Press "Enter" to skip to content

Some observations on Starbucks, Fiat, and their potential impact on future amendments to the arm’s length principle


Jérôme Monsenego (Stockholm University)/September 28, 2019

The General Court of the European Union has issued two awaited rulings in the Starbucks[1] and Fiat[2] cases. The length and the depth of the analysis made by the judges of the General Court should be acknowledged, even if certain key issues are perhaps too rapidly dealt with. Although the Commission lost in Starbucks, the cases can be interpreted as a victory for the Commission and several of the arguments it has been relying on since these cases were initiated. These cases also point to potential conflicts between the State aid rules and the amendments to the arm’s length principle currently considered by the OECD so as to move part of the corporate tax base to market jurisdictions. It will, however, be necessary to wait until the CJEU rules on these issues to have a clearer view on the relation between the State aid rules and both the arm’s length principle and other tax rules. This blog post aims at emphasising certain important points made in the two judgments, together with considering their potential impact on the ongoing project in relation to the tax challenges of the digitalization of the economy.

To start with, the General Court confirmed the right of the European Commission to assess the correct application of the arm’s length principle by the Member States on the basis of article 107(1) of the TFEU, even if the measure assessed is an advance tax ruling or an advance pricing agreement. This point was already made in the judgment issued on 14 February 2019 on the excess profit regime of Belgium[3], and is now clearly motivated by the traditional statement that “while direct taxation, as EU law currently stands, falls within the competence of the Member States, they must nonetheless exercise that competence consistently with EU law”[4]. It is then interesting to wonder how the State aid control should be performed, in particular concerning the benchmark with which a national measure must be assessed, ie the notion of “normal taxation”. Here the General Court made an important statement, as it considered that “the Commission does not, at this stage of the development of EU law, have the power autonomously to define the ‘normal’ taxation of an integrated undertaking, disregarding national tax rules”[5]. This statement might come as a relief to lawmakers and other observers, including the parties to the Apple case, as it can be wondered whether or not the State aid rules, as such, imply an obligation to legislate in a certain manner and include certain rules in a tax system. In other words, this statement seems to indicate that the reference system should only be made of the domestic law of a Member State. That would contradict the argument made by the Commission in several decisions as well as in the 2016 notice as to the intrinsic obligation of the State aid rules to apply the arm’s length principle.

However, the reasoning of the General Court is somewhat confusing as it did not strictly stick to domestic tax law, at least not in a literal manner, both to determine the objective of the reference system,[6] and to give a material content to the arm’s length principle.[7] Concerning the objective of the reference system (with respect to which the comparability analysis has to be conducted), in Fiat the Court seemed to have rather assumed that the Luxembourg tax system did generally pursue the objective of treating independent and associated enterprises in a similar manner,[8] so as to ensure a ”reliable approximation of a market-based outcome”[9] for the latter; the arm’s length principle would then be the correct tool to reach this objective. Even if several rules of the Luxembourg tax system might pursue this objective, it is not necessarily true of all rules (certain anti-avoidance rules, for example, may only apply to associated enterprises). Also, even if a tax system pursues the objective of treating independent and associated enterprises alike, there may be different methods to achieve such a treatment. In relation to the material content of the arm’s length principle, the General Court stated that “suffice it to note that it is apparent from the contested decision that that principle is a tool for checking that intra-group transactions are remunerated as though they had been negotiated between independent undertakings”[10]. This statement is too little motivated, and the link to the reference system is too weak, although the question is central: the expression “arm’s length principle” does not have a clear material content in itself, apart from the conceptual idea of requiring some form of similar pricing or profit margins between independent and associated enterprises. The arm’s length principle can be interpreted in different manners and be given different material contents. The ever-increasing number of transfer pricing disputes, the differences between the OECD Guidelines and the UN Manual, or between domestic legislations as well as court cases, evidence the variety of views potentially embedded in the expression “arm’s length principle”. The weak motivation by the General Court of the material content of the arm’s length principle seems to confirm the argument of the Commission that article 107(1) of the TFEU in itself includes an obligation to apply the arm’s length principle, or rather an arm’s length principle, with no clear view on what material content to give to this principle. Although the wordings used by the General Court are somewhat puzzling,[11] it is difficult not to interpret these cases as a confirmation that the arm’s length principle is inherent to the State aid rules: the General Court found that it was right for the Commission to state that “the arm’s length principle was a ‘benchmark’ for establishing whether an integrated company was receiving, pursuant to a tax measure determining its transfer pricing, an advantage within the meaning of Article 107(1) TFEU.”[12] The French expression used for the word bencharmark at paragraph 143 of the Fiat case, “critère de référence”, strengthens the impression that article 107(1) of the TFEU intrinsically includes the arm’s length principle as a principle of equal or similar treatment between independent and associated enterprises. The acceptance of the reference made to the Forum 187 case[13] confirms this view.

