The use of the interquartile range in transfer pricing1

“…because transfer pricing is not an exact science, there will also be many occasions when the application of the most appropriate method or methods produces a range of figures, all of which are relatively equally reliable.”2

The establishment of a range of acceptable values, a so-called arm's length range in transfer pricing jargon, is commonplace and typically results from the identification of a number of comparables when conducting a benchmarking analysis.

While the concept of the arm's length range has widespread recognition in international guidance, domestic legislation and within transfer pricing, how this range is determined and applied in practice can differ significantly practitioner by practitioner and country to country.

The most notable difference is with respect to the practice of narrowing the range, using statistical tools such as the interquartile range. The interquartile range is defined as the middle part of a range of data, i.e. the data points found between the first and third quartiles.3

Often, transfer pricing practitioners, both in the private sector and government, default to the use of the interquartile range without due consideration as to why it is being applied.4 Alternatives are very rarely considered.

This article first provides an overview of international guidance, after which a comparative analysis of a sample of countries' legislation is provided. The authors then conclude with their thoughts and considerations regarding the current practice.

International guidance

In this section we provide a summary of the international guidance available on the determination of the arm's length range and the use of statistical tools, such as the interquartile range, to narrow the range.

In the vast majority of transfer pricing cases, the application of the most appropriate transfer pricing method results in the establishment of a range of values. In a limited number of cases, a single value may be determined; however, such cases are extremely rare. Furthermore, guidance at international level highlights the importance of the comparability analysis and reliability of the data when conducting a benchmarking analysis. Once these are taken into account, international guidance typically suggests that “any point in the range satisfies the arm's length principle.”5

While the concept of the arm's length range is clearly embedded in the OECD Guidelines and in the United Nations (UN) Practical Manual on Transfer Pricing for Developing Countries, 2017 Edition (UN Manual)6, the concept of the interquartile range is not as well recognised as it is commonly used.

OECD

The use of statistical techniques such as the interquartile range, are referred to in the OECD Guidelines as statistical tools that "might" help to enhance the reliability of the analysis, but only under specific circumstances:7

• after “every effort has been made to exclude points that have a lesser degree of comparability”;
• where there are limitations in information available on comparable transactions, such that some comparability defects remain that cannot be identified and/or quantified, and are therefore not adjusted; and
• where the range “includes a sizable number of a observations.”

No further guidance is provided, and most importantly the OECD Guidelines do not advocate the automatic application of the interquartile range, or other statistical methods, in all cases.

United Nations

The concept of the arm's length range is also recognised in the UN Manual. In regards to the use of statistical techniques to “enhance reliability”, no specific guidance is given as to which methods could or should be applied or when, other than references in the following examples:

• Example 5 (UN Manual, Part B, B.3.2.12.5) states that if “available data is sufficiently complete and accurate to conclude that it is likely that all material differences between the controlled and uncontrolled transactions have been identified; such differences have a definite and reasonably ascertainable effect; and reliable adjustments are made to account for such differences, the results of each of the uncontrolled distributors may be used to establish an arm’s length range.”
• Example 6 (UN Manual, Part B, B.3.2.12.6) states that where insufficient data is available to determine whether a specific difference that may impact comparability exists (in the example; information on warranties) the use of a statistical measure may enhance the results.
• Example 1 (UN Manual, Part B, B.3.3.5.1) states that when the data is not sufficiently complete to conclude that all material differences between the tested party and the uncontrolled transactions have been identified, the tax authority can measure the arm’s length result by using the interquartile range.

Comparative analysis

The table below provides a summary of the comparative analysis of a sample of countries' legislation and guidance regarding the determination of the arm's length range.8

For most countries reviewed, we have found no reference in the legislation to the automatic use of the interquartile range, nor to an adjustment being made (by the tax authority) when the transfer price falls outside of the interquartile range, as opposed to the arm's length range. 9 This position has been recognised in Courts, for example the Swedish Supreme Administrative Court ruled on transfer pricing benchmarking analyses in July 201910, specifically recognising support in the OECD Guidelines for the application of the full range of arm's length results, rather than the interquartile range. A notable exception is Ukraine, where the use of the interquartile range is mandatory when more than one comparable is found.11

The use of a statistical technique to narrow a range does not necessarily produce a more reliable result. Countries in group C, the majority of our sample, do not mandate the use of the interquartile range per se, but rather specify it as a possibility when comparability assessment proves the set not to be highly reliable.

Conclusion

The mere fact that there are numerous observations does not mean a statistical tool is needed to narrow the range, unless required by law. Where there is concern over a wide range, the authors suggest that instead of automatically defaulting to a statistical tool, one should first look at the reason behind such a wide range. Only after a thorough comparability analysis, necessary adjustments and subsequent steps have been taken into account, should one consider the application of a statistical technique such as, inter alia, the interquartile range or other tests for outliers (e.g. Tukey or standard deviation methods), and even then it should be applied with caution.

A potential problem with the automatic use of the interquartile range is the possibility of discarding otherwise acceptable comparables which fall within the full range but outside the interquartile range. The interquartile range is in fact “a descriptive statistic defined entirely by the values observed in the range and has no relationship to the provenance of those numbers.” Canada also enforces its view on the reliability of the data rather than on the use of statistical tools, stating that “the use of statistical tools may actually result in the loss of information and a reduction in comparability.”

Based on the authors research and experience, the general practice to determine the arm's length range and the widespread, and seemingly automatic, use of statistical tools to narrow the range, does not appear to be reflected in the guidance available at international level and at country level.

Contact the authors or your regular DLA Piper advisors to discuss the use of interquartile range in transfer pricing.

---

1 This article has been written to the best knowledge of the authors with information available as of September 2019. While all reasonable care has been taken to ensure the content is accurate, it reflects the authors' perspective and it is intended for use as a guide only and should not be relied on upon without seeking formal advice.
2 The Organisation for Economic Co-operation and Development Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations, 2017 Edition (the “OECD Guidelines”), Chapter III, paragraph 3.55.
3 Introduction to Descriptive Statistics, J. Nicholas, Mathematics Learning Centre University of Sydney NSW 2006.
4 In particular whether it is legally required to be applied, and whether it is appropriate to be applied.
5 OECD Guidelines Chapter III, paragraph 3.62.
6 UN Manual, Part B, B.1.4, Concepts in Transfer Pricing, recognising the 'associated enterprises' and related definitions as per the UN Model Tax Convention Article 9.
7 OECD Guidelines Chapter III, paragraph 3.57.
8 This comparability analysis has been performed with information available as of September 2019. This should not to be relied upon as legal advice.
9 The arm's length range is referred to as the range of reliable variables, rather than the interquartile range.
10 Judgment of the Swedish Supreme Administrative Court on 19 June 2019 (1913-18), upholding the decision of the Administrative Court of Appeal in Stockholm in January 2018. Judgement available in Swedish, last accessed October 2019.
11 Art.39.3.2.2 of the Tax Code of Ukraine.