as well as several aspects of corporate income taxation.28See, e.g., Council Directive 2006/112/EC of Nov. 28, 2006 on the Common System of Value Added Tax, 2006 O.J. (L 347) 1 (providing harmonized VAT rules); Council Directive 2011/96/EU of Nov. 30, 2011 on the Common System of Taxation Applicable in the Case of Parent Companies and Subsidiaries of Different Member States, 2011 O.J. (L 345) 8 (eliminating double taxation of cross-border direct dividends).

State aid and the fundamental freedoms are two doctrines that limit Member States’ retained tax powers. The prohibition of state aid forbids subsidies that treat cross-border and domestic commerce unjustifiably differently from each other; the fundamental freedoms forbid taxes that do so.29See generally Micheau, supra note 24 (state aid); Ruth Mason & Michael S. Knoll, Waiting for Perseus: A Sur-Reply to Graetz and Warren, 67 Tax L. Rev. 375 (2014) (fundamental freedoms and state aid); Ruth Mason & Michael S. Knoll, What is Tax Discrimination?, 121 Yale L.J. 1014 (2012) (fundamental freedoms). Together they guarantee the free movement across state borders of goods, people, business activity, services, and capital.30The fundamental freedoms permit freedom of movement of goods and productive factors across the EU. See TFEU, supra note 1, art. 26 (internal market); id. art. 28 (goods); id. art. 45 (workers); id. art. 49 (establishment); id. art. 56 (services); id. art. 63 (capital).

The state-aid prohibition prevents states from interfering with market competition by selecting winners—such as particular industries, large multinationals, or national champions—and conferring on them unfair advantages in the form of discriminatory subsidies. I direct the arguments in this Article to a particular aspect of the prohibition of state aid, namely, its function of promoting the free flow of commerce across the states. Commentators use different terms to refer to this value, including “competitive neutrality,” “commerce neutrality,” a “level playing field,” “capital ownership neutrality,” and “nondiscrimination.”31The Commission refers to the state-aid rules as “limiting distortions of competition, preserving a level playing field and combating protectionism.” Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions: EU State Aid Modernisation, ¶ 15, COM (2012) 209 final (May 8, 2012) [hereinafter Commission Communication on State Aid Modernisation]. Various formulations of this idea exist in the literature. Mihir A. Desai & James R. Hines Jr., Evaluating International Tax Reform, 56 Nat’l Tax J. 487, 487 (2003) (taxes are “capital ownership neutral[]” when they do not “distort the ownership of capital assets” and such neutrality “promotes global efficiency whenever the productivity of an investment differs based on its ownership”). To emphasize its connection to facilitating private competition and to make clear that it applies outside the capital-investment context, Michael Knoll and I adopted the term “competitive neutrality” to refer to this idea. Mason & Knoll (2012), supra note 29, at 1021 (“A tax law is competitively neutral when it does not distort the matching of owners with investments (or workers with jobs). Viewed in this way, a tax system is competitively neutral if it maintains a level tax playing field between resident and nonresident taxpayers.”). By using “competitive neutrality,” we also sought to avoid ambiguity. See Michael S. Knoll, Reconsidering International Tax Neutrality, 64 Tax L. Rev. 99, 110 n.47 (2011) (noting that economists refer to competitive neutrality as “capital ownership neutrality,” whereas tax policymakers refer to it as “capital import neutrality,” which is itself a term economists reserve for distortions of savings neutrality). Referring to competitive neutrality, Ryan Lirette and Alan Viard adopted the term “commerce neutrality” to emphasize the role of the concept in U.S. dormant Commerce Clause doctrine. See Lirette & Viard, supra note 23, at 482–83 (“commerce neutral” tax systems are “those for which prices can change to preserve relative incentives for interstate transactions”); see alsoKnoll & Mason (2017), supra note 23, at 319, 336 n.126 (continuing to use “competitive neutrality” when discussing the principle in the dormant Commerce Clause context). The Supreme Court refers to both “discrimination” and taxes that “operate[] as tariffs.” See, e.g., Comptroller of Treasury v. Wynne, 135 S. Ct. 1787, 1804 (2015) (“Maryland’s tax scheme is inherently discriminatory and operates as a tariff.”). Regardless of terminology, the underlying idea is simple: states should not use their spending, regulatory, or tax systems to distort competition between economic actors of different states.32See references in supra note 31. Put differently, within the EU, the prohibition of state aid prevents protectionism and retentionism;33“Whereas taxes or regulations are protectionist when they discourage outsiders from engaging in economic activities within a state, taxes or regulations are retentionist when they discourage in-state economic actors from engaging in out-of-state activities.” Knoll & Mason (2017), supra note 23, at 309. it prevents taxes that function equivalently to tariffs or import/export subsidies.34Mason & Knoll (2012), supra note 29, at 1046 n.121, 1061 n.151 (defining competitive neutrality for labor taxes). In this Article, I focus on differences in treatment between cross-border and domestic commerce, and, therefore, I do not take up other types of state aid, including sectoral discrimination. Although the techniques proposed here may be helpful in sectoral discrimination cases, I do not focus on them.

Equally critical to my analysis is the claim that EU law generally preserves state tax autonomy, including autonomy to set tax bases, tax rates, and methods for taxing cross-border income.35See TFEU, supra note 1, art. 4 (“[C]ompetences not conferred upon the Union in the Treaties remain with the Member States.”); see also Alexandre Saydé, One Law, Two Competitions: An Enquiry into the Contradictions of Free Movement Law, in 13 Cambridge Y.B. Eur. Legal Stud. 365 (Catherine Barnard & Okeoghene Odudu eds., 2011) (explaining that the EU values both regulatory and market competition). The tax state-aid doctrine of the European Union’s highest court, the Court of Justice of the European Union (CJEU or “Court of Justice”), is still nascent. But the Court has decided many cases under the fundamental freedoms that help explicate the relative tax competences of the EU and the Member States and the ways that EU law limits Member State tax powers.36See generally Mason & Knoll (2012), supra note 29, at 390–99 (highlighting cases where the CJEU has interpreted the fundamental freedoms as promoting competition).

Although Member States may not discriminate against cross-border commerce or against nationals from other Member States, the TFEU generally does not prohibit nondiscriminatory tax laws, even if those laws distort where EU taxpayers work, invest, and do business.37See generally id. at 425–28 (providing examples of laws that the TFEU has allowed, despite the locational distortions they created). Significant consequences follow from this interpretation. Most importantly, neither the fundamental freedoms nor the state-aid rules require allocational efficiency, or what tax policymakers refer to as “capital export neutrality” or “locational neutrality.”38For a remarkably clear explanation of the various benchmarks, see Rosanne Altshuler, Recent Developments in the Debate on Deferral, 20 Tax Notes Int’l 1579 (2000). For a general discussion of the neutrality benchmarks for international tax versus EU law, see Elizabeth F. Donald, Michael S. Knoll & Ruth Mason, Tax Discrimination (manuscript on file). For implications of the neutrality benchmarks for common markets, see generally Mason & Knoll (2012), supra note 29; Knoll & Mason (2017), supra note 23. For the argument that the fundamental freedoms do not coherently enforce any of the neutrality benchmarks, see Michael J. Graetz & Alvin C. Warren, Jr., Income Tax Discrimination: Still Stuck in the Labyrinth of Impossibility, 121 Yale L.J. 1118 (2012). Although the EU’s tax nondiscrimination rules contribute to efficient allocation of productive factors across the EU, that is not their primary purpose.

This limit on EU law is crucial for maintaining Member State tax autonomy. For example, the Court of Justice consistently regards “tax disparities” as compatible with the TFEU.39See, e.g., Case C-403/03, Schempp v. Finanzamt München V, ECLI:EU:C:2005:446 ¶ 34 (July 12, 2005). Tax disparities are cross-border tax differences that “result from divergences existing between the various Member States, so long as they affect all persons subject to them in accordance with objective criteria and without regard to their nationality.”40Id. Thus, as long as they do not discriminate, Member States may define their tax bases as they wish, even though tax disparities may inefficiently distort location of productive factors across the European Union. To survive EU law challenges, divergent laws must apply the same way to Member State nationals and nationals of other EU Member States. They also must apply the same way to cross-border and purely domestic commerce. In keeping with tax vernacular, I will refer to such disparities as “tax mismatches.”41See, e.g., Organisation for Economic Co-operation and Development (OECD), Neutralising the Effects of Hybrid Mismatch Arrangements, Action 2: 2015 Final Report 11 (2015) (developing unilateral defensive measures for combatting mismatches); Mitchell A. Kane, Strategy and Cooperation in National Responses to International Tax Arbitrage, 53 Emory L.J. 89, 114 (2004) (referring to mismatches as creating “international tax arbitrage” opportunities). For more on mismatches under EU law, see Mason, supra note 21.

Mismatches in tax rates are the classic example. A taxpayer from a high-tax state who invests in a low-tax state may secure a cross-border tax advantage. A taxpayer from a low-tax state who invests in a high-tax state may suffer a cross-border tax disadvantage. Because states retain autonomy over their rates, however, EU law does not prohibit such cross-border tax advantages or disadvantages, as long as each state applies its tax rules evenhandedly to all.42See, e.g., Case C-336/96, Gilly v. Directeur des Services Fiscaux du Bas-Rhin, ECLI:EU:C:1998:221 ¶ 1 (May 12, 1998)(declining to remediate a cross-border tax disadvantage arising from a rate mismatch). Thus high tax rates do not violate the fundamental freedoms as discriminatory taxes, and low tax rates do not constitute illegal state aid. This is true even though differences in tax rates may cause locational distortions and perhaps even a race to the bottom on taxes. According to the CJEU, EU law “offers no guarantee to a citizen of the Union that transferring his activities . . . will be neutral as regards taxation.”43Case C-365/02, In re Lindfors, ECLI:EU:C:2004:449 ¶ 34 (Mar. 4, 2004). EU law does not prevent tax mismatches because, according to the Court of Justice, requiring tax harmonization “would . . . encroach on [Member State] sovereignty in matters of direct taxation.”44See, e.g., Gilly, ECLI:EU:C:1998:221 ¶ 48.

Taxpayers frequently argue that double taxation arising from mismatches violates EU law.45See, e.g., Schempp, ECLI:EU:C:2005:446 ¶¶ 9–10. The Court of Justice has consistently rejected such claims under the fundamental freedoms, reasoning that because the TFEU provides no rules for setting tax rates, tax bases, or methods to divide income among the states, the Court cannot resolve double taxation arising from the application of different, but nondiscriminatory, laws adopted by two or more states.46See id. The CJEU has come to this conclusion notwithstanding that double taxation may chill cross-border economic activity.47See Case C-376/03, D. v. Inspecteur van de Belastingdienst, ECLI:EU:C:2004:663 ¶ 85 (opinion of Advocate General Ruiz-Jarabo Colomer) (“[T]he fact that a taxable event might be taxed twice is the most serious obstacle there can be to people and their capital crossing internal borders.”).

For example, in Block,48Case C-67/08, Block v. Finanzamt Kaufbeuren, 2009 ECLI:EU:C:2009:92 (Feb. 12, 2009). Germany considered debt instruments to be taxable German assets under its inheritance tax if the creditor resided in Germany.49Id. ¶ 7, 9. In contrast, Spain had the opposite rule. It considered bonds to be Spanish if the debtor resided in Spain.50See id. ¶ 9. When a German creditor owned debt in a Spanish company, both Germany and Spain would include the debt for inheritance tax purposes. A taxpayer complained that the resulting double taxation violated the fundamental freedoms, but the Court of Justice disagreed. It observed that “Community law . . . does not lay down any general criteria for the attribution of areas of competence between the Member States in relation to the elimination of double taxation.”51See id. ¶ 30.

To fix the overlap that resulted from differences in the Spanish and German rules, the Court would have had to choose between the German and Spanish rule. But the Court acknowledged that absent EU legislation, the EU Treaties do not require any state to conform its tax laws to that of any other.52Id. ¶¶ 30–31 (“[T]he Member States . . . are not obliged therefore to adapt their own tax systems to the different systems of tax of the other Member States in order, inter alia, to eliminate the double taxation arising from the exercise in parallel by those Member States of their fiscal sovereignty . . . .”). With unanimous agreement of the Member States, the EU central government could set a common rule for the states. But the Court acknowledged that, in the absence of such legislation, states may implement the tax policies that their voters prefer.53Id. (“[N]o uniform or harmonisation measure designed to eliminate double taxation has as yet been adopted at Community law level.” (citation omitted)). This interpretation accommodates tax competition; under it, states may compete with each other to attract business, investment, and residents to their jurisdiction.

If the facts of Block had been reversed so that a Spanish creditor owned bonds in a German company, neither country would include the asset in its inheritance tax. This mismatch would create a tax gap or a double benefit that would only occur in cases of cross-border debt holdings. Just as Block did not violate the fundamental freedoms as tax discrimination, reverse-Block would not violate the state-aid rules as an illegal tax subsidy.

A consequence of this division of tax powers between the EU central government and the Member States is that some forms of tax competition (including tax-rate competition) must be tolerated, even though the impact of those policies spills over, affecting residents of other states.54Tax competition risks a race-to-the-bottom in which states lower tax burdens until they cannot adequately fund public goods and services or must shift taxes to less mobile factors such as labor or land. Empirical studies arrive at mixed conclusions on the impact of tax competition. For a literature review, see Michael Keen & Kai A. Konrad, The Theory of International Tax Competition and Coordination, in 5 Handbook of Public Economics 257 (Alan J. Auerbach et al. eds., 2013). Indeed, the Court of Justice repeatedly has emphasized that regulatory competition is a value protected by the EU Treaties.55See Case C-212/97, Centros, ECLI:EU:C:1999:126 ¶ 27 (Mar. 9, 199) (affirming regulatory competition as an EU value). Similarly, gaps and overlaps arising from mismatches may distort cross-border economic activity and intensify tax competition. For example, the German-Spanish mismatch approved in Block may distort taxpayers’ decisions about where to reside and whether to hold debt or other types of assets. It may impact taxpayers’ preferences for holding cross-border versus domestic debt. Nevertheless, the TFEU does not require conformity in state tax laws; it only prohibits states from discriminating among nationals or enterprises of different states or between cross-border and domestic commerce.56See, e.g., TFEU, supra note 1, art. 107 (also forbidding sectoral discrimination). This approach is not unusual in the common-market context. The Supreme Court has held that nondiscriminatory tax mismatches among U.S. states do not violate the dormant Commerce Clause and that only Congress—not courts—may eliminate them.57See, e.g., Comptroller of Treasury v. Wynne, 135 S. Ct. 1787, 1802 (2015); Moorman Mfg. Co. v. Bair, 437 U.S. 267, 278 (1978). See generally Knoll & Mason (2017), supra note 23 (discussing Wynne).

Appropriately mindful of states’ reserved tax powers, the Commission has not interpreted the prohibition of state aid to forbid mismatches, require tax harmonization, or require minimum taxation.58See 2016 Notice, supra note 1, ¶¶ 132–89. Rather, the Commission has interpreted the state-aid rules to promote competitive neutrality, that is to “preserv[e] a level playing field” among private enterprises and to “combat[] protectionism.”59Commission Communication on State Aid Modernisation, supra note 31, ¶ 15. As Advocate General Jääskinen put it, “[H]armful institutional or tax competition between Member States clearly does not fall within the mechanism for controlling State aid established by the Treaty, even though there are cases where measures are liable to amount both to harmful tax competition and to State aid incompatible with the common market.” Joined Cases C-106/09 P & C-107/09 P, Comm’n v. Gibraltar, ECLI:EU:C:2011:215 ¶ 143 (Apr. 7, 2011) (opinion of Advocate General Jääskinen), http://curia.europa.eu/juris/celex.jsf?celex=62009CC0106 [https://perma.cc/93QB-ZA2Q]. It is worth emphasizing once more that the rest of this Article assumes that, like the fundamental freedoms, the state-aid rules, at least in part60This Article does not cover some aspects of state aid, including its prohibition of sectoral discrimination., are designed to promote competitive neutrality, and they are not properly interpreted to eliminate interstate tax (or other regulatory) competition, except incidentally to promote competitive neutrality. If my assumptions are unjustifiably narrow, or if the state-aid doctrine expands to encompass more general limits on tax competition, then the arguments in this Article would apply only to a subset of cases, namely those implicating differences in tax treatment between cross-border and domestic commerce.61EU law expressly recognizes that the state-aid concept evolves with the common market, so that Member State policies previously judged compliant can become illegal aid due to further EU integration. See, Joined Cases C-182/03 & C-217/03, Belgium & Forum 187 v. Comm’n, ECLI:EU:C:2005:266 (June 22, 2006) (finding to confer state aid a regime that the Commission had previously approved as notified aid); Ryan Finley, Commission’s State Aid Powers Are Too Broad, German Report Says, Tax Notes Int’l (Nov. 27, 2017) (quoting the technical advisory board on state aid of the German Ministry of Finance as observing that the state-aid rules have a reputation for “permanent expansion”).

B.   Subsidy Control: Practice and Procedure

1.      Elements of state aid

The EU Commission bears the burden to prove the core element of tax state aid: the presence of a so-called “selective advantage.”62See, e.g., Joined Cases C-393/04 & C-41/05, Air Liquide Indus. Belg. SA v. Ville de Seraing, ECLI:EU:C:2006:403 ¶ 30 (June 15, 2006); 2016 Notice, supra note 1, ¶ 118. The Commission bears the burden of proof for other issues, including showing that the aid expended state resources and impacted the EU market. Proving the former is trivial in tax cases, and the CJEU permits the Commission to assume the latter. Selective advantage therefore represents the Commission’s main burden in any tax state-aid case. The “advantage” is the financial benefit that accrues to the enterprise on account of the state’s action. In cases involving state aid delivered through the tax system, it is the tax savings.63See, e.g., Commission Notice on the Application of the State Aid Rules to Measures Relating to Direct Business Taxation, 1998 O.J. (C 384) 3, ¶ 22 [hereinafter 1998 Notice]; 2016 Notice, supra note 1, ¶ 127. This can include tax savings delivered through deductions, exemptions, credits, deferral, and the like.64See 2016 Notice, supra note 1, ¶¶ 126–41; 1998 Notice, supra note 63, ¶ 9.

The Commission also must show that the state granted the subsidy “selectively.” Selectivity is state aid’s discrimination requirement; the Commission must show that the state granted the benefit to some firms, but not to similarly situated others.65See, e.g., Air Liquide, ECLI:EU:C:2006:403 ¶¶ 31–32; 2016 Notice, supra note 1, ¶ 118. Selectivity derives from Article 107’s requirement that, to be illegal, aid must favor “certain undertakings or the production of certain goods.”66TFEU, supra note 1, art. 107(1).

Although the CJEU does not refer to “suspect classes” the way the U.S. Supreme Court does, analysis of state-aid cases reveals that state aid’s “selective classes” include nationality, region, and whether an enterprise engages in domestic or cross-border economic activity.672016 Notice, supra note 1, ¶ 121; see also Ruth Mason, An American View of State Aid, 157 Tax Notes 645, 647, 649 (2017) (cataloging selective classifications and linking them to state-aid’s goal to prevent discrimination between cross-border and domestic commerce). State aid includes other selective classes not considered here, including sector. The Commission sometimes claims that categories having nothing to do with sector or cross-border commerce constitute selective classes, which suggests that the application of the state aid prohibition could be very broad. For example, although the CJEU rejected it, the Commission made the argument that companies making losses constitute a selective class. See, e.g., Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 ¶ 77 (Nov. 15, 2011) (annulling the Commission’s decision that a tax limited to fifteen percent of a company’s profits favored companies that made losses). In another recent case, the Commission determined that a Member State conferred state aid when it restricted loss offsets to “insolvent or over-indebted” companies, while denying offsets to other companies. Commission Decision 2011/527 of Jan. 26, 2011 on State Aid Implemented by Germany (Sanierungsklausel), 2011 O.J. (L 235) 33, ¶ 73. This decision suggests that the Commission regards distressed companies as a selective class. Such an expansive interpretation would unmoor state aid from its goal to facilitate market integration. The CJEU reversed the Commission’s decision on other grounds. Case C‑203/16 P, Andres v. Comm’n, ECLI:EU:C:2018:505 ¶ 93 (June 28, 2018) (holding that the Commission used the wrong reference base for determining selective advantage). Additionally, when a state confers an advantage to only a single company, the Commission is entitled to presume that such “individual aid” is selective.68Case C-15/14 P, Comm’n v. MOL, ECLI:EU:C:2015:362 ¶ 60 (June 4, 2015). Most of the cases discussed in this Article involved individual aid.

That these classifications would draw the focus of the Commission and CJEU makes sense because states could use these classifications to enact policies that undermine the European common market. The state-aid rules promote free trade by prohibiting protectionism and policies that function equivalently to import or export subsidies.69See, e.g., Joined Cases C-20/15 P & C-21/15 P, Comm’n v. World Duty Free Grp. SA, ECLI:EU:C:2016:981 (Dec. 21, 2016) (capital export subsidy); Joined Cases C-182/03 & C-217/03, Belgium & Forum 187 v. Comm’n, ECLI:EU:C:2005:266 (June 22, 2006) (subsidy for increasing economic activity within the challenged state). It is not the goal of this Article to summarize the Commission’s state-aid doctrine. Many others have ably done so. Micheau, supra note 24 (emphasizing the similarity to WTO rules). Again, while I acknowledge that the state-aid rules prohibit sectoral subsidies, the principal focus of this Article is the impact of the prohibition of state aid on Member States’ ability to treat cross-border and domestic commerce differently. If the Commission were to expand its conception of state aid, we would expect to add to my list of “selective classes.”

