Addressing the Economic Downturn Under Existing Transfer Pricing Methods
The period from the early 1980s through mid-2008, during which many transfer pricing methods and their application matured, was one of generally strong profitability for multinational companies. Although there were downturns, most were shallow and short-lived and created no unusual transfer pricing challenges. Companies could develop transfer pricing policies with the knowledge that compliance issues would mainly involve selecting the best method for measuring results and identifying uncontrolled comparables used to benchmark those results. In an environment characterized by healthy profits and steady economic growth in most countries, there was little concern that increasing the profits from controlled transactions in one jurisdiction might create a loss in another jurisdiction. Business cycles continued and companies did incur losses, but downturns generally could be dealt with through existing transfer pricing principles.
The current economic downturn, in contrast, is bringing several basic transfer pricing issues to the fore. These include the ability of controlled taxpayers to modify existing advance pricing agreements to take account of adverse business conditions and to change their existing business structures, including long-standing allocations of risk. Controlled taxpayers and practitioners are also devoting greater attention to the transfer pricing analysis of events (or costs) such as plant closures, scale-down of operations, and employee separation. To some extent, these issues dovetail with a broader debate that is under way concerning taxpayers’ ability to modify their existing business arrangements in conformity with the arm’s-length principle.1
Traditional transfer pricing and business restructuring issues will likely continue to intersect. On a more practical and immediate level, however, taxpayers need to evaluate their transfer pricing policies and compliance strategies in light of deteriorating financial results. For example, how does a durable goods manufacturer price transactions with its distribution subsidiary, when pricing at a break-even cost-plus margin would still result in a loss for the distributor? If a transfer price less than the cost of production is necessary for a positive distribution return, will that pricing expose the distributor to a charge of dumping or unfair trade practices in the local market? Assuming that a taxpayer has not changed its basic U.S. business structure or its allocation of risk, and assuming that it wants to keep the same basic transfer pricing methods or analysis that it used in previous years, the best approach may be to apply multiple-year analysis and to perform a more thorough analysis of comparability with uncontrolled companies.
The U.S. transfer pricing regulations and corresponding OECD guidance recognize that multiple-year data may be useful in evaluating whether the results of a taxpayer’s controlled transactions are arm’s length.2 These rules acknowledge that, particularly when applying profit-based transfer pricing methods, it is necessary to take account of business cycles and special circumstances, such as the startup of a business, that cannot be fully addressed through comparability adjustments. Absent those rules, a profit-based method might indicate that a transfer pricing adjustment is called for simply because the tested party is at the bottom of its business cycle when the uncontrolled comparables are at the peak of theirs. That is, pricing of the controlled transactions might be arm’s length, but the taxpayer’s results could be outside the (single-year) arm’s-length range determined by reference to the comparables. A multiple-year analysis, which is primarily intended to address such cyclical effects, may also be relevant when a broader downturn affects the whole economy.
The Rules in Detail
Multiple-year averaging is commonly used when applying profit-based methods such as the comparable profit method.3 In fact, under the CPM, data for the uncontrolled comparables generally ‘‘should encompass at least the taxable year under review and the preceding two taxable years.’’4 The regulations also provide for use of a longer period if necessary to evaluate business cycles, life cycles, or the product under examination.5 The underlying rationale is that the period subject to review should be long enough to ‘‘reduce the effect of short-term variations that may be unrelated to transfer pricing.’’6 A multiple-year analysis evaluates the taxpayer’s results and the corresponding arm’slength ranges for the tax year under examination and for the applicable multiple-year period.7 If the taxpayer’s results for the single year fall outside the arm’slength range for that year, the taxpayer’s multiple-year results are compared with the multiple-year range.8 If the taxpayer’s multiple-year results fall outside the corresponding multiple-year range, the question is whether an allocation for the single year would move the taxpayer’s multiple-year result closer to the arm’s-length range for the multiple-year period or any point within the range.9 If so, an allocation for the single year is permissible. This analysis is illustrated in Example 4 in Treas. reg. section 1.482-1(f)(2)(iii)(E) and examples 2 and 3 in Treas. reg. section 1.482-5(e).
