Bloomberg Tax
Free Newsletter Sign Up
Bloomberg Tax
Advanced Search Go
Free Newsletter Sign Up

INSIGHT: Comparability Adjustment of Market-Specific Features for Transfer Pricing

Oct. 22, 2019, 7:00 AM

In dealing with transfer pricing regulations, discussion of market premiums, location savings, and other local market features (Location-Specific Advantages; LSAs) can be controversial between taxpayers and tax authorities, and their financial impact tends to be significant.

Many taxpayers may argue that LSAs are the primary reason when the tested party’s result falls outside the arm’s-length range. However, taxpayers’ attempts to develop concrete arguments often lack persuasiveness by only providing qualitative explanations or referencing materials that lack direct relevance.

We saw, for example, some companies recording very high profit margins at subsidiaries in certain regions, despite reporting low profits or even losses at the group-wide level. Taxpayers generally attempt to provide qualitative explanations about the background and reasons for such situations; however, they often lack quantitative data and analysis to support their qualitative explanation and are unable to develop arguments that can withstand anticipated objections by the tax authorities.

This article explains an objective and rational quantitative method that can be important in providing an LSA-based explanation in transfer pricing analysis. As an example, we explain a case study using a statistical approach called regression analysis to calculate market premiums.

Quantifying LSAs for Comparability Adjustments

United Nations Practical Manual on Transfer Pricing for Developing Countries (United Nations, 2017, hereinafter, “The UN Manual”) identifies the following five kinds of LSAs (pp. 103-104):

  • Highly specialized skilled manpower and knowledge;
  • Proximity to growing local/regional market;
  • Large customer base with increased spending capacity;
  • Advanced infrastructure (e.g. information/communication networks, distribution system); or
  • Market premium.

The UN Manual defines the LSA as being measured as the net value of location savings (net cost savings derived from relocation to a lower cost country) and other location-specific benefits (additional benefits derived from a specific local market).

OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations (OECD, 2017. Hereafter, “The OECD Guidelines”) does not clearly define LSAs but refers to location savings and other local market features. Primarily, the OECD Guidelines provide that these factors should be considered in the comparability analysis, and further states that if LSAs cause significant comparability differences, such differences should be adjusted to reliably improve comparability.

Now, what methods are reliable in evaluating LSAs that improve comparability, as described in the OECD Guidelines? As discussed earlier, many taxpayers lack quantitative explanation. Thus, we explain an example quantitative approach using regression analysis in the following section.

Regression analysis is not only widely used in academic studies but also an established technique used in regulatory and litigation practices, including tax cases. In the U.S. APA application, for instance, in some cases taxpayers propose comparability adjustment using regression analysis. In addition, according to an academic paper in the U.S., there were approximately 100 cases in the database that asserted evidence with the application of regression analysis in tax-related litigation.

In Japan, though not a tax case, regression analysis was adopted as evidence in the High Court decision on the stock purchase price in the Intelligence, Ltd. case. In the trial, the Tokyo High Court said, “[t]he regression analysis is a reasonable method based on scientific evidence, and its accuracy can be objectively verified.” In addition, Japan Fair Trade Commission (JFTC) regularly uses economic analyses including regression analysis, in practice.

Regression analysis results are highly reproducible and verifiable because the same results can be derived by anyone if the same data and method are applied. Therefore, comparability adjustments using regression analysis would be an appropriate method in reliably improving comparability as the OECD Guidelines describes. Thus, it would be worthwhile considering the use of regression analysis for taxpayers facing comparability issues arising from LSAs.

Case Study: Quantifying Market Premiums Arising from Different Customer Segments

Company A manufactures and sells a variety of widgets in various regions. Company A did not have a group-wide control over the transfer pricing operations and urgently needed to establish its transfer pricing management system to comply with transfer pricing regulations.

In one segment of Company A’s, there are two types of group companies: (i) manufacturing/distribution entities, and (ii) distribution entities. Manufacturing/distribution entities develop, design, and manufacture certain product lines and sell them in the region where manufacturing/distribution entities locate. On the other hand, distribution entities purchase products from manufacturing/distribution entities and resell in the region where distribution entities locate.