As the last step in state-aid analysis, if the Commission carries its burden to show selective advantage, the accused state has a chance to show that public policy reasons justified the aid.70MOL, ECLI:EU:C:2015:362, ¶ 60.

2.      State-aid procedure

The TFEU assigns the Commission the responsibility to monitor state aid and abolish it where appropriate.71TFEU, supra note 1, art. 108. The prohibition of state aid is subject to many treaty and statutory exceptions. See, e.g., id. art. 107(2)–(3) (exceptions to raise standards of living in less developed regions, address natural monopolies and other market failures, and aid to remedy natural disasters). For more on state-aid procedure, see Ruth Mason, Tax Rulings as State Aid FAQ, 154 Tax Notes 451 (2017). Due to its association with what Europeans call competition policy (what Americans call antitrust), the Commission’s Directorate-General for Competition handles most state-aid enforcement, including tax state aid.

The Commission enjoys wide investigatory discretion and broad compulsory powers—it can obtain information from any state and any firm, including a putatively aided company and its competitors.72Council Regulation 2015/1589 of 13 July 2015, Laying Down Detailed Rules for the Application of Article 108 of the TFEU, 2015 O.J. (L 248) 9 [hereinafter 2015 Regulation] (information gathering powers). The challenged state, other EU states, and other interested parties (including competitors of the putatively favored enterprise) may participate in the investigation.73Id. arts. 1(h), 12(1).

Adverse Commission decisions lead to modification or termination of the aid, as well as monetary recovery of the subsidy from the aided enterprise, going back ten years from the Commission’s action.74The aiding state determines the exact amount of recovery under guidance from the Commission. See, e.g., Apple, supra note 2, ¶¶ 445–51 (describing how Ireland should calculate the recovery from Apple “to restore the position to the status quo ante”). The enterprise repays the aid to the state that originally granted it, a remedy intended to put the enterprise in the position it would be in had it not received aid.752015 Regulation, supra note 72, art. 16. That is why the blockbuster case against Ireland resulted in an order to recover billions in back taxes from Apple. The taxpayer must pay the recovery amount regardless of the Member State’s statute of limitations for taxes.76Id. pmbl. ¶ 25, art. 16. Contrary to other anti-subsidy rules, the offending state keeps the recovery.77Rather than disgorgement, WTO law permits members to impose countervailing duties on goods from WTO states judged to have provided specified kinds of subsidies. General Agreement on Tariffs and Trade, Oct. 30, 1947, 61 Stat. A-11, 55 U.N.T.S. 188, art. VI, ¶ 2.

The aided taxpayer, the aiding Member State, or both may appeal the Commission’s state-aid determination, first to the European Union’s lower court and then to the CJEU. All the cases involving U.S. multinationals discussed in this Article that resulted in a recovery order have been appealed.

II.   Identifying Subsidies Using Tax-Expenditure Analysis

The familiar problem at the heart of tax-expenditure analysis—whether performed for purposes of legislative accountability or identifying illegal subsidies—is that it requires a baseline against which “special” deviations can be measured. Unfortunately, it has never been clear how to construct the baseline.78See generally Boris I. Bittker, Accounting for Federal “Tax Subsidies” in the National Budget, 22 Nat’l Tax J. 244, 247 (1969). Until recently, the Commission exclusively used reference law as the benchmark for identifying tax state aid. Reference-law benchmarking regards facially neutral rules—including structural rules—as conferring no tax advantages (and therefore no state aid), and it regards all facially selective rules as conferring tax advantages that may be state aid. Because facially neutral rules can violate competitive neutrality and facially selective rules can comply with competitive neutrality, however, reference-law benchmarking is both under- and overinclusive. The main alternative approach to reference-law benchmarking favored by the Commission in recent cases—external benchmarking—has problems of its own. For example, selecting an external benchmark for evaluating Member State rules requires the Commission to make tax policy decisions that exceed the powers conferred to it by the TFEU. This leads to tax-expenditure’s double bind: reference-law benchmarking causes the Commission to err; benchmarking by norms causes the Commission to invade the reserved tax powers of the states.

A.   The Ineluctable Baseline Problem

Tax-expenditure analysis did not begin as a tool to identify illegal subsidies under trade law or constitutional law. Instead, it began as a tool for promoting legislative accountability. Many countries, including the United States, calculate tax-expenditure budgets to keep lawmakers and the public informed about how the government spends and regulates through the tax system. For example, the Congressional Budget Act of 1974 defines tax expenditures as “revenue losses attributable to provisions of the Federal tax laws which allow a special exclusion, exemption, or deduction from gross income or which provide a special credit, a preferential rate of tax, or a deferral of tax liability.”792 U.S.C. § 622(3) (2006).

The Joint Committee on Taxation and the Treasury Department typically use reference law as the baseline for calculating tax expenditures.80Staff of J. Comm. on Taxation, 115th Cong., Estimates of Federal Tax Expenditures for Fiscal Years 2018-2022 at 2(Comm. Print 2018). This choice means that some provisions are not labeled as tax expenditures, even though they confer tax savings.81For example, the Joint Committee and Treasury regard the exclusion of imputed income and the realization rule to be part of the reference base, so they do not score as tax expenditures.

As an alternative to using reference law as the benchmark, estimators can identify an external benchmark and then deem any provisions that vary from the external benchmark to be tax expenditures. The external benchmark could be a normative benchmark consisting of an idealized view of tax policy, or it could consist of an international standard, dominant state practice, or another norm.822008 JCT Report, supra note 6, at 39–40; see also David Kamin, Basing Budget Baselines, 57 Wm. & Mary L. Rev. 143 (2015) (providing a taxonomy of baselines for federal budgeting purposes). For example, depreciation schedules can be benchmarked against economic depreciation under financial accounting rules.

An obvious drawback of external benchmarking is the need to select an external norm. Whereas a reference-law baseline has political legitimacy, it is not clear where the budget estimator or subsidy adjudicator derives the authority to choose a norm other than the reference law, that is, other than the tax law as determined by democratically elected lawmakers. Boris Bittker famously observed that “every man can create his own set of ‘tax expenditures,’ but it will be no more than his collection of disparities between the income tax law as it is, and as he thinks it ought to be.”83Bittker, supra note 78, at 244.

Setting aside the legitimacy of benchmark choices, both reference-law and external-norm benchmarking leave important questions unanswered due to ambiguity or indeterminacy.84Cf. 2008 JCT Report, supra note 6, at 41 (acknowledging that for some tax provisions, tax-expenditure analysis “arbitrarily select[s] one taxing pattern or another as the general rule”). The appropriateness of a tax rule or administrative practice can be ambiguous under either approach. For example, one could treat corporate debt as a favorable deviation from equity just as easily as one could treat equity as a tax penalty relative to debt. Or both corporate debt and equity could be measured against some normative tax treatment of all corporate capital. Likewise, one could understand a tax exemption for pass-through entities as a favorable deviation from corporate taxation, yet one could just as easily regard corporate taxation as a tax penalty relative to pass-through treatment. The equivalent for individual taxpayers would be to decide whether married couples are penalized relative to singles, or whether singles are advantaged relative to couples. Because there exists neither a clear reference-law baseline nor a clear normative answer to these questions, the proper baseline for tax provisions touching on these questions is ambiguous under either reference-law benchmarking or external-norm benchmarking. Selecting a baseline in such cases is necessarily arbitrary. For other tax provisions, both the proper (in a normative sense) and the reference-law tax treatment may be indeterminate. Suppose we suspect that the annual filing period and the marginal rate system confer tax advantages. Against what baseline would we measure such advantages?

The problem of ambiguity or indeterminacy of the baseline is particularly acute for so-called structural provisions. The distinction between structural and non-structural rules basically tracks the distinction between the background features of the tax system and its particular provisions, but the notion of structural tax provisions is not well-defined and may be illustrated best by examples.85See 2008 JCT Report, supra note 6, at 2 (quoting Surrey as defining structural provisions as those “necessary to implement the income tax” and distinguishing them from tax expenditures); see also Daniel N. Shaviro, Rethinking Tax Expenditures and Fiscal Language, 57 Tax L. Rev. 187, 212 (2004) (defining “structural rules” as those that are “usefully distinguished both from theoretically pure income tax rules and from conventional tax expenditures that seem more politically interchangeable with appropriations”). One category of structural provisions involves tax-mix questions that policymakers do not typically intend to revisit. For example, selection of an income-tax base necessarily implies rejection of a consumption-tax base.

Another type of structural provision results from practical necessity; for simplicity and ease of administration, the tax system may, for example, ignore inflation, exempt imputed income, defer tax on gains until realization, and choose an arbitrary reporting period (such as one year).86Imputed income is the value of services one provides to oneself or the value of using one’s own assets. For example, if a taxpayer lives in a house she owns, she has imputed rental income, which the United States excludes from taxation. Under the realization rule, states defer inclusion of gains and deduction of losses until disposition of the related asset. An alternative would be a mark-to-market regime, which would account for annual appreciation and depreciation of assets. Ari Glogower, Taxing Capital Appreciation, 70 Tax L. Rev. 111, 111, 113 (2016). Other building blocks of the tax system that generally lie outside tax-expenditure analysis involve definitions or classifications (such as entity classification, filing status, or definitions of debt and equity), progressive rates, and, importantly for this Article, methods for dividing income from cross-border transactions among states with jurisdiction to tax it.

That a provision is structural does not mean it conveys no tax advantages. For example, the realization rule favors long-term capital investments over wage income, and differences in the tax treatment of debt and equity distort capitalization of companies.87See Tax Reform and the Tax Treatment of Capital Gains: Joint Hearing Before the H. Comm. on Ways and Means & S. Fin. Comm., 112th Cong. 36 (2012) (statement of Leonard E. Burman, Daniel Patrick Moynihan Professor of Public Affairs, Syracuse University). The main characteristic of structural tax provisions that is relevant to tax-expenditure analysis is uncertainty regarding the baseline against which they could be tested.88See Victor Thuronyi, Tax Expenditures: A Reassessment, 1988 Duke L.J. 1155, 1165–70 (1988) (noting that the Haig-Simons definition of income is “ambiguous or silent regarding some basic structural features of the tax system,” including rates, taxable unit, corporate integration, and the treatment of capital gains). For example, is it helpful to conceive of an income tax as a set of tax expenditures and penalties compared to a consumption-tax base? How confident could we be that resulting tax expenditures represent relevant subsidies?

Importantly for the recent state-aid cases, the normative basis for splitting international income among states is contested.89See id. at 1200; see also 2008 JCT Report, supra note 6, at 10 (concluding that the “general rule for taxing foreign earnings” “can fairly be said to be ambiguous”); Mitchell A. Kane, A Defense of Source Rules in International Taxation, 32 Yale J. on Reg. 311 (2015) (discussing various justifications for source rules). Before we conclude that normative indeterminacy should lead us to a reference-law approach, it is worth noting that for many countries, it may not be possible to identify with confidence a state’s principal income-allocation rule.90See infra note 143.

B.   Easy Cases Under Reference-Law Benchmarking

The Commission traditionally has identified illegal state aid by comparing the tax treatment of a selective class (or particular company) to a “reference system . . . composed of a consistent set of rules that generally apply.”91See 2016 Notice, supra note 1, ¶ 133; see also id. ¶ 128 (describing a “three-step analysis” involving (1) identifying the “system of reference,” (2) “determin[ing] whether a given measure constitutes a derogation from that system,” and (3) determining whether that derogation is selective). The Commission has identified some provisions that it regards as part of the state’s reference baseline, including “the taxable persons, the taxable event and the tax rates.” Id. ¶ 134. “Purely technical” provisions, such as depreciation rules, usually do not constitute state aid, even if they “var[y] from one Member State to another.” Id. ¶¶ 177–80. But see Joined Cases C‑78/08 & C‑80/08, Ministero dell’Economia e delle Finanze v. Paint Graphos, ECLI:EU:C:2011:550 ¶ 50 (Sept. 8, 2011) (agreeing with the Commission that the benchmark for evaluating the taxation of exempt cooperative association was the corporate tax base but holding that the tax exemption for cooperatives could be justified). This is a tax-expenditure approach.921998 Notice, supra note 63, ¶ 9 (explaining that illegal state aid can consist of tax savings delivered through “special deductions,” exemptions, credits, deferral, or cancellation or rescheduling of tax debts); see also 2016 Notice, supra note 1, ¶ 128 (confirming this approach). Crucially, it uses the state’s own domestic law as the baseline for measuring the tax advantage.932016 Notice, supra note 1, ¶ 129. Although this method finds no direct support in the TFEU, the Court of Justice accepts it,94See, e.g., Case C‑203/16 P, Andres v. Comm’n, ECLI:EU:C:2018:505 ¶ 93 (June 28, 2018) (describing the approach in detail). and in the late 1990s, the Commission committed to using it as a best practice.95See 1998 Notice, supra note 63, ¶¶ 9–10. The Commission is far from alone in using reference-law benchmarking to identify illegal subsidies, and the approach works well most of the time. The WTO Dispute Resolution Body and Supreme Court under the dormant Commerce Clause both use the approach to scrutinize subsidies.96See, e.g., Coenen & Hellerstein, supra note 4 (dormant Commerce Clause); Paul R. McDaniel, Trade Agreements and Income Taxation: Interactions, Conflicts, and Resolutions, 57 Tax L. Rev. 275 (2004) (WTO).

The Commission typically has encountered two types of easy cases under tax-expenditure analysis that uses a reference-law benchmark. The first category of easy cases centers on special tax provisions that benefit only members of selective classes. An example might be an overt export subsidy. Tax-expenditure analysis that uses reference law as a baseline properly identifies such provisions as conferring selective advantages. As a result, overtly selective tax expenditures are state aid unless they are justified.97Neither the Commission nor CJEU has been clear on what it takes to justify a selective tax advantage. The CJEU has said that a tax advantage can be justified by the “nature or general scheme of the [tax] system.” Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 ¶ 36 (Nov. 15, 2011). The Commission has declared that “intrinsic basic or guiding principles . . . [and] inherent mechanisms necessary for the functioning and effectiveness of the [tax] system” can justify differences in treatment. 2016 Notice, supra note 1, ¶ 138 (distinguishing “external policy objectives”). While it is unclear what justifies selective tax advantages, selective provisions that are justified for public policy reasons must be narrowly tailored. Id. ¶ 140 (justified selectivity must “not go beyond what is necessary to achieve the legitimate objective being pursued, in that the objective could not be attained by less far-reaching measures”). The second category of easy cases involves derogating provisions that do not benefit selective classes.98See 2016 Notice, supra note 1, ¶¶ 117–18. An example might be a one-year-only expensing rule that is available to all companies. Although such derogating provisions may be regarded as conferring tax advantages as judged by a reference-law benchmark, they are not relevantly discriminatory because they do not treat cross-border commerce differently from domestic commerce. Thus, derogating provisions that benefit no selective class do not constitute state aid. Because they involve tax subsidies delivered via overt derogations, both types of cases are easy to identify, and because the two types of cases—selective and nonselective—are typically easy to distinguish from each other, reference-law benchmarking has, until recently, produced predictable and uncontroversial outcomes.

1.      Unjustified facially selective rules

Countless cases illustrate the Commission’s reference-law benchmarking approach to state-aid analysis. For example, the CJEU affirmed the following as illegal aid: special deductions for foreign branch start-up and promotional costs when those costs were incurred by steel exporters, but not when incurred by other enterprises;99Case C-501/00, Spain v. Comm’n, ECLI:EU:C:2004:438 ¶ 120 (July 15, 2004). special income-calculation rules available to the coordination centers of multinational groups, but not other kinds of companies;100Joined Cases C-182/03 & C-217/03, Belgium & Forum 187 v. Comm’n, ECLI:EU:C:2005:266 (June 22, 2006). exemption from a tax on stopovers available to local, but not foreign, vessel owners;101Case C-169/08, Presidente del Consiglio dei Ministri v. Regione Sardegna, ECLI:EU:C:2009:709 ¶ 66 (Nov. 17, 2009). and favorable goodwill depreciation deductions available for purchases of foreign, but not domestic, stock.102Joined Cases C-20/15 P & C-21/15 P, Comm’n v. World Duty Free Group, ECLI:EU:C:2016:981 ¶ 22 (Dec. 21, 2016).

In each case, the Commission used deviations from the state’s own domestic law to establish the existence of a tax advantage. In each case, the Commission compared the state’s treatment of expressly favored taxpayers (e.g., exporters or purchasers of foreign stock) to that same state’s treatment of taxpayers outside the selective class (i.e., non-exporters or purchasers of domestic stock).103See, e.g., id. ¶¶ 68, 76–77, 98, 106. In these cases, the Commission focused on facially selective (i.e., discriminatory) tax rules.104See, e.g., id.

Each of these cases upheld a traditional value underlying the prohibition of state aid. The Commission held that protectionist tax benefits available only to steel exporters105Case C-501/00, Spain v. Comm’n, ECLI:EU:C:2004:438 ¶¶ 120–25 (July 15, 2004) (holding that a start-up cost deduction for steel exporters operated as an export subsidy even though it was not linked directly to exports). or local vessel owners locals were state aid.106See Regione Sardegna, ECLI:EU:C:2009:709 ¶ 66 (holding that limiting a tax exemption to Sardinian vessels discriminated against foreign vessel owners). It also held that capital export subsidies in the form of faster cost-recovery for foreign than domestic investments constituted state aid.107See World Duty Free Grp. SA, ECLI:EU:C:2016:981 ¶¶ 71–77 (holding that favorable goodwill depreciation available for purchases of foreign, but not domestic, stock constituted state aid). Each of these cases straightforwardly enforced the values undergirding the prohibition of state aid; they also all represent “easy cases” under the reference-law benchmarking approach. In each case, the state clearly derogated from its own generally applicable law to favor one of state aid’s selective classes. Before Part III introduces hard cases, the next Subsection reviews derogating tax provisions that were not state aid because they did not benefit selective classes.

2.      Nonselective advantageous rules

Not all derogations from generally applicable law confer tax advantages that require scrutiny as possible illegal state aid. Tax expenditures are permissible, provided they do not benefit a selective class. Thus, the following cases did not involve state aid: a tax on all companies limited to fifteen percent of their profits;108Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 ¶¶ 77–84 (Nov. 15, 2011). a procedural rule allowing companies with tax disputes that were more than ten years old to settle those disputes by paying five percent of the deficiency;109See Case C-417/10, Ministero dell’Economia e delle Finanze v. 3M Italia SpA, ECLI:EU:C:2011:550 ¶ 26 (Sept. 8, 2011). and an accelerated depreciation rule limited to certain tangible assets custom-built for leasing.110Case C-100/15 P, Neth. Mar. Tech. Ass’n v. Comm’n, ECLI:EU:C:2016:254 ¶¶ 61, 71, 74, 77, 84, 88 (Apr. 14, 2016) (affirming the Commission, which had decided not to pursue a formal state-aid case against Spain because its rule was not selective). Although the states in these cases drew distinctions—including distinctions related to profitability, the age of the tax dispute, or the intended use of assets—the states avoided use of selective classifications.

Not every state would make the same policy choices as the ones reflected in this set of cases, but none of the challenged policies involved covert subsidies to darling enterprises or industries; none involved preferences to national taxpayers; none functioned equivalently to an import or export subsidy. Because the TFEU leaves states free to pursue their own tax policy goals (absent protectionism or other discrimination between domestic and cross-border commerce), the state-aid rules do not prohibit these policies, even if they are foolish or inefficient.

III.   Hard Cases Under Tax-Expenditure Analysis

Most state-aid cases do not raise thorny baseline questions. Instead, they involve overt tax advantages conferred to selective classes of taxpayers in cases in which the state can provide no justification for the selective advantage. When a state overtly favors a selective class, both the deviation and baseline are clear: the deviation is the treatment of the selective class, and the baseline is the treatment of everyone else. Because past cases involved such unjustified facial selectivity, the Commission’s decisions employing tax-expenditure analysis received little political or academic criticism—or even attention. The widespread acceptance of the Commission’s approach to state aid was probably due to observers’ familiarity with tax-expenditure analysis as a technique for identifying tax subsidies and due to the Commission’s use of the state’s own law as its reference benchmark, which seemed either neutral or properly deferential to state tax policy decisions.111The Court of Justice itself observed that the tax-expenditure approach is sufficient to identify state-aid in cases involving facially selective derogations. Case C‑203/16 P, Andres v. Comm’n, ECLI:EU:C:2018:505 ¶ 93 (June 28, 2018) (referring to cases in which “two categories of operators are distinguished and a priori treated differently”).

Thus, until recently, the Commission’s tax-expenditure approach to state-aid review performed well. Using a domestic-law reference base allowed the Commission to avoid accusations that it exceeded its authority. Nor did such deference render the standard toothless. Instead, it branded as illegal any rules that facially derogated in favor of selective classes. But, in my view, recent cases posed two puzzles that tax-expenditure analysis is ill-equipped to handle: Can facially neutral rules convey state aid? and Can facially selective rules confer tax advantages that are not state aid? This Part explains that these questions are arising now because the Commission has stumbled into the tax-expenditure trap. Recent cases expose that not all tax provisions can be neatly categorized as either tax expenditures or normal provisions. For cases in the grey area, tax-expenditure analysis is unreliable.

To confront these emerging challenges, the Commission altered its approach in recent cases. Failure to adequately explain its departure from the reference-law approach left the Commission vulnerable to criticism that it sought to harmonize state taxes outside the European Union’s legislative process or that it sought to impose a single-tax requirement.112See Advisory Bd., Fed. Ministry of Fin., supra note 10, at 12. Either of these goals falls outside the Commission’s institutional authority.