Opinions differ on how the multiple-year provisions should be applied. Taxpayers have historically viewed the regulations as providing two chances to pass — that is, if the taxpayer’s results fall within either the single-year range or the multiple-year range, no section 482 allocation is appropriate. However, some IRS agents evidently view the regulation as providing transfer prices two chances to fail. Under that interpretation, a section 482 allocation is appropriate unless the taxpayer’s results pass both the single-year and the multiple-year tests. Although not as critical, some confusion also surrounds the requirement that the allocation for the single year must move the taxpayer’s multiple-year average result closer to the multiple-year average range for the uncontrolled comparables.10
Apart from the regulations, Treasury and the IRS have issued no formal guidance on multiple-year averaging. FSA 199945011, however, does provide some insight into the thinking of the IRS National Office.11 That field service advice concluded that an allocation involving multiple-year data should place the taxpayer’s operating profits at the median (or the mean) of the comparable operating profits for the single tax year at issue — that is, the midpoint of the arm’s-length range for that year. It also concluded that a section 482 allocation is appropriate only if the taxpayer’s results are outside both arm’s-length ranges (single-and multiple-year) and the single-year adjustment moves the taxpayer’s multiple-year average results in the ‘‘right direction.’’ The field service advice, although not an official statement of IRS position, thus adopted the (generally) taxpayer-friendly ‘‘two chances to pass’’ interpretation.12
The diagram shows a hypothetical application of the multiple-year averaging concept. The taxpayer’s results are outside the arm’s-length range for year 3, and its average results are also outside the arm’s-length range on a multiple-year basis. The section 482 allocation indicated in the diagram is to the median of the arm’slength range for year 3, although as noted above, the size of the adjustment that should be made in those cases is unclear.
Application in Broader Economic Downturn
One scenario that is likely to occur is as follows. A foreign durable goods manufacturer operates a subsidiary in the United States. The results of the controlled transactions with the U.S. subsidiary are tested under the CPM using a rolling three-year average, and these results are within the arm’s-length range for each of the years 2004 through 2007. In 2008, although pricing of the controlled transactions is unchanged, the tested party experiences sharply reduced demand, leading to reduced revenues and operating profits. The uncontrolled comparables that were historically used to benchmark profitability may or may not have experienced similar declines in their operating results during the same period. Business cycles or factors specific to the tested party may result in the uncontrolled comparables experiencing the downturn earlier or later than the tested party, or to a different degree than the tested party.13
The question facing a taxpayer in this situation is how much taxable income to report for 2008. If the taxpayer’s multiple-year results for 2006 through 2008 are within the arm’s-length range for those years, FSA 199945011 suggests that no section 482 allocation is required for 2008. The taxpayer would then presumably file its tax return based on the transfer prices ‘‘actually charged’’ during the year.
More challenging issues arise if the taxpayer’s results for 2008 and for the multiple-year period fall outside the corresponding arm’s-length ranges. In that case, an adjustment is probably indicated, but the amount is unclear. The informal guidance in FSA 199945011 suggests that the taxpayer should adjust its results for 2008 to the median (or the mean) of the arm’s-length range for that year, although a taxpayer in this situation might often wish to make a smaller adjustment to some other point within the (single-year) arm’s-length range.
By their terms, the multiple-year averaging provisions appear to deal only with section 482 allocations made by the IRS — allocations that by definition take place in an examination setting.14 They do not explicitly address self-initiated section 482 allocations by the taxpayer, which, at least in the case of a reduction of U.S. income, need to be made on a timely filed U.S. tax return for the year in question. If a taxpayer makes a self-initiated section 482 allocation in ‘‘real time’’ (before the filing of the tax return), it appears that the taxpayer can adjust its results to any point in the single-year range, assuming the other conditions for the allocation are satisfied. But the regulations are not clear on this point.
Multiple-year averaging can reduce the harshness of the results produced by a ‘‘one-sided’’ transfer pricing method such as the CPM. But it does not address what some might see as a more fundamental issue: How would similarly situated uncontrolled parties deal with a major change in economic conditions, such as a prolonged downturn? Occasionally taxpayers have used evidence concerning the response of uncontrolled parties to events such as currency devaluations or commodity price spikes in an effort to determine whether and how uncontrolled parties modify their prices in response to economic shocks. Applying such a two-sided analysis or otherwise trying to simulate the type of bargaining that takes place between uncontrolled parties presents substantial challenges. Also, such approaches may be difficult to reconcile with a one-sided transfer pricing method such as the CPM.
