Comparability Analysis in Transfer Pricing and Intellectual Property: A Quantitative Approach

Comparability analysis has been suggested by government regulatory agencies and is widely applied in transfer pricing and intellectual property litigation. The idea behind comparability analysis is to find similar uncontrolled transactions to benchmark the subject transaction and use the royalty rates, profit, or price measures indicated in the uncontrolled transactions to estimate the arm’s-length royalty rate or income of the subject transaction.

REVIEW OF APPLICABLE LAWS

In transfer pricing cases, courts often rely on comparable analyses to determine the arm’s-length transfer pricing of the intangible property being transferred between two related parties. In intellectual property cases, courts often rely on comparability analyses to determine the fair, reasonable, and non-discriminatory royalty rate of the subject patent in dispute, which can be applied to estimate damages from patent infringement.

Tax code Section 482 and the Organization for Economic Cooperation and Development (OECD) transfer pricing guidelines provide guidance to conduct comparability analyses and describe five comparability factors. This article first reviews the U.S. tax laws with which the OECD guidelines are consistent in terms of comparability analyses and comparability factors.

Treasury Regulation Section 1.482-1(d)(1) provides:

“Whether a controlled transaction produces an arm’s length result is generally evaluated by comparing the results of that transaction to results realized by uncontrolled taxpayers engaged in comparable transactions under comparable circumstances. For this purpose, the comparability of transactions and circumstances must be evaluated considering all factors that could affect prices or profits in arm’s length dealings (comparability factors). While a specific comparability factor may be of particular importance in applying a method, each method requires analysis of all of the factors that affect comparability under that method. Such factors include the following – (i) Functions; (ii) Contractual terms; (iii) Risks; (iv) Economic conditions; and (v) Property or services.”

In implementing comparability analyses for regulatory laws, practitioners often seek criteria to define the “most” comparable transactions, to select the relevant comparability factors, and to measure the degree of comparability. This article deeply examines the topic of comparability analysis.

CURRENT LIMITATIONS OF COMPARABILITY ANALYSES

Current litigation generally focuses on providing qualitative explanations to identify comparable uncontrolled transactions. It is often dependent on each expert to pick what they think is the most comparable one to benchmark. Opinions often differ, which causes each side to disagree with the opposing side’s selection and thus the arm’s-length estimation. In addition, it is rare that a perfect comparable exists that matches all dimensions of the comparability factors to the subject transaction, and it is often necessary to decide which factors are the important ones to match while making justifications for the other unmatched ones.

Even in the rare case in which an uncontrolled transaction perfectly matches the subject one in all observed dimensions, it may still not be the best benchmark. There may be unobserved factors and idiosyncratic errors, which may affect the results. For example, there could be hidden factors that affect the underlying market at a particular time or in a specific region.

CONTRIBUTION OF THIS ARTICLE

This article proposes a principled and fully data-driven econometric methodology for improving comparability analyses in transfer pricing and intellectual property cases. The article looks for refined conditional structures in comparability analyses, where the proposed regression model can include all targeted comparability factors and the potential unobserved factors in a single large model.

Specifically, the regression framework models the conditional mean of the dependent variable given the values of comparability factors. The regression coefficients provide a smooth interpolation of models of various dimensions and the linear combination of comparability factors leads to a weighted average of those factors, where the weights contain the information of the conditional structure that cannot be captured by the unweighted average. The importance of comparability factors can be interpreted in terms of the estimated regression coefficients. Statistical optimality can be obtained for deciding the factor-importance or selecting comparability factors to match.

OBSERVED/UNOBSERVED COMPARABILITY FACTORS

This section gets down to a more practical perspective of applying regression techniques in comparability analyses in the context of transfer pricing and intellectual property cases. Regression methodology is an acknowledged technique in regulatory and litigation practices. In the Intelligence Ltd. case in Japan, the court commented that regression analysis is a reasonable method based on scientific evidence and that its accuracy can be objectively verified. Gregory Sidak’s 2017 paper claims that “an expert economic witness can use well-established statistical techniques to estimate a regression model relating the passage of time and other explanatory factors to the observed royalties that third parties have willingly agreed to pay for the patent in suit in comparable licenses negotiated at arm’s length.”