A.   Facially Neutral Rules

State-aid enforcement targets not only overtly selective classifications, such as nationality, but also facially neutral classifications that correlate strongly with a selective classification. For example, under long-standing precedent, discriminating against nonresidents is tantamount to discriminating against non-nationals.113See, e.g., Case C-330/91, Queen v. Inland Revenue Comm’rs ex parte: Commerzbank AG, ECLI:EU:C:1993:303 ¶ 1, 14–18 (July 13, 1993). Tax-expenditure analysis does a good job of analyzing covert discrimination, at least when the Commission is correct in its conclusion that a particular facially neutral classification impermissiblyproxies a selective classification. Yet, as this Section explains, tax-expenditure analysis provides no traction in cases that employ neither facially discriminatory classifications nor proxy classifications. To address the inability of tax-expenditure analysis to handle facially neutral rules, in recent cases the Commission turned to both impact analysis and external benchmarking.

1.      Vodafone

A recent referral114See, e.g., Request for a Preliminary Hearing from the Fővárosi Közigazgatási és Munkaügyi Bíróság (Hungary) Lodged on 6 February 2018—Vodafone Magyarország Mobil Távközlési Zrt. v. Nemzeti Adó-és Vámhivatal Fellebbviteli Igazgatósága, 2018 O.J. (C 221) 2, 3 [hereinafter, Vodafone reference]. concerns the application of Hungary’s turnover-tax regime to Vodafone. The regime imposed graduated tax rates that increased with companies’ turnover, and it included a complete exemption for companies with turnover below a statutory threshold. Because high-turnover companies tend to be foreign, while low-turnover companies tend to be domestic, graduated turnover taxes raise indirect discrimination questions.115Id. (also raising the question of whether the higher tax on higher-turnover companies violates the fundamental freedoms). The Hungarian regime does not employ overtly selective classes. Instead, Vodafone raises the question of whether the state’s use of turnover size indirectly discriminates on the basis of company nationality. The CJEU typically analyzes such cases according to the impact of the proxy classification on the legally relevant class.116See generally Ruth Mason & Leopoldo Parada, Company Size Matters, Brit. Tax Rev. (forthcoming Dec. 2019). For company-size cases, the Court recently held that the relevant question was whether a majority of those subject to higher taxes were foreign.117Case C-385/12, Hervis Sport-és Divatkereskedelmi Kft. v. Nemzeti Adó-és Vámhivatal Közép-dunántúli Regionális Adó Fóigazgatósága, ECLI:EU:C:2014:47 ¶ 46 (Feb. 5, 2014).

Whether turnover-size constitutes an impermissible proxy for nationality remains an open legal question. But, assuming that the Commission can accurately identify proxies for selective classes, the Commission can easily adapt its traditional tax-expenditure approach for use with proxy classifications. Instead of measuring the tax advantage as a favorable deviation from the state’s regular tax base that benefits a selective class, the Commission simply would measure the tax advantage as a favorable deviation from the regular tax that benefits a proxy class.

Thus, if the subsidy adjudicator “correctly”118The question of how the Commission and EU courts ought to identify such proxies remains for another day. Note also that due to its specific procedural history, the CJEU, not the Commission, will conduct the state-aid analysis in Vodafone. concludes that a facially neutral classification proxies a selective class, application of traditional tax-expenditure analysis becomes straightforward. What this means in Vodafone is that if CJEU determines that turnover size proxies nationality, the Hungarian law would derogate in favor of small (national) companies. If Hungary could offer no convincing justification for such favoritism, its graduated turnover tax would constitute illegal state aid. Under these conditions, Vodafone would become an “easy” case under tax-expenditure analysis; the proxy classification (turnover size) would be treated as a facially selective classification.119In contrast, if the subsidy adjudicator were to conclude that size does not proxy nationality, then there would be no derogation from domestic law that favored a proxy or a selective class, and therefore there would be no state aid under current methods of state-aid analysis. Section II.A.2, infra, analyzes scenarios like this.

2.      Gibraltar

Tax regimes can confer benefits to selective classes without using selective classifications or proxy classifications. The 2011 Gibraltar120Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 (Nov. 15, 2011). case involved a proposal for a facially neutral tax regime that, if adopted, would have had disproportionate cross-border impacts.121See id. EU law governs Gibraltar as part of the United Kingdom. See id.; see also TFEU, supra note 1, art. 355(3). Gibraltar submitted a proposed tax regime to the Commission for advance determination of whether the regime would be state aid if implemented.122See Gibraltar, ECLI:EU:C:2011:732 ¶¶ 8–14. The proposal was for a payroll and property tax capped at a percentage of corporate income.123Id. ¶¶ 77–84. The regime also envisioned a corporate registration tax. The Commission originally argued that the profit cap constituted a selective advantage for unprofitable companies. See Aid C 66/2002 (ex N 534/2002)—Gibraltar Government Corporation Tax Reform, 2002/C 300/02, ¶ 38, 2002 O.J. (C 300) 7 (noting the profit caps would “appear self-evidently to grant an advantage to those companies that make no profit”). The CJEU rejected this argument. See Gibraltar, ECLI:EU:C:2011:732 ¶ 77. The descriptions of the proposed Gibraltar regime in the Commission’s decisions and the EU courts’ judgments makes it difficult to determine whether the rule was facially discriminatory (because it only applied to Gibraltar companies that employed people in Gibraltar or occupied business property in Gibraltar) or facially neutral (applied to all companies, regardless of whether they were registered in Gibraltar, that employed people in Gibraltar or occupied business property there). See, e.g., Commission Invitation to Submit Comments Pursuant to Article 88(2) of the EC Treaty on Gibraltar Government Corporation Tax Reform, 2002 O.J. (C 300) 2, ¶ 7 (Dec. 4, 2002) (describing the payroll tax as applicable to “Gibraltar employers,” which are “companies incorporated or registered in Gibraltar”). It is not clear whether companies incorporated outside Gibraltar would have to register there to have employees there. See, e.g., id. ¶ 12 (suggesting that all non-public employees in Gibraltar would generate tax for their employers). It appears that the property tax would have applied to both resident and nonresident companies. See id. ¶ 12 (property tax would apply to “all companies occupying property in Gibraltar for business purposes); see also Commission Decision 2005/261 of Mar. 30, 2004 on Gibraltar Corporation Tax Reform, 2005 O.J. (L 85) ¶ 63 (noting that “the reform abolishes any distinction between resident and nonresident companies”). Although the discussion in the text assumes facial neutrality, infra note 291 considers how the analysis would change if the regime was facially selective such that it applied only to certain companies with payroll and property in Gibraltar, such as companies incorporated there.

The Commission refused to approve the proposal, even though the regime would nominally apply the same way to all taxpayers.124Gibraltar, ECLI:EU:C:2011:732 ¶¶ 85–110. The Commission reasoned that the regime would confer state aid as applied because it would favor “offshore companies,” which tended not to have payroll and property in Gibraltar, over onshore companies that tended to have property and payroll in Gibraltar.125Id. ¶ 106. Although the Commission did not define “offshore companies,” the term seemed to refer to “letterbox” companies, companies formally incorporated in Gibraltar, but that did not have much substantive activity there. Id. ¶ 17 (referring to “letterbox” or “asset management companies”).

It was not just the expected outcome but surely also the intended result that the proposed Gibraltar regime would tax offshore companies lightly while taxing onshore companies more heavily. A tiny British territory with a population of about 30,000, Gibraltar thrives by offering foreign companies a low-tax berth for their profits.126See Daniel Bunn, Fiscal Fact No. 623, Corporate Income Tax Rates Around the World (2018), Tax Found. (Nov. 2018), https://files.taxfoundation.org/20190603100114/ Tax-Foundation-FF623.pdf [https://perma.cc/LUE4-CV5M] (listing Gibraltar’s ten percent tax rate as one of the twenty lowest in the world and noting the worldwide average corporate income tax as twenty-three percent). Prior to the proposed regime, Gibraltar had another offshore-friendly tax regime, but the Commission had invalidated it as state aid.127Gibraltar, ECLI:EU:C:2011:732 ¶¶ 4–6. The old regime was facially selective.128See id. ¶ 4. The proposed new regime was Gibraltar’s clever attempt to duplicate the result of the old regime while avoiding facial selectivity, and thereby, Gibraltar hoped, avoiding an adverse state-aid decision.

The Commission saw through Gibraltar’s gambit; it understood that the goal of the regime was to favor offshore companies, a selective class. But the Commission could not establish that Gibraltar conferred a selective advantage using its traditional tax-expenditure approach because that approach depended on identifying the tax advantage as a selective deviation from generally applicable law. Since the Gibraltar regime was facially neutral and applied to all companies the same way, the Commission could not point to any deviating provision that conferred tax advantages only to offshore companies. Unlike in a typical indirect discrimination case, the Commission also could not point to any classification in the Gibraltar law that clearly proxied a selective class. The Gibraltar regime was ingenious—it was engineered to exempt offshore companies as applied, but not only was it facially neutral, it also employed no overt proxy classifications. It achieved its goal to exempt offshore companies by taxing factors that offshore companies tended not to possess in Gibraltar—namely, payroll and property.

Tax-expenditure analysis provides no traction in a case like Gibraltar because the state used neither a selective class nor a proxy class. Lacking a preexisting doctrinal rule to defeat the proposed Gibraltar regime, the Commission invented a new one. It asserted that the Commission was not required to measure state aid as a deviation from a domestic-law reference-law base. According to the Commission, Gibraltar’s facially neutral rule constituted illegal state aid because it would benefit a selective class (offshore companies) as applied.129See id. ¶¶ 14, 20–22.

The lower EU court vacated the Commission’s decision on appeal because the Commission failed to use the traditional tax-expenditure approach.130The Court of First Instance, now known as the General Court, is the European Union’s lower court; its decisions are appealable to the Court of Justice of the European Union, the European Union’s highest court. The court held that EU law required the Commission to show that Gibraltar deviated from its own domestic law to advantage a selective class.131See Joined Cases T-211/04 & T-215/04, Gibraltar v. Comm’n, ECLI:EU:T:2008:595 ¶ 143 (Dec. 18, 2008). According to the lower court, dispensing with the domestic-law reference base would invade Member State competences; it would “enable the Commission to assume the role of the Member State with regard to determination of that State’s tax system.”132Id. ¶ 145. The court noted that states have the power “to devise systems of corporate taxation which they consider the best suited to the needs of their economies,” and the state-aid rules do not “prejudice . . . the power of the Member States to decide on their economic policy and, therefore, on the tax system . . . they consider the most appropriate and, in particular, to spread the tax burden as they see fit across the different factors of production.”133Id. ¶ 146.

On appeal, however, the Court of Justice sided with the Commission to overrule the lower court. The CJEU agreed with the Commission that, as applied, the proposed Gibraltar regime would favor offshore companies, a selective class for state-aid purposes.134Again, the European Union does not use the term “selective class;” I use it for the convenience of U.S. readers. The Court further accepted that the Commission could show selectivity not only by establishing a deviation from a domestic-law reference base but also by looking to the practical effect of the regime.135Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 ¶¶ 85–110 (Nov. 15, 2011). The high court emphasized that relying exclusively on domestic-law deviations to identify tax advantages could create loopholes in state-aid enforcement because a facially neutral regime could achieve the same result as a facially discriminatory regime “by adjusting and combining the tax rules in such a way that their very application results in a different tax burden for different undertakings.”136Id. ¶ 93. The CJEU seemed to be asserting that facial neutrality could not immunize intentional discrimination.137See Mason & Parada, supra note 116 (considering the role of legislative intent in EU tax discrimination cases).

In addition to evaluating the likely impact of the proposed rule on offshore companies, the Court of Justice asserted that the Commission had, in fact, employed a reference-base approach in Gibraltar. Instead of a domestic-law reference base, however, the Commission devised a new benchmark for establishing selectivity, namely whether as applied the regime deviated from its apparent objective. The CJEU observed that despite being “founded on criteria that are in themselves of a general nature, in practice [the proposed regime] discriminates between companies which are in a comparable situation with regard to the objectiveof the proposed tax reform, namely to introduce a general system of taxation for all companies established in Gibraltar.”138Gibraltar, ECLI:EU:C:2011:732 ¶ 101 (emphasis added).

According to the Court, because the proposed regime would not generate tax for “all companies,” it did not achieve its objective, and it therefore constituted illegal state aid.139Id. ¶¶ 100–01. We can think of this analysis as evaluating the anticipated impact of the proposed Gibraltar regime against an external reference base consisting of “a general system of taxation for all companies.”140Id. ¶ 101. Although neither the Commission nor CJEU was explicit about what that tax base might look like, they presumably had in mind a more typical comprehensive corporate income tax base of the type seen in other Member States.

B.   Facially Selective Rules

In my view, Gibraltar posed the question of whether facially neutral rules that do not employ proxy classifications nevertheless can confer tax advantages that warrant state-aid scrutiny. Other recent cases—including Apple—posed essentially the inverse question: can facially selective rules that confer advantages nevertheless avoid conferring state aid? Under the traditional tax-expenditure approach, derogating tax provisions that benefit a selective class are state aid unless justified. But, as this Subsection explains, there are good reasons for states to write and enforce tax rules that distinguish between members of certain selective classes. The most obvious example is income-allocation rules; rules that determine which part of a multinational’s income will be taxed by a particular state. Such rules regularly distinguish between resident and nonresident taxpayers, but they do so by necessity. Facial selectivity, therefore, does not always signal state aid. Although this observation rings true, traditional tax-expenditure analysis cannot separate out facially selective rules that do not confer state aid from those that do. Because it could not resolve recent income-allocation cases using traditional tax-expenditure analysis, the Commission turned to external benchmarking.

1.      Transfer pricing

The inability of traditional tax-expenditure analysis to separate problematic uses of selective classes from nonproblematic ones can be seen in controversial recent cases, which involved a particular kind of allocation rule, namely, arm’s-length transfer pricing.141See infra note 154 and accompanying text; see also Avi-Yonah, supra note 19, at 105 (describing the rise of arm’s-length transfer pricing).

To understand the recent cases, we need some background on international tax. Rules for taxing cross-border income often employ selective classifications—particularly classifications based on tax residence, geographic location of economic activities, and so on. States need these classifications to accommodate differences in their jurisdiction to tax resident versus nonresident taxpayers. International law permits states to tax resident companies on their worldwide income, but states may tax nonresidents only on income “sourced” in their territory.142See, e.g., Restatement (Fourth) of Foreign Relations Law § 411 (Am. Law Inst. 2018). Whether a company resides in a state depends on whether the company meets the state’s definition of corporate tax residence. As the Apple tax-residence mismatch shows, there is no global definition of tax residence; this lack of consensus can lead to dual-resident companies or companies that reside nowhere. Disputes have arisen over the proper way to source income; it is sufficient for our purposes to observe that although many source rules are widely accepted, others are not, and there is no agreed normative framework for determining source.143See, e.g., 2008 JCT Report, supra note 6, at 10 (concluding that the pre-2017 “general rule for taxing foreign earnings . . . can fairly be said to be ambiguous”); J. Clifton Fleming, Jr. & Robert J. Peroni, Reinvigorating Tax Expenditure Analysis and Its International Dimension, 27 Va. Tax Rev. 437, 528–61 (2008) (scoring U.S. income-allocation rules as tax expenditures (or not) against various benchmarks, including a normative ability-to-pay benchmark of the author’s own devising, a reference-law benchmark consisting of the tax treatment of similar domestic income, and a normative anti-deferral benchmark). Kane, supra note 89, at 314–15, 317–18 (arguing that source rules make practical sense but do not follow clear economic principles); Wolfgang Schön, State Aid in the Area of Taxation, in EU State Aids 321, 323–24, 333 (Leigh Hancher et al. eds., 2016) (discussing the absence of a common international tax system to form a baseline for transfer pricing).

Reflecting limits on tax jurisdiction, income-allocation rules—including “source” rules, apportionment rules, methods of relieving double taxation, and so on—typically distinguish between residents and nonresidents or between domestic and foreign income.144See, e.g., Brian J. Arnold, Future Directions in International Tax Reform, 5 Austl. Tax F. 451, 452 (1988) (describing the interplay between resident and nonresident allocation schemes in the international context). Such distinctions constitute selective classifications for state-aid purposes. Moreover, income-allocation rules often seem to favor selective classes. For example, nonresidents are subject to a narrower tax base than are residents, and some types of income that would be taxed if earned domestically may be exempt when earned abroad.145See I.R.C. § 245A (2017) (exempting certain foreign dividends).

Transfer-pricing rules for determining the income of entities that are part of a multinational group likewise may employ selective classifications. Vertically integrated multinationals have affiliates and business activities in many states, and each state must determine how much of the group’s income it should tax. There exist many theoretical frameworks for taxing the cross-border income of multinationals, but separate accounting paired with arm’s-length transfer pricing dominates the world.146In contrast, some subnational jurisdictions, including the U.S. states and the Canadian provinces, use unitary taxation with formulary apportionment. I will discuss formulary in infra Section IV.A. Under separate accounting with arm’s-length transfer pricing, states separately determine the income of each taxable unit of the business (typically, each domestic corporation and/or each taxable branch).

Related companies can secure advantages under separate accounting by manipulating which entity in the group is deemed for tax purposes to earn the integrated enterprise’s income. For example, a group can shift income from a subsidiary in a high-tax country to a subsidiary in a low-tax country by causing the high-tax subsidiary to buy goods or services from the low-tax subsidiary at an artificially inflated price. The high-tax subsidiary gets a larger expense deduction, and the low-tax subsidiary has a larger income inclusion. The group’s overall profit remains the same, but it saves taxes due to the rate differential between the two countries.

Moreover, companies may do better than lowering their tax rates. By exploiting differences between states’ definitions of tax residence, they may be able to escape tax entirely. For example, under laws applicable at the time the facts of Apple arose, Ireland regarded certain companies to be tax resident in Ireland only if they were managed and controlled in Ireland.147Apple, supra note 2, ¶ 7 n.12 (describing old and new Irish tax-residence rules). Ireland has since changed its law to deem as an Irish resident any company incorporated in Ireland, but not regarded by any other state as resident there by virtue of management and control. See Revenue Commissioners, Company Residency Rules, https://www.revenue.ie/en/companies-and-charities/corporation-tax-for-companies/corporation-tax/company-residency-rules.aspx [https://perma.cc/D3ZD-23GU]. In contrast, the United States long has used a place-of-incorporation rule for tax residence.148I.R.C. § 7701(a)(30). By incorporating subsidiaries in Ireland that Apple managed and controlled in the United States, Apple was able to create companies that resided nowhere for tax purposes.149Offshore Profit Shifting and the U.S. Tax Code—Part 2 (Apple Inc.): Hearing Before the Permanent Subcomm. on Investigations of the S. Comm. On Homeland Sec. & Governmental Affairs, 113th Cong. 1–4, 6–7 (2013) (detailing the U.S., European, and Member State tax rules that Apple exploited to effectuate this profit shift). By shifting profits to these nowhere companies, Apple escaped tax on them.150See id. at 42.

The arm’s-length standard counters abusive profit-shifting by requiring both group members to report a market, or arm’s-length, transfer price on the transaction. The arm’s-length standard has faced significant and warranted criticism, but lack of politically viable alternatives has led to its widespread adoption.151See Avi-Yonah, supra note 19, at 28–29. Although the idea of a market price may seem straightforward, determining transfer prices in related-party transactions and allocating income among related companies is no simple matter. Over decades, states have cooperated through the OECD to develop sophisticated and voluminous guidance on acceptable ways to determine income in related-party settings.152Of the twenty-eight EU members (including the United Kingdom), twenty-three are among the thirty-six OECD Member States. Compare Where: Global Reach, OECD, http://www.oecd.org/about/members-and-partners [https://perma.cc/L6 BY-4TQW] (listing OECD Member countries) with Countries, Eur. Union, https://europa.eu/european-union/about-eu/countries_en [https://perma.cc/K9 8X-F3LV]. Memorialized in the OECD Transfer Pricing Guidelines and other OECD reports,153OECD Comm. On Fiscal Affairs, OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations (2017) [hereinafter 2017 OECD Transfer Pricing Guidelines]. most countries, including non-OECD countries, follow this guidance. Typically, states incorporate into domestic law both the arm’s-length standard and OECD guidance for interpreting the arm’s-length standard, but each state modifies the standard and guidance as it sees fit.154See Commission Decision 2017/502 of Oct. 21, 2015 on State Aid Implemented by the Netherlands to Starbucks, 2017 O.J. (L 83) 38, ¶¶ 409–12 [hereinafter Starbucks] (reporting that the Netherlands incorporated the OECD arm’s-length standard into domestic law); Commission Decision 2016/2326 of Oct. 21, 2015 on State Aid Luxembourg Granted to Fiat, 2016 O.J. (L 351) 1, ¶¶ 219–311 [hereinafter Fiat] (reporting that Luxembourg incorporated the OECD standard into domestic law); Amazon, supra note 13, ¶ 129 (same). In contrast, Ireland did not expressly endorse the arm’s-length standard for allocating income to Irish branches of nonresident companies. See Apple, supra note 2, ¶ 152. The Commission argued that arm’s-length was Ireland’s de facto standard. Id. ¶¶ 369–78.

Application of the arm’s-length standard is notoriously complex, and it involves considerable discretion.155See generally Avi-Yonah, supra note 19 (providing an overview of the rise of the arm’s-length standard). Given this environment of legal uncertainty, it is no surprise that multinationals seek confirmation from governments about how to apply the standard to their particular facts. Many countries, including the United States, supply this confirmation in the form of administrative rulings.156See IRS, Rev. Proc. 2015-41, Procedures for Advance Pricing Agreements (2015). Such rulings are usually prospective and cover several years—they describe the taxpayer’s situation and how the taxpayer intends to report income.157See id. at 2 (noting that advance pricing agreements can apply to past or “rollback” years). Because arm’s-length determinations require taxpayers to disclose sensitive business information, they are confidential, although they may be forwarded to tax administrators in other countries in which the multinational has activity.158See IRS, Announcement and Report Concerning Advance Pricing Agreements (2018), https://www.irs.gov/pub/irs-drop/a-18-08.pdf [https://perma.cc/MKT2-L6S9].

