By Kash Mansori, WTP Advisors
Posted: January 16, 2015

Today the EU Commission released a document detailing its reasoning behind its decision to consider Luxembourg’s tax treatment of Amazon to be a violation of EU competition rules. The full document can be found here (pdf).

The crux of the EU Commission’s argument is that Luxembourg gave Amazon permission to set up a transfer pricing structure that was inconsistent with the arm’s length standard, and that in so doing it conferred a “selective advantage upon Amazon in so far as it results in a lowering of its tax liability in Luxembourg.” And that’s where the problems start in the EU’s reasoning.

To begin with, one might think that in order for Amazon to have received a “selective advantage” it must be the case that Luxembourg would have rejected the same structure for another taxpayer. But there’s no evidence presented that Luxembourg would have interpreted its own tax code differently for any other taxpayer.

But even putting that issue aside, how does the Commission believe that the Amazon structure was inconsistent with the arm’s length standard? Its argument is based on the following points:

1. Luxembourg did not provide the Commission with Amazon’s transfer pricing analysis supporting its transfer pricing policies.
2. The method used to calculate Amazon’s transfer prices in Luxembourg (specifically royalty payments to the owner of the Amazon IP in Europe) “does not seem to correspond to any of the methods listed in the OECD Guidelines”.
3. Amazon’s calculation of the royalty payments “is not related to output, sales, or to profit.”
4. Luxembourg accepted a margin of 4-6% over cost for the Amazon entity without a comparability analysis, and that margin may be too low.
5. Luxembourg also accepted a floor on the Amazon entity’s profitability of 0.45% of Amazon’s EU sales revenue, along with a cap equal to all of Amazon’s European profits.

There are problems with each of these points. Taking them in turn:

1. The fact that Luxembourg did not provide the Commission with Amazon’s transfer pricing analysis says nothing about whether or not the arrangement is inconsistent with the arm’s length standard.

2. The method described in the Commission report is easily recognizable as the income method (i.e. the TNMM as specified in the OECD Guidelines) for the calculation of payment for intangibles. The income method is a frequently used approach in determining an appropriate level of royalty payments, and in fact is the method preferred by many tax authorities, such as those in the US.

3. Amazon’s calculation methodology is in fact directly related to sales and/or profit. This is obscured somewhat by the two-step nature of the calculation, in which the royalty payments are calculated such that they leave Amazon’s Luxembourg entity with profits equal to either 4-6% of its cost base or at least 0.45% of total Amazon sales revenue in Europe. But it’s clear that as Amazon’s sales and profitability in Europe go up, the income attributable to Amazon’s Luxembourg entity will also rise. So it’s hard to see where the Commission arrived at this particular conclusion.

4. The 4-6% range of profitability for the Amazon Luxembourg entity is not obviously too low; a 5% margin is considered very typical for a wide range of service providers. Whether or not it’s the right range in this case depends on the specific functions performed by the Luxembourg entity. A full functional analysis and set of comparable companies would be necessary to form a final opinion on this issue, but given the widespread use of 5% margins for intragroup services it seems oddly ill-informed to pick on this figure. Regardless, this point is largely irrelevant due to the second portion of the calculation methodology, in which the profitability reported by the Luxembourg entity actually depends on Amazon’s European revenues, not its cost base, as discussed in the next point below…

5. Yes, Luxembourg accepted Amazon’s rather unusual policy of having its Luxembourg entity’s taxable income depend not just on its cost base, but on total sales revenue in Europe. But there are two important points to be made about this. First, it’s not at all unusual to see arm’s length licensing arrangements where two or more alternative calculations are used to arrive at the amount of the payment, so this is not an obvious flag that the arrangement is inconsistent with the arm’s length standard. Second, given reasonable assumptions it’s entirely possible that the second, sales-based condition could actually increase Amazon’s tax bill in Luxembourg over what it would otherwise pay. Consider that Amazon’s sales in Europe were around €13.6 bn in 2013. Applying the 0.45% floor on the Amazon entity’s taxable income means that it must have declared income of at least €60m in Luxembourg in that year. Amazon’s Luxembourg entity has about 1,000 employees, suggesting that its direct operating expenses are probably in the neighborhood of a few hundred million euro. Even if its operating costs were €600m in 2013, it would have earned a margin of 10% over its costs – arguably significantly more than would be expected in an arm’s length arrangement. The point is that it’s not at all clear that Amazon’s transfer pricing policy yields a lower-tax result than would be expected in an arm’s length arrangement.

More fundamentally, the weaknesses in the EU Commission’s argumentation remind us that the EU competition authority is taking on a vast new area of inquiry with its recent push to address transfer pricing issues. Are the EU’s competition regulators now going to try to seriously assume a new role as a review body for each member country’s interpretation and enforcement of transfer pricing policies? If so, they have a lot of work in front of them.

UPDATE: For a bit more on the issue of “selective advantage”, take a look at this post by Renata Ardous. In it, she points out that the General Court of the EU has recently taken issue with the Commission’s attempts to characterize aspects of national tax laws as providing selective advantage, and that similar reasoning may apply to this case.