Ding-Dong! The EU Arm’s-Length Standard Is Dead

Ruth Mason

Ruth Mason is the Edwin S. Cohen Distinguished Professor of Law at the University of Virginia School of Law and director of the Virginia Center for Tax Law. She thanks Stephen Daly for his helpful comments.

In this article, Mason argues that the recent decision by the Court of Justice of the European Union vacating the European Commission’s state aid decision in Fiat is a positive development for the rule of law and member state tax sovereignty.

On November 8 the Court of Justice of the European Union decided the appeal in Fiat,¹ one of the state aid transfer pricing cases. Sitting en banc, the Court vacated the European Commission’s decision finding that Fiat owed Luxembourg tax.² More importantly, the Court rejected the commission’s most audacious theory in the recent transfer pricing cases — that the commission could apply its own conception of arm’s length to check member state transfer pricing rulings, regardless of member state domestic law. This reasoning was crucial to the commission’s decision in Apple,³ so the CJEU’s decision in Fiat suggests that the commission will lose its appeal in Apple.

I. Arguments on Appeal

In 2015 the commission decided that Luxembourg had granted illegal state aid to Fiat’s Luxembourg financing subsidiary, FFT. As was typical of these cases, the form of the aid was a ruling that supposedly allowed FFT to report too little income to Luxembourg. But how did the commission know FFT reported too little income? Too little compared to what?

The main issue in the transfer pricing state aid cases, including Fiat (and Apple), was the baseline the commission used to determine whether the companies paid too little tax. In the more run-of-the-mill decisions that preceded the transfer pricing decisions, the commission consistently used a tax-expenditure-style approach to identify state aid. Specifically, it took the accused member state’s domestic tax law as a reference base and then asked whether a particular provision of domestic law deviated from that base in a manner that conferred a selective advantage. To make a long story short, tax savings that deviated from the domestic law reference base and were available only to a particular company (as through a ruling) or to only specified companies would constitute a selective advantage.⁴ Of course, we all know that tax expenditure analysis is more art than science, and states and the commission wrangled over whether particular provisions should be considered part of the reference base or a deviation from it. But the commission’s commitment to judging member state provisions only by reference to domestic law kept those disputes local; it prevented them from gaining wider attention.

Enter the transfer pricing disputes, which raised questions of whether member states conferred tax advantages through rulings for individual taxpayers. Under the traditional approach to state aid, all the commission needed to do was review whether the member state properly applied its allocation rules — whatever they were — to the case at hand. Fatefully, that’s not what the commission did, and why it didn’t do so requires an explanation. My surmise is that the commission really set its sights on Apple, not Fiat, Starbucks, or Amazon. But the problem with Apple was that because that dispute involved a branch, rather than a subsidiary, it was unclear whether the arm’s-length standard applied. Ireland claimed that it did not use the arm’s-length standard to attribute profit to the branches of nonresident companies at the time the facts of Apple arose. Instead, it used a method that considered only the activities in the Irish branch (which were not substantial) and remunerated them accordingly. For the commission to allocate to those Irish branches the large portion of Apple’s global profit that it thought should be there, the commission needed to go beyond Irish tax law.⁵

That’s how I figure we got to what I have called sui generis arm’s length.⁶ Hoping to find that Ireland violated the state aid rules in Apple, the commission tried to firmly establish the notion of EU arm’s length in all the other (lower-stakes) cases. Thus, in anticipation of Apple, the commission insisted on applying the EU arm’s-length standard, even in cases in which the member state had expressly incorporated arm’s length into its domestic law.⁷ After asserting its prerogative to apply its own arm’s-length standard, which applied independently of domestic law, the commission actually ended up judging the various rulings by OECD standards.⁸

By the time of the Fiat appeal, the commission had essentially two legal arguments remaining for its independent arm’s-length standard.