Ruling procedures are generally open to any taxpayer, and the Commission acknowledges that rulings usually constitute a legitimate means of increasing legal certainty.159Commissioner Vestager Speech to TAXE, supra note 19, at 2. Although it is valuable, legal certainty is not a relevant advantage for state-aid purposes. But rulings can do more than confirm the application of laws to the taxpayer’s specific facts. Due to their obscurity, the complexity of the laws they apply, their confidentiality, and their application to only a single taxpayer, tax rulings represent an ideal mechanism for governments to deliver benefits to a favored taxpayer while denying similar treatment to the taxpayer’s competitors. When such rulings are secret and unilateral—and all the recent state-aid cases involved secret unilateral rulings—states can use them to impose different, and more favorable, tax rules than those available under domestic law. Disclosure in 2014 by whistleblowers of a large cache of Luxembourg’s secret unilateral rulings for multinationals, an event known as LuxLeaks, seemed to reveal that Luxembourg deviated from its own laws and guidance in secret rulings.160See Omri Marian, The State Administration of International Tax Avoidance, 7 Harv. Bus. L. Rev. 1, 3 (2017) (“Luxembourg ignored its own administrative guidance, binding intergovernmental legal procedures, and well-established principles of international tax law.”). The leak ultimately prompted the Commission to demand that every Member State submit for Commission inspection all tax rulings issued from 2010 to 2012.161Background to the High Level Forum on State Aid of 3 June 2016 ¶ 6 (DG Competition, Working Paper).

2.      Apple and other transfer-pricing cases

Following inspection of these rulings, the Commission initiated several cases against small EU Member States for rulings they granted to large, mostly U.S., multinationals. The Commission found that tax rulings by Ireland for Apple, by the Netherlands for Starbucks, and by Luxembourg for Amazon and Fiat-Chrysler all conferred state aid.162Apple, supra note 2, ¶ 452; Starbucks, supra note 154, ¶ 450; Amazon, supra note 13, ¶ 616; Fiat, supra note 154, ¶ 371. It also found that Luxembourg aided French energy giant Engie. European Commission Press Release IP/18/4228, Commission Finds Luxembourg Gave Illegal Tax Benefits to Engie; Has to Recover Around €120 Million (Jun. 20, 2018). Cases involving IKEA and Nike are pending. European Commission Press Release IP/17/5343, State Aid: Commission Opens In-Depth Investigation into the Netherlands’ Tax Treatment of Inter IKEA (Dec. 18, 2017). The essential question in all four cases was the same: did the Member State illegally favor the company when determining what portion of its income would be taxable by the state? The Commission’s answer in all four cases was “Yes.”

In one strand of reasoning in these cases, the Commission employed its traditional tax-expenditure approach, under which a state confers aid only if it deviates from domestic law to favor a selective class. For example, the Netherlands had incorporated the OECD arm’s-length standard and Transfer Pricing Guidelines as its own domestic standard for allocating income. The Commission in Starbucks,163Starbucks, supra note 154. however, found that the Netherlands deviated from that standard to relieve Starbucks of taxes.164Id. ¶¶ 409–12. The Commission took exactly the same analytical approach in Amazon and Fiat,165Fiat, supra note 154. the two cases involving Luxembourg.166Amazon, supra note 13, ¶ 599; Fiat, supra note 154, ¶¶ 315–17. To this extent, all three decisions employed a traditional tax-expenditure approach. All three cases involved facial selectivity—the states deviated from their regularly applicable law to favor a particular company, and particular companies are a selective class.

Starbucks, Amazon, and Fiat all involved transfer-pricing rulings that determined the taxable income of a resident corporation. In contrast, Apple involved rulings that determined how much income to allocate to the Irish branch of one of Apple’s Irish-incorporated, but nowhere-resident subsidiaries. Because Ireland had not adopted into its domestic law the OECD arm’s-length standard to allocate income to Irish branches of foreign companies, it was unclear whether it was appropriate for the Commission to evaluate the Apple rulings against that standard. Ireland asserted that the Commission was required to evaluate the rulings against a benchmark consisting only of Irish law, and Ireland furthermore insisted that it had followed its own law, which required allocation of “taxable profits commensurate with the value of the contribution made by the Irish branch in each case to overall company profitability.”167Apple, supra note 2, ¶ 159. Irish domestic law provides that nonresident companies are subject to tax only on Irish-source income, unless they have an Irish branch, in which case, the branch is taxable on “all its chargeable profits wherever arising.” Id. ¶ 73. An Irish branch’s chargeable profits are “any trading income arising directly or indirectly through or from the branch or agency, and any income from property or rights used by, or held by or for, the branch or agency, [not including] distributions received from [resident] companies.” Id. Beyond that, Ireland’s descriptions of its attribution rule were vague. Ireland stated that its attribution rule was fair, consistent, appropriate, and not the result of a bargaining process. Id.¶¶ 157–58. Ireland reasoned that since the Irish branches contributed only a small amount to Apple’s global profitability, it was appropriate and consistent with Irish law to allocate only a small amount of Apple’s global profit to Ireland.168Ireland argued that not only was it not obliged to tax the billions of profits made by Apple’s nowhere-resident subsidiaries, but that it had no authority under Irish law to tax that profit. Ireland’s obligation, in its view, was limited to determining and taxing the profits of the Irish branch of the nowhere companies (and therefore it had no responsibility to determine, much less tax, the companies’ total profits). Id. ¶¶ 276–308.

The Commission advanced three alternative accounts as to how Ireland illegally aided Apple. First, it argued that, notwithstanding Ireland’s protestations to the contrary, arm’s length was—de facto, if not de jure—the standard Ireland used to allocate income to branches.169After examining of all of Ireland’s branch-profit allocation rulings over ten years (amounting to only eleven such rulings), the Commission determined that Ireland consistently accepted income allocations to branches that relied on the OECD arm’s-length standard. Id. ¶¶ 371–72, 372 n.310. Armed with this de facto reference-law benchmark, the Commission could use traditional tax-expenditure analysis to find that Ireland deviated from its own law to favor Apple.170See id. ¶ 377. Under this approach, Ireland—like the Member States in Starbucks, Amazon, and Fiat—violated its own domestic allocation rules to favor an individual company, a selective class for state-aid purposes.171See id. ¶¶ 361–78. On this account, all the recent cases fit easily with the Commission’s prior decisions.

The Commission’s second theory was that Ireland had no pre-determined allocation rules; it just made special deals with taxpayers.172See id. ¶¶ 396–403 (highlighting Ireland’s inconsistent application of allocation rules). Such tax negotiation would be state aid under long-standing doctrine because it would allow Irish officials too much discretion.173Permitting excessive administrative discretion can lead to adverse state-aid decisions under long-standing guidance. 2016 Notice, supra note 1, ¶¶ 123–25.

But what if, as Ireland claimed, Ireland had income-allocation rules other than the arm’s-length standard, and what if, as Ireland also claimed, it applied those rules uniformly to all taxpayers, including Apple? The ruling for Apple was technically selective because tax rulings apply only to a particular company, and particular companies constitute a selective class. And because states’ tax jurisdiction over nonresidents is limited compared to their jurisdiction over residents, nonresidents typically face less tax than residents; this difference technically may qualify as a tax advantage under tax-expenditure analysis. But selectivity in income-allocation rules, even when paired with advantageous tax treatment, does not seem to be enough to make a reliable finding that a state conferred the type of tax advantage that the state-aid rules ought to care about. Thus, unlike in the easy cases discussed in Part II, the presence of a selective tax advantage in the income-allocation cases cannot be dispositive.174See supra Section II.B.1.

But to simply approve of the result of Apple’s tax plan—tens of billions of income in an Irish-incorporated company that would face no current tax anywhere in the world—must have seemed untenable to the Commission. Apple and Ireland argued that the income eventually would be taxed by the United States when distributed as dividends to Apple’s U.S. parent,175The Irish-incorporated Apple subsidiaries would trigger this tax by distributing profits to their U.S. parent. Because the timing of the U.S. tax was, at the time of the case, controlled entirely by Apple, Apple could indefinitely defer it. Since the Commission’s final decision in Apple, U.S. law changed so that rather than indefinite deferral, previously retained foreign profits will be taxed whether distributed or not. See I.R.C. § 965 (2017). but no one knew when that would be, and in the meantime, if Ireland did not tax Apple, no one would.

To resolve the dilemma, the Commission advanced its controversial third theory for how Ireland violated the state-aid rules in Apple. The Commission concluded that the state-aid rules themselves require all states to allocate income according to the arm’s-length standard, regardless of domestic law.176See Apple, supra note 2, ¶ 255. In other words, the Commission used arm’s length as an external benchmark from which to measure the advantage Ireland conferred to Apple.

Earlier, I described benchmarks for tax-expenditure analysis as falling into several categories.177See supra Section II.A. Any of the following could serve as the benchmark for measuring tax expenditures: reference law, idealized tax policy views, or commonly accepted standards, such as financial accounting rules. Supporting the notion that it saw arm’s-length as a normative standard, the Commission asserted that the standard applied in all cases as a matter of EU law.178Starbucks, supra note 154, ¶¶ 264–65; Fiat, supra note 154, ¶ 229; Apple, supra note 2, ¶¶ 255–57. The Commission even left open the possibility that a ruling that complied with OECD guidance nevertheless could violate the EU arm’s-length standard.179See, e.g., 2016 Notice, supra note 1, ¶ 173 (noting that rulings complying with the OECD Transfer Pricing Guidelines are “unlikely to give rise to State aid”).

At the same time, the Commission also used language that suggested it applied the arm’s-length standard because it viewed arm’s length as a dominant state practice.180Apple, supra note 2, ¶ 322 (internal citations and footnotes omitted) (noting that, aside from OECD guidance, “no other alternative detailed and comprehensive analyses on methods of attributing profits are available to tax administrations and multinational enterprises to assist them in establishing arm’s length conditions”); see also 2016 Notice, supra note 1, ¶ 173 (explaining that the OECD “guidelines do not deal with matters of State aid per se, but they capture the international consensus on transfer pricing and provide useful guidance to tax administrations and multinational enterprises on how to ensure that a transfer pricing methodology produces an outcome in line with market conditions”). Additionally, it fleshed out the standard by reference to OECD guidance. In Apple, it used both OECD guidance that was in effect at the time Ireland rendered the Apple rulings and OECD guidance issued thereafter. The Commission thus retroactively applied modern OECD guidance to a country that had adopted into its domestic law neither the OECD guidance itself nor the arm’s-length standard it modified.181Apple, supra note 2, ¶¶ 88–89, 272–73, 323 (relying heavily on the 2010 OECD branch profits attribution report, even though Ireland granted the later of the two contested rulings in 2007); see also Starbucks, supra note 154, ¶¶ 174, 177 (citing both the 1995 OECD Transfer Pricing Guidelines, which the Netherlands had incorporated into domestic law, and the 2010 OECD Transfer Pricing Guidelines, which post-dated the tax rulings at issue in the case).

Using this external benchmark for allocating income, the Commission concluded that all the income that Apple had booked in its nowhere-resident subsidiaries should have been taxed by Ireland as part of the profits of those subsidiaries’ Irish branches. That interpretation led to the counter-intuitive result that half of Apple’s global profit was Irish.182Apple, supra note 2, ¶¶ 305–07. Apple’s aggressive tax planning drove this counter-intuitive result because Apple, believing that the profits of its nowhere subsidiaries would not be taxed, shifted as much of Apple’s global profit to those subsidiaries as possible. That is why the tax bill in Apple was so high. In contrast with the Irish approach, which focused only on the contribution of the Irish branches to Apple’s global profitability, the Commission reasoned that because the head offices of the nowhere subsidiaries (of which the Irish branches formed part) had no substance, whereas the Irish branches had substance, all the profit of the nowhere subsidiaries had to be allocated only to the branches, and none of the profit could be allocated to the head offices. See id.¶¶ 276–307. It is not a goal of this paper to analyze the correctness of the Commission’s application of the arm’s-length standard to Apple. Rather, this Article concerns the Commission’s legal justifications for applying that standard in a state-aid case.

Perhaps to shore up its claims in Apple about the state-aid arm’s-length standard, the Commission claimed in its nearly contemporaneous decisions in Starbucks, Amazon, and Fiat that, in addition to violating their own law, the Netherlands and Luxembourg also violated the EU arm’s-length standard that applied regardless of domestic law.183Fiat, supra note 154, ¶ 228 (“The arm’s length principle therefore necessarily forms part of the Commission’s assessment under Article 107(1) of the TFEU of tax measures granted to group companies independently of whether a Member State has incorporated this principle into its national legal system.”); see Amazon, supra note 13, ¶¶ 402–08 (establishing the arm’s-length principle as a requirement under state-aid law); Starbucks, supra note 154, ¶ 245 (“The arm’s length principle therefore necessarily forms part of the Commission’s assessment under Article 107(1) of the Treaty of tax measures granted to group companies independently of whether a Member State has incorporated this principle into its national legal system”); id. (the state-aid arm’s-length standard “is not that derived from Article 9 of the OECD Model”). The Commission regards only its own arm’s-length standard as producing a state-aid compatible “market-based outcome.” See generally id. In each of these cases, the Commission argued that the state’s application of what the Commission regarded as a deviating arm’s-length standard was selective because it unjustifiably favored a selective class. Either it constituted individual aid to a particular company, which was selective per se,184See Amazon, supra note 13, ¶ 399; Apple, supra note 2, ¶ 244; Starbucks, supra note 154, ¶ 254; Fiat, supra note 154, ¶ 218. or it favored multinational companies engaged in cross-border commerce over domestic companies engaged in domestic commerce.185Amazon, supra note 13, ¶ 599; Starbucks, supra note 154, ¶ 236; Fiat, supra note 154, ¶¶ 198–209.

3.      Belgian Excess Profits

Like the tax rulings cases, Belgian Excess Profits186Commission Decision 2016/1699 of Jan. 11, 2016 on the Excess Profit Exemption State Aid Scheme S.A. 37667 Implemented by Belgium, 2016 O.J. (L 260) 61 [hereinafter Belgian Excess Profits]. involved a facially selective regime. The Belgian regime exempted from tax so-called excess profits earned by multinational corporate groups. Excess profits were those deriving from “synergies, economies of scale or other benefits drawn from . . . participation in a multinational group and which would not exist for a comparable standalone company.”187Id. ¶ 14. The exclusion applied to the Belgian portions of all multinational groups, whether those groups were headquartered in Belgium or not, but it did not apply to domestic groups.188Id. ¶ 13 (describing the benefits as applying to Belgian companies or Belgian permanent establishments that were part of a multinational group). Despite Belgium’s protests that the excess-profits regime should be regarded as part of its domestic-law reference base,189Id. ¶ 123. and that the exclusion ensured that multinationals would be taxed—like domestic companies—only on arm’s-length profits,190According to Belgium, “the rationale for the Excess Profit exemption is to ensure that a Belgian group entity is only taxed on its arm’s length profit.” Id. ¶ 14. the Commission took a normative approach. It claimed that the “objective of the Belgian corporate income tax system is to tax all corporate taxpayers on their actual profits, whether they are a standalone or group company.”191Id. ¶ 126. Having identified “tax[ing] . . . actual profits” as Belgium’s policy objective, the Commission concluded, as it had in prior cases, that the state-aid specific arm’s-length method of allocating income formed a necessary part of the reference system, regardless of whether Belgium adopted it into domestic law.192Id. ¶¶ 126, 150. The Commission concluded that the excess-profits regime conferred tax advantages by deviating from this purposive benchmark.193Id. ¶¶ 132–33. Belgium’s expressed policy preference as to taxability of excess profits was simply irrelevant under the benchmark the Commission chose.

C.   Hard-to-Classify Rules

Some rules may be hard to classify as either facially selective of facially neutral, making it particularly difficult under tax-expenditure analysis to determine whether the state has conferred state aid. This Subsection provides two examples.

1.      McDonald’s

McDonald’s194Commission Decision of Sept. 19, 2018 on Tax Rulings in Case SA.38945 (ex 2015/NN) (ex 2014/CP) Granted by Luxembourg in Favour of McDonald’s Europe, 2018 O.J. (L 195) 20, http://ec.europa.eu/competition/state_aid/cases/261647/ 261647_2033697_264_2.pdf [https://perma.cc/4D8N-YRK9] [hereinafter McDonald’s Final Commission Decision]. was the only recent case against a U.S. multinational in which the Commission found no state aid.195Id. It involved a tax-treaty question. Under the U.S.-Luxembourg tax treaty, Luxembourg undertook an obligation to exempt income that, under the treaty, “may be taxed in the United States.”196Id. ¶ 60. The case involved the U.S. branch of a Luxembourg subsidiary of the U.S. franchise giant McDonald’s.197Id. ¶¶ 20–22. As is typical under tax treaties, the United States could tax the branch of a Luxembourg company only if the company had a nexus and business profits in the United States, as defined in the tax treaty and under U.S. law.198Id. ¶¶ 111–23 (the conflict involved interpreting the tax-treaty terms “permanent establishment” (i.e., nexus), “business” and “business profits”).

The tax treaty directs a state to interpret unclear treaty terms by referring to the law of the state applying the treaty.199Id. ¶ 113 (referring to Article 3(2) of the U.S.-Luxembourg tax treaty). Following this prescription, the United States concluded that the branch did not meet U.S. nexus and business-profits conceptions, and, consequently, the United States could not tax the branch.200Id. ¶¶ 39–41, 46, 66­–67, 73. In contrast, when Luxembourg considered the same question, it came to a different conclusion. Luxembourg interpreted the tax treaty terms by reference to its own law.201Id. ¶¶ 34, 38. Since U.S. law and Luxembourg law differed on the meaning of the terms used in the tax treaty, the United States concluded it could not tax the branch, but Luxembourg, applying its own domestic-law interpretations of these terms, concluded that the United States could have taxed the branch.202Id. ¶ 114.

In reality, the United States did not tax the branch, and Luxembourg knew the United States would not tax it.203Id. ¶ 74. Nevertheless, Luxembourg took the position that because the branch’s income “may be taxed in the United States,” the tax treaty obliged Luxembourg to exempt the branch’s income.204State Aid SA. 38945 (2015/C) (ex 2015/NN) Luxembourg Alleged Aid to McDonald’s ¶ 84 (Dec. 12, 2015) [hereinafter McDonald’s Opening Commission Decision] (explaining that in its ruling for McDonald’s, Luxembourg accepted McDonald’s argument that “since the US Franchise Branch constitutes a PE under Luxembourg tax law, the profits attributed to it should be exempt from Luxembourg corporate income tax, irrespective of whether it also constitutes a PE under US tax law”). The result, which involved a known tax-planning technique, was that the branch—to which McDonald’s had shifted as much income as possible—was taxed by neither Luxembourg nor the United States.205Id. ¶ 85 (Dec. 12, 2015).

Luxembourg interpreted its treaty in a way that resulted in nontaxation. It presumably did so deliberately to attract business to Luxembourg. Other countries might have interpreted the same language not to require exemption.206See Fadi Shaheen, Tax Treaty Aspects of the McDonald’s State Aid Investigation, 86 Tax Notes Int’l 331, 340 (2017) (arguing in part that because the relevant paragraph of the treaty begins “in Luxembourg double taxation shall be eliminated as follows,” Luxembourg was obliged to exempt only in cases where failure to do so could result in double taxation; since the U.S. branch of McDonald’s Luxembourg subsidiary was not taxed by the United States, there would be no double taxation, and therefore the treaty did not require Luxembourg to exempt). Both interpretations are plausible.207For discussion of the controversy, see generally Klaus Vogel on Double Taxation Conventions § 1.1.2.4 (Ekkehart Reimer et al. eds., 4th ed. 2015). Some commentators would argue that Luxembourg’s interpretation was unreasonable. See, e.g., Shaheen, supra note 206, at 340. What matters for the analysis in this Article is not whether Luxembourg’s interpretation was correct, but that the Commission regarded it as part of the reference base.

While Luxembourg may not have used a facially selective classification or a proxy classification, tax-treaty interpretations by their nature apply only to cross-border situations. Thus, if they confer systematic benefits, those benefits will redound only to cross-border taxpayers, a selective class. But if multiple potential treaty interpretations are plausible, and the state chooses the one most favorable to cross-border taxpayers or to a particular taxpayer, does it thereby confer state aid?

Rather than apply either of the two new approaches introduced in the earlier Gibraltar case, in McDonald’s, the Commission hewed to its traditional reference-law approach and concluded that Luxembourg conferred no state aid. The Commission reasoned that Luxembourg’s reference law included both the tax treaty itself and Luxembourg’s reasonable interpretation of that tax treaty.208McDonald’s Final Commission Decision, supra note 194, ¶ 109 (“It is not established that the contested tax rulings constitute a derogation from the rules set by the double taxation treaty. Such a derogation would exist if the contested tax rulings misapplied (i.e. deviated from) a rule of the double taxation treaty . . . .”). In doing so, the Commission emphasized that there were twenty-five rulings for other companies in which the Luxembourg tax authorities took the same position, amounting to “a coherent interpretation.”209See id. ¶ 123. Since Luxembourg’s tax-treaty interpretation for McDonald’s did not derogate from Luxembourg’s reference law, the Commission concluded that Luxembourg conferred no aid, notwithstanding that McDonald’s paid no tax anywhere on the income of its branch.210See id. ¶ 124. The Commission concluded that the double nontaxation arose from the failure of the United States to “make use of its right to tax assigned to it under the double taxation treaty,”211Id. ¶ 117. This conclusion by the Commission calls into question whether it actually understood the legal background in the case. The United States did not regard itself as having authority to tax the income, which it then affirmatively exempted. Rather, the United States concluded that it had no authority to tax the income under the tax treaty. While the possibility that the Commission misunderstood the law is concerning, it does not affect the analysis here because I am using this case for the principle that the Commission announced in it, which was that a state’s tax-treaty interpretation forms part of its reference law. a gap for which Luxembourg could not be held responsible. In reality, of course, the United States was no more at fault than was Luxembourg; under its own interpretation of the tax treaty and U.S. law, the United States lacked authority to tax the income.

