The EU’s arm’s length principle – does it exist and what are its effects?
The General Court confirmed that, where domestic tax law subjects both integrated enterprises and stand-alone enterprises to a tax on profits in a similar way, the domestic “reference system” should be considered to be aimed at taxing profits of group companies as if related transactions were carried out at market prices. This means that if the Commission can show that a tax ruling determines taxable profits at a lower amount than what would have followed from applying market prices, an advantage that usually gives rise to state aid is present. There can be little doubt that this must be seen as confirming the existence of an arm’s length principle directly derived from EU law. At face value, this is not exactly surprising, given the European Court of Justice’s judgment on Gibraltar’s corporate tax system in 2011. That decision – and, according to the General Court, the judgment in the Forum 187 case regarding the Belgian coordination centre regime –had already confirmed that the autonomy of member states to determine its corporate tax base rules is not unrestricted.
It appears, however, that the effects are limited for member states insofar as they apply OECD-compliant transfer pricing rules. The General Court attaches great importance to the OECD guidelines because they reflect international consensus, although it confirms that the Commission is not formally bound by them. So, for many rulings and pending state aid cases involving these member states, the primary question regarding transfer pricing aspects continues to be whether or not the ruling deviates from the OECD guidelines and in doing so benefits the taxpayer.
The confirmation of the EU arm’s length principle raises interesting questions for other matters. Do EU state aid rules de facto supplement domestic corporate tax systems in respect of other features, too? This may, for instance, be relevant for cases like Apple. The Apple case is not about transfer pricing between group companies, but about allocating assets, and consequently, taxable profits to a permanent establishment (which is not a legal entity, but a “fiction” for tax purposes) located in the relevant member state (in the Apple case, Ireland) of an entity that is tax resident of another country. The OECD has issued extensive guidelines on this as well, endorsing the “functionally separate entity” approach which, in essence, requires two steps. First, assets and risks are attributed between the head office of the legal entity and the permanent establishment based on the location of the significant people functions. Second, the overall profits are allocated between the head office and the branch on the basis of the first step, applying OECD transfer pricing guidelines. An important difference with transfer pricing between group companies is that in the first step, assets and risks are attributed based on functions. For determining transfer prices between group companies, the legal arrangements are the starting point. So the question is, is there also an “EU functionally separate entity approach”?
For member states that generally apply OECD-compliant standards, but which deviate from those in some respects (for example, safe harbours or the characterisation of legal arrangements for tax purposes), the question may arise of whether these deviations can be subject to state aid scrutiny.
In general, the aforementioned aspects reconfirm that EU state aid rules are applied dynamically. The state aid rules date back to the 1957 Treaty of Rome, while the arm’s length principle has only been widely endorsed after 1957, and the original version of the OECD guidelines is from 1995 (based on a 1979 report).
Which version of the OECD guidelines should apply?
The OECD adopted the initial version of its transfer pricing guidelines in 1995 and issued further versions in 2010 and 2017, with multiple supplements and modifications made in between. The guidelines have evolved significantly over the years, both in terms of the level of detail and substance. Examples of the latter are: (i) the endorsement and increased importance of DEMPE (development, enhancement, maintenance, protection and enforcement) and other functions compared to legal ownership of assets; and (ii) the application of substance of form approaches when characterising legal arrangements for profit allocation purposes. The OECD considers the revisions (mostly) as clarifications only, suggesting a dynamic application. That is, posterior guidance may be applied to previous years. This is debatable since applying different versions can, depending on the facts, result in different outcomes.
The General Court does not explicitly address this point. It refers to both the 1995 and 2010 versions in the Starbucks judgment, as the Commission did in its decisions, while the relevant tax ruling was issued in 2008. This seems inconsistent with the statement by the General Court that the Commission must base its analysis on information available when the measure – for example, the tax ruling – was taken (see also below). Why shouldn’t the same logic apply when applying guidelines?
Are double taxation or double non-taxation outcomes relevant?
Luxembourg and FIAT put forward arguments that the Commission should have investigated the tax treatment of the FIAT group as a whole, suggesting that not doing so would lead to double taxation. The General Court rejected these arguments.
First, it stated that FIAT and Luxembourg did not demonstrate that tax reductions in Luxembourg were neutralised by lower deductions by – and, therefore, higher taxable profits of – group companies. This may be seen as confirming that there are boundaries to the Commission’s task to evaluate all possible aspects, rather than selectively addressing specific deficiencies. In our experience, it is often difficult to reconcile the transfer pricing applied by other group companies with the indirect, and aggregated, transfer pricing methods applied in the relevant tax rulings.