First, the commission reasoned that if Luxembourg sought to tax both stand-alone and group companies on their profits (a claim to which no party objected) and Luxembourg taxed stand-alone companies on their market profits, the commission was justified in applying the arm’s-length standard to determine the income of group companies because arm’s length is also a market-based standard.⁹ Under this reasoning, it didn’t matter whether Luxembourg expressly incorporated the arm’s-length standard into its domestic law. That Luxembourg’s purpose was to tax profits triggered sui generis arm’s length. The only relevant question was whether Luxembourg sought to tax group companies on their profits. If so, those profits had to be determined by the EU arm’s-length standard despite anything to the contrary in domestic law. Specifically, it didn’t matter that Luxembourg had its own regulations governing arm’s length.

In contrast, Luxembourg argued that the reference base should include national law and practice, including administrative guidance on how to determine arm’s-length remuneration for intragroup financing. But the commission insisted on its own arm’s-length standard. Moreover, as part of the sui generis state aid standard, the commission considered itself entitled to apply OECD guidance retroactively in some cases,¹⁰ and it reserved the right to depart altogether from OECD guidance.¹¹ Thus, the commission urged that state aid review was limited by neither domestic law nor internationally agreed standards.

The commission had a second argument that it said supported sui generis arm’s length. It argued that precedent in an old Belgian case, Forum 187,¹² supported application of the arm’s-length standard in state aid cases. Although it was true that the CJEU had approved the commission’s application of arm’s length in Forum 187, it was also true that Belgium had incorporated the arm’s-length standard into its domestic law. Thus, the case did not obviously stand for the proposition that the commission could apply the arm’s-length standard even in cases in which the accused member state had not incorporated it into domestic law (or had incorporated a different version).¹³

The General Court of the European Union (GCEU) had largely ratified the conclusions of the commission in Fiat.¹⁴ The high court described that reasoning as follows:

Moreover, the GCEU had also agreed with the commission’s reading of Forum 187.¹⁶

Ireland, Luxembourg, and Fiat all appealed the GCEU’s decision, and their cases were consolidated before the CJEU.

II. CJEU Decision

The most important holding in Fiat was the Court’s rejection of the sui generis arm’s-length standard. The CJEU made it crystal clear that the commission couldn’t invent an arm’s-length standard and apply it to states, ignoring the states’ own allocation rules. On the contrary, except in unusual circumstances, like Gibraltar,¹⁷ the commission had to use the state’s domestic law as the reference base for determining whether the state conferred aid. In this way, the CJEU returned the doctrine to the status quo before the commission began its ill-fated review of transfer pricing cases.¹⁸ The states are sovereign in tax; they determine their tax bases, including their allocation rules, and the commission’s review in state aid cases is (mostly) limited to evaluating whether the member state departed from its own law (rather than some normative concept of “taxing profit”) to confer a selective tax advantage.

The CJEU faulted the lower court for approving a commission decision that applied “an arm’s length principle different from that defined by Luxembourg law.”¹⁹ It criticized the notion that an “abstract expression of . . . principle” — namely that a state seeks to tax profits — could relieve the commission of the obligation to take into account “the way in which the said principle has actually been incorporated into that law.”²⁰ The CJEU further held that the General Court infringed on the TFEU when it accepted “that the Commission may rely on rules which were not part of Luxembourg law, even though . . . [the commission] did not . . . have the power autonomously to define the ‘normal’ taxation of an integrated company.”²¹ Although acknowledging that many member states use arm’s length and the OECD guidelines, the CJEU also recognized that states are entitled to vary from those standards.²² The CJEU concluded that “parameters and rules external to the national tax system at issue cannot therefore be taken into account in the examination of the existence of a selective tax advantage within the meaning of Article 107(1) TFEU.”²³ This result stemmed not only from the state’s reserved tax powers but also from:

Because the commission had not actually analyzed (even in the alternative) whether Luxembourg had also violated its own domestic law conception of the arm’s-length standard, the CJEU held that the commission’s error in legal analysis was fatal.²⁵

Fiat is a momentous decision because it is the first high court decision to reject the notion that the commission could supply an external reference base by which to measure state aid. The General Court in Fiat and Apple had tried to find a middle path under which a member state’s use of a market-based method of allocation in domestic law enabled the commission to use the OECD guidelines as a tool to check whether the member state had applied its own market-based allocation rule correctly. The CJEU, however, did not approve the use of such external tools. On the contrary, it returned to traditional doctrine, holding that except in cases like Gibraltar, the commission had to judge member state tax rules by a reference base consisting of its own law. This reverts the doctrine to where it was before the transfer pricing cases began. As such, it is a crucial decision. And it is also correct.