2.      Irish tax-residence rules

Finally, it is worth noting a claim that the Commission did not pursue in Apple. Although the Commission was well aware that differences in the U.S. and Irish definitions of tax residence could result in the formation of stateless companies, it did not pursue the Irish residence rules as potential state aid.212See Apple, supra note 2, ¶ 277 (“Commission’s investigation did not examine either the compatibility of [Ireland’s residence rules] with the State aid rules, or [the Apple subsidiaries’] lack of tax residency resulting from an application of . . . those rules.”). The Irish tax residence rule is hard to classify as either selective or not. As part of Ireland’s definition for tax residence, the provision was technically open to all, but it conferred benefits only in cross-border situations.213Id. ¶ 49 (providing Irish definition of corporate tax residence). The Commission did not explain why it did not challenge the Irish tax-residence rule, but perhaps it regarded the rule as part of the Irish reference law.

D.   Analysis of Recent Cases

This Part identified two types of subsidy cases in which tax-expenditure analysis that relies on a domestic-law reference base fails.

The first type of case involves a facially neutral rule. Tax-expenditure analysis that relies exclusively on domestic law for the reference base regards facially neutral rules as incapable of conferring state aid. But facially neutral rules that convey cross-border tax advantages should only survive state-aid scrutiny when the benefits they confer arise from mismatches—that is, differences between two states’ nonselective rules. Mismatches include differences in generally applicable tax rates, tax definitions, and tax bases.

Second, the Commission’s traditional tax-expenditure approach also fails when it labels facially selective tax laws as state aid even when they do not discriminate between cross-border and domestic commerce. For example, although income-allocation rules are often facially selective, it is not clear that it is appropriate to compare the tax treatment of nonresidents (over whom the state has limited tax jurisdiction) to residents (over whom the state has unlimited tax jurisdiction). Tax-expenditure analysis that uses a domestic-law reference base is incapable of distinguishing selective provisions that discriminate between cross-border commerce and domestic commerce and those that do not. For example, the transfer-pricing cases involving U.S. multinationals may have involved state aid, or they may not, but the Commission’s varying approaches in those cases and its lack of explanation for why its approaches varied inspire little confidence that it properly disposed of them. And to the extent that the outcomes in those cases depended on benchmarking by external norms, they provide little guidance as to how future cases might be resolved.

I also identified a third set of cases, which I described as difficult to classify. This set of cases further highlights the difficulties with an approach that requires the Commission to divide all tax rules into either “normal” rules or deviating provisions that benefit selective classes. For example, Luxembourg may have applied its tax treaty the same way to all taxpayers, but by their nature, tax-treaty questions only arise for cross-border taxpayers. Should we understand the tax treaty and its interpretation part of the normal base, or does the treaty deviate from domestic law? And if tax treaties and/or their interpretations are not part of the baseline, then what isthe baseline? Likewise, the Irish definition of corporate tax residence ostensibly applied to everyone, but the benefits of nowhere companies were only available to multinational companies, specifically, those managed outside Ireland. Should we consider the Irish tax-residence rule to be part of the baseline or a deviation from it? These difficulties highlight that concepts like facial selectivity or derogations from the normal base cannot resolve all state-aid cases.

By adopting a mode of analysis that requires it to be able to classify tax rules as either part of the normal tax base or derogations from the normal tax base, the Commission stumbled into the tax-expenditure trap. If there is one fundamental lesson that critics of tax-expenditure analysis have taught us, it is that what constitutes the normal tax law is often unclear. Although there may be significant agreement or even unanimity about some provisions, other provisions spur endless debate. Obviously, to the extent that it is debatable whether to classify a provision as a tax expenditure or not, state-aid analysis that depends on tax-expenditure analysis will generate results that are likewise debatable.

Indeed, the states’ main defense in all of the cases discussed so far has been that the Commission ought to regard the challenged rule as part of the normal reference base. If a rule is part of the normal base, then by definition under tax-expenditure analysis, it does not confer derogating advantages. Hungary will no doubt argue that variability of the rates in its turnover-tax regime must be regarded as part of the reference base; Gibraltar made similar arguments about its payroll-and-property tax.214Joined Cases T-211/04 & T-215/04, Gibraltar v. Comm’n, ECLI:EU:T:2008:595 ¶ 119 (Dec. 18, 2008) (according to Gibraltar the proposed property-and-payroll tax “constitutes a tax scheme in its own right, founded on the criteria of employment and property occupation, and is not a derogation from any profit-based tax scheme”). Ireland and the other states involved in the transfer-pricing cases all argued that their own, domestic transfer-pricing rules must be regarded as part of the baseline from which state aid should be measured.215Apple, supra note 2, ¶ 178; Amazon, supra note 13, ¶¶ 290–94; Fiat, supra note 154, ¶ 148; Starbucks, supra note 154, ¶ 183. Likewise, Belgium argued that its exemption of excess profits must be regarded as part of the reference base.216Belgian Excess Profits, supra note 186, ¶ 123.

The problem caused by the need to divide “normal” provisions from “special” provisions is particularly acute for structural tax provisions. Structural tax rules are background features of the tax system; they include the tax base (e.g., income versus consumption), the tax rate, the taxable unit, the taxable period, the depreciation rate, the realization rule, how the tax regime handles inflation, losses, income allocation, and so on. The classic example of a structural provision involves tax rates for individual taxpayers. Suppose you believe that a given tax system should be more progressive. You might consider its marginal tax rate regime to generate tax subsidies for high-income earners. Your view of the size of the subsidy delivered by the rate regime would depend on your view of the ideal rate structure. It is obvious that people will disagree about ideal tax rates, and therefore, a reasonable response to such disagreement might be to forgo trying to estimate the tax expenditures (or penalties) that progressive income tax rates impose.217See 2008 JCT Report, supra note 6, at 19 (observing that tax-expenditure analysts typically regard the marginal tax rates as part of the baseline). At the same time, however, it is equally obvious that governments can target benefits to select taxpayers by reducing their tax rates.218Congressional Budget Act of 1974, 2 U.S.C. § 622(3) (2006) (defining a tax expenditure to include “a preferential rate of tax”). The problem with the traditional tax-expenditure approach to state aid is that it requires the Commission to make arbitrary decisions about when to include structural rules in the reference base.

Structural provisions thus pose a special challenge for tax-expenditure analysis (and, in turn, for state-aid determinations that rely on tax-expenditure analysis). First, structural provisions may be difficult to identify. Second, even once identified, it is unclear what the Commission should do with them. As this Part showed, the Commission developed three approaches to handle structural rules; the Commission has (1) ignored structural provisions by automatically regarding them to be part of the reference-law benchmark; other times it (2) devised an external benchmark for determining whether they conferred tax advantages. As an alternative to tax-expenditure analysis, the Commission has (3) conducted an impact analysis to determine if, as applied, structural rules would benefit selective classes. Any subsidy adjudicator using tax-expenditure analysis, including the WTO, would face the same choices. As this Subsection explains, each approach has implications for the scope of subsidy control.

1.      Traditional tax-expenditure analysis

To date, most state-aid cases have involved overtly selective tax advantages for which states could offer no justification. Tax-expenditure analysis works well for such easy cases. It even works well in cases, like Vodafone, in which a state employs a facially neutral proxy that highly and obviously correlates with a selective classification. Instead of measuring the tax advantage as a favorable deviation from the normal tax base that benefits a selective class, in proxy cases, the Commission would measure the tax advantage as a favorable deviation from the normal tax base that benefits a proxy class.

Structural rules posed a new challenge to the Commission’s traditional approach. Under the traditional approach, which relied exclusively on domestic law to formulate the benchmark, the Commission would automatically incorporate structural tax provisions into the reference base. As such, the traditional approach would regard no structural provisions as subsidies. For example, the Commission implicitly took an automatic-incorporation approach towards Ireland’s tax-residence rule by not challenging it (despite being acutely aware of its existence).219See Apple, supra note 2, ¶ 277. Likewise, in McDonald’s, the Commission essentially incorporated Luxembourg’s tax-treaty interpretation into the reference base.220See supra Section III.C.1.

Unless we think that structural tax provisions are incapable of conferring subsidies that discriminate between cross-border and domestic commerce, the incorporation approach is underinclusive.221Structural provisions can convey tax benefits. See, e.g., 2008 JCT Report, supra note 6, at 10 (giving deferral of tax on foreign income under prior law and the debt-equity distinction as examples of structural rules that “materially affect economic decisions”); see also infra Section IV.C.3. (arguing that one-sided transfer-pricing methods are structural rules that convey cross-border advantages that violate competitive neutrality and therefore may constitute state aid). The Commission’s view that structural rules could convey state aid is precisely what drove it to adopt external-benchmarks in Gibraltar and Apple.

2.      External benchmarking

Benchmarking by external norms carries its own risks. The first problem with external benchmarking is that it is unpredictable. No one knows when the Commission will substitute its own benchmark for domestic law; nor does anyone know what the Commission will choose as its external benchmark.222Tax experts would have no trouble recognizing the benchmarks applied in recent cases. Comprehensive income taxation represents an idealized norm; it is a typical baseline for tax-expenditure analysis. See 2008 JCT Report, supra note 6, at 25–26 (distinguishing between reference-law and normative benchmarking). And arm’s length represents an international consensus standard. But the notion that widespread use or familiarity of such benchmarks entitles the Commission to use them (or makes it predictable that the Commission will use them) as benchmarks in state-aid cases requires further justification. In the rulings cases, the Commission specifically purported to apply a state-aid-specific arm’s-length standard, not the OECD arm’s-length standard, notwithstanding that it relied extensively on OECD guidance in elaborating the state-aid arm’s-length standard. See Ruth Mason, Whose Arm’s Length Standard?, 155 Tax Notes 947, 951 (2017). Finally, the OECD arm’s-length standard has been so thoroughly criticized that even if some aspects of it constituted customary international law binding on states, real questions would remain about the advisability of the Commission further entrenching it in state-aid law. Reuven Avi-Yonah, Altera, the Arm’s-Length Standard, and Customary International Tax Law, 38 Mich. J. Int’l L. OpinioJuris 1, 4 (2017) (arguing that arm’s-length constitutes customary international law). Even if the arm’s-length standard represents customary international law, it would not bind objectors. Restatement (Third) of Foreign Relations Law § 102 cmt. d (Am. Law Inst. 1987). To bind Ireland, at the time the facts of Apple arose, arm’s-length would have had to be customary international law for attributing profits to a permanent establishment. The first ruling in Apple was granted twelve years before the OECD had released its 2010 PE Profits Attribution Report, which report Ireland did not adopt into its law. Apple, supra note 2, ¶¶ 155, 385. As far as I know, no one argues that OECD reports are customary international laws that bind any country, much less that they bind objectors. Id. ¶ 79 (acknowledging that such guidance is “non-binding” and OECD members are “simply encouraged to follow the OECD’s framework”). For example, the Commission did not use external benchmarking in McDonald’s, but it easily could have. The Commission had at least two options for external bases in McDonald’s. In place of Luxembourg’s interpretive approach, the Commission could have substituted the OECD’s interpretive approach, which recommends taxation by Luxembourg in a case like McDonald’s.223See McDonald’s Final Commission Decision, supra note 194, ¶¶ 120–22 (referring to OECD commentary that recommends states avoid “double non-taxation” by interpreting the “may be taxed” predicate for exemption as satisfied only when the source state regards itself as entitled to tax, whether or not it does tax). As an alternative, as it had in Gibraltar, the Commission could have tried to determine the purpose of Luxembourg’s tax regime and used that as the reference baseline. The Commission hinted at this when it observed early on in its investigation that Luxembourg’s interpretation “seems neither faithful to the wording of [the provisions of the tax treaty] nor their objective,”224McDonald’s Opening Commission Decision, supra note 204, ¶ 86. which presumably was to eliminate double taxation, not to facilitate non-taxation.

Second, external benchmarking may erroneously mistake mismatches for state aid. If the main distinction between discrimination and mismatches is that a mismatch arises from the combination of two or more states’ different rules, whereas discrimination arises from only a single state’s laws, then McDonald’s involved a mismatch, and the Commission was correct to hold that Luxembourg did not confer state aid, despite the fact that its tax-treaty interpretation enabled McDonald’s to escape tax by any state. Thus, external benchmarking would have led the Commission astray in McDonald’s.

Third, external benchmarking also faces a problem of legitimacy. From where does the Commission derive the authority to devise the external baseline? In Gibraltar, the Commission appealed to the overarching objective of the proposed tax regime in order to formulate an external baseline consisting of “a general system of taxation for all companies.”225Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 ¶ 101 (Nov. 15, 2011). But a goal to tax “all companies” cannot be counted among Gibraltar’s intended purposes for the regime. Indeed, Gibraltar had gone to great lengths to construct a regime that would exempt certain companies. Thus, the Commission’s conclusion that the Gibraltar regime would generate inadequate liabilities must have rested on an implicit comparison by the Commission to some other possible regime that would generate greater liabilities. The requirement to tax “all companies” could represent a normative benchmark—that is, it could implicitly represent the view of the Commission and Court of Justice that states ought to impose tax regimes that tend to generate liability for most or nearly all companies.226Of course, tax regimes often fail to generate liability for all companies. For example, companies that earn a net loss typically have no current tax liability, and a state’s failure to tax companies operating at a loss has not been regarded as state aid. See 1998 Notice, supra note 63, ¶ 25. In the alternative, the baseline choice could reflect best practices; the Commission could have arrived at the baseline by observing that most Member States have comprehensive corporate income taxes, and so comprehensive corporate income taxation is a reasonable benchmark against which to judge Gibraltar’s regime. As the Commission did not explicitly describe the baseline against which Gibraltar failed to measure up, we are left to speculate. One thing is clear, however, such a norm cannot be derived from the proposed Gibraltar regime or from the TFEU. Similar questions could be posed about the arm’s-length standard, which the Commission claimed applied to the states even when they did not adopt the arm’s-length standard into domestic law.227See, e.g., Apple, supra note 2, ¶ 257. The Commission’s version of the arm’s-length standard is similar, but not identical, to the OECD arm’s-length standard.2282016 Notice, supra note 1, ¶ 173. Again, it is unclear where the Commission derives authority to impose a mandate on the states for how to allocate income.

When the Commission propounds an external norm, it invades reserved Member State tax authority. The TFEU provides a legislative procedure to adopt tax laws at the EU level; that procedure requires unanimous agreement of all the states.229See TFEU, supra note 1, art. 115. Fidelity to the Treaty demands that state aid not serve as a backdoor method of tax lawmaking in which the states not only lose the legislative veto the Treaty provides them, but in which they have no direct participation whatsoever.

Fourth, and closely related to the question of legitimacy, is the question of whether the Commission has the capacity to make the tax policy decisions inherent in external benchmarking. Tax policymaking is a complex affair with profound political and social implications.230See John R. Brooks, The Definitions of Income, 71 Tax L. Rev. 253, 290 (2018) (describing a variety of income tax bases and concluding “income . . . is not a pure concept, but is rather a constructed idea” that necessarily embodies policy, and therefore political, goals). Major tax policy questions, such as whether a state should tax on the basis of payroll and property, are not susceptible to Commission or judicial resolution. Even if it were generally agreed that tax policy decisions should be made at the EU level, for example, to save costs or reduce spillovers, there are more expert and more accountable bodies at the European Union that could do it.231The Commission’s Directorate-General for Taxation, not the Directorate-General for Competition that analyzes state aid, takes the leading role in drafting EU tax policies and legislation for consideration at the EU level. Tax legislation at the EU level is itself problematic from a democratic perspective because, although it requires consultation with European Parliament, it mainly involves the agreement of each state’s finance minister.

Fifth, it is unclear what justifies calculating monetary recoveries in two different ways under the two different benchmarks in state-aid cases. In cases involving facial selectivity, the state recovers from the taxpayer the difference between what the favored company actually paid and what it would have paid under domestic law absent the selective rule. In contrast, under an external benchmark, the state recovers the difference between what the favored company actually paid and what it would have paid under the Commission’s selected norm. Use of multiple baselines—even arbitrary baselines—may be acceptable in tax-expenditure budgeting because such budgets are meant only to convey information to lawmakers and voters about the distributive effects of tax policy choices.232Federal budget estimators meticulously report their baseline choices, enabling them to cabin thorny questions about ideal tax policies. In the most recent tax-expenditure budget, Treasury spent twenty pages reviewing how it estimated tax expenditures, carefully explaining its baseline choices. See Office of Mgmt. & Budget, Exec. Office of the President, Analytical Perspectives, Budget of the U.S. Government, Fiscal Year 2018 127 (2018). But the choice to use a baseline other than reference law is harder to accept in a context like state-aid enforcement that will generate liabilities for taxpayers. If the baseline is illegitimate because it is external to the challenged state’s law, then so is the recovery amount as measured by that baseline.

On the other hand, the Commission cannot simply say “anything goes” for structural rules, including income-allocation rules. Accepting that states have unrestrained autonomy to define their income-allocation rules would enable states to confer significant benefits to a selective class—multinationals—by adopting allocation rules that systematically under-tax multinationals’ income relative to that of domestic companies. It would, for example, mean affirming the regime in Belgian Excess Profits, despite the fact that the special benefits Belgium conferred to multinational groups compared to domestic groups were unrelated to Belgium’s need to allocate between itself and other states the income from cross-border economic activities. Likewise, there can be little doubt that Gibraltar intentionally arranged its tax system to favor offshore companies. Must the Commission accept this arrangement simply because Gibraltar cleverly avoided using overtly selective provisions that would be regarded as state aid under the traditional tax-expenditure approach? Apple put the Commission in a similar quandary, but for different reasons. Because selective advantages conferred by income-allocation rules may be motivated by limitations on tax jurisdiction, it would be overbroad to regard all facially selective rules that confer tax advantages as state-aid. That approach would imply, for example, that nearly all income-allocation rules as applied to nonresidents confer state aid. This follows from the fact that international law grants states fewer powers to tax nonresident, as opposed to resident, companies. As a result, the rules that apply to nonresident companies typically result in less tax than the rules that apply to resident companies.

The Commission therefore needs a reliable way to distinguish allocation rules that confer illegal state aid by violating competitive neutrality from those that do not. The problem with tax-expenditure analysis is that it provides no insight into how to distinguish structural rules that discriminate between cross-border commerce and domestic commerce from structural rules that do not do so.

3.      Impact analysis

That leaves impact analysis. Although Part IV argues that impact analysis can be effective when used in combination with the internal consistency test, without internal consistency, impact analysis is overinclusive. All mismatches disproportionately affect cross-border taxpayers for the simple reason that, by definition, mismatches arise only in cross-border cases. For example, impact analysis would lead to the conclusion that the Irish residence rule involved state aid since it benefitted exclusively cross-border cases.233As in Gibraltar, in McDonald’s the Commission could have reasonably inferred that the challenged state designed the regime to help multinationals avoid tax. But the Commission correctly declined to challenge the Irish residence rule because the benefits it conferred arose from a mismatch with U.S. law. Like external benchmarking, disproportionate-impact analysis cannot distinguish mismatches from provisions that discriminate between cross-border and domestic commerce.

4.      Justification

One might respond to objections of overbreadth by saying that they could be addressed in the justification phase of state-aid inquiries. For example, faced with an initial finding that its income-allocation rule disproportionately favored cross-border taxpayers, a state could argue that lighter tax for cross-border taxpayers simply reflects the state’s limited tax jurisdiction over those taxpayers and is therefore justified.

EU law places the burden to show a selective advantage on the Commission; it places only the burden of justification on the Member State. But adopting an overbroad method for identifying subsidies would tend to shift the burden of proof on selectivity to the Member State. The claim I am making in this Article is that at least some facially-selective rules and some rules that generate disproportionate impacts do not discriminate because they do not treat cross-border commerce relevantly differently from domestic commerce; they do not violate competitive neutrality. Thus, states should not have to justify them, which is important because neither the Commission nor CJEU has provided clear guidance on what it takes to justify selective tax advantages.234See supra note 97. For example, we have no guidance on how to distinguish between cases in which the lighter tax was attributable to narrower jurisdiction and cases in which the state deliberately favored outsiders by taxing them less than comparable insiders.

* * *

This Part has shown that tax-expenditure analysis is underinclusive. But external benchmarking and disparate-impact analysis are both overinclusive; they may erroneously regard as state aid Member State tax rules that confer benefits on cross-border taxpayers only through mismatches. To the extent that the Commission mistakes mismatches for state aid, it unnecessarily constrains tax diversity in the European Union and invades Member State tax competence. For example, to resolve the U.S.-Luxembourg tax-treaty mismatch, the Commission would have had to decide a contentious tax policy question—how to interpret the nexus requirement in tax treaties. Likewise, to resolve the U.S.-Irish tax-residence mismatch, the Commission would have had to decide a similarly contentious tax policy question—how to define corporate tax residence—that has no normatively correct answer and upon which states differ. Like the question of whether to impose a comprehensive income tax or a payroll and property tax, these questions are structural and relate to how to allocate taxable income among members of society and among the Member States, a task left unaddressed by the TFEU.235The CJEU has consistently avoided passing judgment on the content of states’ allocation rules on the grounds the TFEU provides no insight into income allocation. See supra Section I.A.