In any event, the multiple-year averaging provisions are likely to play an important role as controlled taxpayers apply the CPM and other profit-based methods to years affected by the economic downturn. The analytical and procedural questions presented are substantial. In the scenario described above, if the single-year 2008 results move the taxpayer’s three-year average outside the multiple-year range, could a four-or five-year average be used instead? What if the taxpayer historically analyzed prices on a single-year basis? Can the taxpayer now adopt a multiple-year analysis for compliance purposes?
Even more complex issues are likely to surface as these cases go through the IRS examination process. To name one, assuming the common situation in which the tested party’s sales and profits are both depressed, the multiple-year results under the CPM may differ substantially depending on whether the profit-level indicator used to determine comparable operating profits is based on sales or assets.15 The multiple-year averaging rules may produce unanticipated results, particularly when the effect of the economic downturn on the tested party differs from its effect on the uncontrolled comparables. Treasury and the IRS should consider updating the rules in this area, or at least providing more detailed illustrations of how the existing rules would apply in common settings.
Alternative Approach: Pooling of Results
Another approach that is sometimes applied to multiple-year data involves pooling the results derived from the uncontrolled comparables. This approach treats each observation (for example, the comparable operating profits derived from the operating margin of an uncontrolled company for a particular year) as a discrete data point, rather than averaging or weight-averaging the available observations for each comparable over the multiple-year period, which is the standard approach. Pooling becomes relevant when a complete data set is unavailable — for example, when a large segment of the industry ceases operations or several companies otherwise become unsuitable for use as comparables.16 The IRS doesn’t favor pooling. The APA program training materials observe that averaging is the preferred approach to multiple-year data, but note that pooling has sometimes been used when shown to produce more reliable results than averaging.17
Pooling can produce a very different arm’s-length range than averaging, and this may benefit either the taxpayer or the IRS. Pooling can also have undesirable effects, such as giving undue weight to individual observations of comparable operating profits under a sales-based profit-level indicator when the tested party’s sales show large variations across the multiple-year period. Because pooling is relevant when the time series of data is incomplete, it may come into play when comparable companies are excluded because of persistent operating losses, bankruptcy, or termination of operations, or when taxpayers examine a multiple-year period longer than the standard three-year period specified in the CPM regulations. As developed further below, all of these scenarios are likely to be encountered more frequently in coming years, as the full impact of the economic downturn is felt.
The section 482 regulations apply to both the IRS and taxpayers. A taxpayer cannot compel the IRS to apply section 482, but the regulations permit a taxpayer to adjust its U.S. taxable income if necessary to reach an arm’s-length result (reductions must be made on a timely filed U.S. income tax return, but increases can be made regarding any open tax year).18 This allows taxpayers to take into account unanticipated developments or to make other true-ups necessary to bring the tested party’s results within the arm’s-length range.19
Recent declines in revenue and profits in many sectors of the U.S. economy suggest that taxpayer-initiated adjustments may become more prevalent. For example, assume that a controlled U.S. distributor of heavy construction equipment was on pace in mid-2008 to meet its target operating margin of 5 percent for the full year. As economic conditions deteriorated in August 2008, several of the company’s customers cancelled orders as they sold down existing inventories. Price reductions and extended payment terms were adopted to maintain sales volume, but revenue for the balance of 2008 continued to decline, resulting in a fourth-quarter operating margin of -15 percent. To meet its target operating margin of 5 percent for 2008, the company would need to report more U.S. taxable income than would result under the transfer prices charged.
Conversely, the economic downturn might suggest the need for a taxpayer-initiated adjustment that reduces U.S. taxable income. For example, assume a U.S. subsidiary provides a budgeted amount of services — say $10 million per year — to its foreign parent. Historically, the return to the subsidiary under the CPM20 was set at 9 percent, by reference to the profits earned by uncontrolled providers of similar services. Updating of the service provider comparables in 2008 indicates that because of deteriorating economic conditions, the top of the interquartile range is now 6 percent. In that situation, the taxpayer would likely consider reducing the return to the U.S. service provider in 2008 to 6 percent at the most.
