INSIGHT: An Unbiased Approach to PLI Selection in Transfer Pricing Studies

The Comparable Profits Method (CPM) is the most widely used method in transfer pricing. Aggregated data from the U.S. Treasury’s Advance Pricing Agreement (APA) Program shows that the CPM was used in 89% of tangible and intangible property APAs and 76% of service APAs to which the U.S. was a party in 2016.

One critical decision in the implementation of the CPM is the selection of the profit level indicator (PLI). However, in place of careful analysis, PLI selection is often performed using simple rules, or heuristics, which are generally correct overall, but may be incorrectly relied upon in certain situations. Heuristics are often used to select the PLI based on the functions of the tested party. For example, the operating margin (OM) PLI is often selected for testing the results of distributors in related party transactions without further examination of the implied arm’s-length range of profit from the use of other PLIs.

Within this article we empirically examined the use of heuristics in PLI selection. To evaluate the validity of some common heuristics, we created a framework for measuring the quality of each PLI for a given set of comparable companies and tested party. Comparing the empirical reliability of PLIs versus their heuristic merits then provided a framework for an unbiased approach to PLI selection.

ANALYSIS FRAMEWORK

We constructed our method for evaluating and comparing PLI accuracy based on the principle that the most reliable PLI is the one that results in the narrowest range of arm’s-length prices, measured indirectly as the range of arm’s-length profit under the CPM.

This principle is consistent with the “best method rule” in Treasury Regulation Section 1.482-1(c)(1) which calls for the selection of a method, and by extension a PLI in cases where the CPM is selected as the best method, that “provides the most reliable measure of an arm’s-length result.” The regulations reiterate that the selected PLI should be “likely to produce a reliable measure of the income that the tested party would have earned had it dealt with controlled taxpayers at arm’s-length.” (Treas. Reg. Section 1.482-5(b)(4): “Profit level indicators.”) Tax authorities are often worried about tax payers incorrectly concluding that a result was arm’s-length, when in fact it was not (a “type II error”). The best PLI would be one that minimizes the frequency of such a type II error.

The concept of constructing methods to analyze PLI selection is not an entirely novel idea. For example, Chapter 9 of “Practical Guide to U.S. Transfer Pricing, 3rd Edition” introduces the “Newlon Ratio” as a tool for examining the selection of PLI in CPM analyses. This method specifically calculates the implied variance for the interquartile ranges of PLIs and asserts that the more reliable PLI is the one which results in the least variance. (The Newlon Ratio is calculated as (75^th Percentile – 25^th Percentile) / 50^th Percentile.) However, the Newlon Ratio’s shortfall lies in the fact that it only examines percentiles and never takes into account the denominator (basis) of each individual PLI, nor their potential for measurement issues.

The purpose of the PLI selection process is to select a PLI that minimizes the flaws of comparability under a CPM analysis. The PLI selected should do this notwithstanding the issues that practitioners face in PLI selection, such as asset measurement or asset intensity considerations. As such, we proposed a method that determines the most reliable PLI for a specific scenario through measurement of absolute profit. To our knowledge, our approach of determining a best PLI through an absolute profit perspective has not been formally studied or written about by transfer pricing practitioners before.

All else equal, a smaller range of implied arm’s-length profit is likely to be more reliable than a wider range. This is partly because a narrower range of implied arm’s-length profits would reduce variance in the range of arm’s-length results. A result with less variance would produce a more reliable measure of the income that the tested party would have earned in a controlled transaction than a result with higher variance.

Assuming that all of the profit ranges are equally likely to be “correct,” the PLI producing the narrowest range of implied arm’s-length profit would have the lowest probability of committing a type II error and would indicate the presence of a preferred PLI (referred throughout this paper as the “more reliable PLI”). We believe our method to be superior to the Newlon Ratio because our framework analyzed each PLI from an objective, common basis (absolute profit), rather comparing variance between PLI without taking into account their unique bases.

In application, our method took a selection of comparable companies and computed the interquartile ranges for these companies for two common PLIs; Return on Assets (ROA) and OM. For simplicity, we selected ROA and OM. Net cost plus markup (NCPM) was not selected since NCPM is a mathematical derivative of OM. Other PLIs were considered but ultimately not selected for discussion within our study. For example, many publicly traded companies have negative return on capital employed (ROCE) or implement different accounting procedures in cost classification, which makes the interpretation for the ROCE, BR, and Gross Margin (GM) PLIs challenging.

We evaluated interquartile ranges for PLI calculations, rather than full ranges because (i) in any given case, our resulting set of comparable companies was imperfect and using the interquartile range reduced the impact of outliers, and (ii) such valid statistical methods are suggested under the regulations and often employed to adjust ranges to increase reliability of PLI results. (Treas. Reg. Section 1.482-1(e)(2)(iii)(B): “Adjustment of range to increase reliability.”)

Subtracting the 25th quartile PLI result from the 75th quartile PLI result for each PLI created a measure of PLI spread (PLI Differential). We then constructed two methods that provided an indication of a more reliable PLI by studying the resulting range of profit.

Method One: Narrowest Aggregate Profit Range From a Set of Comparable Companies

Under this method, we applied the PLI Differential to the median basis, or denominator, from the set of comparable companies for each of the four PLIs selected for evaluation. This resulted in a single measure of the range of profit that each PLI would produce under a given set of comparable companies at the median basis level from the set. The table below shows an example of the calculation for Method One using an interquartile range of 5-10% for each PLI and assuming a basis of 100 and 200, respectively:

Table 1: Method One Calculation Example

In the example calculation above, PLI A produced the narrowest range of profit (E), measured in absolute terms (i.e., dollars) for this comparable set. This is interpreted as PLI A being the more reliable PLI for this under Method One in this example.