Second, the General Court categorically dismissed the argument as unfounded. This seems to be a logical outcome because the state aid assessment focuses on the domestic reference system. A logical consequence is that double non-taxation resulting from transfer pricing mismatches would not construe state aid in and of itself either except, arguably, if the relevant domestic laws only allows a downward adjustment if there is a corresponding pick-up. While logical, this conclusion appears to be at odds with the Commission’s comments in its press release regarding the Apple case. There, the Commission suggested that if countries other than Ireland required Apple to pay more tax on the relevant profits, this would reduce the amount of underpaid tax in Ireland. The judgments may also be relevant for cases like the Huhtamaki case. The latter regards the question of whether Luxembourg can allow deemed interest deduction by a Luxembourg taxpayer on an interest-free loan made by an Irish group company that did not have to impute interest income in Ireland. While the General Court’s reasoning makes sense, we think that the transfer pricing analysis of group companies of the taxpayer will still be relevant evidence in the – often unlikely – scenario that it could be reconciled with the transfer pricing analysis applied for the tax ruling.
What should I know about the technical transfer pricing considerations?
The General Court carried out a comprehensive review and an exhaustive analysis in arriving at its decisions, especially on the technical transfer pricing aspects. Much of the analysis is case-specific, but there are number of interesting points with broader relevance:
– Arm’s length analyses are of an approximate nature; this means that there is a margin of error for local tax authorities. But, does this also mean that tax authorities and taxpayers can pick any outcome that falls within the range of arm’s length outcomes as typically determined in a transfer pricing study?
– The General Court recognised that a taxpayer is in principle free to choose a transfer pricing method deemed appropriate, and the Commission cannot easily dismiss the transfer pricing method applied in the ruling (often the TNMM);
– Challenging the identification of the taxpayer – party to the relevant tax ruling – as “tested party” for the purpose of applying an one-sided indirect transfer pricing approach (often the TNMM) is not easily accepted. The Commission had not shown that more reliable data was available in respect of the other party– in Starbucks and other pending cases, the non-taxed IP owner to which substantial royalties were paid – or that the selection of the tested party had resulted in inappropriate outcome.
Evidence: burden of proof and intensity of review
A key element in the discussions before the General Court, especially in the Starbucks case, centred on: (i) the burden of proof, and (ii) the intensity of the review by the General Court. In other words: the evidence. The judgments reconfirm that the Commission must demonstrate that all requirements for a finding of State aid have been met. The Commission must carry out a comprehensive review on whether a measure qualifies as state aid, looking at both the specific features of the case before them and at the technical or complex nature of the Commission’s assessments.
The real question is how much evidence is enough for the Commission to “prove” state aid. The General Court recognises that determining arm’s length transfer prices is not an exact science. To this end, it ruled that the Commission should verify whether the prices applied are a reliable approximation of a market-based outcome, and whether there is any variation that goes beyond inaccuracies inherent in the methodology used to obtain that approximation.
In addition, the General Court points out that a mere finding of a methodological error does not in itself demonstrate state aid. It also cannot be concluded that, as a rule, one method prevents a reliable approximation of an arm’s length outcome from being reached. Therefore, the General Court held that the Commission should demonstrate that methodological errors, or real or alleged lack of analysis, led to a reduction of the tax burden. Similarly, a suggested lack of justification – by the member state or company – for a step that was taken in the transfer pricing analysis (for example, adjustments to the “basis” for applying the relevant profit level indicator) does not suffice to demonstrate, as such, that the step was erroneous, or that it led to a reduction in the tax burden.
Briefly put, the Commission should not only prove that the transfer pricing accepted by the member state was “flawed”. It must also demonstrate that these flaws led to reduced taxes. In the Starbucks case, the General Court carried out a comprehensive assessment of the Commission’s decision and underlying evidence, and pointed out various findings by the Commission that were not supported by evidence. In contrast, the Commission’s findings in the FIAT case were upheld in their entirety. Therefore, the specific evidence in each case was a decisive factor between “win” or “lose.”
A specific point was raised about the time-period from which evidence may be used. The General Court is clear on this: if the Commission considers that the adoption of a prior tax agreement, such as an APA, itself constitutes state aid, the Commission may only base its decision on information that was available or reasonably foreseeable at the time of the APA’s conclusion and may not rely on elements after its conclusion. However, this does not prevent the Commission from finding state aid outside of what was in the APA. In this respect, the General Court confirmed that if the relevant tax authorities did not terminate or require an amendment to a tax ruling in connection with new information becoming available, this could construe a state aid measure.