More generally, since the General Court seems to have established the objective of the Luxembourg tax system on the basis of certain of its characteristics (in particular the separate entity approach), and given that the Court accepted that the Commission relied on the arm’s length principle as a “tool” to check whether this objective was correctly met, one possible interpretation of these cases is that as long as a certain objective can be identified in a corporate income tax system, the Commission has the right to rely on principles or “tools”[14] to test if that objective has been reached without granting illegal State aid.

Next, a key element in these cases concerns the possible comparability between independent and associated enterprises. Should they not be in a factual and legal comparable situation for State aid purposes, the claims of the Commission would fail as the Member States would have the right to treat these categories of enterprises differently. The General Court did not analyse this issue at depth, but considered that the two categories of enterprises were comparable,[15] thus accepting the arguments of the Commission. I have argued in favour of a similar view, when the comparison is made in the light of the objective of a corporate income tax system.[16]

It can also be observed that the Court correctly acknowledged the imprecision that is inherent to the arm’s length principle,[17] ie its “approximate nature”.[18]  This means that the General Court should logically accept the notion of range, advocated since long in the OECD transfer pricing guidelines. In other words, it is a recognition that transfer pricing is not an exact science. Therefore, if the notion of range is accepted under the State aid rules, there cannot be a claim that only a certain value within the range, eg the median, would be the “right” market value, a deviation from which would necessarily imply an illegal State aid. Nevertheless, the General Court does not clearly precise how to determine the arm’s length range in a manner that is compliant with the State aid rules, ie how to distinguish between an inaccuracy “inherent in the application of a method designed to obtain a reliable approximation of a market-based outcome”, and one that constitutes an incorrect application of the arm’s length principle and a potential illegal State aid. Again, this issue brings us to the determination of the reference system as well as its interpretation. The General Court was also correct in finding that the choice of a transfer pricing method does not, as such, imply an illegal State aid, since the methods described in the OECD transfer pricing guidelines all aim at implementing the same principle.[19]

When it comes to determining which provisions to take into account to assess a possible deviation from the reference system, the General Court made clear that only information existing at the time of an APA should be taken into account.[20] This should reasonably be valid both for arguments in law and arguments in facts, since a correct transfer pricing analysis implies to determine a certain set of facts and apply certain transfer pricing principles to this set of facts. Therefore, it should be the reference system as it was at the time of the enactment or the issuance of a tax measure that is used as a benchmark to determine what should be the “normal” taxation. This precludes, among others, the use of a later version of the OECD transfer pricing guidelines to assess the correctness of an earlier transfer pricing measure, at least if it is established that there is a material difference between different versions of the guidelines. That is the case, for example, when it comes to hard-to-value intangibles and the procedure described at paragraphs 6.192 and 6.193 of the 2017 guidelines.

From a transfer pricing perspective, without going into the details of the Starbucks and Fiat transfer pricing models, it must be acknowledged that the General Court really tried to analyse the issues at depth. This level of ambition is seldom found in the rulings issued by administrative or tax courts around the globe, although such courts are more used to dealing with tax law and transfer pricing. To take one example, the denial of deduction claimed by the Commission for the royalties paid by the Dutch Starbucks manufacturer for the use of roasting intangible property could hardly be entirely justified, given the apparent lack of functions related to the development of such intangible property in the Netherlands. Even if the recipient of the royalties may not either have entirely performed the functions related to the development of the intangible property, this is not a sufficient reason for fully denying deduction of the payment and thus taxing the Dutch entity as if it was the IP owner, at least not without demonstrating that the Dutch Starbucks manufacturer did have some substance supporting the development of roasting intangible property, that the recipient was not the IP owner and that the royalty payments were not passed on to the IP owner. The General Court was thus right in finding that “SMBV’s payment of a royalty to use the roasting IP is not devoid of all economic rationality”.[21]

To conclude, since transfer pricing is found by the General Court to be able to be assessed in the light of the State aid rules, an important matter will – similarly to traditional transfer pricing cases – relate to the burden of proof as to the correctness of the transfer prices or profit margins applied. However, no matter how well future cases will be argued, a crucial issue will remain with respect to how to determine the norm used as a benchmark, as well as how to interpret this norm. The General Court did not clearly answer this question, although it found the OECD transfer pricing guidelines non-binding but still relied on them.[22] Should the CJEU confirm the findings of the General Court, it is hoped that it will provide more precisions as to the determination of the material content of the reference system, in particular whether or not the reference system should be strictly limited to the domestic law of a Member State.