Ironically, the multiple-year averaging provisions may limit the taxpayer’s ability to make a self-initiated adjustment of its U.S. taxable income — in either direction.21 Under the regulations, a taxpayer may apply section 482 only if necessary to reflect an arm’s-length result.22 Based on the prevailing view that multiple-year averaging gives transfer prices two chances to pass, a self-initiated adjustment might be precluded by reference to the taxpayer’s multiple-year average results,23 although that same adjustment would be appropriate if the results for the single year were examined in isolation. In preparing income tax returns for the first year in which the full impact of the economic downturn is felt, taxpayers should keep in mind that the multiple-year averaging provisions might limit their ability to use the self-initiated adjustment mechanism.
The discussion above focused on several issues that are likely to persist during the economic downturn. Even after economic recovery begins, the downturn may have lingering effects in the form of reduced availability of uncontrolled comparables. Taxpayers and tax administrations have recently come to rely more on the CPM and other profit-based methods for determining an arm’slength result.24 Application of the CPM and similar methods is based on the ability to identify a sufficient number of uncontrolled companies that are comparable to the tested party and that can be used to construct a reliable range of comparable operating profits.
In most cases, the tested party under a CPM performs routine functions characterized by relatively low risk, and consequently the tested party is expected to earn positive returns. Correctly or incorrectly, uncontrolled companies that have persistent operating losses or that show other signs of distress are often eliminated from the set of potential comparables, almost as a threshold matter. An uncontrolled company with operating losses in one year or in two successive years might be kept as a comparable, but a company with operating losses over an extended period, or whose status as a going concern is in doubt, is generally rejected. This approach reflects an instinctive reaction — one that is often borne out by a more detailed analysis of the facts — that an uncontrolled company with persistent losses has a risk profile different from that of the tested party (which is generally low-risk). Companies operating under chapter 11 or chapter 7 are also excluded, for different reasons. The prevailing view is that a bankrupt company operates under conditions (for example, ongoing judicial supervision or maximization of returns to secured creditors rather than shareholders) that distinguish it from the tested party, which is deemed to operate under traditional free-market conditions with the goal of maximizing profits.25
Over the past year or so, companies across a broad swath of the U.S. economy have been affected by the economic downturn. This may call into question the ability to apply the CPM and similar profit-based methods, if the screening techniques described in the previous paragraph are applied in the traditional manner. And when economic distress is particularly acute in an industry segment, it may be impossible to identify a suitable set of comparables in that sector, again assuming that such screening techniques are applied in the usual way.
In some settings, such as a contested IRS examination, an APA negotiation, or a mutual agreement proceeding, screening of comparables may yield to a more in-depth analysis of comparability.26 There it is common to review annual reports and other public financial data to determine why a company had adverse operating results. A detailed analysis of this type might show that the level of comparability between the uncontrolled company and the tested party is acceptable.
When an uncontrolled company generates operating losses in successive years, the underlying reasons for those losses might, on closer analysis, be found to apply equally to the controlled taxpayer. Or it may be possible to use a company with adverse operating results as a comparable, provided that reliable adjustments can be made to account for differences between that company and the tested party. Unfortunately, pragmatic considerations tend to win out over the results of a comparability analysis. For example, it is not unheard of for loss companies to be excluded because they move the arm’s-length range into very low (or negative) operating profits.
An in-depth comparability analysis should indicate whether an uncontrolled company in bankruptcy is operating under conditions that are dissimilar to those faced by the controlled party.27 If material differences in contractual terms or other conditions are identified, it may be possible to make adjustments that reliably account for those differences.28 Analyzing specific companies at this level of detail is costly, at least in comparison with screening techniques that simply eliminate all bankrupt companies. In this economic environment, however, the blanket exclusion of all uncontrolled companies operating in bankruptcy may reduce the overall reliability of a CPM analysis.
Most public U.S. companies will survive the economic downturn, and most will earn operating profits (though perhaps diminished) despite more challenging business conditions. Some companies, however, will have sharply reduced profits or losses, and some of those may eventually declare bankruptcy or cease operations altogether. Under these conditions, the composition of the database of uncontrolled comparables, particularly the decision whether to exclude companies on the basis of negative operating results or bankruptcy, takes on greater importance.
Historically, the IRS and many other tax administrations have assumed that a taxpayer that has a guaranteed flow of transactions from another controlled party should earn non-de-minimis operating profits, without regard to adverse operating conditions that may have arisen in the taxpayer’s industry. (There are exceptions — for example, taxpayers in high-risk sectors, start-up/ shutdown scenarios, or market penetration.) Even if that assumption is valid in the abstract, it will nonetheless come under scrutiny in this environment, particularly in sectors especially hard hit by the economic downturn. The operating assumption that all parties that do business on an arm’s-length basis must earn profits may be suspect when most or all of the uncontrolled companies operating in the industry sector are generating losses.