A. Conceptual Failure

First, beyond Forum 187, which was at best ambiguous on the question, the commission never offered any real support for the notion that it was entitled to introduce its own arm’s-length standard. Pared down, the commission’s best argument was that countries aim to tax profits, and arm’s length determines taxable profits. But that doesn’t follow. Taxable profits are a matter of the tax base, whereas arm’s length is more a matter of allocation, of “a division of the tax base,” as the CJEU so often puts it. A desire to tax profits has no necessary connection to any particular allocation rule, although it probably implies a desire to avoid gaps and overlaps in member state rules for dividing the tax base.

Because profits can be divided among states through numerous (indeed infinite) methods, the notion that an aim to tax profits necessarily encompasses an aim to use arm’s length to divide the taxable profits was never particularly convincing. Then, using that weak argument, the commission dramatically overplayed its hand, grasping for $14 billion in back taxes from Apple on the theory that the commission — and only the commission — was entitled to say what arm’s length meant. This approach evoked the ire (and worse, the interest) of tax lawyers around the world. The commission’s reasoning just couldn’t hold up to scrutiny.

B. Separation and Reservation of Powers

The TFEU reserves to member states the entitlement to determine their tax bases and allocation rules. Although the EU can enact direct tax legislation, it can do so only unanimously, by the process provided in TFEU article 115. By declaring itself capable of substituting its own allocation rules for those chosen by the states, the commission in the recent decisions usurped the member states’ reserved powers. In his opinion advising the CJEU to annul the commission’s decision and overrule the General Court’s decision in Fiat, Advocate General Priit Pikamäe rightly characterized the commission’s use of the sui generis arm’s-length standard as “undue interference in the Member States’ tax autonomy which the Court has always carefully condemned until now.”²⁶

Hence Ireland’s complaint in Fiat that by imposing an EU-wide arm’s-length mandate, the commission engaged in impermissible harmonization outside the methods contemplated in the text and structure of the TFEU. Tax harmonization is not impossible under the TFEU, but it must take place through article 115 or through the coordinated actions of the member states. In either alternative, each member state would have a veto. That’s why there is so little EU tax legislation and why pillar 2 hasn’t passed. Application of arm’s length through article 107’s prohibition of state aid, by contrast, would represent an end run around these textual and structural protections for member state tax autonomy.

As Pikamäe noted, the CJEU has steadfastly disclaimed any entitlement (or desire) to interfere with member states’ reserved tax powers. The Court said it plainly in Fiat:

Put simply, barring unusual circumstances like those in Gibraltar, state aid analysis of transfer pricing rulings must use a domestic law reference base.

C. Legal Certainty

One of the many problems with sui generis arm’s length was that no one knew what it meant.²⁸ Because the commission insisted that it was different from domestic law and also at least potentially different from OECD guidance,²⁹ no one — not states and not taxpayers — could be certain when it was satisfied. Indeed, had the Court approved the commission’s sui generis conception of arm’s length, more problems would have ensued. Even taxpayers self-assessing under a state’s domestic law version of arm’s length could be understood to receive state aid to the extent that the EU arm’s-length standard differed from the domestic law standard. Thus, acceptance of the sui generis standard might have led to the absurd result that taxpayers against which the commission opened state aid cases would pay tax according to a different legal regime than those the commission left alone. That result would be unfair, uncertain, and untenable.

