Building the Range: IQR, Outliers and Significance

Chapter 11 of the OECD Guidelines is the benchmarking-studies chapter: how the comparables are searched and selected, how the range is built from them, and what makes the result reliable. Its most-cited anchor — 11.100-11.107 — is the provision that settles the arm’s length range: the interquartile range as the default, the presumption that a result within it is at arm’s length, and the reliability factors that qualify the whole output.

The chapter at a glance

Subject	What it covers
Search and selection	The design of the comparables search and the selection of the set, under the Chapter 3 factors and screens
The arm’s length range	How the range of comparable results is constructed from the set, and the single most representative point (the median) within it
11.100-11.107 — the interquartile range	The standard range as the interquartile range — the 25th to the 75th percentile of the comparable results — and the presumption that a result within it is at arm’s length
Outliers	The treatment of the comparable results that sit outside the normal distribution, and the test for whether they are removed or carried
Statistical significance	The minimum reliability of the set — the sample size and the dispersion that make the range a defensible measure rather than a coincidence
Reliability factors	The factors that qualify the output — the quality of the comparables, the adjustments, the period — that make the study reliable: 11.107

11.100-11.107 — the range and the presumption

In paraphrase, the provision directs that the arm’s length range be the interquartile range — the range between the 25th and 75th percentile of the comparable results — and that a result falling within that range is presumed to be at arm’s length. The working content:

• The range is the interquartile range, not the full range. The default is the 25th-to-75th percentile band, which strips the tails of the distribution and gives a range the typical comparable earns. The IQR vs full range discussion is the practical argument for why the tails are excluded, and the interquartile range entry carries the definition.
• Within the range, the result is presumed arm’s length. The presumption is the provision’s core: a tested-party result inside the IQR does not need a further defence, and moving it out of the range is the adjustment.
• The median is the most representative point. Where a jurisdiction proposes moving the result to a specific point of the range rather than accepting any point within it, the provision cited is the one that names the median as the single most representative point of the comparables.

Outliers and significance

The two provisions that qualify the range before it is used:

• Outliers. The results that sit at the extremes of the distribution are tested before they are carried: where the outlier is an error in the data or an unmeasurable difference, it is removed; where it is a genuine comparable with a measured difference, the adjust-versus-exclude rule at 3.30-3.31 decides. The outlier entry covers the test.
• Statistical significance. The range is a defensible measure only where the set is large enough and tight enough that the band is not an artefact of a few entities. A small, dispersed set produces a range that looks authoritative and is not — the statistical significance test is the provision that says so, and it is the factor 11.107 carries into the reliability assessment.

The reliability factors at 11.107

In paraphrase, the provision lists the factors that make a benchmarking study reliable: the quality and comparability of the set, the adjustments applied, the period of the data, and the significance of the result. The file must address them — a range built on a defensible set, with the adjustments recorded and the significance checked, is the output the documentation tier is required to carry.

What it means in practice

The reader’s map names Chapter 11 as the chapter that settles how ranges are built and what makes a study reliable. In an Indian proceeding the IQR is the range the TPO works from, and the arguments that recur are the same three the provisions frame: is the set significant, are the outliers handled, and is the median the right point to move the result to. The working mechanics are covered in the IQR vs full range guide.

Where this takes you

• The range in practice: IQR vs full range.
• The definition: interquartile range.
• The significance test: statistical significance.
• The point the range is measured from: median.
• The data the range is built on: OECD comparability.