A More Difficult Case
In some cases, the economic downturn may exacerbate a process of contraction or consolidation already under way in an industry sector. In the extreme case, most or all uncontrolled companies in a sector might exit the business, be acquired, or no longer be viewed as appropriate comparables (for example, because of going concern issues, operating losses, or bankruptcy).
Consider the case of a vertically integrated manufacturer that historically tested discrete elements of its operations under the CPM. It is common for such a manufacturer to use CPM comparables to benchmark one or more low-value production activities — for instance, tabletting by a pharmaceutical company or metal-stamping by a manufacturer of home appliances. The economic downturn may mean the end of the line for uncontrolled companies that have historically performed such routine functions on a stand-alone basis. In these situations, it may be necessary to reapply the best method rule, taking into account the changed conditions in the industry.29 Ultimately, a more pragmatic view of comparability and reliability may also be necessary. For example, one might consider using as comparables uncontrolled companies that are generating persistent losses or are in bankruptcy. Or one may need to use segment data (if available) for an uncontrolled acquirer of the former comparable or to identify uncontrolled companies that have less direct comparability to the relevant business activities of the tested party.
For their part, the IRS and other tax administrations should use flexibility in evaluating whether results are arm’s length under these circumstances. Tax administrators should avoid second-guessing a controlled taxpayer’s decision to continue performing specific low-value activities that constitute elements of an integrated production process. A controlled party’s decision to perform a function should not be questioned because it has become more difficult to evaluate the arm’s-length return to that function, due to changes in industry composition.
Taxpayers will continue to face substantial challenges as they report U.S. taxable income under the arm’s-length standard. Given the increased reliance in recent years on the CPM and similar profit-based methods, it may be necessary to reconsider some of the fundamental principles that have informed application of those methods. These methods should continue to provide reliable results in most cases, if taxpayers, practitioners, and tax administrators apply them in a flexible and pragmatic manner, taking into account the changed business and economic environment.
1. See generally OECD, ‘‘Public Discussion Draft on Transfer Pricing Aspects of Business Restructurings’’ (Sept. 19, 2008),
available at http://www.oecd.org/document/7/0,3343,en_2649_33753_41328775_1_1_1_1,00.html. The draft was released for comment in 2008, but the project began in 2005.
2. Treas. reg. section 1.482-1(f)(2)(iii); OECD, Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations, paras. 1.49-1.51.
3. See also discussion of the transactional net margin method in the OECD transfer pricing guidelines, paras. 3.26 and following.
4. Treas. reg. section 1.482-5(b)(4).
5. See Treas. reg. section 1.482-1(f)(2)(iii)(B).
6. Treas. reg. section 1.482-1(f)(2)(iii)(D).
7. Treas. reg. section 1.482-1(f)(2)(iii)(B).
8. Treas. reg. section 1.482-1(f)(2)(iii)(D).
10. Some interpret ‘‘only to the extent that’’ to mean the singleyear allocation must bring the taxpayer to the edge of the interquartile range for the multiple-year period. In contrast, FSA 199945011 (infra) suggests that the allocation for the single year must have the ‘‘correct sign’’: It cannot move the taxpayer’s multiple-year results (results that by definition are outside the interquartile range for the multiple-year period) further away from the multiple-year range. See also Treas. reg. section 1.482- 1(f)(2)(iii)(D).
11.The main issue in the field service advice was the appropriate analysis under the CPM of antidumping duty cash deposits that a U.S. subsidiary had posted under 19 U.S.C. section 1673(e) — deposits that the U.S. government refunded to the subsidiary in a subsequent tax year.
12. As developed further below, this interpretation may cause difficulties when the taxpayer seeks to make a ‘‘self-initiated’’ adjustment under Treas. reg. section 1.482-1(a)(3).
13. As a practical matter, taxpayers must cope with the fact that financial data for the uncontrolled comparables may lag the data for the tested party by at least six months. In developing a compliance strategy, taxpayers must coordinate data availability with the due dates for income tax returns and contemporaneous transfer pricing documentation — due dates that often differ from one jurisdiction to another.