Luckily for the EU, the CJEU reined in the commission and confirmed that state aid had to be established by reference to a domestic law reference base. It also definitively put to rest the idea that Forum 187 supported the commission’s use of sui generis arm’s length by stating that Forum 187 “does not support the position that the arm’s length principle is applicable where national tax law is intended to tax integrated companies and stand-alone companies in the same way, irrespective of whether, and in what way, that principle has been incorporated into that law.”³⁰

These conclusions — along with the CJEU’s citation to what it called “the principle of legality of taxation,” which entitles the taxpayer to notice of its liability, and which the CJEU described as a part of “the legal order of the European Union”³¹ — will help constrain the commission in future state aid decisions. In these respects, Fiat must be understood as a significant win for the rule of law.

III. Implications of Fiat

A. Apple (and Amazon)

Apple and Ireland won their case at the General Court, but the commission appealed.³² Similar to Fiat, in Apple the lower court accepted that the commission could apply the OECD’s conception of arm’s length. But the General Court still vacated the commission’s decision because the commission had failed to prove that any deviation Ireland had made from arm’s length conferred a tax advantage on Apple. Thus, in the lower court’s view, it was the application, rather than the applicability, of the arm’s-length standard that the commission got wrong in Apple.³³ The lower court in Apple also imposed an extremely high burden of proof on the commission to prove that the member state’s misapplication of the law resulted in tax savings for the taxpayer. The commission in its Apple appeal before the CJEU now faces a doubly hard road: According to the court of justice in Fiat, its theory was wrong, and according to the General Court in Apple, its application was wrong.

When it comes to theory, the CJEU’s decision in Fiat makes clear that the commission cannot use the arm’s-length standard unless it is part of the challenged state’s domestic law. Thus, the decision in Fiat improves the chances of success for Apple and Ireland on appeal because the commission in that case applied its sui generis arm’s-length standard. Fiat is a major victory for Apple and Ireland because, as I mentioned earlier, Ireland claims that domestic law allocates relatively little of Apple’s system profit to Ireland. Thus, if the CJEU sticks to its holding in Fiat that transfer pricing rulings must be judged by domestic law, Apple and Ireland should win their appeal. Although Ireland may have made errors in applying its own branch profit allocation rule to Apple, rectifying any such errors could not result in a deficiency anywhere near €14 billion. Moreover, for Ireland to secure any recovery at all from Apple, the commission would have had to prove that Ireland failed to properly apply its own standard (rather than the sui generis arm’s-length standard). Thus, the commission’s big gamble — that pushing sui generis state aid in the low-value cases would enable it to win Apple — seems as if it will result in defeat in both. Because the commission used the same logic in Amazon to apply its sui generis arm’s-length standard, the commission presumably also will lose its appeal in that case.³⁴

B. Gibraltar

The case that set the commission on its disastrous path in the transfer pricing cases was Gibraltar.³⁵ After Gibraltar, one could almost forgive the commission for concluding that state aid was whatever the commission wanted it to be. In Gibraltar, the CJEU approved the commission’s decision that an entire corporate tax regime was state aid. Gibraltar’s regime would have taxed companies on their profits based on the proportion of the company’s payroll and property in Gibraltar. Although the regime was not facially selective, as applied, offshore companies would pay very little tax (because they had very little payroll and property in Gibraltar), while onshore companies would pay more. This was not by accident. The regime was Gibraltar’s replacement for an expressly discriminatory (ring-fenced) regime that the commission had already found to convey illegal state aid.

The problem with Gibraltar from the commission’s perspective was that the traditional reference base approach produced no selective advantage. Under the terms of the law, the same regime applied to all taxpayers the same way. There was no derogation from the domestic law reference base that advantaged offshore companies. But as applied, offshore companies would benefit, and this was by design. Faced with Gibraltar’s attempt to flout the state aid rules, the commission dispensed with the reference base, and it declared the entire regime to be state aid. The General Court vacated the commission decision on the grounds that the commission had not shown a derogation from a domestic law reference base.³⁶ However, the CJEU reinstated the commission’s decision, concluding that the commission could show that a state conveyed state aid not only by derogating from its domestic law but also by pointing to the practical effect of domestic law.³⁷ Given the commission’s success in Gibraltar, it should perhaps have been no surprise that it dispensed with the domestic law reference base in other cases.