In this segment, product performance and quality are the key value drivers. As a result, manufacturing/distribution entities are positioned as strategic and risk-taking entities (principals), and the distribution entities are positioned as limited risk distributors (LRD). Each manufacturing/distribution entity develops different product lines and valuable intangible assets associated with the development and manufacture of those products. For this reason, principals and LRDs differ by different product lines.

In other words, a manufacturing/distribution entity (BSub) which develops Product X is the principal for Product X, and other distribution entities reselling Product X are LRDs. On the other hand, for Product Y, the manufacturing/distribution entity (CSub) which develops Product Y is the principal, and other distribution entities including BSub which resells Product Y are positioned as LRDs of Product Y.

Therefore, Company A adopted a transfer pricing policy applying the Transactional Net Margin Method (TNMM) as the transfer pricing method for each product line (for each principal) with LRDs selected as the tested parties.


Company A’s products generally function by incorporating them into other devices. They are utilized in a wide variety of uses, including testing of automated devices, semiconductor manufacturing, and X-ray generators among others. On the other hand, Product X developed by BSub, a manufacturing/distribution entity, is a high value-added product. It can function as a stand-alone device, equipped with communication interfaces that receive external signals and control functions. Main applications of Product X are for test and measurement and manufacturing (hereinafter referred to as “T&M” and “MFG”, respectively).

Company A considers the electric power (wattage (W)) which their products can accommodate to be an important pricing factor. The higher the electric power is, the higher the prices tend to be.


Customers for BSub’s products are broadly categorized into T&M and MFG depending on their use. Customers for T&M account for a large proportion of the customers of the products.

Customers for T&M look for high-quality in products, such as product functionality and reliability, and tend to demand short delivery times. The primary negotiation counterparties for T&M uses are engineers in the company’s R&D departments or research institutions. It is important that product performance conforms to the customer’s intended use, and there is a tendency that price discounts are not required unless the price exceeds the customer’s (e.g., engineers) budget. As a result, they are often traded at regular prices (full price). In addition, due to the nature of its use, the sales volume per transaction is often small.

Customers for MFG place the greatest importance on price competitiveness, and product quality is lower in priority than those for T&M customers. The negotiation counterparty is the purchasing staff of the company for MFG. Because MFG customers usually obtain quotes from Company A’s competitors, it is usual to discount a certain amount from the regular price. In addition, sales volume per transaction for MFG customer is large, which is a key difference from transactions with T&M customers.

With respect to the competitive landscape, there are many small to medium-sized competitors in the T&M market, whereas there are many large competitors in the MFG market.

With these facts given, it is assumed that the return on investment will be low for new entrants in the T&M market, BSub’s primary market. The market will be relatively small in scale, and the customers’ expectation in quality is high. New entrants must spend significant investment in developing competitive products. Thus, the market is characterized as niche with high entry barriers, which is unattractive especially for large players.


In developing the Company A’s transfer pricing policy, comparable companies selected in benchmark analysis to establish target profit margins (arm’s-length range) included only those companies in business that requires price negotiation like MFG market. This is because it is very unlikely to be able to identify a sufficient number of comparables in niche markets even if a benchmark analysis is attempted with a focus on the market for T&M. It would be desirable to perform a benchmark analysis for each customer segment (market), but this was not a realistic approach for Company A under BSub’s circumstances.

BSub’s products are often used for T&M and are priced higher than for MFG customers. As a result, LRDs of BSub’s products achieve high operating margins. Comparing operating margins of comparable companies with those of LRDs (i.e., tested parties), operating margins of the tested parties are higher effected by sales price to the T&M customers. Therefore, LRDs’ operating margins exceed the arm’s-length range, and Company A is exposed to transfer pricing risk.

Company A believed that, as objective facts, there were comparability differences between T&M markets and MFG markets, and that the tax authorities would agree with making comparability adjustments for the impact of the premium caused by market differences. However, Company A could not find a solution to quantitatively evaluate the market premium and adjust the operating margins of tested parties. As a result, Company A could only focus on a qualitative explanation of the market premium.

