Benchmarking analyses – authorities may err

A benchmarking analyses is a key part of the transfer pricing documentation – it confirms the arm’s length nature of intra-group transactions. It is an important tool for taxpayers and tax authorities. The latter can use it to examine the arm’s length nature of transaction terms agreed between related entities.

Do the authorities always correctly prepare benchmarking analyses?

The matter was addressed in a decision of the District Administrative Court of 9 June 2022 (file no. I SA/Go 103/22).

What did tax authorities establish?

The dispute was triggered by a customs and fiscal control. The first instance authority challenged the arm’s length level of the taxpayer’s settlements in 2016. It accused the company of selling goods to related entities below the cost of their production. The terms of the transaction differed from those that would have been established by third parties, so the authority prepared a benchmarking analysis. It served as the basis for estimating the taxpayer’s income. This position was upheld during the appeal proceedings, but the company officially complained against this decision.

So, what did the court find?

The court had no doubts that the taxpayer calculated sales prices incorrectly. According to the court decision, the company-applied conditions for related entities were inconsistent with the arm’s length principle. The result: lower taxable income. The court stated the authority was right to launch the procedure of additional income assessment. Nevertheless, the court overturned the authority’s decision. Why? The tax authority made benchmarking analysis incorrectly.

Authorities do err

The court challenged the fact that the authority had made an additional estimation based on the arithmetic average of the median. The tax authority justified it by referring to the fact that the analysis is only an approximation of prices applied in transactions between third parties. However, the court emphasized that in practice the interquartile range is used to verify the arm’s length level. Consequently, arm’s length values should be those in the range between the lower and upper quartiles. Furthermore, the court accused the authority of adopting an incorrect level of revenues as the basis for estimating the company’s additional income. Namely: the revenue from transactions with related entities accounted for 95% of the total value of the taxpayer’s revenues, while the authority made an additional assessment at the level of revenues from total operations. According to the court, extending the scope of additional income assessment to cover transactions with third parties is inappropriate.

Unusual approach by the court

These allegations are justified and substantiated, but it seems incomprehensible that the court challenged the fact of the analysis being based on data for 2014-2016. According to the court, the analysis should have been performed only on the basis of data relating to the year covered by the control: 2016. The approach may seem controversial, especially since the tax authority made the analysis based on the transaction net margin method (TNMM) which is most often based on data from many years. As a rule, multi-annual data is used to increase data comparability: mitigate the impact of business cycles, product life cycles, or increasing the sample size.

The court’s position may come as a surprise since making an analysis based on long-term data is supported both in the OECD Guidelines, explanations of the Ministry of Finance, the recommendations by TP Forum and is well-established in practice.

Currently, the benchmarking analysis is a mandatory part of transfer pricing documentation. It must be made properly. However, merely having a benchmarking analysis in place may not be enough in the case of a possible inspection. It is necessary to have underlying high-quality documents that would confirm the arm’s length level of the settlements made. If you wish to be on the safe side in the area of transfer pricing, please contact us.


Powiązane treści