An important challenge for state-aid enforcement, then, is to distinguish discriminatory subsidies from perfectly legal cross-border tax advantages generated by nondiscriminatory mismatches. The principal difference between the two is whether the advantage arises from differences between different states’ laws or whether the advantage arises irrespective of other states’ laws.236See Case C-336/96, Gilly v. Directeur des Services Fiscaux du Bas-Rhin, ECLI:EU:C:1998:221 ¶ 2 (May 12, 1998). The former constitutes a mismatch that, absent EU legislation, must be tolerated, even if it creates distortions and intensifies tax competition. In contrast, cross-border tax advantages that arise irrespective of other states’ laws are discriminatory, and they may violate EU law as state aid.237See id.; see also Mason, supra note 21 (discussing tax disparities and how to distinguish them from discriminatory taxes that violate the fundamental freedoms). None of the Commission’s three approaches—reference-law benchmarking, external benchmarking, or impact analysis—is capable of distinguishing mismatches from potentially discriminatory provisions. The next Part argues that the Supreme Court’s internal consistency test can help.

IV.   Using Internal Consistency to Identify State Aid

This Part offers the internal consistency test as an alternative to tax-expenditure analysis for identifying tax subsidies that undermine competitive neutrality by treating cross-border and domestic commerce differently. The Supreme Court developed the internal consistency test to identify discriminatory state tax rules. The test has two advantages over prior approaches to state-aid analysis. First, it definitively identifies mismatches, preventing “false positive” cases in which the Commission mistakes mismatches for state aid. Second, it identifies tax rules, including structural tax rules, that violate competitive neutrality. Therefore, to the extent that the Commission and CJEU continue to enforce the prohibition of state aid to promote competitive neutrality, that is, to promote a level playing field between enterprises from different Member States, the test can make the Commission’s decisions more predictable and reliable.

A.   Internal Consistency Identifies Mismatches

This Subsection introduces the internal consistency test and shows that it accurately identifies cross-border tax mismatches. As a result, if the Commission adopted it, the Commission would cease mistaking mismatches for state aid.

The Supreme Court developed the internal consistency test to evaluate state income-allocation rules under the dormant Commerce Clause.238Much has been written on internal consistency. See, e.g., Walter Hellerstein, Is “Internal Consistency” Foolish?: Reflections on an Emerging Commerce Clause Restraint on State Taxation, 87 Mich. L. Rev. 138 (1988); Knoll & Mason (2017), supra note 23, Lirette & Viard, supra note 23; Mason, supra note 21; Adam B. Thimmesch, Comptroller v. Wynne and the Futile Search for Nondiscriminatory State Taxation, 67 Vand. L. Rev. En Banc 283, 283–84 (2014); Edward A. Zelinsky, The Enigma of Wynne, 7 Wm. & Mary Bus. L. Rev. 797, 797 (2016). Because the Constitution forbids tax discrimination, but does not require tax harmonization or single taxation, the Supreme Court needs a method to reliably distinguish cross-border tax disadvantages that arise from discrimination from those that arise from simple mismatches between two states’ different, but nondiscriminatory, tax laws.

Understanding the internal consistency test requires background on U.S. state income-allocation rules. Whereas countries use separate-entity accounting and the arm’s-length standard to allocate a multinational’s income among themselves, the U.S. states use unitary accounting with formulary apportionment to divide the income of a multistate business among the states where it is active. In contrast with separate accounting, formulary apportionment aggregates the income of a unitary business across all its entities and then divides that aggregated income among states according to a formula that considers the presence in each state of the enterprise’s productive factors—its workers, property, and sales.239Barclays Bank PLC v. Franchise Tax Bd., 512 U.S. 298, 312 n.10 (1994) (“[T]he theory underlying unitary taxation is that certain intangible ‘flows of value’ within the unitary group serve to link the various members together as if they were essentially a single entity.’”). Use of formulary apportionment is not limited to the U.S. states. The Canadian provinces use it,240Ruth Mason, Delegating Up: State Conformity with the Federal Tax Base, 62 Duke L.J. 1267, 1319 n.223 (2013). and in 2016, the Commission re-introduced a legislative proposal for the EU states to move from separate accounting to formulary apportionment.241Proposal for a Council Directive on a Common Consolidated Corporate Tax Base (CCCTB), at 2, 35–37, COM (2016) 683 final (Oct. 10, 2016) [hereinafter CCCTB Relaunch]. Unlike international tax rules, which typically employ facially selective source and residence rules, apportionment formulas can be facially neutral. But as this Subsection shows, facially neutral rules can discriminate between cross-border and domestic commerce.

The Supreme Court has frequently considered the compatibility of facially neutral rules with the dormant Commerce Clause. In the 1978 case Moorman Manufacturing Co. v. Bair,242437 U.S. 267 (1978). a company complained that it suffered double taxation in violation of the dormant Commerce Clause because Illinois had an apportionment formula that differed from that of Iowa, another state in which the company did business.243Id. at 269, 271. At the time of Moorman, Illinois (and nearly all other states) equally weighed payroll, property, and sales in their apportionment formulas, but Iowa apportioned only by sales.244See id. at 269–70, 276. Iowa presumably designed its sales-only apportionment formula to entice companies to locate their property and payroll in Iowa by exempting those factors from tax.245See id. at 283–84 (Powell, J., dissenting).

To evaluate Moorman’s dormant Commerce Clause claim while avoiding “extensive judicial lawmaking,” including dictating a common allocation formula, the Supreme Court employed the internal consistency test.246Id. at 277, 278 (majority opinion). Under the test, the Court imagines that all states apply the challenged state’s tax rule.247Id. at 277; see Okla. Tax Comm’n v. Jefferson Lines, Inc., 514 U.S. 175, 185 (1995). It then asks whether the complained-of tax disadvantage would persist under this hypothetical harmonization.248Moorman, 437 U.S. at 277. If the answer is yes, the rule unconstitutionally discriminates against interstate commerce compared to domestic commerce.249Wynne, 135 S. Ct. at 1802. In contrast, if the tax disadvantage disappears under hypothetical harmonization, it must be due to variation between different states’ laws, a mismatch for which no single state can be held responsible.250Id.; see Moorman, 437 U.S. at 277 n.12 (noting that “whatever disparity may have existed is not attributable to the Iowa statute . . . . [T]he alleged disparity can only be the consequence of the combined effect of the Iowa and Illinois statutes, and Iowa is not responsible for the latter”). Such mismatches can be resolved by Congress, but not the courts.251Moorman, 437 U.S. at 280 (holding Congress could fix the overlap through the “enactment of legislation requiring all States to adhere to uniform rules for the division of income. It is to that body, and not this Court, that the Constitution has committed such policy decisions”).

In Moorman, if all fifty states adopted Iowa’s sales-only formula, tax gaps and overlaps from state-to-state formula diversity would disappear. Under such hypothetical harmonization, every company would pay tax on exactly all its income, allocated only according to sales, and interstate commerce would face no more or less tax than would purely domestic commerce.252To be internally consistent, the hypothetically harmonized state law must result in equal taxation of cross-border and domestic transactions. For application of this prescription to allocation rules, see infra Section IV.B. The Court found that the double tax Moorman complained about would disappear under internal consistency, so the Iowa rule was valid, notwithstanding any actual double taxation that may have occurred due to lack of harmonization between the Iowa and Illinois rules.253Moorman, 437 U.S. at 277 (noting that because the record did not show which of the appellant’s profits had been earned in Iowa and which had been earned in Illinois, the dormant Commerce Clause claim could not “rest on the premise that profits earned in Illinois were included in [Moorman’s] Iowa taxable income and therefore the Iowa formula was at fault” for the double taxation). Notice that the internal consistency test relieved the Supreme Court of the need to consider what an ideal apportionment formula might be. Internal consistency also allowed the Court to avoid deciding whether Iowa’s rule should prevail over Illinois’s or vice versa. The test enabled the Court to evaluate a facially neutral rule that seemed to disproportionately impact cross-border commerce without having to appeal to normative benchmarks or other values external to the challenged state’s law. Although it does not employ tax-expenditure analysis, internal consistency allows the Court to evaluate state allocation rules using a reference-law benchmark, rather than an external norm.

Suppose Moorman had sued Illinois instead of Iowa, and the Court had applied the internal consistency test to Illinois’s three-factor apportionment rule. That formula also passes the internal consistency test. The Moorman Court acknowledged that adoption by two different states of different but internally consistent formulas could result in double tax (or tax gaps), but it concluded that because “[t]he Constitution . . . is neutral with respect to the content of any uniform [apportionment] rule,” only Congress could fix that problem.254Id. at 279. No particular state, in the Supreme Court’s view, could be held responsible for tax disadvantages that arise from state-to-state rule diversity.255Id. at 277, 277 n.12 (noting that “[e]ven assuming some overlap, we could not accept appellant’s argument that Iowa, rather than Illinois, was necessarily at fault in a constitutional sense”).

The internal consistency test would “pass” both sales-only and three-factor formulary apportionment. But, unlike an approach that considers all facially neutral laws to be part of the reference-law baseline, the internal consistency test does not simply pass allincome-allocation rules. For example, suppose Illinois had adopted a formula that allocated income to Illinois on the basis of in-state and out-of-state sales. If every state adopted that formula, companies would face systematically more tax on cross-border sales than on domestic sales. That protectionist formula would be internally inconsistent and unconstitutional under the dormant Commerce Clause.

By imagining that every state applied the challenged rule, and then considering what the impact of such a harmonized rule would be on interstate commerce, the internal consistency test eliminates tax advantages and disadvantages that stem from mismatches between the challenged state’s rules and the rules of other states. As the Supreme Court put it in a 2015 decision,

[the test] allows courts to distinguish between (1) tax schemes that inherently discriminate against interstate commerce without regard to the tax policies of other States, and (2) tax schemes that create disparate incentives to engage in interstate commerce . . . only as a result of the interaction of two different but nondiscriminatory and internally consistent schemes.256Comptroller of Treasury v. Wynne, 135 S. Ct. 1787, 1802 (2015).

Thus, internal consistency is a hypothetical-impact approach, but one that avoids mistaking mismatches for discrimination. Internal consistency enables identification of cross-border tax differences that derive exclusively from the accused state’s law, including its structural provisions, such as tax rates and income-allocation rules. Internal consistency thus makes structural tax provisions tractable without automatically “passing” all structural provisions nor relying on external benchmarks.

B.   Internal Consistency Identifies Discrimination

Internal consistency does more than identify mismatches. It also identifies income-tax provisions that violate competitive neutrality by operating equivalently to tariffs and import/export subsidies. Because the state-aid rules prohibit Member States from conferring subsidies that violate competitive neutrality,257See infra Section I.A. the internal consistency test can identify state-aid violations. Economic analysis confirms that to maintain competitive neutrality, the tax burden a state applies to residents’ in-state income must be the same as the sum of the burdens it imposes on residents’ out-of-state income and nonresidents’ in-state income.258Space does not permit me to delve deeply into this economic analysis, but several sources establish that the internal consistency test identifies state tax provisions that violate competitive neutrality, meaning they function equivalently to tariffs or import and export subsidies. See, e.g., Knoll & Mason Wynne Brief, supra note 23, at 8; Tax Economists’ WynneBrief, supra note 23, at 4, Knoll & Mason (2017), supra note 23, at 326; Lirette & Viard, supra note 23, at 471. This argument can be made most simply in cases involving tax rates. Knoll & Mason (2017), supra note 23, at 323 (establishing that in cases involving tax rates, maintaining competitive neutrality demands that residents’ domestic-source income, Td, must equal or exceed the combined tax imposed on domestic residents’ out-of-state income, To, and nonresidents’ in-state income, Ti ). Thus, maintaining competitive neutrality requires Td ≥ To + Ti – (To x Ti). Id. The internal consistency test reveals whether a state satisfies this requirement. It does so by implicitly comparing the burden a state places on domestic economic activity to the same state’s cumulative burden on interstate economic activity.259In our amicus brief in Wynne, Michael Knoll and I argued that “the hypothetical harmonization performed under the internal consistency test has the effect of applying to residents that earn out-of-state income and to nonresidents that earn in-state income both the state’s treatment of nonresidents’ in-state income and the state’s treatment of residents’ out-of-state income. The internal consistency test thus calculates the combined tax rate from assessing both of those taxes and then compares the combined tax to the rate the state imposes on residents’ domestic income. If the rates are the same, then there is no violation. If the rates are different, then taxes distort competition. Thus, the internal consistency test asks the same question as does economics in determining whether a state’s tax system distorts competition between residents of different states.” Knoll & Mason Wynne Brief, supra note 23, at 16–17 (citations omitted); see also Tax Economists Wynne Brief, supranote 23, at 6 (“[T]he internal consistency test succeeds as a substantive indicator of such discrimination because it correctly measures the combined tax burden on inbound and outbound transactions and compares it to the tax burden on within-state transactions.”); Knoll & Mason (2017), supra note 23, at 331 (“understanding the impact of [a state’s] tax regime on cross-border commerce requires examining its entire tax regime, comprised of inbound, outbound, and domestic taxes. The internal consistency test provides a court with a simple tool to examine the whole regime”).

A virtue of the internal consistency test is that it reveals violations of competitive neutrality even when they are obscured by facially neutral rules. For example, the 1987 Supreme Court case American Trucking Ass’n v. Scheiner260Am. Trucking Ass’n v. Scheiner, 483 U.S. 266 (1987). involved a facially neutral flat tax on trucks that traveled within or through a state. Because the state did not apportion the tax by miles driven in-state versus out-of-state (or by some other method), the tax was internally inconsistent because, if adopted by all the states, interstate trucking would face duplicative burdens, but intrastate trucking would not.261Id. at 284 (“If each State imposed flat taxes for the privilege of making commercial entrances into its territory, there is no conceivable doubt that commerce among the States would be deterred.”). But see Am. Trucking Ass’n v. Mich. Pub. Serv. Comm’n, 545 U.S. 429, 437–38 (2005) (declining to invalidate a similar unapportioned and internally inconsistent trucking tax). Likewise, a facially neutral sales tax that applied to bus tickets that either originated or terminated in a state would be internally inconsistent. If every state adopted such a rule, cross-border bus trips would suffer double tax compared to in-state trips.262Cf. Okla. Tax Comm’n v. Jefferson Lines, Inc., 514 U.S. 175, 185 (1995) (upholding as internally consistent such an allocation rule when paired with a requirement that the bus ticket be sold in the taxing state). The internal consistency test reveals that, despite their facial neutrality, these rules overtax cross-border commerce. The test allows the reviewer to “isolate”263Comptroller of Treasury v. Wynne, 135 S. Ct. 1787, 1802 (2015) (“By hypothetically assuming that every State has the same tax structure, the internal consistency test allows courts to isolate the effect of a defendant State’s tax scheme.”). and evaluate the impact of a given rule on cross-border commerce as compared to domestic commerce. It does so without requiring the reviewer to otherwise pass judgment on the challenged tax. For example, the reviewer need not decide how interstate trucking or bus tickets ought to be taxed.

The Supreme Court’s most recent case applying internal consistency was Comptroller of Treasury v. Wynne264135 S. Ct. 1787 (2015). in 2015.265Id. at 1803. Maryland’s income-allocation regime in Wynne consisted of three rules: it taxed (1) Maryland residents on income earned in Maryland, (2) Maryland residents on income earned from other states, and (3) nonresidents on income earned in Maryland.266Id. at 1792. Assume the rate for all three taxes was identical.267In the actual case, not all the rates were the same. Rates (1) and (2) were equal to each other, but (3) was lower. See id. I changed the third rate to emphasize that rate neutrality does not immunize tax regimes from challenge as protectionist. Analyzing the same scheme with any lower, but still positive, rate for tax (3) reveals its internal inconsistency. This regime is facially neutral; it treats residents’ income the same no matter where they earn it; and it taxes nonresidents and residents the same. The case arose when the Wynnes, who were Maryland residents, complained that they faced double taxation in violation of the dormant Commerce Clause because they had to pay both source taxes to the states where they earned income and residence taxes to Maryland, which offered them no credit for other states’ source taxes.268Id. at 1792–93.

Maryland responded by correctly observing that the Constitution does not forbid double taxation; Maryland further argued that the Wynnes’ tax disadvantage arose from a mismatch between Maryland’s law and that of other states.269Id. at 1798. Maryland asserted that it taxed all its residents at the same rate no matter where they earned their income, and Maryland could not be held responsible for taxes imposed by other states, even if the result was double taxation.270Id.

But the internal consistency test revealed the constitutional infirmity in Maryland’s facially neutral regime. If every state enacted Maryland’s regime, cross-border commerce would always be taxed at twice the rate of domestic commerce.271For example, universalize the Maryland regime and then suppose a resident of Maryland earned income in Virginia. The Marylander would pay tax to Maryland as a resident and to Virginia as a nonresident who earned income there. Because the Maryland regime allowed no credit for other states’ taxes, there would be no relief for the double tax. In contrast, a Marylander who earned income in Maryland would be taxed only once, as would be a Virginian who earned income only in Virginia. The internal consistency test showed that Maryland imposed more tax on cross-border commerce than on domestic commerce, thereby violating competitive neutrality. This extra burden on interstate commerce arose from Maryland’s regime alone; if all states enacted identical rules, it would persist. Moreover, the impact on interstate commerce would persist even if, in the actual world, none of the other states imposed taxes.272Knoll & Mason (2017), supra note 23, at 339. Under dormant Commerce Clause jurisprudence, failure of internal consistency is fatal as a matter of law, so Maryland lost its case in Wynne.273Okla. Tax Comm’n v. Jefferson Lines, Inc., 514 U.S. 175, 185 (1995) (“A failure of internal consistency shows as a matter of law that a State is attempting to take more than its fair share of taxes from the interstate transaction, since allowing such a tax in one State would place interstate commerce at the mercy of those remaining States that might impose an identical tax.”).

Although the Supreme Court originally arrived at the internal consistency test by intuition, in Wynne it acknowledged the “economic bona fides” of the test.274Wynne, 135 S. Ct. at 1803 (2015); see also id. at 1804 (“[T]he internal consistency test reveals what the undisputed economic analysis shows: Maryland’s tax scheme is inherently discriminatory and operates as a tariff.”); Knoll & Mason (2017), supra note 23, at 313 (explaining that the Supreme Court renewed its commitment to the internal consistency test in Wynne due to the ability of the test to ferret out violations of competitive neutrality). Economic analysis shows that internally consistent rules do not violate competitive neutrality because they do not place different tax burdens on cross-border and domestic commerce, but internally inconsistent tax rules do.275Knoll & Mason (2017), supra note 23, at 332. Except in limited circumstances, the internal consistency test can identify when income-tax rules function equivalently to tariffs and import/export subsidies. Lirette & Viard, supra note 23, at 500 (noting that except in cases where one state expressly makes its tax treatment depend on the actions of another state—for example, by granting a credit—the “internal consistency test is functionally identical” to competitive neutrality). Tax rules that depend on actions or conditions in another state are more complicated to resolve using internal consistency. See id.; see also Mason, supra note 21, at 1323–25 (discussing conditionality). None of the cases discussed in this Article involved such conditionality. The test works for taxes,276See generally Lirette & Viard, supra note 23; Mason & Knoll (2012), supra note 29; Knoll & Mason Wynne Brief, supra note 23; Tax Economists’ Wynne Brief, supra note 23. subsidies277Lirette & Viard, supra note 23, at 473 (“The validity of subsidies can be evaluated under the same internal consistency test that applies to taxes.”); see id. at 528–36 (showing that internal consistency identifies subsidies that treat domestic and cross-border commerce differently). and structural rules, including income-allocation rules.278Knoll & Mason (2015), supra note 23, at 925–29 (analyzing allocation rules under competitive neutrality). Thus, as long as the Commission and CJEU continue to interpret the state-aid rules to prohibit tax rules that violate competitive neutrality by discriminating between cross-border and domestic commerce, they can benefit from the internal consistency test. Because state aid is essentially the inverse of tax discrimination—it concerns cross-border tax advantages rather than disadvantages—the internal consistency test can work for both.279See supra notes 21–23, 29 and accompanying text.

The treatment of allocation rules under internal consistency deserves special attention because so many of the recent state-aid cases involved allocation rules. Allocation rules that create tax overlaps when all states adopt them are equivalent to tariffs. Such rules prefer domestic to cross-border commerce.280Cf. Okla. Tax Comm’n, 514 U.S. 175, 185 (1995) (noting that because all bus ticket sales would be subject to a single state tax if all states applied the challenged state’s rule, it was not unconstitutional); Container Corp. of Am. v. Franchise Tax Bd., 463 U.S. 159, 169 (1983) (holding that a state’s apportionment formula “must be such that, if applied by every jurisdiction, it would result in no more than all of the unitary business’s income being taxed”). Allocation rules that create tax gaps when all states adopt them are equivalent to import or export subsidies.281See Knoll & Mason (2015), supra note 23. Such rules prefer cross-border to domestic commerce. Thus, internally inconsistent allocation rules violate competitive neutrality. Since allocation rules always implicate a selective class—namely, multinationals—income-allocation rules that create tax gaps and overlaps under hypothetical harmonization would be state aid unless justified for public policy reasons.

