In Transfer Pricing benchmarking, the median has become the dominant reference point to test compliance with the arm’s length principle. Favoured by many tax authorities, it is perceived as stable, neutral, and administratively efficient.

However, this convenience often overlooks a key reality: profitability data in benchmarking analyses are rarely normally distributed. In many industries, margins are skewed, influenced by outliers, and driven by high-performing companies. As a result, mean and median represent different economic perspectives.

Mean vs Median: two different economic messages

• Median | captures the “typical” performer, insulating effects of extreme values
• Mean | reflects the “expected market return”, incorporating superior performers

In asymmetric distributions, systematic reliance on the median may produce a downward structural bias, leading to benchmarks that underestimate genuine market profitability.

Practical insights

The authors demonstrate through empirical examples that:

• carefully handling outliers
• using trimmed or adjusted means
• analysing the statistical nature of datasets

can significantly improve reliability without losing robustness.

Policy and administrative considerations

Although OECD Guidelines allow flexibility in selecting any appropriate point within the range, many tax authorities still default to the median for reasons of simplicity and perceived neutrality.

The key takeaway:
The choice of central indicator should not be mechanical. It must be analytically justified, aligned with:

• statistical distribution characteristics
• industry dynamics
• taxpayer’s functional and risk profile

Conclusion

A more mature Transfer Pricing approach requires rigorous statistical thinking, transparency, and economic realism. Moving beyond automatic reliance on the median is essential for ensuring truly arm’s length outcomes.