While I do provide a recommendation on the selection of control variables, more weight should be given to the proposal of the regression design and the generalized five-step algorithm to conduct comparability analyses in a systematic way. Given that each case has a different nature, the selection of control variables could vary.

To design the regression, I recall the comparability factors suggested by the IRC and the OECD guidelines: (1) functions, (2) contractual terms, (3) risks, (4) economic conditions, and (5) characteristics of property or services. I shall discuss each factor category following the orders of the steps in the proposed algorithm.

Characteristics of Property or Service

Property or service, whether tangible or intangible, may have differing characteristics, which may lead to a difference in their values in the open market. These differences should be accounted for in comparability analyses. Characteristics that may be important to consider are:

• for tangible property: physical features, quality, reliability, availability, and the volume of supply;
• for service: nature and extent of the service; and
• for intangible property: form of the transaction (for example, licensing or sale), type and form of property, duration and degree of protection, and anticipated benefits from use of the property.

Industry classification systems are widely applied to serve as the first step in screening the products or services of similar characteristics. The standard industry classification systems often bring together a group of establishments primarily engaged in producing or handling the same product or group of products or in rendering the same service. For example, the North American Industry Classification System (NAICS) uses a production-oriented conceptual framework to group establishments into industries based on the activity in which they are primarily engaged. The Global Industry Classification Standard (GICS) is found by empirical study to group companies with similar stock returns together. Depending on the nature of the property or service, NAICS, GICS, Standard Industrial Classification, or Industry Classification Benchmark are common, standardized industry classification codes for the initial screening.

Function Terms

Function terms involves tracing the flows of subject transaction in different stages which may include:

• research and development;
• product design and engineering;
• manufacturing, production, process engineering, and design work;
• purchasing, materials management, and other procurement activities;
• manufacturing, production, or assembly work;
• transportation, warehousing and inventory;
• marketing, advertising, publicity, and distribution;
• market intelligence on technological developments; and
• intra-group services, for example managerial, legal, accounting and finance, credit and collection, training, and personnel management services.

Many current contract databases group transactions into different categories based on the functions performed in the agreements. For example, Intangible Spring contains a detailed classification of agreement types, including customer service agreement, marketing agreement, sales or distribution agreement, software agreement, advertising agreement, and others. ktMINE contains 10 agreement categories from distribution, asset purchases, and franchising to joint development, marketing intangibles, and others. Selecting agreements within the same category often further filters the transactions with similar functions.

The two-step screening emphasizes homogeneity among transactions. I now make the case for heterogeneity across transactions by considering other comparability factors: contractual terms, risks, and market conditions and unobserved factors which may affect the regions or the time in a systematic way. I propose a regression design with the selected transactions from the first two steps serving as the sample.

Contractual Terms

I examine contractual terms which are often discussed in the recent litigation work and can be quantified with either numerical values or dummy variables: (1) exclusivity; (2) the duration of agreement; (3) the covered territories; and (4) the rights to updates, revisions, or modifications. Those factors are in line with the examples given in Treas. Reg. Section 1.482-1(d)(3). I include those contractual terms in the regression analysis and keep in mind that the selection of contractual terms could vary based on the transaction nature and could potentially include less or more factors.

Risks

I adjust for risks by breaking down the risk factors into two categories: systematic and unsystematic. Systematic risk refers to the general level of risk associated with any business enterprise—the basic risk resulting from fluctuating economic, political, legal, regulatory and market conditions.

There are three channels to control for systematic risk. The risk associated with transaction functions and product nature can be controlled through the screening process in steps 1 and 2 above. The risk from fluctuating economic, political, legal, and regulatory conditions can be controlled by adding time and location fixed effects in the regression. The risk associated with companies’ investment relative to market performance (for example, S&P 500 index) can be controlled by beta factor (β), which is widely used in Capital Asset Pricing Model (CAPM) and litigation.

Unsystematic risk refers to the risk related to the specific business in which a company is engaged. By filtering out companies that perform similar functions and produce similar products or services, the business and product risk is controlled to some degree. Some other unsystematic risk could depend on the specific market conditions the company faces and can be absorbed in the economic condition control variables to be discussed in the next subsection.