Approach of the Analysis

Based on the assumption that differences in market comparability between the BSub’s major customer segment (T&M) and the customer segment of the comparable companies (MFG) selected in the benchmark analysis cause significant differences in profitability, the market premium arising from the T&M customer segment is calculated and adjusted as explained in the following sections.

This market premium is “other local market features” in the OECD Guidelines, and if the marker premium is causing significant differences in comparability that may affect profitability, adjustments should be made to improve the comparability using reliable methodologies.

In this case study, we apply a statistical technique called regression analysis in calculating market premiums and making comparability adjustment. We take the two-step approach. First, by applying the regression analysis, we calculate the market premium arising from differences in customer segments as the amount in the product unit price. Second, we deduct the premium from the profit or loss of the tested parties (segment P/L).

Details of the Analysis

(1) Summary of Tested Transactions

For the sake of simplicity, only the transaction between BSub in Country B and CSub in Country C is explained in this example.

Figure 1 illustrates the transaction flow under review. BSub is the principal that develops and manufactures the products and owns the related intangible assets. CSub is an LRD reselling BSub’s products to customers in Country C. In Country C, the customers of BSub’s products are divided into two segments: (i) T&M, and (ii) MFG.

Figure 1: Tested Transactions

(2) Need for Identification and Quantification of Market Premium

Based on the facts given, the selling price of BSub’s products may contain the two types of premiums as we define below:

  • Product premium: it constitutes the amount customers would pay for performance, functionality and high quality of the product. BSub’s products provide high added-values, functioning as stand-alone devices and loading communication interfaces to receive external signals. Therefore, it is believed that the products themselves have high added-values. Thus, we call the premium paid by customers for the product characteristics (e.g., performance, functionality, quality) as the product premium.
  • Market premium: it arises from the difference in market conditions. BSub’s products are sold in two customer segments: T&M and MFG. Despite being the same product, they can be sold at premium price for T&M customers. Thus, we call the premium arising solely from the difference in customer segments as the market premium.

Figure 2: Image of Premiums Comprising Product Profits

Figure 2 depicts the profit structure of BSub’s products and other companies’ products. Product premiums are derived from the product developer’s intangible assets such as patent, know-how, design and/or other intangible assets. The degree of the product premium may vary depending on the product but is not affected by types of customers. Therefore, the product premiums for BSub’s products should belong to BSub and are not subject to comparability adjustments in applying the TNMM where CSub is selected as the tested party. Figure 2 depicts no difference in the amount of product premiums between BSub’s products (though customer segments differ). On the other hand, the difference in profits between BSub’s products and other company’s products illustrates difference in the value of product premium derived from intangible assets inherent to respective products.

It should be noted that the market premium here arises not from marketing intangibles, such as brands and marketing campaigns, but merely from sales to customer segments under specific market conditions (i.e., T&M customers). That is, the premium arises from the peculiarity of the market condition, which is one of the comparability factors considered in the benchmark analysis. In applying the TNMM, it is necessary to select comparables from the viewpoint of the functions performed, risks borne, and assets owned, as a benchmark to compare with the tested party (i.e., CSub), but it would be extremely difficult to identify sufficient number of comparables in niche markets. Because this market comparability difference has significant impact on CSub’s profitability, it is necessary to quantify the market premium and make adjustment to improve comparability.

(3) Quantifying the Market Premium

(a) Overview of Regression Analysis

In quantifying market premiums, it is assumed that product prices are influenced by economic factors such as market premiums and product premiums. We explain the regression analysis as one method to capture the relationship between these factors and product prices.

As noted earlier, BSub’s products are niche products and are often sold at regular prices for T&M customers. In the MFG customer segments, however, their products are usually sold at discounted prices. Therefore, there exist market premiums in sales to T&M customers compared to sales to MFG customers. Hence, the focus of this analysis is to quantify market premiums in sales to T&M customers by comparing prices with those for MFG customers.