A last observation concerns the current developments at the OECD and G20 level. It is interesting to wonder how these State aid cases might influence the developments with regard to the modifications to several rules currently envisaged in the OECD inclusive framework so as to adapt the international tax system to the digitalization of the economy. Assuming the CJEU confirms the findings of the General Court, State aid law would probably be requiring a similar treatment between independent and associated enterprises so as to ensure the taxation of associated enterprises on a “market-based outcome”, at least for the Member States that have implemented the arm’s length standard in their domestic laws. However, the OECD is currently working on a fundamental change to the transfer pricing rules, the purpose of which is to allocate a portion of the residual profits to the market jurisdictions, ie the countries where sales are made or where users are located.[23] A relief from tax in the country of origin would per definition not apply to independent enterprises or to domestic groups. Therefore, it can be wondered if the envisaged changes do not amount to a selective advantage in the country of origin to the benefit of multinational enterprises, at least those that export goods or services, or have users abroad. In the country of destination, the new nexus rules will mostly target certain business models, especially if thresholds are being applied. This may result in a selective advantage to the benefit of undertakings that will not have a taxable presence, or for which little or no income will be allocated to the newly created taxable presence. These potential conflicts with the State aid rules are not surprising, since the purpose of pillar 1 of the BEPS 2.0 project is to design tax rules that better capture the income earned by multinationals in the market jurisdictions, especially in the digital sector. There may also be a conflict with the fundamental freedoms, especially in the market jurisdiction where the tax base will increase, since the future amendments to the arm’s length principle will not apply in domestic situations, and might not be justified by the need to prevent tax avoidance or safeguard a balanced allocation of the power to impose tax.[24] In other words, while the neutrality that is intrinsic to the arm’s length principle might prevent an incompatibility with EU law and even be an obligation under the State aid rules, deviations from the arm’s length principle might imply a conflict with EU law.

Jérôme Monsenego is full professor of international tax law at Stockholm University. He is the author of “Introduction to Transfer Pricing” (Wolters Kluwer 2015), and “Selectivity in State Aid Law and the Methods for the Allocation of the Corporate Tax Base” (Wolters Kluwer 2018).


[1] T-760/15 and T-636/15 (hereafter referred to as Starbucks).

[2] T-755/15 and T-759/15 (hereafter referred to as Fiat).

[3] T-131/16 and T-263/16.

[4] Fiat, §104; Starbucks, §142.

[5] Fiat, §112.

[6] Fiat, §141 and §145.

[7] Fiat, §147.

[8] Fiat, §141 (referring to “national tax law” in abstracto), or §145 (assuming that the Luxembourg Tax Code “is intended to tax” in a certain manner, without however substantiating how this intention is established or what material content to confer to it).

[9] This expression, initiated by the Commission, is also used by the Court: see Fiat, § 176.

[10] Fiat, §155.

[11] See e.g. Fiat, §153, where it is mentioned that “the arm’s length principle is being applied in the context of the examination under Article 107(1) TFEU” (emphasis added). The statement is even more puzzling in the French version of the case: “le principe de pleine concurrence intervient dans le cadre de l’examen au titre de l’article 107, paragraphe 1, TFUE” (emphasis added).

[12] Fiat, §143.

[13] C-182/03 and C-217/03.

[14] The meaning of the word ”tool” is rather vague, but is used several times by the General Court (e.g. Fiat, §151 and §159).

[15] Fiat, §141; Starbucks, §149.

[16] Jérôme Monsenego, Selectivity in State Aid Law and the Methods for the Allocation of the Corporate Tax Base (Wolters Kluwer 2018) 75-101.

[17] Starbucks, §199.

[18] Fiat, §207.

[19] Starbucks, §202, §209, and §211. For a similar view see Jérôme Monsenego, Selectivity in State Aid Law and the Methods for the Allocation of the Corporate Tax Base (Wolters Kluwer 2018) 125.

[20] Starbucks, §251.

[21] Starbucks, §262.

[22] Fiat, §173.

[23] See the programme of work: http://www.oecd.org/tax/beps/programme-of-work-to-develop-a-consensus-solution-to-the-tax-challenges-arising-from-the-digitalisation-of-the-economy.pdf

[24] C-382/16.