Method Two: Narrowest Individual Profit Range for Each Comparable Company in a Set

Next, we expanded upon Method One by taking the PLI Differential for each PLI and applying it to the specific basis (denominator) for each of the comparable companies in the set, rather than the median basis for the aggregate set of companies. We then counted the number of companies in the set for which each PLI was the most reliable (produced the narrowest range of arm’s-length profits). The table below shows an example of the calculation for Method Two:

Table 2: Method Two Calculation Example

We used the results of Method Two to corroborate Method One’s indication of the more reliable PLI and test PLI selection at a company level versus at an aggregate level (i.e., Method One). In general, we expected the more reliable PLI in Method One to correspond with the more reliable PLI from Method Two. In the hypothetical example in Table 2 above, PLI B resulted in the narrowest range of profits for both companies. The more reliable PLI under Method One is defined as the one that resulted in the narrowest range of profit for a given set of companies. The more reliable PLI under Method One is defined as the one that resulted in the narrowest range of profit for a given set of companies.

Based on this framework for PLI evaluation, we applied our method for determining the most reliable PLI to empirical data. Using a model that examines all active companies within Compustat’s North American database over the most recent three-year weighted average period of data available at the time of this analysis (2013 to 2015), we empirically examined the use of heuristics in PLI selection, specifically, the use of heuristics to select the PLI based on the functions of the tested party. All active companies in the Compustat North American database between 2013 and 2015 were considered. We applied data sufficiency screens to ensure that all companies had three years of revenue, cost of goods sold (COGS), operating expenditures, and assets. This resulted in a set of 5,772 companies. Additionally, we further refined our data set to exclude companies that did not report all of the reported three-year WAVG PLIs (e.g., reported COGS as null, reported revenue as null, etc.)

RESULTS

According to the Treasury’s APA report, OM was the most-selected PLI in APAs in which the U.S. was a party in 2016 for both tangible and intangible property transactions. Our method for examining the use of heuristics in PLI selection suggested that in certain instances, further attention may be warranted in PLI selection rather than applying conventional wisdom in PLI selection based on function.

Upon initial examination of all Compustat North American companies across all Standard Industrial Classification (SIC) codes (4970 companies) within our dataset, results were mixed for which PLI was the more reliable .

Table 3: PLI Results for All SIC Codes

However, results observed from applying further comparability criteria to our set of companies yielded more determinable results. First, we narrowed the comparable set using SIC codes to examine specific industries. The table below shows the results of our method for different industries when screening by SIC code. SIC code was assumed to be an objective proxy for screens on industry as a whole.

Table 4: PLI Results by SIC (Industry)

Although we recognize that the Treasury Regulations in Section 1.482 state that in measuring arm’s-length results “unadjusted industry average returns themselves cannot establish arm’s-length results” (Tax code Section 1.482-1(d)(2): “Standard of comparability.”), our SIC code-based results were nonetheless compelling based on the observation that ROA consistently the more reliable PLI under Method One and Method Two, with the exception of the results observed in the Manufacturing industry.

We observed the ROA as the more reliable PLI in the retail trade, wholesale distribution, and services industry both under Method One and Method Two. These results are compelling given the preference of transfer pricing practitioners to select the PLI based on function. For example, distributors are often benchmarked with OM based on the heuristic of value creation being best measured by their ability to sell. And hence, revenue as the denominator of the PLI.

To apply our analysis one step further, we compiled two sets of companies generally accepted within the transfer pricing practice as standard service comparables; one set for routine manufacturing and one for routine distribution. The companies in both sets were generally accepted as having limited valuable intangible assets, bearing limited non-routine risks, and performing routine manufacturing or distribution functions, respectively. The companies in these sets are commonly used as comparables in CPM analyses. The results of our method as applied to these two sets are shown in the table below.

Table 5: PLI Results for Standard Sets

Results on our routine comparable sets further reinforced our observations from the results in Table 4 based on general SIC industry screens. Notably, after screening for standard service providers with limited valuable intangible assets, limited non-routine risks, and routine functions, OM was the more reliable PLI for the manufacturing industry and ROA was the more reliable PLI for the distribution industry.

CONCLUSION

Many transfer pricing practitioners rely on “conventional wisdom” and heuristics in their selection of the PLI in transfer pricing analyses. Our method provided an unbiased approach in empirically examining the use of heuristics in PLI selection. Our analysis showed that PLI selection heuristics may lead to selecting the wrong PLI in some cases. Notably, our method implied that for both broadly defined industries (i.e., by SIC code) and routine comparable sets, the OM under both of our methods was the more reliable PLI for the manufacturing industry, while the ROA was the more reliable PLI for the distribution industry. Additionally, despite the OM being the most-selected PLI in U.S. transfer pricing studies, our method made a compelling argument for the ROA as the more reliable PLI in three of the four broadly defined industries (i.e., by SIC code) that are commonly the subject of transfer pricing studies.

Our article only examined heuristics in PLI selection on a high level. Based on our results, we conclude that there is compelling evidence to further examine and refine an unbiased approach to PLI selection in transfer pricing studies. While we want to be careful to avoid replacing current heuristics with new ones, additional research is warranted on the difficulties in PLI selection, since there are important issues that practitioners face which have not specifically been covered by our paper. Examples of topics for further analysis include asset measurement or asset intensity considerations in PLI selection.

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

Andrew Hughes is an economist specializing in transfer pricing, valuation, and risk management in Brussels, Belgium. The author would like to thank Tylor Klein and Bob Clair for their advice, valuable comments and suggestions during the construction of this analysis.