One possible way to quantify market premiums would be a simple comparison of average prices or profits between two customer segments. However, such a method alone may not provide adequately reliable results because market premium is not the only factor causing the difference in average prices or profits between customer segments. For example, depending on the time period analyzed, high functionality products may be sold to MFG customers. In that case, if this high functionality is not appropriately controlled in the analysis, the average price for MFG customers may be inappropriately high (due to the product premium from the high functionality), which may mislead to a conclusion that there is no market premium in the T&M segment.

In addition to differences in market conditions that give rise to market premiums, there will be a variety of factors that may affect prices and profits. Therefore, market premiums cannot be precisely quantified unless impacts of these factors are appropriately controlled in the analysis. This is the reason why we use the regression analysis.

In regression analysis, the relationship between a dependent variable (y, product prices) and independent variables (x, market premiums, etc.) as represented by the following equation (regression line) is estimated based on the data using statistical methods.


Figure 3: Graph Image of Regression Analysis1

y is the value we are estimating, b is the slope of the regression line, x is the value of the independent variable, and a is the intercept. Figure 3 shows observed values (data) with the independent variable on the horizontal axis and the dependent variable on the vertical axis. The line inclined upward to the right corresponds to the equation derived by applying regression analysis to the data. In this figure, as x increases, y increases. Therefore, we can see that there is a positive relationship (i.e., b takes a positive value) between the dependent variable and the independent variable.

In regression analysis, multiple variables may be used as independent variables. For example, when there is an interest in the relationship between a dependent variable and two independent variables, the following equation is formalized.

y = b1x1+b2x2+a

In this case, b1 indicates the effect of changes in x1 on y with changes in x2 appropriately controlled. Similarly, b2 indicates the effect of changes in x2 on y with changes in x1 appropriately controlled.

With data such as observed prices, we can capture a qualitative relationship between variables by drawing scatter plots as in Figure 3. Moreover, we can take one step further with regression analysis and evaluate a quantitative relationship between the variables. In the context of transfer pricing analysis, when there exist LSAs which cause significant comparability differences between a tested party and comparables, we can assess quantitative impacts of LSAs appropriately controlling impacts of other factors.

(b) Data

We need detailed transaction data for BSub’s products which contains product specifications, sales date, sales price, sales volume, profit, etc. The data was also grouped by customer segments (i.e., T&M and MFG). It is best that these data are managed in a single file. If they are sourced from separate files, it is necessary to match the data to integrate into one file.

Data cleaning is usually a necessary process. In most cases, raw transaction data contain data irrelevant to the analysis or data that would result in incorrect results if included in the analysis. For example, the data might include observations for returned products or sales data other than products in interest. It is important to remove these kinds of unrelated data to improve the accuracy and reliability of the analysis.

(c) Application of regression analysis

The regression model used in this case is as follows:

Price = α0 + α1 × T&M dummy + α2 × Country B dummy + α3 × electric power + error

The effect of each factor on prices are represented by parameters α0 through α3. These parameters are called coefficients. T&M dummy and Country B dummy indicate particular events in a transaction, taking a value of 0 or 1. For example, if T&M dummy takes a value of 1, it means that the product is sold to T&M customers. If it takes 0, it means that the product is sold to MFG customers. Similarly, Country B dummy takes the value of 1 if sold in Country B and the value of 0 if sold in Country C. The error term indicates the part of the price change which is not explained by the independent variables used in the regression model.

The coefficient for T&M dummy represents the market premium related to the sales to the T&M customer segment (the effect on the price of the product for T&M compared to the price of the product for MFG), which is the most important coefficient in this analysis. If the hypothesis that a market premium exists for T&M is correct, the market premium (α1) should be calculated as a positive value.

We calculate both point and interval estimates for coefficient. The point estimates provide a single point, and the interval estimates provide a range of values which would include a parameter (the true value of the market premium in this case). For example, if the 95% confidence interval is from 100 to 150, it is certain with 95% probability that the true value of coefficient falls within the range between 100 and 150. Table 1 shows the result of the estimation.

Table 1: Results of Regression Analysis

Market premiums range from JPY10,000 to JPY32,000 for the 95% confidence interval and it is JPY21,000 for the point estimate. In other words, the point estimation suggests that the price of the product for T&M is higher than that of the product for MFG by JPY21,000.