14. This can be seen, for example, in the reg’s cross-reference to section 1.482-1(g)(2)(iii), which defines a prior-year section 482 allocation that is ‘‘finally determined.’’ See Treas. reg. section 1.482-1(f)(2)(iii)(D) (final sentence). None of the events that constitute a final determination correspond exactly to a taxpayer initiated adjustment, although ‘‘payment of the deficiency’’ (Treas. reg. section 1.482-1(g)(2)(iii)(C)) would arguably permit a prior-year self-initiated adjustment to be taken into account in calculating the taxpayer’s multiple-year average results in subsequent tax years.
15. Under a sales-based profit-level indicator, comparable operating profits are weighted to the tested party’s sales, which are likely to be depressed in the first year of the economic downturn, as compared with previous years. In contrast, the level of the tested party’s operating assets or capital employed is less likely to decrease substantially from one year to the next, even during an economic downturn.
16. If the time series of data is complete, pooling and averaging should produce identical (or nearly identical) results.
17. APA Study Guide at 44. The IRS APA training materials are available at http://www.irs.gov/businesses/corporations/
article/0,,id=96186,00.html. The APA office cautions that the training materials should not be relied on or otherwise cited as precedent by taxpayers.
18. Treas. reg. section 1.482-1(a)(3).
19. A taxpayer-initiated adjustment that increases U.S. taxable income may prevent a section 482 allocation by the IRS and may also eliminate associated penalty exposure. But regardless of the direction of the self-initiated adjustment, a U.S. taxpayer seeking a corresponding adjustment in a foreign country generally must request assistance from the U.S. competent authority. The considerations relevant to these adjustments are discussed in greater detail in H. David Rosenbloom, ‘‘Self-Initiated Transfer Pricing Adjustments,’’ Tax Notes Int’l, June 4, 2007, p. 1019, Doc 2007-12382, or 2007 WTD 111-6.
20. Assume that the conditions for use of the CPM for services method are satisfied. See temp. Treas. reg. section 1.482-9T(f). Also assume that the services are not eligible for the services cost method in temp. Treas. reg. section 1.482-9T(b).
21.The section 482 allocation at issue in FSA 199945011 was initiated by the taxpayer, not the IRS.
22.Treas. reg. section 1.482-1(a)(3). See also IRS general legal advice memorandum, AM-2007-007 (Mar. 15, 2007) (addressing the commensurate with income standard).
23. That is, the taxpayer’s multiple-year average results might be within the range of multiple-year average results for the comparables, or the potential allocation for the single year might move the taxpayer’s multiple-year results away from the range of multiple-year average results for the comparables.
24. In the case of a U.S. subsidiary of a foreign parent, the subsidiary’s operating profit generally is tested directly against U.S. comparable companies with similar functions and risks. In the case of a U.S. parent, the results of the foreign subsidiary often are tested against suitable comparables, and if those results are determined to be arm’s length, the residual profit earned by the U.S. parent is deemed to be arm’s length.
25. Other conditions that may affect comparability include financial assistance or loan guarantees from the federal government, both of which have become more common in some sectors. See, e.g., ‘‘U.S. Offers $5 Billion to Car Suppliers,’’ The Wall Street Journal, Mar. 20, 2009, available at http://www.wsj.com.
26. For a general discussion of loss comparables that predates the economic downturn, see OECD, Comparability: Public Invitation to Comment on a Series of Draft Issues Notes, at 72-74 (May 2006). The document is on the OECD Web site, available at http://www.oecd.org/document/12/0,3343,en_2649_37989753_36651660_1_1_1_1,00.html.
27. Paradoxically, a bankrupt company may have contractual terms with third parties that are more favorable than the terms the controlled party has in its contracts with third parties. For example, a bankruptcy court may require the company’s customers to accept passthrough of increases in raw material costs, or it may void a collective bargaining agreement that specifies above-market wages and benefits to the company’s employees.
28. In the case of a bankrupt company that has a more favorable labor contract, the comparable’s labor costs might be adjusted upward to reflect the prevailing wage rates in the applicable industrywide collective bargaining agreement.
29. The OECD transfer pricing guidelines (para. 1.12) observe that in some cases, information necessary to apply the arm’slength principle may be difficult to obtain or may not exist. In this context, the guidelines also note that transfer pricing is not an exact science, but calls for ‘‘the exercise of judgment on the part of both the tax administration and the taxpayer.’’ Id.