Articles

How do you calculate the interquartile range in a transfer pricing study?

Multiply the number of results by 25, 50 or 75 percent; a whole number means the average of two results, otherwise take the next result up.

The interquartile range of a transfer pricing study runs from the 25th to the 75th percentile of the comparables’ results, counted as the US regulation counts: n x 25 percent, and if that is a whole number, the average of that result and the next. With 14 results the range is the 4th to the 11th result and the median is the average of the 7th and 8th: 2.44 to 4.18 percent, median 3.04 percent, in the example below.

Worked in full in The Transfer Pricing Professional by Julian R. Sterling.Get the DRM-free PDF, $19.99 →or the paperback

Free: the transfer pricing benchmark template in Excel, which computes the interquartile range, the median and the true-up by the counting rule of the regulation.

The rule in the US regulation

Under 26 CFR 1.482-1(e)(2)(iii), when the comparables are not close enough to be treated as exact matches, which is the usual case for a profit-based study, the interquartile range ordinarily gives an acceptable arm’s length range. It runs from the 25th to the 75th percentile of the comparables’ results.

The regulation defines the 25th percentile by counting, not by interpolation: it is the lowest result such that at least 25 percent of the results are at or below it. If exactly 25 percent of the results are at or below a result, the 25th percentile is the average of that result and the next higher one. The 75th percentile is found the same way, and so is the median, the 50th percentile, which matters because an adjustment of a result outside the interquartile range is ordinarily made to the median (26 CFR 1.482-1(e)(3)).

In practice: multiply the number of results n by 25, 50 or 75 percent. If the product is a whole number k, take the average of results k and k + 1. If not, take the next result up.

Worked example 1: fourteen distributors

A study for a limited-risk distributor accepted 14 companies, each measured by its weighted average operating margin over three years (total operating profit over total net sales). Sorted from lowest to highest, in percent: 1.28, 1.77, 2.01, 2.44, 2.55, 2.80, 3.00, 3.07, 3.37, 3.84, 4.18, 4.33, 4.63, 5.13.

Fictional study of the book, fiscal 2024.
PointPositionResult
25th percentile14 x 25% = 3.5: 4th result2.44%
Median14 x 50% = 7: average of 7th and 8th3.04%
75th percentile14 x 75% = 10.5: 11th result4.18%

The median is not the result of any company: it is the average of 3.00 and 3.07 percent. A spreadsheet’s built-in quartile function would interpolate and give different values, so the study should show the counting step for each point.

The tested party, Harnwick Europe, earned an operating profit of $1,074,631 on net sales of $118,400,000: $1,074,631 / $118,400,000 = 0.91 percent. That is below the 25th percentile of 2.44 percent, by 1.53 percentage points.

The trap: bringing the result back to the edge of the range is not what the US rule points to. The adjustment ordinarily goes to the median, here $118,400,000 x (3.04% - 0.91%) = $2,521,920, against $118,400,000 x (2.44% - 0.91%) = $1,811,520 to the 25th percentile. The rules of the distributor’s own country may differ and have to be read separately.

Worked example 2: one company fewer

A year later the study is rolled forward. One company was acquired and has no new accounts, so 13 results remain: 1.64, 2.13, 2.17, 2.61, 2.74, 3.04, 3.07, 3.17, 3.34, 4.27, 4.36, 4.60, 5.16.

Fictional study of the book, fiscal 2025.
PointPositionResult
25th percentile13 x 25% = 3.25: 4th result2.61%
Median13 x 50% = 6.5: 7th result3.07%
75th percentile13 x 75% = 9.75: 10th result4.27%

With an odd number of results the median is one company’s result. With 12 results, every product is a whole number (3, 6 and 9), and each point is the average of two results: the book’s manufacturer study of 12 companies gives 5.16, 6.34 and 7.76 percent that way.

What goes in the file

Sources

  1. 26 CFR 1.482-1 (Cornell LII), read 17 November 2026
  2. 26 CFR 1.482-5 (Cornell LII), read 17 November 2026

Rules checked on 17 November 2026 in the text of the regulations. US federal rules only, illustrative, and not tax or legal advice. The group in the examples is fictional and its figures are teaching assumptions.

Questions readers ask

Is the interquartile range the same as Excel's QUARTILE function?

Not always. Excel's quartile functions interpolate between results; 26 CFR 1.482-1(e)(2)(iii)(C) counts results and averages two of them only when exactly 25 percent sit at or below one. The two methods can give different points.

What happens if the tested party falls outside the range?

Under 26 CFR 1.482-1(e)(3), the IRS may adjust the result to any point within the range; when the interquartile range is used, the adjustment is ordinarily made to the median.

Do I always have to use the interquartile range?

No. It applies ordinarily when the comparables are of lesser reliability; the regulation allows a different statistical method if it gives a more reliable measure, and exact comparables can support the full range.

How many comparables does a study need?

The regulation sets no minimum number. A larger set makes each quartile less sensitive to one company, and the rejection log has to show why each company was kept or left out.

Read the whole book

This article is one calculation from The Transfer Pricing Professional, a practitioner’s guide to the work of a transfer pricing manager: the arm’s length principle, the functional analysis, the choice of method, benchmarking studies, the year-end true-up, services, loans and royalties, intercompany agreements, documentation, audits and the mutual agreement procedure.

Also on this site