Additionally, some unsystematic risk will depend on the company’s performance and can be referenced from its financial statements and balance sheets. Depending on the transaction nature, unsystematic risk measures could include the contribution margin ratio, operating leverage ratio, financial leverage ratio, combined leverage ratio, debt-to-capital ratio, debt-to-equity ratio, interest coverage ratio and others.

Economic Conditions

Along with contractual terms and risk factors, differences in underlying economic conditions can affect the bargaining power of the two parties involved in one transaction and hence reflect a different negotiated price. Classical monopoly and competition theories imply that when one party faces a less concentrated market, it will have lower market power and a weaker bargaining position; and in the extreme case when the party faces a perfectly competitive market, it may have zero market power, and becomes a price taker.

Economic conditions can be broadly classified into three categories: (1) global economic trends, (2) regional economic trends, and (3) market positions of the enterprise and surrounding economic conditions. While the first two can be captured in the time and region fixed effects respectively, the third category is more transaction–specific and can be included as explanatory variables in the regression.

While justifying economic factors is not common in transfer pricing and intellectual property cases, it is common in antitrust cases in which practitioners consider market factors to examine the post-merger price effects. The selection of relevant economic factors varies from case to case. Economists generally agree on the following market factors: (1) market share, (2) market concentration, (3) level of demand, (4) cost of production. Those factors are also in line with what has been suggested in Georgia-Pacific Corp. v. U.S. Plywood Corp.

REGRESSION DESIGN OF COMPARABILITY ANALYSES

I combine all the examined factors together with the time and location fixed effects to propose the following regression design for comparability analyses:

where y_j,r,t is the dependent variable of transaction j, at region r and time t.

The dependent variable of interest typically can be the royalty rate or negotiated price from a transaction agreement.

—Contract_j,l,r,t is the l-th contractual term for transaction j, at region r and time t. α_l is the weight on the l-th contractual term. L is the total number of contractual terms to consider.
—Risk_j,k,r,t is the k-th risk factor associated with transaction j, at region r and time t.
—β_k is the weight associated with the k-th risk factor. K is the total number of risk measures to include in the regression.
—Market_j,n,r,t is the n-th market condition measure for transaction j, at region r and time t.
—ϒ_n is the weight on the n-th market condition. N is the total number of market conditions to include.
—ȿ_t is the time fixed effect, which accounts for the unobserved variables that affect all the transactions systematically at time t.
—ȿ_r is the region fixed effect and it accounts for the unobserved variables which exert influence on every transaction in location j.
—ε_j,r,t is the idiosyncratic error term and is assumed to follow a normal distribution.

While the first two steps screen transactions of similar product characteristics and functions, the regression design accounts for variations across similar transactions in contractual terms, risks and underlying economic conditions. The regression evaluates the importance of each factor by estimating the coefficient and the standard errors. By applying the estimated regression results, the arm’s-length behavior of the subject transaction can be inferred by constructing confidence intervals.

ALGORITHM OF COMPARABILITY ANALYSES

This section recaps the discussion and proposes a generalized algorithm to systematically conduct comparability analyses with five steps:

Step 1: Screen transactions with similar products or services using industry classification systems.

Step 2: Screen transactions with similar functions based on agreement classification.

Step 3: Transform contractual terms, risk factors, and economic conditions into explanatory variables.

Step 4: Perform regression analysis with observed comparability factors, unobserved fixed effects and idiosyncratic errors.

Step 5: Apply regression coefficients and standard errors to construct confidence interval to estimate the arm’s-length behavior of the subject transaction.

The algorithm fills the gap between intellectual property and transfer pricing cases with antitrust analysis. Given each transaction can be different and the selection of controls can vary according to case nature, this article is not intended to select fixed controls for all transactions but to provide a general guideline for applying econometric techniques to comparability analyses.

CONCLUSION

This article suggests econometric techniques from a theoretical perspective. It also proposes a regression blueprint to systematically convert qualitative evidence into quantitative results, provides a guideline for its implementations with a generalized five-step algorithm to quantify the comparability analyses and demonstrates an outlook for further comparability analyses.

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

Author Information

Ziyi Qiu is currently a Visiting/Adjunct Assistant Professor of Economics at the University of Illinois Urbana Champaign.

Bloomberg Tax Insights articles are written by experienced practitioners, academics, and policy experts discussing developments and current issues in taxation. To contribute, please contact us at TaxInsights@bloombergindustry.com.