A principal implication of this analysis is that to comply with EU law, Member State allocation rules must be such that, if applied by all the states, they would tax exactly all the income of multinationals, no more and no less.282Id., at 925–28, explaining that, “[a]s with tax rates, with [allocation] rules we compare (a) the domestic tax to (b) the sum of (1) the inbound tax and (2) the outbound tax. The twist is that since we cannot check for higher tax rates (because the constitutional challenge does not involve rates), we must instead look to how much income is subject to taxation. Specifically, we compare (a) the portion of a purely domestic taxpayer’s income allocated to the taxpayer’s state of domicile with (b) the sum of (1) the portion of the taxpayer’s cross-border income allocated to the challenged state and (2) the portion the taxpayer’s of cross-border income allocated to other states when they apply the challenged state’s allocation rule under the 50-state harmony assumption. That is, the court compares the (a) domestic allocation with (b) the sum of (1) the inbound allocation and (2) the outbound allocation . . . . When all of a taxpayer’s economic activity occurs in the state in which the taxpayer resides or is domiciled (that is, in a purely domestic situation), all of the taxpayer’s income from that activity is sourced (allocated) in the state of the taxpayer’s residence or domicile. In other words, (a) is equal to 100 percent. Thus, when the challenge is to the sourcing rules, we almost always compare the sum of the inbound and outbound situations to 100 percent. In these circumstances, the sourcing rule passes the test if in the inbound plus outbound situation no more than 100 percent of income is hypothetically taxed.” “Income” for this purpose would have the meaning given in the tested state’s law. For the cases we have discussed, it would refer simply to the tested state’s corporate tax base, whatever its content and quirks.283Thus, for example, when universalizing Gibraltar’s proposed regime, the Commission would also universalize the fifteen percent profit cap, so that amounts above the cap would not be part of tested “income.” Likewise, when universalizing a regime that completely exempts excess profits for all groups, the exempted excess profits would not be a part of the tested “income.” “Income” then, would be subjective, not normative—it would be defined by the particular challenged state’s rules. This would excuse the Commission from having to develop notions of what it might mean to tax “all companies” on their “actual profits.”

Put simply, if the state’s income-allocation rule, when applied by all the states, would allocate less than 100 percent of the income in cross-border situations, then the allocation rule subsidizes cross-border commerce compared to domestic commerce. It would constitute state aid unless justified. By the same token, if an income allocation rule allocates more than 100 percent of income in cross-border situations, it is a tax penalty for cross-border commerce compared to domestic commerce, which, unless justified, would be both state aid and a violation of the fundamental freedoms.284The prohibition of state aid forbids both favoritism of cross-border over domestic commerce and favoritism for domestic over cross-border commerce. In contrast, the fundamental freedoms only forbid taxes that prefer domestic to cross-border commerce. Because internal consistency relies on the state’s own definition of income and its own structural rules, it uses only domestic law as a benchmark, which is critical for legitimacy and protecting state tax prerogatives.

C.   Analyzing Recent Cases Under Internal Consistency

The internal consistency test does not appeal to norms that are external to the tested state’s law. Internal consistency simply universalizes the challenged state’s own tax rules to determine whether, if every state applied those rules, cross-border and domestic commerce would receive different treatment. Unlike tax-expenditure analysis, internal consistency does not depend for its analytic force on the Commission’s ability to determine whether the challenged rules deviate or not from a reference benchmark. Because it dispenses with the need to categorize rules as “normal” or derogating provisions, the internal consistency test sidesteps the tax-expenditure trap.

Equally importantly for state-aid purposes, the internal consistency test accurately distinguishes cross-border mismatches—cases like Block or reverse-Block285See supra Section I.A.—from tax provisions that generate cross-border tax differences no matter what other states do—cases like Vodafone or Wynne. Its use in the state-aid context, therefore, would allow the Commission to distinguish tax rules that confer tax advantages relative to other states’ regimes from rules that confer tax advantages regardless of other states’ regimes. When we universalize a challenged state’s rule under internal consistency, mismatches disappear because all states apply the same law. Thus, if the purported tax advantage disappears under internal consistency, it must have been caused by a mismatch, not state aid. But if the cross-border advantage remains under internal consistency, it cannot have been caused by a mismatch. Because internal consistency can positively identify mismatches, it can help the Commission avoid false positives in which it mistakes mismatches for covert discrimination.

As noted above, internal consistency also determines whether tax rules, including income-allocation rules, violate competitive neutrality by discriminating between cross-border and domestic commerce. Economic analysis generates precise requirements for allocation rules to avoid violating competitive neutrality, and the internal consistency test provides an easy method to determine whether a state’s rules violate those requirements.286See supra note 282. It does so without the need to identify a benchmark for measuring tax advantages, and it does so as easily for facially neutral income-allocation rules as for facially selective income-allocation rules. Thus, absence of clear reference-law or normative baselines in income-allocation cases would not stymie analysis under internal consistency.

Eschewing the categories presented in Part III as no longer relevant, this Subsection considers how the cases described in Part III would be analyzed under internal consistency.

1.      Vodafone

Recall that in Vodafone, a Hungarian court asked the CJEU whether Hungary’s graduated turnover tax, in which tax rates increased with turnover size, constituted state aid to companies that paid no tax due to low turnover.287Vodafone reference, supra note 115. The Hungarian regime is facially neutral—turnover size is not a selective class. But turnover size correlates strongly with nationality.288See supra Section III.A.1.

The internal consistency test allows us to confirm that neither the tax disadvantage suffered by high-turnover companies nor the tax benefit experienced by low-turnover companies under the Hungarian regime arises from mismatches between two or more state’s different regimes. If every state adopted the Hungarian regime, the high-turnover disadvantage would remain, and so too would the advantage for low-turnover companies.

The internal consistency test only fully resolves cases of “covert” or proxy discrimination if we are confident about the Commission’s judgment that a facially neutral classification impermissibly proxies a selective class. Thus, in addition to turning on the Court’s view of turnover-size as a proxy for nationality, the outcome in Vodafone will depend on whether Hungary can justify the difference in treatment between high- and low-turnover companies. Notwithstanding that internal consistency cannot fully resolve Vodafone or other proxy classification cases, it nevertheless helps the Commission to rule out mismatches, thereby cabining its discretion.

2.      Gibraltar

Gibraltar was similar to Moorman in that it involved a strategically variant, but facially neutral, regime for allocating income from interstate commerce.289Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 (Nov. 15, 2011). In Moorman, Iowa aimed to poach payroll and property from other states by eliminating those factors from its apportionment formula, while Gibraltar presumably hoped to attract corporate residents from other states by limiting its tax to payroll and property factors. The efficacy of both regimes as incentives depended on their uniqueness: For the Iowa and Gibraltar regimes to be attractive, other states had to have different regimes.290Unlike with tax rates, with advantages that arise from rule variation, the deviating state need not have inherently more attractive rules. Both the Iowa and the Gibraltar regimes would tax all income if every state applied them; the advantage inheres in the gaps between different systems, not in the Iowa or Gibraltar system itself. Cf. Kane, supra note 41 (discussing international tax arbitrage). We know that incentives that depend for their effectiveness on state-to-state variation—like low tax rates—are not necessarily state aid because state aid does not require tax harmonization. But how can the Commission distinguish permissible variation from impermissible subsidization? Internal consistency helps by revealing whether any advantage a taxpayer receives arises from variation among states’ regimes or whether it arises exclusively from the accused state’s regime.

If every EU state adopted the proposed Gibraltar regime, so that all states taxed all companies—both resident and nonresident—based on their in-state payroll and property, then although Gibraltar would not collect much tax from offshore companies, the income Gibraltar exempted would nevertheless be taxed by other states, specifically, it would be taxed where the offshore companies had the rest of their payroll and property.291Earlier, I noted that although it was unclear from the cases, the proposed Gibraltar payroll-and-property tax regime may have applied only to Gibraltar-incorporated (or registered) companies, and not to nonresident companies. See supra note 123. Thus, for example, a nonresident company that hired an employee to work in Gibraltar may not have had to pay the payroll tax. See Joined Cases C-106/09 P & C-107/09 P, Gibraltar v. Comm’n, ECLI:EU:C:2011:732 ¶ 21 (Nov. 15, 2011). If this was so, then the proposed regime was internally inconsistent because if every state applied the regime, people employed by out-of-state companies and property occupied by out-of-state companies would not generate tax anywhere for any company. This represents a preference for cross-border over domestic commerce. Such outbound preferences constitute state aid unless justified. See, e.g., Joined Cases C-20/15 P & C-21/15 P, Comm’n v. World Duty Free Group, ECLI:EU:C:2016:981 ¶ 22 (Dec. 21, 2016). Thus, under a universalized Gibraltar payroll-and-property tax, all companies—whether engaged in cross-border or only domestic commerce—would pay tax on all their income (as defined under Gibraltar law), and the putative tax advantage for offshore companies identified by the Commission in Gibraltar would disappear. The proposed Gibraltar regime was therefore internally consistent: the advantage it would have conferred as applied was not inherent to the Gibraltar regime alone, rather it would have arisen only from interactions between Gibraltar’s tax base and the tax bases of other states. As long as they do not themselves interfere with cross-border commerce, such differences are acceptable under an anti-subsidy regime that permits variation in state taxes. Like Moorman, Gibraltar involved a mismatched allocation rule. The internal consistency test enables us to see the mismatch. It shows us that, if adopted universally, the Gibraltar rule would create no tax gaps or overlaps for offshore companies; as much income would be included for offshore companies as for domestic companies.

Unlike the two approaches the Commission took in Gibraltar—external benchmarking and disproportionate impact—the internal consistency test would have enabled the Commission to see that the case involved a nondiscriminatory mismatch. That the Commission failed to recognize Gibraltar as a mismatch suggests that the Commission could benefit from the clarity the internal consistency test provides. Moreover, as the Vodafone example shows, internal consistency also identifies when facially neutral rules generate cross-border tax differences that cannot be attributed to mere mismatches.292The internal consistency test likewise identifies discrimination even in cases in which states do not employ suspect or proxy classifications. Examples from the U.S. jurisprudence include Wynne and American Trucking. See supra Sections IV.A. and B. The ability to correctly trace a cross-border tax advantage to a particular state’s rule, rather than to interactions between different states’ rules, is important both to preventing both false positives and false negatives. False positives improperly limit state tax policy choices, while the false negatives inhibit economic integration among the states.

3.      Apple and the other transfer pricing cases

How would internal consistency handle Apple? Suppose we take seriously Ireland’s argument that it had its own allocation rules that differed from the OECD rules. How can we determine whether any advantage those allocation rules conferred to Apple was due to permissible variation across states in allocation rules or due to impermissible favoritism by Ireland? If income-allocation rules are structural provisions that reflect tax policy choices, how should the Commission evaluate them?

If Apple can point to a consistent and clear set of domestic income-allocation rules that applied at the time the facts of Apple arose, then the Commission could universalize those rules by assuming every state had them. Unlike a tax-expenditure approach to identifying state aid that uses an external baseline, internal consistency would allow the Commission to attribute any resulting tax advantages exclusively to Irish law, rather than to deviations between Irish law and the Commission’s notion of ideal rules.

If the universalized Irish rules would tax all of the global income of multinationals (however defined under Irish law for domestic companies), then they do not confer relevant tax advantages when implemented only by Ireland, and they cannot be state aid.293See supra Section IV.B; see also Knoll & Mason (2015), supra note 23, at 925–26 (explaining how to use the internal consistency test to evaluate allocation rules); Lirette & Viard, supra note 23, at 470, 537–45 (discussing internal consistency and allocation rules). Ireland did not appear to condition its branch allocation rule on tax treatment of the same or related income by any other state. If this is correct, the application of the internal consistency test would be straightforward in Apple. In the contrary case, if Ireland’s allocation rules would not tax all of a multinational group’s income if every state adopted them (again, income would be defined under Irish law for domestic companies), then those allocation rules would function as impermissible subsidies because they would create a preference for cross-border over domestic commerce.294See references cited in supra note 293.

In Apple, Ireland did not clearly articulate its system for allocating income to branches. But Ireland’s system seemed to rely on a “one-sided” transfer-pricing method under which a state allocates income from cross-border economic activities by looking only at the activities that take place in its own territory.295See generally Bret Wells & Cym Lowell, Tax Base Erosion: Reformation of Section 482’s Arm’s Length Standard, 15 Fla. Tax Rev. 737 (2014). Upon examining these activities, a state might use arm’s-length or a routine return on assets, or some other method to assign a return to those activities. A defining feature of one-sided methods is that they do not allocate the “residual” profit.296Id. at 751. The residual profit is the same concept that Belgium referred to as the “excess profit;” it includes returns that cannot be attributed to specific activities in any one state because the returns arise due to operation as a firm; they include synergies, economies of scale, returns to retaining ownership to unique intangibles within the firm, and so on. Any income-allocation method that fails to allocate the residual only for multinationals—and not for domestic groups—would create a tax gap because such income would not be allocated anywhere.297See id. at 738 n.2 (criticizing one-sided transfer pricing methodologies for creating “homeless income”); id. at 789 (calling one-sided methods “the single largest contributor” to homeless income and base erosion and profit shifting). In contrast, two-sided transfer-pricing methods (like profit-splits) take a holistic approach that allocates the residual.298See id. (advocating mandatory two-sided transfer-pricing methods). If applied by all the states, one-sided transfer pricing methods create gaps, but two-sided methods do not.

To the extent that they create gaps for multinationals, but not domestic companies, one-sided transfer-pricing methods violate competitive neutrality by favoring cross-border commerce over domestic commerce.299Id. at 751. They are the income-allocation equivalent of import and export subsidies. The benefits conferred by one-sided methods redound exclusively to a selective class—multinationals. Thus, they are state aid unless justified.

Even if all one-sided transfer-pricing methodologies violate internal consistency by favoring cross-border over domestic commerce, they may not all constitute state aid. The Commission might regard the one-sided methods memorialized in the OECD Transfer Pricing Guidelines as justified because they are consistent with international practice.3002016 Notice, supra note 1, at 38 (noting OECD guidance “capture[s] the international consensus on transfer pricing” and “if a transfer pricing arrangement complies with the guidance . . . [it] is unlikely to give rise to State aid”). EU law also permits justifications to be conditioned on other state’s law, so that if, for example, Apple’s income had been actually taxed by the United States, that U.S. taxation could have justified Ireland in exempting it. See, e.g., Case C-403/03, Schempp v. Finanzamt München V, ECLI:EU:C:2005:446 ¶ 28 (July 12, 2005). But it is not obvious that the Commission would or should approve a one-sided method unique to Ireland or any other single Member State.

Notice that the approach advocated here does not compel any state to tax the residual or excess profit. All that is required to maintain competitive neutrality is to treat the residual the same for cross-border and domestic groups. Thus, internal consistency is agnostic as to the content of the tax base, but it requires that the tax base—whatever its content—apply without discrimination to both cross-border and purely domestic economic actors and activities. The discussion of Belgian Excess Profits in the next Subsection makes this clear.

4.      Belgian Excess Profits

Recall that Belgian Excess Profits involved a facially discriminatory regime that exempted excess profits for multinational, but not domestic, groups.301Belgian Excess Profits Final Commission Decision, supra note 186. The Commission found that the regime conferred state aid by comparing it to a normative tax of all corporations on their “actual profits.”

Belgium won its appeal in the lower EU court on other grounds,302See Case T-131-16, Belgium v. Comm’n, ECLI:EU:T:2019:91 (Feb. 14, 2019) (holding that the Commission had not borne its burden to prove that Belgium had what state-aid law refers to as an “aid scheme”). but the case raises the question of how the Commission can distinguish facial selectivity needed to accommodate limited jurisdiction over nonresidents or a desire to avoid double tax, on the one hand, from facial selectivity that violates competitive neutrality, on the other. Consider the same Belgian regime under the internal consistency test. Every state would apply Belgium’s regime, and, as a result, every state would exclude the excess profits of multinational, but not domestic, groups. That the tax advantage for multinationals persists under hypothetical harmonization shows that it arises from Belgium’s law alone, not from mismatches between multiple states’ different tax bases.

Such internally inconsistent allocation rules violate competitive neutrality; they discriminate between cross-border and domestic commerce.303See supra Section IV.B. The effect does not depend on what other states do. Thus, internal consistency would have enabled the Commission to identify the rule in Belgian Excess Profits as state aid without the need for a normative determination of whether a state ought to tax excess profits and without the need to identify the object or purpose of the Belgian tax system. Similarly, internal consistency relieves the Commission of the need to determine what it would mean to tax “actual profits.”

Finally, avoiding reliance on external norms is crucial for preserving Member State tax sovereignty, including a state’s prerogative to exclude synergistic profits nonselectively. For example, suppose that, contrary to fact, Belgium had applied its excess-profits exclusion to both cross-border and domestic groups.304Assume for purposes of this example, and contrary to the facts of the case, that Belgium had a consistent and reliable method for identifying excess profits that it applied in the same way (nonselectively) to all corporate groups. Such a rule presumably would not tax “actual profits.” Instead, it would exclude a particular item of income, namely, excess profits.

Some might conclude that excluding excess profits is bad tax policy, but the internal consistency test confirms that if Belgium implemented such an exclusion nonselectively, it would not violate competitive neutrality. Under internal consistency, every state would exempt excess profits for both domestic and multinationals groups, and the incentive for groups to engage in cross-border activity over purely domestic activity would disappear.305Under the internal consistency test, the incentive created by the regime to locate companies in Belgium would also disappear. This echoes prior observations that maintenance of competitive neutrality can improve locational efficiency; or, stated inversely, violations of competitive neutrality distort not only whether taxpayers engage in cross-border or domestic economic activities; they also generate locational distortions. See Mason & Knoll (2012), supra note 29, at 1051–71. The overlap of competitive and locational distortions has, in my view, lead some to erroneously conclude that the state-aid rules promote locational neutrality for its own sake, rather than merely incidentally to promoting competitive neutrality. An internally consistent version of the excess profit rule would violate the single-tax principle, and it would favor companies that operate as integrated groups over standalone companies. But it would not violate competitive neutrality because it would not, by itself, discriminate between cross-border and domestic commerce. Since groups—including fully domestic groups—earn higher profits than do standalone companies engaged in similar activities, one could imagine states adopting policies to encourage group formation.306Ronald H. Coase, The Nature of the Firm, 4 Economica 386, 387, 390–91 (1937). Indeed, eliminating barriers to the formation of trans-European groups was one of the motivations for the European Union in the first place.307See, e.g., Robert Schuman, Declaration Proposing Creation of European Community of Steel and Coal (May 9, 1950) (“[T]he organization will ensure the fusion of markets and the expansion of production.”). To be sure, if implemented in the real world only by Belgium, the rule we are imagining would encourage groups to locate in Belgium when combined with other states’ adoption of different tax bases. But such locational distortions do not violate EU law as long as they do not discriminate between cross-border and domestic commerce.

In contrast with internal consistency, if the Commission evaluates state-aid cases by simply assuming in every case that each state’s goal is to tax “all companies” or “actual profits,” then the Commission will have to develop and enforce notions about what satisfies those goals. It will, therefore, have to resolve tax policy questions.

5.      McDonald’s

Universalizing Luxembourg law and Luxembourg canons of tax-treaty interpretation likewise confirms the Commission’s analysis in McDonald’s. The tax advantage in that case arose due to differences between U.S. and Luxembourg law and interpretation. If every state—including the United States—used Luxembourg’s law and interpretations, the McDonald’s branch would be taxed by the United States, and the tax gap would disappear, as would the incentive for cross-border over domestic commerce. Disappearance of the tax gap upon universalization of Luxembourg’s regime shows that the gap arose from mismatches between U.S. and Luxembourg law, not from Luxembourg law alone. Because the advantage in McDonald’s arose from a mismatch and mismatches are not illegal, McDonald’s was rightly decided.

6.      Irish tax-residence rule

Return to the Irish tax-residence rule and apply the internal consistency test. Under internal consistency, every state, including the United States, would apply the Irish management-and-control test to determine tax residence. Applying Irish law, the United States would conclude that Apple’s Irish-incorporated subsidiaries were managed and controlled in the United States and therefore tax resident in the United States. Under internal consistency, the opportunity to create stateless companies would therefore disappear. The disappearance of the advantage upon universalization of the Irish rule shows that the advantage arose from differences between U.S. and Irish law, not from Irish law alone. If all states adopted the Irish rule, the incentive to prefer cross-border economic activity to domestic activity would disappear. Internal consistency therefore confirms that the Commission was correct not to bring a state-aid claim against Ireland for its tax-residence rule, even though in the real world, the Irish rule when combined with the U.S. rule led directly to significant tax avoidance. Mismatches may create cross-border advantages or disadvantages, but, under current doctrine, they do not violate EU law.

D.   Superiority of Internal Consistency over Tax-Expenditure Analysis

The internal consistency test has more to recommend it than simplicity and an attractive resemblance to Kant’s categorical imperative. This Subpart reviews the advantages of internal consistency when compared to tax-expenditure analysis.

First, internal consistency correctly identifies cross-border tax differences that arise from mismatches between states’ laws. Because hypothetical harmonization eliminates mismatches, any tax differences that persist under hypothetical harmonization can be attributed exclusively to the accused state, and the desire to preserve tax diversity among states, therefore, need not shield such rules from further state-aid scrutiny. Because it confirms them as nondiscriminatory mismatches, internal consistency makes it easy to see why certain tax provisions never have been considered state aid, notwithstanding that they are widely understood to confer significant cross-border tax benefits. Such mismatched provisions include generally applicable low tax rates, narrow tax-residence rules, fast depreciation schedules, and the like. Internal consistency also explains both the outcome in McDonald’s and why the Commission did not challenge the Irish tax-residence rules that facilitated Apple’s tax planning: Both involved mismatches. The test also confirms that some states could have separate accounting with arm’s-length, while others implement unitary accounting with formulary apportionment. This is important because the mandate the Commission announced in recent cases requiring states to apply the arm’s-length standard regardless of their domestic law would seem to prohibit movement by a subset of states to formulary.

Second, internal consistency can reliably identify income tax rules that violate competitive neutrality because they are protectionist or function equivalently to import or export subsidies. The internal consistency test can identify rules that violate competitive neutrality regardless of whether those rules are facially neutral or facially selective, and regardless of whether they are structural. Internal consistency provides the largest advantages over tax-expenditure analysis in cases involving either facially neutral rules with a disparate impact or facially selective rules where the selection (such as distinguishing between residents and nonresidents in income-allocation) seems to be prima facie justified.