Although the CJEU has previously limited Gibraltar, it cited the case favorably in Fiat. The Court observed that “even a measure which is not formally a derogation and founded on criteria that are in themselves of a general nature may be selective, if it in practice discriminates between companies which are in a comparable situation in the light of the objective of the tax system concerned.”³⁸ In my view, Gibraltar was wrongly decided³⁹ and led to the current chaos, but the CJEU makes abundantly clear that it is still good law. What this endorsement means remains to be seen, but one possibility, considered next, is that a member state’s allocation rules (as compared with its application of those rules) or other structural aspects of its tax base could constitute state aid.

C. Structural Rules That Confer State Aid

The recent cases did not consider whether the allocation rules themselves conferred aid.⁴⁰ Rather, the question was whether the states’ application of those rules to particular taxpayers conferred tax advantages to those taxpayers. But as I have previously observed in these pages, if a member state’s allocation rules “reliably lower income compared with what would be due from a wholly domestic company engaged in identical activities [so that] multinationals can achieve systematically lower taxes than domestic companies, there is a serious problem from a common market perspective.”⁴¹

In Fiat, the CJEU made clear that member state allocation rules do not receive a free pass simply by virtue of their status as domestic law. One of the Court’s final observations in Fiat was that a state confers illegal aid when:

This dictum suggests that allocation rules themselves can constitute state aid. An example of an underallocating rule might be a one-sided transfer pricing method. Application of one-sided transfer pricing methods (including the type evidently used by Ireland when the facts of Apple arose) could systematically undertax multinational groups relative to domestic groups and companies.

But analyzing transfer pricing rules as delivering systematic aid would be challenging because allocation rules are what tax expenditure analysts refer to as structural rules.⁴³ In the United States, estimators typically assume that allocation rules are part of the reference base, and therefore they do not confer tax expenditures. But automatically assuming that structural rules are part of the reference base would give structural rules (including allocation rules) a free pass under EU state aid law, which the CJEU does not want to do. Thus, the commission will need to develop new methods to analyze allocation rules themselves for consistency with article 107.⁴⁴

IV. Conclusion

If you have been following my articles on state aid so far, you know that I am no fan of the commission’s sui generis state aid standard. Fiat represents an important loss for the commission, but I prefer to see it as a significant win for the rule of law in the EU. With this decision, the CJEU showed that it was not beholden to politics but rather took a principled approach that respects the TFEU, member state sovereignty, and fundamental fairness issues related to notice to taxpayers.

At the same time that I have opposed the commission’s sui generis state aid standard, I have also acknowledged that member states may confer illegal subsidies through both rulings and allocation rules. I have simply disagreed with the commission regarding how those state aid violations should be proved. For particular rulings, I have long argued for the approach taken by the CJEU in Fiat: Rulings should be judged by domestic law. I have argued elsewhere that structural rules, including allocation rules, should be judged by a standard of internal consistency.⁴⁵ Stephen Daly has argued that the underlying problem in the transfer pricing cases is not that the administration might make a mistake in applying the rights but rather the problem is administrative impropriety, so the commission’s approach ought to target impropriety directly.⁴⁶ Others have suggested fruitful ways forward. Thus, it has never been my contention that the commission should get out of the business of reviewing tax state aid. And I agree with the CJEU’s statement in paragraph 122 of its Fiat decision that tax rules themselves can constitute state aid.

Indeed, I think the prohibition of state aid will be more effective and promote European interests better if its underlying values were clearly articulated, giving both taxpayers and member states notice. To that end, I observe that the CJEU in Fiat did not take the opportunity to articulate what the prohibition of state aid is meant to do. While it’s clear that article 107 aims to reduce protectionist subsidies and subsidies that discriminate on the basis of sector, less clear is whether it also aims to reduce corporate tax avoidance and/or curb member state tax competition. I hope the upcoming decision in Apple will answer these crucial questions.