Next, for Country B dummy, a point estimate of JPY2,000 is obtained, but the result is not statistically significant. That is, there is no statistically significant difference between the price in Country B and the price in Country C as the p-value is high.

Finally, the point estimate for electric energy is JPY80. In other words, one kilowatt increase in electric power that the product can accommodate would increase the price by JPY80.

It was confirmed that these results were in line with Company A sales personnel’s educated guess.

Adjustment to the Tested Party Segment P/L

In the regression model, we calculated the market premium as the difference from the base price, which is the average unit price of products sold to MFG customers. In applying the TNMM, this difference from the base price (i.e., the market premium) must be reflected in the (unadjusted) segment P/L of the tested party (i.e., CSub). Adjustments can be made by subtracting the market premium, obtained by multiplying the number of units sold at premium price (i.e., sales to T&M customers) and the per-unit market premium, from CSub’s net sales.

For example, Table 2 shows the result of adjustment for the segment P/L of CSub with the upper limit of the interval estimate (JPY32,000).

Table 2: Calculation of Adjusted Segment P/L

The operating margin decreased by 12% after the adjustment, which represents the impact of the market premium on operating margin.

Summary (Applicability of Regression Analysis)

LSAs are one of the comparability factors between the tested party and comparables to be considered. The difference in comparability should be adjusted in a reliable manner if the LSA would significantly impact on profitability when comparing the tested party and comparables. In tax investigation and litigation cases which profitability from such comparability differences is at issue, it is important to evaluate LSAs quantitatively in a reasonable and objective manner to defend the taxpayer’s position. From the perspective of the OECD Guidelines and based on the fact that regression analysis is adopted as evidence in litigations, the regression analysis would be a legitimate and useful method to quantitatively evaluate LSAs.

In the case study, we have examined the effectiveness of regression analysis as a way of quantifying the market premium caused by differences in customer segments (markets). We believe that there are many other situations in which similar approach can be taken.

For example, it is easily imagined that sales of products such as air conditioners and ice cream would closely correlate with climate conditions. When extreme weather conditions lead to significant increases or decreases in revenue, regression analysis will be useful to quantify the impact by analyzing the relationship between temperature and revenue. Similarly, regression analysis may be applied to quantify the effect of social phenomena, booms, or reputational damage that are unrelated to company’s marketing activities but potentially affect its sales.
The regression analysis covers a broad range of areas for application and is useful in demonstrating taxpayers’ qualitatively asserted positions. In situations which significant differences between the tested party and comparables may exist because of market premiums exposing the taxpayer to significant tax risks, it is worthwhile considering the use of this approach as a mean to demonstrate the existence and the impact of market premiums.

Finally, in the assessment of LSAs by regression analysis, it is advisable to pay attention to the following points: (i) whether it is a reasonable analytical model in light of the facts; (ii) whether variables and sample sizes used for regression analysis are statistically reasonable; and (iii) whether the initial conclusions do not change even when additional analyses are performed based on requests or feedback from the tax authorities. Because these will be critical issues to address in tax investigations and litigation, it would be beneficial to involve experts with sufficient experience and expertise in conducting the regression analysis in the transfer pricing contexts.

This column does not necessarily reflect the opinion of The Bureau of National Affairs, Inc. or its owners.

Author Information

Takahiro Yamada is an experienced transfer pricing expert in Japan and the U.S. He provides advice and supports in complex transfer pricing investigations and negotiations with tax authorities in APA application to clients in a wide range of industries including automobiles, electronic components, apparel, and entertainment among others.

Keita Fukunaga has expertise in economic analysis related to transfer pricing, antitrust laws, patent laws, and other laws and regulations. For more than a decade, he has been involved in a number of lawsuits and examinations as an economist at consulting firms and the Secretariat of the Japan Fair Trade Commission.

Yasushi Kudo is an economic analysis consultant with a decade of experience at the Japan Fair Trade Commission. As an economist, he planned and implemented economic analysis in joint research reports and reviews of business combinations in many complex cases.