At the same time as helping to resolve such “hard cases” without resorting to external norms, the test also works on what I have been calling “easy cases,” namely those involving facially selective rules for which there is no prima facie justification. Consider a sectoral discrimination case in which the Court of Justice held that Italy conferred state aid when it excused the textile companies, but not other companies, from making social-security contributions on behalf of their employees.308Case 173/73, Italy v. Comm’n (Italian Textile Case), ECLI:EU:C:1974:71 ¶ 15. This was an easy case because it involved facial sectoral discrimination. So far, this Article has set aside questions of sectoral discrimination, but the internal consistency test can be helpful in such cases, too, at least by confirming whether the challenged tax advantage arises from mismatches or not. The test confirms that if every state applied the Italian rule, the textile industry would be preferred everywhere to other sectors, and thus the benefit Italian law conferred was not due to mismatches. Of course, this case also could be easily resolved under traditional methods that scrutinize state rules for facial selectivity.

But a subtler case would be harder to resolve under traditional selectivity analysis. Suppose Italy had excused all industries from social security taxes. If other Member States continued to require social security contributions, Italy’s non-standard rule might cause locational distortions—specifically, companies might prefer to employ workers in Italy compared to workers in other states. Would this be a case, like Gibraltar, where the Commission would be tempted to evaluate Italy’s rule against a normative base that included social-security taxes?

Once we universalize the rule under internal consistency so that no state imposes social security taxes, the advantage for employing Italian workers disappears, alerting the Commission to the fact that the tax advantage derives from the interaction of Italian law with the laws of other states; it does not arise from Italian law alone. We would not necessarily approve, as a policy matter, of Italy’s choice to dispense with social-security taxes, but it is not the job of the Commission to second-guess state tax policies.

In the absence of actual harmonization, the envisioned Italian rule might cause locational distortions. This is because the Italian rule would confer an effective tax rate reduction. But the prohibition of state aid does not prevent a state from making its tax or regulatory regime more attractive than that of its neighbors. It only prevents states from adopting policies that discriminate on the basis of sector or that violate competitive neutrality because they are protectionist or confer tax advantages that are equivalent to import or export subsidies.

Third, internal consistency increases legal certainty by avoiding the tax-expenditure trap. Because the Commission’s traditional tax-expenditure approach to state aid requires it to identify deviations from the state’s own domestic tax laws, the Commission and the states often dispute whether particular provisions constitute part of the domestic-law reference base or a deviation from the reference base.309See, e.g., Case C‑203/16 P, Andres v. Comm’n, ECLI:EU:C:2018:505 (June 28, 2018) (disputing whether a provision framed as an exception from a loss-carryover rule, which was itself framed as an exception to a general rule against loss offsets, constituted part of the reference base or a deviation). As the famous debate between Professors Surrey and Bittker about identifying tax expenditures amply illustrates, such baseline questions are not susceptible to easy answers.310See Bittker, supra note 78; see also Stanley S. Surrey & William F. Hellmuth, The Tax Expenditure Budget: Response to Professor Bittker, 22 Nat’l Tax J. 528, 530–31 (1969) (discussing the challenge of establishing the baseline). Because the internal consistency test can universalize the challenged state’s entire tax regime, it does not require the Commission to determine which parts of the regime constitute the baseline, and which parts constitute the deviations. It simply asks whether, if all the states adopted the same regime, a cross-border advantage would accrue to a selective class. If so, the state would be required to offer a justification for the selective advantage to avoid an adverse state-aid determination.

Fourth, the requirement under internal consistency to tax all income (as defined under domestic law) for both domestic and cross-border taxpayers is different from a single-tax requirement. Although it requires any particular state’s rule to fully allocate the income of multinational enterprises if adopted by all the states, because it allows states’ internally consistent regimes to differ from each other, internal consistency does not require single taxation in the real world. Thus, Illinois’s and Iowa’s internally consistent apportionment rules can exist side-by-side, and Gibraltar’s payroll-and-property tax regime can coexist with other states’ comprehensive corporate taxes using separate counting and arm’s-length. All four regimes can even operate simultaneously in different EU Member States without giving rise to illegal state aid, even though their simultaneous application to the same taxpayer might result in under-taxation or over-taxation in the real world.

Fifth, where internal consistency reveals a cross-border tax advantage, it also reveals the extent of that advantage without relying on an external norm. While this may seem like a small point, it is important for fair treatment of taxpayers. In cases involving facial selectivity, the Commission estimates the recovery due from the taxpayer by reference to the offending state’s generally applicable domestic law.311See, e.g., McDonald’s Final Commission Decision, supra note 194, ¶¶ 111–12. The amount recovered from the favored taxpayer is the difference between taxation under regularly applicable domestic law and taxation under the special rule. But in recent cases in which the Commission used an external norm, it calculated recovery by reference to the external norm, not domestic law.312See, e.g., Amazon, supra note 13, ¶ 244; Apple, supra note 2, ¶ 447. Fairness requires the recovery to be calculated in similar fashion for all taxpayers. Because internal consistency never uses an external norm, any tax advantage it identifies is simply the difference in tax treatment under domestic law between two groups. If one of those groups is a selective class, and the difference is unjustified, the state must recover that difference from the favored class. But the amount of that difference would be established by reference to the tax treatment of the non-favored class, precisely as is currently the Commission’s practice with facially selective rules.

Sixth, the internal consistency test increases predictability because states and taxpayers can readily determine whether taxes violate it without wondering about the content of an unstated external norm. In contrast, wider use of external benchmarking would increase states’ and taxpayers’ vulnerability to the Commission’s I-know-it-when-I-see-it approach to state-aid enforcement, in which, for example, the Commission has felt free to retroactively apply OECD guidance in Ireland.313See supra note 180 and accompanying text.

V.   Beyond State-Aid Enforcement

Compared to benchmarking by external norms, internal consistency may seem unsatisfying because it allows states to retain regimes that create distortions through mismatches, including those that poach business from neighboring states. For example, internal consistency confirms that the Commission was correct to allow the Irish tax-residence rule and the Luxembourg tax-treaty interpretation to stand. But both of these rules resulted in significant tax advantages for multinationals in the real world.314See supra Section III.C.1. And, because of both countries’ well-known reputations for writing business-friendly tax laws, we can safely assume that Ireland and Luxembourg not only were aware of these tax savings but that they also arranged their tax systems to create them.

Moreover, internal consistency would change the result in Gibraltar. The Commission found state aid, and the Court of Justice affirmed, but the proposed Gibraltar payroll-and-property tax was internally consistent. If applied by all the states, it would allocate all the income of multinationals, leaving no gaps or overlaps. The regime therefore does not violate competitive neutrality by functioning equivalently to a tariff or an import/export subsidy. Instead, the Gibraltar regime confers advantages only through mismatches with other states’ laws. The CJEU has consistently held that mismatches do not violate EU law. Notwithstanding its internal consistency, however, the very purpose of the proposed Gibraltar regime was to make Gibraltar more attractive to multinationals relative to other Member States. Under the approach advocated in this Article, internally consistent tax gaps and overlaps represent a cost of tax diversity, and no particular state can be held responsible for higher or lower taxes that result from the interaction of two states’ internally consistent rules. Moreover, the most obvious ways to resolve such mismatches—legislation or international coordination—face political obstacles. This Part reviews potential responses to this problem.

A.   Accept Tax Competition

One approach to tax competition arising from tax base diversity is simply to accept it. When they enter into anti-subsidy agreements, states typically reserve for themselves significant control over their tax systems because they value tax autonomy, diversity, and competition.315See supra Section I.A. Thus, it is proper to interpret the prohibition of state aid to limit, but not completely quash, tax competition. An interpretation of state aid that facilitates some tax competition would ensure that, as economists advise, small states that have trouble competing with large ones for tax base and location of productive factors can adopt uniformly applicable tax incentives to improve their relative positions.316See Keen & Konrad, supra note 54, at 273–74 (reviewing tax competition literature). Similarly, states that keep their tax rates low by governing efficiently will have comparative advantages in competing for productive factors, as will states whose voters prefer smaller public sectors. In addition to fidelity to the language of the Treaty and its interpretation by the CJEU,317See supra Section I.A. an interpretation of state aid that accommodates some tax competition is also consistent with EU Parliament’s tax policy preferences. As a 2002 report by the Parliament observed,

tax competition is not at odds with the completion of the internal market, which does not entail a total levelling-out of competitive conditions in each country and certainly not those relating to taxation . . . [taxation] is an issue which is internal to each country.318Report on the Commission Communication on Tax Policy in the European Union—Priorities for the Years Ahead, 6, 7 (Feb. 22, 2002), http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//NONSGML+REPORT+A5-2002-0048+0+DOC+PDF+V0//EN [https://perma.cc/U4KQ-64WY] (“[T]ax competition may in itself be an effective instrument for reducing a high level of taxation.”).

In contrast, by requiring conformity with an external norm, the Commission’s new approach to state-aid enforcement would eliminate not only harmful tax competition but also legitimate variation in tax policy choices, inhibiting the expression of voter preferences through tax diversity and experimentation among the states.

It is also important to note that internal consistency shares important features with tax-expenditure analysis that uses a reference-law benchmark. Although neither approach can eliminate tax competition, both increase the costs to states from implementing competitive tax policies. Compare two policies. First, a state might provide an accelerated depreciation rule that encourages companies to increase economic activity by purchasing and putting into service new depreciable assets. To attract economic activity from its neighbors, the country would want to limit the accelerated depreciation rule to nonresidents. In international tax circles, a regime that is available only to nonresidents or that otherwise insulates the domestic economy from its effects is “ring-fenc[ed].”319OECD, Harmful Tax Competition: An Emerging Global Issue 21 (1998). Ring-fencing helps keep the costs of competitive tax practices—in terms of tax revenue losses—low. Alternatively, the state could grant accelerated depreciation to everyone—foreign and domestic. This alternative regime might still attract business from higher-tax states, but it costs more. The first regime would be state aid under both reference-law benchmarking and internal consistency, while the second would not be state aid under either approach.320The first regime would provide a tax advantage (accelerated depreciation) to a selective class (foreign companies). The second would provide a tax advantage, but because it would be open to all, it would not be state aid.

Similarly, under either benchmark, to be able to offer multinationals a 12.5 percent corporate tax rate, Ireland must collect that—and no more—from domestic companies. To completely repeal its corporate tax, a state would need to either raise taxes by other means, borrow, or cut spending. By requiring attractive tax regimes to apply equally across selective and non-selective classes, both reference-law benchmarking and internal consistency discourage tax competition by increasing its costs. But neither approach completely bans tax competition; each bans only competitive tax practices that discriminate between cross-border and domestic commerce.

B.   Limit Tax Competition by Law

Adopting internal consistency for identifying subsidies need not imply capitulation on all internally consistent tax practices. It simply acknowledges that the appropriate venue for combatting tax competition arising from them is legislation and tax treaties, rather than state-aid enforcement.

Readers that support both closing loopholes and limits on the discretion of supranational subsidy adjudicators will find recent international tax developments encouraging. For example, as part of the G20/OECD Base-Erosion and Profits-Shifting (BEPS) Project, a large number of EU and non-EU countries321The G20/OECD BEPS Project’s “inclusive framework” includes 115 countries, including all the EU countries. See About the Inclusive Framework on BEPS, OECD, https://www.oecd.org/tax/beps/beps-about.htm [https://perma.cc/K8CR-LFQB]. Not all 115 countries in the inclusive framework have implemented the BEPS recommendations. committed to adopt hard-law reforms to counter widespread corporate tax avoidance, including certain mismatches322OECD/G20 Base Erosion and Profit Shifting Project Explanatory Statement: 2015 Final Reports, OECD 9 (2016) https://read.oecd-ilibrary.org/taxation/beps-project-explanatory-statement_9789264263437-en [https://perma.cc/7MS8-8QQA#page1]. and transfer-pricing abuses.323Reforms directed at income-allocation abuse include peer reviews of state tax regimes to determine whether they contain harmful tax practices, automatic exchange among national tax authorities of certain types of formerly secret tax rulings (including the type at issue in the recent state-aid cases), amendments to the OECD Transfer Pricing Guidelines to make them more robust against taxpayer strategies, country-by-country income-reporting rules that will prevent companies from keeping states in the dark about their activities and reported income in other states, and expanded transfer-pricing documentation requirements for taxpayers Id. at 6–9. None of these reforms was in effect at the time the facts of the cases discussed in this Article arose. Many of the BEPS recommendations have already been adopted into EU law, tax treaties, and domestic law.324See BEPS Actions Implementation Matrices, Deloitte (2019), https://www. deloitte.com/global/en/pages/tax/articles/beps-action-implementation-matrices.html [https://perma.cc/J6EH-PKU8] (providing BEPS implementation by country). Whereas the Commission lacks both authority and capacity to address corporate tax avoidance arising from mismatches between internally consistent rules, state legislatures possess both. That the EU states have unanimously agreed to coordinate legislative responses to corporate tax avoidance suggests that back-door tax legislating by the Commission using its state-aid powers is not necessary.

States may also resolve mismatches unilaterally. Ireland has addressed the tax-residence mismatch in Apple. New Irish rules tax formerly stateless Irish-incorporated subsidiaries on their worldwide income.325Apple, supra note 2, at 7 n.12 (describing new Irish tax-residence rules). And U.S. rules passed in 2017 address income-shifting by imposing minimum taxation on the income of U.S.-parented foreign subsidiaries, even when those subsidiaries pay no dividends to their U.S. parents.326I.R.C. § 951A (Supp. V 2013-2018) (minimum taxation of global intangible low-taxed income); § 965 (repatriation tax). Under these new rules, the United States will tax the past and future income of stateless subsidiaries of U.S. companies. Luxembourg has even agreed to revise its approach to tax-treaty interpretation to prevent the kind of nontaxation at issue in McDonald’s.327European Commission Press Release IP/18/5831, Commission Investigation Did Not Find that Luxembourg Gave Selective Tax Treatment to McDonald’s (Sept. 19, 2018) (quoting Commissioner Vestager as stating, “I very much welcome that the Luxembourg Government is taking legislative steps to address the issue that arose in this case and avoid such situations in the future”).

C.   Adopt an “Undue Benefit” Safety Valve, or, How to Solve a Problem like Gibraltar

Another way to handle internally consistent, but distortive tax rules—including mismatches—might be to give the subsidy reviewer limited discretion to eliminate them in carefully prescribed circumstances. For example, dissenting in Moorman, Justice Powell argued that the Supreme Court should have imposed three-factor apportionment on Iowa to harmonize Iowa’s allocation rule with that of the other states. In his view, if one state’s rule was “out of line” with those of other states, such that it burdened interstate commerce, the Court should invalidate the variant rule unless the state could justify it with a legitimate local purpose that outweighed the harm.328Moorman Mfg. Co. v. Bair, 437 U.S. 267, 295–96 (1978) (Powell, J., dissenting) (internal citations omitted). Justice Powell conceded that he was unsure how much uniformity would be required before the Court could impose a dominant rule on an outlier state. He stated that “there can be no rule of 26 States, of 35, or of 45.”329Id. Nevertheless, Justice Powell argued that “it is not only proper but essential to determine the validity of the Iowa formula against the background of practices in the other States.”330Id. at 295.

Justice Powell’s suggestion was far from unprecedented.331Among other cases, Justice Powell cited Bibb v. Navajo Freight Lines, Inc., 359 U.S. 520, 529–30 (1959). In Bibb, the Court invalidated under the dormant Commerce Clause a nondiscriminatory state mudguard regulation that differed from the mudguard regulations of 45 other states; the Court concluded that the deviating regulation imposed an “undue burden” on interstate commerce. Under so-called Pike balancing, also known as “undue burden” analysis, the Supreme Court sometimes strikes down rules that it concludes unduly burden interstate commerce, even if those rules do not discriminate.332Pike balancing” comes from Pike v. Bruce Church, Inc., in which the Court found a statute that would require a company to locally pack produce grown in Arizona unduly burdened interstate commerce. 397 U.S. 137, 145 (1970). The Court found that even though the law did not discriminate on its face against interstate commerce, the burden the regulation placed on interstate commerce was not justified by the state’s interest, which was to ensure proper identification of the state of origin of the produce. Id. at 146. One could imagine the Commission or EU courts regarding as state aid an internally consistent rule that was so far out of line with what the majority of other states were doing that the rule created “undue benefits.” An undue-benefits approach would provide the EU with a non-legislative fix for extreme disparities, perhaps including that seen in Gibraltar.333This could be an important safety-value because EU tax legislation requires unanimity. As in Bibb, the U.S. Supreme Court occasionally has resolved facially neutral mismatches by bringing outlier states into conformity with dominant rules. See, e.g., Kassel v. Consolidated Freightways Corp., 450 U.S. 662, 671 (1981) (holding a state could not divert interstate trucking traffic by adopting a shorter truck-length limit than that of its neighbors even if the length limit did not facially discriminate). In Bibb and Kassel , litigants challenging state rules provided ample evidence that the minority rule was either more dangerous or, at best no safer, than the dominant rule. Id.

Careful consideration would have to be given to the proper use of any such doctrine because it would transfer tax power to the Commission. Too liberal a use of undue-benefits analysis could swamp the primary definition of state aid as involving discrimination, essentially recreating the type of excessive Commission lawmaking criticized in this Article. Tellingly, for example, even though forty-four states used three-factor apportionment at the time of Moorman, a majority of the Supreme Court was unwilling to impose that formula on Iowa.334See Moorman, 437 U.S. at 283. At that time, only forty-five states had comprehensive income taxes. Id. This suggests the need for great care before interfering with state income-allocation rules. Indeed, the Court of Justice seems uncomfortable with the implications of its own decision in Gibraltar. It even took the unusual step of limiting its precedential value.335Joined Cases C-20/15 P & C-21/15 P, Comm’n v. World Duty Free Group, ECLI:EU:C:2016:981 ¶¶ 71–77 (Dec. 21, 2016) (limiting the precedential value of Gibraltar to cases in which the Commission challenged an entire tax regime).The CJEU’s emphasis on legislative intent in Gibraltar suggests one way to cabin Commission discretion under undue-benefit analysis; it could be limited to cases in which the statute on its face suggests discriminatory intent.

Conclusion

This Article establishes that identifying tax subsidies through tax-expenditure analysis creates a double bind. When it relies on a domestic-law reference base, tax-expenditure analysis is underinclusive. For example, under a domestic-law benchmark, the subsidy adjudicator would regard no structural rules as conferring illegal subsidies, no matter their actual impact on cross-border commerce. But when tax-expenditure analysis relies on a reference base that is external to the challenged state’s domestic law, it is overinclusive. Benchmarking by external norms mistakes mismatches for state aid. Benchmarking by norms also substitutes the tax policy views of the subsidy adjudicator for those of the legislator. This double bind is not unique to the European Union. It attaches to any legal regime that uses tax-expenditure analysis to identify subsidies, including WTO subsidy review.

In contrast, the internal consistency test poses no such double bind. Developed by the Supreme Court for identifying tax discrimination that violates the dormant Commerce Clause, internal consistency accurately distinguishes nondiscriminatory tax diversity from discriminatory subsidies that affect the free flow of commerce across state borders. Moreover, the test does so without relying on benchmarks external to the challenged state’s law. It therefore properly defers to Member State tax policy choices.

Now is a defining moment for state-aid. As this Article explains, rejecting the Commission’s new external-benchmarking approach does not necessarily require vacating the Commission’s recovery orders in Amazon, Apple, Fiat, or Starbucks, because in those cases, the Commission claimed that in addition to deviating from the Commission’s chosen external norm, the defendant states also deviated from their own domestic allocation rules. To the extent that the EU courts agree that the Commission carried its burden to show that the states deviated from their own laws, the EU courts can affirm the outcome in those cases while rejecting the Commission’s new approach. This Article argued, however, that the EU courts should reject the Commission’s imposition of external norms as the reference base for determining whether Member States conferred state aid.

Rejecting the external benchmarking approach the Commission introduced in Apple and the other recent cases has a happy byproduct; it may save the Commission’s star tax legislative proposal. For many years, the Commission’s tax department—which is different from the competition department that enforces the prohibition of state aid—has advocated for the states to harmonize their corporate tax bases and move from separate accounting to formulary apportionment, which would make EU taxation resemble state-level taxation in the United States. The proposal would require all the states to use formulary apportionment, rather than separate accounting at arm’s length, to allocate income from transactions within the European Union.336See CCCTB Relaunch, supra note 241, at 3. But if we take seriously the Commission’s assertion in Apple and the other cases that the state-aid rules require states to allocate income according to the Commission’s arm’s-length standard, then, at a minimum, unilateral state adoption of formulary apportionment would be illegal, at least if (as contemplated in the Commission’s proposal) formulary apportionment would apply selectively to multinational, but not domestic, companies.337See id. at 17–18 (proposing to apply the CCCTB only to multinational corporate groups).

It is worth emphasizing that this Article does not contend that Member States or U.S. multinationals are innocent victims of unwarranted state-aid investigations by the Commission. Information revealed in public hearings held by the U.S. Senate and U.K. Parliament provided ample reason for the Commission to be suspicious that EU Member States might be granting U.S. multinationals sweetheart tax deals. Rather than defending the states and multinationals, this Article took issue with the Commission’s novel legal approach to establishing state aid in those cases. It argued that the Commission’s approach was unnecessary to reach the results in the recent cases, unpredictable, illegitimate, and excessively constrained state tax autonomy. There is a better solution: the European Union could adopt the Supreme Court’s internal consistency test for identifying tax subsidies.